diff options
Diffstat (limited to 'vendor/github.com/vishvananda')
42 files changed, 2657 insertions, 768 deletions
diff --git a/vendor/github.com/vishvananda/netlink/README.md b/vendor/github.com/vishvananda/netlink/README.md index 0b61be217..a88e2f418 100644 --- a/vendor/github.com/vishvananda/netlink/README.md +++ b/vendor/github.com/vishvananda/netlink/README.md @@ -89,3 +89,4 @@ There are also a few pieces of low level netlink functionality that still need to be implemented. Routing rules are not in place and some of the more advanced link types. Hopefully there is decent structure and testing in place to make these fairly straightforward to add. + diff --git a/vendor/github.com/vishvananda/netlink/addr_linux.go b/vendor/github.com/vishvananda/netlink/addr_linux.go index f33242a7c..d59c3281e 100644 --- a/vendor/github.com/vishvananda/netlink/addr_linux.go +++ b/vendor/github.com/vishvananda/netlink/addr_linux.go @@ -2,13 +2,13 @@ package netlink import ( "fmt" - "log" "net" "strings" "syscall" "github.com/vishvananda/netlink/nl" "github.com/vishvananda/netns" + "golang.org/x/sys/unix" ) // IFA_FLAGS is a u32 attribute. @@ -23,7 +23,7 @@ func AddrAdd(link Link, addr *Addr) error { // AddrAdd will add an IP address to a link device. // Equivalent to: `ip addr add $addr dev $link` func (h *Handle) AddrAdd(link Link, addr *Addr) error { - req := h.newNetlinkRequest(syscall.RTM_NEWADDR, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_NEWADDR, unix.NLM_F_CREATE|unix.NLM_F_EXCL|unix.NLM_F_ACK) return h.addrHandle(link, addr, req) } @@ -36,7 +36,7 @@ func AddrReplace(link Link, addr *Addr) error { // AddrReplace will replace (or, if not present, add) an IP address on a link device. // Equivalent to: `ip addr replace $addr dev $link` func (h *Handle) AddrReplace(link Link, addr *Addr) error { - req := h.newNetlinkRequest(syscall.RTM_NEWADDR, syscall.NLM_F_CREATE|syscall.NLM_F_REPLACE|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_NEWADDR, unix.NLM_F_CREATE|unix.NLM_F_REPLACE|unix.NLM_F_ACK) return h.addrHandle(link, addr, req) } @@ -49,7 +49,7 @@ func AddrDel(link Link, addr *Addr) error { // AddrDel will delete an IP address from a link device. // Equivalent to: `ip addr del $addr dev $link` func (h *Handle) AddrDel(link Link, addr *Addr) error { - req := h.newNetlinkRequest(syscall.RTM_DELADDR, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_DELADDR, unix.NLM_F_ACK) return h.addrHandle(link, addr, req) } @@ -65,7 +65,7 @@ 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, _ := addr.Mask.Size() + prefixlen, masklen := addr.Mask.Size() msg.Prefixlen = uint8(prefixlen) req.AddData(msg) @@ -76,7 +76,7 @@ func (h *Handle) addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error localAddrData = addr.IP.To16() } - localData := nl.NewRtAttr(syscall.IFA_LOCAL, localAddrData) + localData := nl.NewRtAttr(unix.IFA_LOCAL, localAddrData) req.AddData(localData) var peerAddrData []byte if addr.Peer != nil { @@ -89,7 +89,7 @@ func (h *Handle) addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error peerAddrData = localAddrData } - addressData := nl.NewRtAttr(syscall.IFA_ADDRESS, peerAddrData) + addressData := nl.NewRtAttr(unix.IFA_ADDRESS, peerAddrData) req.AddData(addressData) if addr.Flags != 0 { @@ -103,16 +103,34 @@ func (h *Handle) addrHandle(link Link, addr *Addr, req *nl.NetlinkRequest) error } } - if addr.Broadcast != nil { - req.AddData(nl.NewRtAttr(syscall.IFA_BROADCAST, addr.Broadcast)) + if family == FAMILY_V4 { + if addr.Broadcast == nil { + calcBroadcast := make(net.IP, masklen/8) + for i := range localAddrData { + calcBroadcast[i] = localAddrData[i] | ^addr.Mask[i] + } + addr.Broadcast = calcBroadcast + } + req.AddData(nl.NewRtAttr(unix.IFA_BROADCAST, addr.Broadcast)) + + if addr.Label != "" { + labelData := nl.NewRtAttr(unix.IFA_LABEL, nl.ZeroTerminated(addr.Label)) + req.AddData(labelData) + } } - if addr.Label != "" { - labelData := nl.NewRtAttr(syscall.IFA_LABEL, nl.ZeroTerminated(addr.Label)) - req.AddData(labelData) + // 0 is the default value for these attributes. However, 0 means "expired", while the least-surprising default + // value should be "forever". To compensate for that, only add the attributes if at least one of the values is + // non-zero, which means the caller has explicitly set them + if addr.ValidLft > 0 || addr.PreferedLft > 0 { + cachedata := nl.IfaCacheInfo{ + IfaValid: uint32(addr.ValidLft), + IfaPrefered: uint32(addr.PreferedLft), + } + req.AddData(nl.NewRtAttr(unix.IFA_CACHEINFO, cachedata.Serialize())) } - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -127,11 +145,11 @@ func AddrList(link Link, family int) ([]Addr, error) { // Equivalent to: `ip addr show`. // The list can be filtered by link and ip family. func (h *Handle) AddrList(link Link, family int) ([]Addr, error) { - req := h.newNetlinkRequest(syscall.RTM_GETADDR, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(unix.RTM_GETADDR, unix.NLM_F_DUMP) msg := nl.NewIfInfomsg(family) req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWADDR) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWADDR) if err != nil { return nil, err } @@ -183,21 +201,21 @@ func parseAddr(m []byte) (addr Addr, family, index int, err error) { var local, dst *net.IPNet for _, attr := range attrs { switch attr.Attr.Type { - case syscall.IFA_ADDRESS: + case unix.IFA_ADDRESS: dst = &net.IPNet{ IP: attr.Value, Mask: net.CIDRMask(int(msg.Prefixlen), 8*len(attr.Value)), } addr.Peer = dst - case syscall.IFA_LOCAL: + case unix.IFA_LOCAL: local = &net.IPNet{ IP: attr.Value, Mask: net.CIDRMask(int(msg.Prefixlen), 8*len(attr.Value)), } addr.IPNet = local - case syscall.IFA_BROADCAST: + case unix.IFA_BROADCAST: addr.Broadcast = attr.Value - case syscall.IFA_LABEL: + case unix.IFA_LABEL: addr.Label = string(attr.Value[:len(attr.Value)-1]) case IFA_FLAGS: addr.Flags = int(native.Uint32(attr.Value[0:4])) @@ -232,17 +250,36 @@ 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 addrSubscribe(netns.None(), netns.None(), ch, done) + return addrSubscribeAt(netns.None(), netns.None(), ch, done, nil, false) } // AddrSubscribeAt works like AddrSubscribe plus it allows the caller // to choose the network namespace in which to subscribe (ns). func AddrSubscribeAt(ns netns.NsHandle, ch chan<- AddrUpdate, done <-chan struct{}) error { - return addrSubscribe(ns, netns.None(), ch, done) + return addrSubscribeAt(ns, netns.None(), ch, done, nil, false) +} + +// AddrSubscribeOptions contains a set of options to use with +// AddrSubscribeWithOptions. +type AddrSubscribeOptions struct { + Namespace *netns.NsHandle + ErrorCallback func(error) + ListExisting bool } -func addrSubscribe(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-chan struct{}) error { - s, err := nl.SubscribeAt(newNs, curNs, syscall.NETLINK_ROUTE, syscall.RTNLGRP_IPV4_IFADDR, syscall.RTNLGRP_IPV6_IFADDR) +// AddrSubscribeWithOptions work like AddrSubscribe but enable to +// provide additional options to modify the behavior. Currently, the +// namespace can be provided as well as an error callback. +func AddrSubscribeWithOptions(ch chan<- AddrUpdate, done <-chan struct{}, options AddrSubscribeOptions) error { + if options.Namespace == nil { + none := netns.None() + options.Namespace = &none + } + return addrSubscribeAt(*options.Namespace, netns.None(), ch, done, options.ErrorCallback, options.ListExisting) +} + +func addrSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-chan struct{}, cberr func(error), listExisting bool) error { + s, err := nl.SubscribeAt(newNs, curNs, unix.NETLINK_ROUTE, unix.RTNLGRP_IPV4_IFADDR, unix.RTNLGRP_IPV6_IFADDR) if err != nil { return err } @@ -252,30 +289,59 @@ func addrSubscribe(newNs, curNs netns.NsHandle, ch chan<- AddrUpdate, done <-cha s.Close() }() } + if listExisting { + req := pkgHandle.newNetlinkRequest(unix.RTM_GETADDR, + unix.NLM_F_DUMP) + infmsg := nl.NewIfInfomsg(unix.AF_UNSPEC) + req.AddData(infmsg) + if err := s.Send(req); err != nil { + return err + } + } go func() { defer close(ch) for { msgs, err := s.Receive() if err != nil { - log.Printf("netlink.AddrSubscribe: Receive() error: %v", err) + if cberr != nil { + cberr(err) + } return } for _, m := range msgs { - msgType := m.Header.Type - if msgType != syscall.RTM_NEWADDR && msgType != syscall.RTM_DELADDR { - log.Printf("netlink.AddrSubscribe: bad message type: %d", msgType) + if m.Header.Type == unix.NLMSG_DONE { 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 + } + 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 + } addr, _, ifindex, err := parseAddr(m.Data) if err != nil { - log.Printf("netlink.AddrSubscribe: could not parse address: %v", err) - continue + if cberr != nil { + cberr(fmt.Errorf("could not parse address: %v", err)) + } + return } ch <- AddrUpdate{LinkAddress: *addr.IPNet, LinkIndex: ifindex, - NewAddr: msgType == syscall.RTM_NEWADDR, + NewAddr: msgType == unix.RTM_NEWADDR, Flags: addr.Flags, Scope: addr.Scope, PreferedLft: addr.PreferedLft, diff --git a/vendor/github.com/vishvananda/netlink/bpf_linux.go b/vendor/github.com/vishvananda/netlink/bpf_linux.go index 533743987..6631626bf 100644 --- a/vendor/github.com/vishvananda/netlink/bpf_linux.go +++ b/vendor/github.com/vishvananda/netlink/bpf_linux.go @@ -1,49 +1,12 @@ package netlink -/* -#include <asm/types.h> -#include <asm/unistd.h> -#include <errno.h> -#include <stdio.h> -#include <stdint.h> -#include <unistd.h> +import ( + "unsafe" -static int load_simple_bpf(int prog_type, int ret) { -#ifdef __NR_bpf - // { return ret; } - __u64 __attribute__((aligned(8))) insns[] = { - 0x00000000000000b7ull | ((__u64)ret<<32), - 0x0000000000000095ull, - }; - __u8 __attribute__((aligned(8))) license[] = "ASL2"; - // Copied from a header file since libc is notoriously slow to update. - // The call will succeed or fail and that will be our indication on - // whether or not it is supported. - struct { - __u32 prog_type; - __u32 insn_cnt; - __u64 insns; - __u64 license; - __u32 log_level; - __u32 log_size; - __u64 log_buf; - __u32 kern_version; - } __attribute__((aligned(8))) attr = { - .prog_type = prog_type, - .insn_cnt = 2, - .insns = (uintptr_t)&insns, - .license = (uintptr_t)&license, - }; - return syscall(__NR_bpf, 5, &attr, sizeof(attr)); -#else - errno = EINVAL; - return -1; -#endif -} -*/ -import "C" + "golang.org/x/sys/unix" +) -type BpfProgType C.int +type BpfProgType uint32 const ( BPF_PROG_TYPE_UNSPEC BpfProgType = iota @@ -55,8 +18,36 @@ const ( BPF_PROG_TYPE_XDP ) -// loadSimpleBpf loads a trivial bpf program for testing purposes -func loadSimpleBpf(progType BpfProgType, ret int) (int, error) { - fd, err := C.load_simple_bpf(C.int(progType), C.int(ret)) - return int(fd), err +type BPFAttr struct { + ProgType uint32 + InsnCnt uint32 + Insns uintptr + License uintptr + LogLevel uint32 + LogSize uint32 + LogBuf uintptr + KernVersion uint32 +} + +// loadSimpleBpf loads a trivial bpf program for testing purposes. +func loadSimpleBpf(progType BpfProgType, ret uint32) (int, error) { + insns := []uint64{ + 0x00000000000000b7 | (uint64(ret) << 32), + 0x0000000000000095, + } + license := []byte{'A', 'S', 'L', '2', '\x00'} + attr := BPFAttr{ + ProgType: uint32(progType), + InsnCnt: uint32(len(insns)), + Insns: uintptr(unsafe.Pointer(&insns[0])), + License: uintptr(unsafe.Pointer(&license[0])), + } + fd, _, errno := unix.Syscall(unix.SYS_BPF, + 5, /* bpf cmd */ + uintptr(unsafe.Pointer(&attr)), + unsafe.Sizeof(attr)) + if errno != 0 { + return 0, errno + } + return int(fd), nil } diff --git a/vendor/github.com/vishvananda/netlink/bridge_linux.go b/vendor/github.com/vishvananda/netlink/bridge_linux.go index a65d6a131..350ab0db4 100644 --- a/vendor/github.com/vishvananda/netlink/bridge_linux.go +++ b/vendor/github.com/vishvananda/netlink/bridge_linux.go @@ -2,9 +2,9 @@ package netlink import ( "fmt" - "syscall" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) // BridgeVlanList gets a map of device id to bridge vlan infos. @@ -16,12 +16,12 @@ func BridgeVlanList() (map[int32][]*nl.BridgeVlanInfo, error) { // BridgeVlanList gets a map of device id to bridge vlan infos. // Equivalent to: `bridge vlan show` func (h *Handle) BridgeVlanList() (map[int32][]*nl.BridgeVlanInfo, error) { - req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_DUMP) - msg := nl.NewIfInfomsg(syscall.AF_BRIDGE) + req := h.newNetlinkRequest(unix.RTM_GETLINK, unix.NLM_F_DUMP) + msg := nl.NewIfInfomsg(unix.AF_BRIDGE) req.AddData(msg) - req.AddData(nl.NewRtAttr(nl.IFLA_EXT_MASK, nl.Uint32Attr(uint32(nl.RTEXT_FILTER_BRVLAN)))) + req.AddData(nl.NewRtAttr(unix.IFLA_EXT_MASK, nl.Uint32Attr(uint32(nl.RTEXT_FILTER_BRVLAN)))) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWLINK) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWLINK) if err != nil { return nil, err } @@ -35,7 +35,7 @@ func (h *Handle) BridgeVlanList() (map[int32][]*nl.BridgeVlanInfo, error) { } for _, attr := range attrs { switch attr.Attr.Type { - case nl.IFLA_AF_SPEC: + case unix.IFLA_AF_SPEC: //nested attr nestAttrs, err := nl.ParseRouteAttr(attr.Value) if err != nil { @@ -63,7 +63,7 @@ func BridgeVlanAdd(link Link, vid uint16, pvid, untagged, self, master bool) err // BridgeVlanAdd adds a new vlan filter entry // Equivalent to: `bridge vlan add dev DEV vid VID [ pvid ] [ untagged ] [ self ] [ master ]` func (h *Handle) BridgeVlanAdd(link Link, vid uint16, pvid, untagged, self, master bool) error { - return h.bridgeVlanModify(syscall.RTM_SETLINK, link, vid, pvid, untagged, self, master) + return h.bridgeVlanModify(unix.RTM_SETLINK, link, vid, pvid, untagged, self, master) } // BridgeVlanDel adds a new vlan filter entry @@ -75,19 +75,19 @@ func BridgeVlanDel(link Link, vid uint16, pvid, untagged, self, master bool) err // BridgeVlanDel adds a new vlan filter entry // Equivalent to: `bridge vlan del dev DEV vid VID [ pvid ] [ untagged ] [ self ] [ master ]` func (h *Handle) BridgeVlanDel(link Link, vid uint16, pvid, untagged, self, master bool) error { - return h.bridgeVlanModify(syscall.RTM_DELLINK, link, vid, pvid, untagged, self, master) + return h.bridgeVlanModify(unix.RTM_DELLINK, link, vid, pvid, untagged, self, master) } func (h *Handle) bridgeVlanModify(cmd int, link Link, vid uint16, pvid, untagged, self, master bool) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(cmd, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(cmd, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_BRIDGE) + msg := nl.NewIfInfomsg(unix.AF_BRIDGE) msg.Index = int32(base.Index) req.AddData(msg) - br := nl.NewRtAttr(nl.IFLA_AF_SPEC, nil) + br := nl.NewRtAttr(unix.IFLA_AF_SPEC, nil) var flags uint16 if self { flags |= nl.BRIDGE_FLAGS_SELF @@ -107,7 +107,7 @@ func (h *Handle) bridgeVlanModify(cmd int, link Link, vid uint16, pvid, untagged } nl.NewRtAttrChild(br, nl.IFLA_BRIDGE_VLAN_INFO, vlanInfo.Serialize()) req.AddData(br) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) if err != nil { return err } diff --git a/vendor/github.com/vishvananda/netlink/class_linux.go b/vendor/github.com/vishvananda/netlink/class_linux.go index 91cd3883d..a4997740e 100644 --- a/vendor/github.com/vishvananda/netlink/class_linux.go +++ b/vendor/github.com/vishvananda/netlink/class_linux.go @@ -5,6 +5,7 @@ import ( "syscall" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) // NOTE: function is in here because it uses other linux functions @@ -50,7 +51,7 @@ func ClassDel(class Class) error { // ClassDel will delete a class from the system. // Equivalent to: `tc class del $class` func (h *Handle) ClassDel(class Class) error { - return h.classModify(syscall.RTM_DELTCLASS, 0, class) + return h.classModify(unix.RTM_DELTCLASS, 0, class) } // ClassChange will change a class in place @@ -64,7 +65,7 @@ func ClassChange(class Class) error { // Equivalent to: `tc class change $class` // The parent and handle MUST NOT be changed. func (h *Handle) ClassChange(class Class) error { - return h.classModify(syscall.RTM_NEWTCLASS, 0, class) + return h.classModify(unix.RTM_NEWTCLASS, 0, class) } // ClassReplace will replace a class to the system. @@ -82,7 +83,7 @@ func ClassReplace(class Class) error { // If a class already exist with this parent/handle pair, the class is changed. // If a class does not already exist with this parent/handle, a new class is created. func (h *Handle) ClassReplace(class Class) error { - return h.classModify(syscall.RTM_NEWTCLASS, syscall.NLM_F_CREATE, class) + return h.classModify(unix.RTM_NEWTCLASS, unix.NLM_F_CREATE, class) } // ClassAdd will add a class to the system. @@ -95,14 +96,14 @@ func ClassAdd(class Class) error { // Equivalent to: `tc class add $class` func (h *Handle) ClassAdd(class Class) error { return h.classModify( - syscall.RTM_NEWTCLASS, - syscall.NLM_F_CREATE|syscall.NLM_F_EXCL, + unix.RTM_NEWTCLASS, + unix.NLM_F_CREATE|unix.NLM_F_EXCL, class, ) } func (h *Handle) classModify(cmd, flags int, class Class) error { - req := h.newNetlinkRequest(cmd, flags|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(cmd, flags|unix.NLM_F_ACK) base := class.Attrs() msg := &nl.TcMsg{ Family: nl.FAMILY_ALL, @@ -112,12 +113,12 @@ func (h *Handle) classModify(cmd, flags int, class Class) error { } req.AddData(msg) - if cmd != syscall.RTM_DELTCLASS { + if cmd != unix.RTM_DELTCLASS { if err := classPayload(req, class); err != nil { return err } } - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -141,12 +142,12 @@ func classPayload(req *nl.NetlinkRequest, class Class) error { var rtab [256]uint32 var ctab [256]uint32 tcrate := nl.TcRateSpec{Rate: uint32(htb.Rate)} - if CalcRtable(&tcrate, rtab, cellLog, uint32(mtu), linklayer) < 0 { + if CalcRtable(&tcrate, rtab[:], cellLog, uint32(mtu), linklayer) < 0 { return errors.New("HTB: failed to calculate rate table") } opt.Rate = tcrate tcceil := nl.TcRateSpec{Rate: uint32(htb.Ceil)} - if CalcRtable(&tcceil, ctab, ccellLog, uint32(mtu), linklayer) < 0 { + if CalcRtable(&tcceil, ctab[:], ccellLog, uint32(mtu), linklayer) < 0 { return errors.New("HTB: failed to calculate ceil rate table") } opt.Ceil = tcceil @@ -169,7 +170,7 @@ func ClassList(link Link, parent uint32) ([]Class, error) { // Equivalent to: `tc class show`. // Generally returns nothing if link and parent are not specified. func (h *Handle) ClassList(link Link, parent uint32) ([]Class, error) { - req := h.newNetlinkRequest(syscall.RTM_GETTCLASS, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(unix.RTM_GETTCLASS, unix.NLM_F_DUMP) msg := &nl.TcMsg{ Family: nl.FAMILY_ALL, Parent: parent, @@ -181,7 +182,7 @@ func (h *Handle) ClassList(link Link, parent uint32) ([]Class, error) { } req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWTCLASS) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWTCLASS) if err != nil { return nil, err } diff --git a/vendor/github.com/vishvananda/netlink/conntrack_linux.go b/vendor/github.com/vishvananda/netlink/conntrack_linux.go index ecf044565..a0fc74a37 100644 --- a/vendor/github.com/vishvananda/netlink/conntrack_linux.go +++ b/vendor/github.com/vishvananda/netlink/conntrack_linux.go @@ -6,9 +6,9 @@ import ( "errors" "fmt" "net" - "syscall" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) // ConntrackTableType Conntrack table for the netlink operation @@ -85,8 +85,8 @@ func (h *Handle) ConntrackTableList(table ConntrackTableType, family InetFamily) // conntrack -F [table] Flush table // The flush operation applies to all the family types func (h *Handle) ConntrackTableFlush(table ConntrackTableType) error { - req := h.newConntrackRequest(table, syscall.AF_INET, nl.IPCTNL_MSG_CT_DELETE, syscall.NLM_F_ACK) - _, err := req.Execute(syscall.NETLINK_NETFILTER, 0) + req := h.newConntrackRequest(table, unix.AF_INET, nl.IPCTNL_MSG_CT_DELETE, unix.NLM_F_ACK) + _, err := req.Execute(unix.NETLINK_NETFILTER, 0) return err } @@ -102,10 +102,10 @@ func (h *Handle) ConntrackDeleteFilter(table ConntrackTableType, family InetFami for _, dataRaw := range res { flow := parseRawData(dataRaw) if match := filter.MatchConntrackFlow(flow); match { - req2 := h.newConntrackRequest(table, family, nl.IPCTNL_MSG_CT_DELETE, syscall.NLM_F_ACK) + req2 := h.newConntrackRequest(table, family, nl.IPCTNL_MSG_CT_DELETE, unix.NLM_F_ACK) // skip the first 4 byte that are the netfilter header, the newConntrackRequest is adding it already req2.AddRawData(dataRaw[4:]) - req2.Execute(syscall.NETLINK_NETFILTER, 0) + req2.Execute(unix.NETLINK_NETFILTER, 0) matched++ } } @@ -127,8 +127,8 @@ func (h *Handle) newConntrackRequest(table ConntrackTableType, family InetFamily } func (h *Handle) dumpConntrackTable(table ConntrackTableType, family InetFamily) ([][]byte, error) { - req := h.newConntrackRequest(table, family, nl.IPCTNL_MSG_CT_GET, syscall.NLM_F_DUMP) - return req.Execute(syscall.NETLINK_NETFILTER, 0) + req := h.newConntrackRequest(table, family, nl.IPCTNL_MSG_CT_GET, unix.NLM_F_DUMP) + return req.Execute(unix.NETLINK_NETFILTER, 0) } // The full conntrack flow structure is very complicated and can be found in the file: diff --git a/vendor/github.com/vishvananda/netlink/filter.go b/vendor/github.com/vishvananda/netlink/filter.go index 938b28b0b..c2cf8e4dc 100644 --- a/vendor/github.com/vishvananda/netlink/filter.go +++ b/vendor/github.com/vishvananda/netlink/filter.go @@ -2,8 +2,6 @@ package netlink import ( "fmt" - - "github.com/vishvananda/netlink/nl" ) type Filter interface { @@ -19,7 +17,7 @@ type FilterAttrs struct { Handle uint32 Parent uint32 Priority uint16 // lower is higher priority - Protocol uint16 // syscall.ETH_P_* + Protocol uint16 // unix.ETH_P_* } func (q FilterAttrs) String() string { @@ -184,14 +182,6 @@ func NewMirredAction(redirIndex int) *MirredAction { } } -// Constants used in TcU32Sel.Flags. -const ( - TC_U32_TERMINAL = nl.TC_U32_TERMINAL - TC_U32_OFFSET = nl.TC_U32_OFFSET - TC_U32_VAROFFSET = nl.TC_U32_VAROFFSET - TC_U32_EAT = nl.TC_U32_EAT -) - // 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. @@ -235,6 +225,21 @@ func (filter *U32) Type() string { return "u32" } +// MatchAll filters match all packets +type MatchAll struct { + FilterAttrs + ClassId uint32 + Actions []Action +} + +func (filter *MatchAll) Attrs() *FilterAttrs { + return &filter.FilterAttrs +} + +func (filter *MatchAll) Type() string { + return "matchall" +} + type FilterFwAttrs struct { ClassId uint32 InDev string diff --git a/vendor/github.com/vishvananda/netlink/filter_linux.go b/vendor/github.com/vishvananda/netlink/filter_linux.go index dc0f90af8..f0eac6b78 100644 --- a/vendor/github.com/vishvananda/netlink/filter_linux.go +++ b/vendor/github.com/vishvananda/netlink/filter_linux.go @@ -9,6 +9,15 @@ import ( "unsafe" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" +) + +// Constants used in TcU32Sel.Flags. +const ( + TC_U32_TERMINAL = nl.TC_U32_TERMINAL + TC_U32_OFFSET = nl.TC_U32_OFFSET + TC_U32_VAROFFSET = nl.TC_U32_VAROFFSET + TC_U32_EAT = nl.TC_U32_EAT ) // Fw filter filters on firewall marks @@ -47,7 +56,7 @@ func NewFw(attrs FilterAttrs, fattrs FilterFwAttrs) (*Fw, error) { if police.Rate.Rate != 0 { police.Rate.Mpu = fattrs.Mpu police.Rate.Overhead = fattrs.Overhead - if CalcRtable(&police.Rate, rtab, rcellLog, fattrs.Mtu, linklayer) < 0 { + if CalcRtable(&police.Rate, rtab[:], rcellLog, fattrs.Mtu, linklayer) < 0 { return nil, errors.New("TBF: failed to calculate rate table") } police.Burst = uint32(Xmittime(uint64(police.Rate.Rate), uint32(buffer))) @@ -56,7 +65,7 @@ func NewFw(attrs FilterAttrs, fattrs FilterFwAttrs) (*Fw, error) { if police.PeakRate.Rate != 0 { police.PeakRate.Mpu = fattrs.Mpu police.PeakRate.Overhead = fattrs.Overhead - if CalcRtable(&police.PeakRate, ptab, pcellLog, fattrs.Mtu, linklayer) < 0 { + if CalcRtable(&police.PeakRate, ptab[:], pcellLog, fattrs.Mtu, linklayer) < 0 { return nil, errors.New("POLICE: failed to calculate peak rate table") } } @@ -90,7 +99,7 @@ func FilterDel(filter Filter) error { // FilterDel will delete a filter from the system. // Equivalent to: `tc filter del $filter` func (h *Handle) FilterDel(filter Filter) error { - req := h.newNetlinkRequest(syscall.RTM_DELTFILTER, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_DELTFILTER, unix.NLM_F_ACK) base := filter.Attrs() msg := &nl.TcMsg{ Family: nl.FAMILY_ALL, @@ -101,7 +110,7 @@ func (h *Handle) FilterDel(filter Filter) error { } req.AddData(msg) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -115,7 +124,7 @@ func FilterAdd(filter Filter) error { // Equivalent to: `tc filter add $filter` func (h *Handle) FilterAdd(filter Filter) error { native = nl.NativeEndian() - req := h.newNetlinkRequest(syscall.RTM_NEWTFILTER, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_NEWTFILTER, unix.NLM_F_CREATE|unix.NLM_F_EXCL|unix.NLM_F_ACK) base := filter.Attrs() msg := &nl.TcMsg{ Family: nl.FAMILY_ALL, @@ -128,9 +137,11 @@ func (h *Handle) FilterAdd(filter Filter) error { req.AddData(nl.NewRtAttr(nl.TCA_KIND, nl.ZeroTerminated(filter.Type()))) options := nl.NewRtAttr(nl.TCA_OPTIONS, nil) - if u32, ok := filter.(*U32); ok { + + switch filter := filter.(type) { + case *U32: // Convert TcU32Sel into nl.TcU32Sel as it is without copy. - sel := (*nl.TcU32Sel)(unsafe.Pointer(u32.Sel)) + sel := (*nl.TcU32Sel)(unsafe.Pointer(filter.Sel)) if sel == nil { // match all sel = &nl.TcU32Sel{ @@ -158,63 +169,71 @@ func (h *Handle) FilterAdd(filter Filter) error { } sel.Nkeys = uint8(len(sel.Keys)) nl.NewRtAttrChild(options, nl.TCA_U32_SEL, sel.Serialize()) - if u32.ClassId != 0 { - nl.NewRtAttrChild(options, nl.TCA_U32_CLASSID, nl.Uint32Attr(u32.ClassId)) + if filter.ClassId != 0 { + nl.NewRtAttrChild(options, nl.TCA_U32_CLASSID, nl.Uint32Attr(filter.ClassId)) } actionsAttr := nl.NewRtAttrChild(options, nl.TCA_U32_ACT, nil) // backwards compatibility - if u32.RedirIndex != 0 { - u32.Actions = append([]Action{NewMirredAction(u32.RedirIndex)}, u32.Actions...) + if filter.RedirIndex != 0 { + filter.Actions = append([]Action{NewMirredAction(filter.RedirIndex)}, filter.Actions...) } - if err := EncodeActions(actionsAttr, u32.Actions); err != nil { + if err := EncodeActions(actionsAttr, filter.Actions); err != nil { return err } - } else if fw, ok := filter.(*Fw); ok { - if fw.Mask != 0 { + case *Fw: + if filter.Mask != 0 { b := make([]byte, 4) - native.PutUint32(b, fw.Mask) + native.PutUint32(b, filter.Mask) nl.NewRtAttrChild(options, nl.TCA_FW_MASK, b) } - if fw.InDev != "" { - nl.NewRtAttrChild(options, nl.TCA_FW_INDEV, nl.ZeroTerminated(fw.InDev)) + if filter.InDev != "" { + nl.NewRtAttrChild(options, nl.TCA_FW_INDEV, nl.ZeroTerminated(filter.InDev)) } - if (fw.Police != nl.TcPolice{}) { + if (filter.Police != nl.TcPolice{}) { police := nl.NewRtAttrChild(options, nl.TCA_FW_POLICE, nil) - nl.NewRtAttrChild(police, nl.TCA_POLICE_TBF, fw.Police.Serialize()) - if (fw.Police.Rate != nl.TcRateSpec{}) { - payload := SerializeRtab(fw.Rtab) + nl.NewRtAttrChild(police, nl.TCA_POLICE_TBF, filter.Police.Serialize()) + if (filter.Police.Rate != nl.TcRateSpec{}) { + payload := SerializeRtab(filter.Rtab) nl.NewRtAttrChild(police, nl.TCA_POLICE_RATE, payload) } - if (fw.Police.PeakRate != nl.TcRateSpec{}) { - payload := SerializeRtab(fw.Ptab) + if (filter.Police.PeakRate != nl.TcRateSpec{}) { + payload := SerializeRtab(filter.Ptab) nl.NewRtAttrChild(police, nl.TCA_POLICE_PEAKRATE, payload) } } - if fw.ClassId != 0 { + if filter.ClassId != 0 { b := make([]byte, 4) - native.PutUint32(b, fw.ClassId) + native.PutUint32(b, filter.ClassId) nl.NewRtAttrChild(options, nl.TCA_FW_CLASSID, b) } - } else if bpf, ok := filter.(*BpfFilter); ok { + case *BpfFilter: var bpfFlags uint32 - if bpf.ClassId != 0 { - nl.NewRtAttrChild(options, nl.TCA_BPF_CLASSID, nl.Uint32Attr(bpf.ClassId)) + if filter.ClassId != 0 { + nl.NewRtAttrChild(options, nl.TCA_BPF_CLASSID, nl.Uint32Attr(filter.ClassId)) } - if bpf.Fd >= 0 { - nl.NewRtAttrChild(options, nl.TCA_BPF_FD, nl.Uint32Attr((uint32(bpf.Fd)))) + if filter.Fd >= 0 { + nl.NewRtAttrChild(options, nl.TCA_BPF_FD, nl.Uint32Attr((uint32(filter.Fd)))) } - if bpf.Name != "" { - nl.NewRtAttrChild(options, nl.TCA_BPF_NAME, nl.ZeroTerminated(bpf.Name)) + if filter.Name != "" { + nl.NewRtAttrChild(options, nl.TCA_BPF_NAME, nl.ZeroTerminated(filter.Name)) } - if bpf.DirectAction { + if filter.DirectAction { bpfFlags |= nl.TCA_BPF_FLAG_ACT_DIRECT } nl.NewRtAttrChild(options, nl.TCA_BPF_FLAGS, nl.Uint32Attr(bpfFlags)) + case *MatchAll: + actionsAttr := nl.NewRtAttrChild(options, 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)) + } } req.AddData(options) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -229,7 +248,7 @@ func FilterList(link Link, parent uint32) ([]Filter, error) { // Equivalent to: `tc filter show`. // Generally returns nothing if link and parent are not specified. func (h *Handle) FilterList(link Link, parent uint32) ([]Filter, error) { - req := h.newNetlinkRequest(syscall.RTM_GETTFILTER, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(unix.RTM_GETTFILTER, unix.NLM_F_DUMP) msg := &nl.TcMsg{ Family: nl.FAMILY_ALL, Parent: parent, @@ -241,7 +260,7 @@ func (h *Handle) FilterList(link Link, parent uint32) ([]Filter, error) { } req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWTFILTER) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWTFILTER) if err != nil { return nil, err } @@ -277,6 +296,8 @@ func (h *Handle) FilterList(link Link, parent uint32) ([]Filter, error) { filter = &Fw{} case "bpf": filter = &BpfFilter{} + case "matchall": + filter = &MatchAll{} default: filter = &GenericFilter{FilterType: filterType} } @@ -301,6 +322,11 @@ func (h *Handle) FilterList(link Link, parent uint32) ([]Filter, error) { if err != nil { return nil, err } + case "matchall": + detailed, err = parseMatchAllData(filter, data) + if err != nil { + return nil, err + } default: detailed = true } @@ -530,6 +556,28 @@ func parseBpfData(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error) return detailed, nil } +func parseMatchAllData(filter Filter, data []syscall.NetlinkRouteAttr) (bool, error) { + native = nl.NativeEndian() + matchall := filter.(*MatchAll) + detailed := true + for _, datum := range data { + switch datum.Attr.Type { + case nl.TCA_MATCHALL_CLASSID: + matchall.ClassId = native.Uint32(datum.Value[0:4]) + case nl.TCA_MATCHALL_ACT: + tables, err := nl.ParseRouteAttr(datum.Value) + if err != nil { + return detailed, err + } + matchall.Actions, err = parseActions(tables) + if err != nil { + return detailed, err + } + } + } + return detailed, nil +} + func AlignToAtm(size uint) uint { var linksize, cells int cells = int(size / nl.ATM_CELL_PAYLOAD) @@ -552,7 +600,7 @@ func AdjustSize(sz uint, mpu uint, linklayer int) uint { } } -func CalcRtable(rate *nl.TcRateSpec, rtab [256]uint32, cellLog int, mtu uint32, linklayer int) int { +func CalcRtable(rate *nl.TcRateSpec, rtab []uint32, cellLog int, mtu uint32, linklayer int) int { bps := rate.Rate mpu := rate.Mpu var sz uint diff --git a/vendor/github.com/vishvananda/netlink/fou.go b/vendor/github.com/vishvananda/netlink/fou.go new file mode 100644 index 000000000..71e73c37a --- /dev/null +++ b/vendor/github.com/vishvananda/netlink/fou.go @@ -0,0 +1,21 @@ +package netlink + +import ( + "errors" +) + +var ( + // ErrAttrHeaderTruncated is returned when a netlink attribute's header is + // truncated. + ErrAttrHeaderTruncated = errors.New("attribute header truncated") + // ErrAttrBodyTruncated is returned when a netlink attribute's body is + // truncated. + ErrAttrBodyTruncated = errors.New("attribute body truncated") +) + +type Fou struct { + Family int + Port int + Protocol int + EncapType int +} diff --git a/vendor/github.com/vishvananda/netlink/fou_linux.go b/vendor/github.com/vishvananda/netlink/fou_linux.go new file mode 100644 index 000000000..62d59bd2d --- /dev/null +++ b/vendor/github.com/vishvananda/netlink/fou_linux.go @@ -0,0 +1,215 @@ +// +build linux + +package netlink + +import ( + "encoding/binary" + "errors" + + "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" +) + +const ( + FOU_GENL_NAME = "fou" +) + +const ( + FOU_CMD_UNSPEC uint8 = iota + FOU_CMD_ADD + FOU_CMD_DEL + FOU_CMD_GET + FOU_CMD_MAX = FOU_CMD_GET +) + +const ( + FOU_ATTR_UNSPEC = iota + FOU_ATTR_PORT + FOU_ATTR_AF + FOU_ATTR_IPPROTO + FOU_ATTR_TYPE + FOU_ATTR_REMCSUM_NOPARTIAL + FOU_ATTR_MAX = FOU_ATTR_REMCSUM_NOPARTIAL +) + +const ( + FOU_ENCAP_UNSPEC = iota + FOU_ENCAP_DIRECT + FOU_ENCAP_GUE + FOU_ENCAP_MAX = FOU_ENCAP_GUE +) + +var fouFamilyId int + +func FouFamilyId() (int, error) { + if fouFamilyId != 0 { + return fouFamilyId, nil + } + + fam, err := GenlFamilyGet(FOU_GENL_NAME) + if err != nil { + return -1, err + } + + fouFamilyId = int(fam.ID) + return fouFamilyId, nil +} + +func FouAdd(f Fou) error { + return pkgHandle.FouAdd(f) +} + +func (h *Handle) FouAdd(f Fou) error { + fam_id, err := FouFamilyId() + if err != nil { + return err + } + + // setting ip protocol conflicts with encapsulation type GUE + if f.EncapType == FOU_ENCAP_GUE && f.Protocol != 0 { + return errors.New("GUE encapsulation doesn't specify an IP protocol") + } + + req := h.newNetlinkRequest(fam_id, unix.NLM_F_ACK) + + // int to byte for port + bp := make([]byte, 2) + binary.BigEndian.PutUint16(bp[0:2], uint16(f.Port)) + + attrs := []*nl.RtAttr{ + nl.NewRtAttr(FOU_ATTR_PORT, bp), + nl.NewRtAttr(FOU_ATTR_TYPE, []byte{uint8(f.EncapType)}), + nl.NewRtAttr(FOU_ATTR_AF, []byte{uint8(f.Family)}), + nl.NewRtAttr(FOU_ATTR_IPPROTO, []byte{uint8(f.Protocol)}), + } + raw := []byte{FOU_CMD_ADD, 1, 0, 0} + for _, a := range attrs { + raw = append(raw, a.Serialize()...) + } + + req.AddRawData(raw) + + _, err = req.Execute(unix.NETLINK_GENERIC, 0) + if err != nil { + return err + } + + return nil +} + +func FouDel(f Fou) error { + return pkgHandle.FouDel(f) +} + +func (h *Handle) FouDel(f Fou) error { + fam_id, err := FouFamilyId() + if err != nil { + return err + } + + req := h.newNetlinkRequest(fam_id, unix.NLM_F_ACK) + + // int to byte for port + bp := make([]byte, 2) + binary.BigEndian.PutUint16(bp[0:2], uint16(f.Port)) + + attrs := []*nl.RtAttr{ + nl.NewRtAttr(FOU_ATTR_PORT, bp), + nl.NewRtAttr(FOU_ATTR_AF, []byte{uint8(f.Family)}), + } + raw := []byte{FOU_CMD_DEL, 1, 0, 0} + for _, a := range attrs { + raw = append(raw, a.Serialize()...) + } + + req.AddRawData(raw) + + _, err = req.Execute(unix.NETLINK_GENERIC, 0) + if err != nil { + return err + } + + return nil +} + +func FouList(fam int) ([]Fou, error) { + return pkgHandle.FouList(fam) +} + +func (h *Handle) FouList(fam int) ([]Fou, error) { + fam_id, err := FouFamilyId() + if err != nil { + return nil, err + } + + req := h.newNetlinkRequest(fam_id, unix.NLM_F_DUMP) + + attrs := []*nl.RtAttr{ + nl.NewRtAttr(FOU_ATTR_AF, []byte{uint8(fam)}), + } + raw := []byte{FOU_CMD_GET, 1, 0, 0} + for _, a := range attrs { + raw = append(raw, a.Serialize()...) + } + + req.AddRawData(raw) + + msgs, err := req.Execute(unix.NETLINK_GENERIC, 0) + if err != nil { + return nil, err + } + + fous := make([]Fou, 0, len(msgs)) + for _, m := range msgs { + f, err := deserializeFouMsg(m) + if err != nil { + return fous, err + } + + fous = append(fous, f) + } + + return fous, nil +} + +func deserializeFouMsg(msg []byte) (Fou, error) { + // we'll skip to byte 4 to first attribute + msg = msg[3:] + var shift int + fou := Fou{} + + for { + // attribute header is at least 16 bits + if len(msg) < 4 { + return fou, ErrAttrHeaderTruncated + } + + lgt := int(binary.BigEndian.Uint16(msg[0:2])) + if len(msg) < lgt+4 { + return fou, ErrAttrBodyTruncated + } + attr := binary.BigEndian.Uint16(msg[2:4]) + + shift = lgt + 3 + switch attr { + case FOU_ATTR_AF: + fou.Family = int(msg[5]) + case FOU_ATTR_PORT: + fou.Port = int(binary.BigEndian.Uint16(msg[5:7])) + // port is 2 bytes + shift = lgt + 2 + case FOU_ATTR_IPPROTO: + fou.Protocol = int(msg[5]) + case FOU_ATTR_TYPE: + fou.EncapType = int(msg[5]) + } + + msg = msg[shift:] + + if len(msg) < 4 { + break + } + } + + return fou, nil +} diff --git a/vendor/github.com/vishvananda/netlink/fou_unspecified.go b/vendor/github.com/vishvananda/netlink/fou_unspecified.go new file mode 100644 index 000000000..3a8365bfe --- /dev/null +++ b/vendor/github.com/vishvananda/netlink/fou_unspecified.go @@ -0,0 +1,15 @@ +// +build !linux + +package netlink + +func FouAdd(f Fou) error { + return ErrNotImplemented +} + +func FouDel(f Fou) error { + return ErrNotImplemented +} + +func FouList(fam int) ([]Fou, error) { + return nil, ErrNotImplemented +} diff --git a/vendor/github.com/vishvananda/netlink/genetlink_linux.go b/vendor/github.com/vishvananda/netlink/genetlink_linux.go index a388a8700..ce7969907 100644 --- a/vendor/github.com/vishvananda/netlink/genetlink_linux.go +++ b/vendor/github.com/vishvananda/netlink/genetlink_linux.go @@ -5,6 +5,7 @@ import ( "syscall" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) type GenlOp struct { @@ -130,9 +131,9 @@ func (h *Handle) GenlFamilyList() ([]*GenlFamily, error) { Command: nl.GENL_CTRL_CMD_GETFAMILY, Version: nl.GENL_CTRL_VERSION, } - req := h.newNetlinkRequest(nl.GENL_ID_CTRL, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(nl.GENL_ID_CTRL, unix.NLM_F_DUMP) req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_GENERIC, 0) + msgs, err := req.Execute(unix.NETLINK_GENERIC, 0) if err != nil { return nil, err } @@ -151,7 +152,7 @@ func (h *Handle) GenlFamilyGet(name string) (*GenlFamily, error) { req := h.newNetlinkRequest(nl.GENL_ID_CTRL, 0) req.AddData(msg) req.AddData(nl.NewRtAttr(nl.GENL_CTRL_ATTR_FAMILY_NAME, nl.ZeroTerminated(name))) - msgs, err := req.Execute(syscall.NETLINK_GENERIC, 0) + msgs, err := req.Execute(unix.NETLINK_GENERIC, 0) if err != nil { return nil, err } diff --git a/vendor/github.com/vishvananda/netlink/gtp_linux.go b/vendor/github.com/vishvananda/netlink/gtp_linux.go index 7331303ec..f5e160ba5 100644 --- a/vendor/github.com/vishvananda/netlink/gtp_linux.go +++ b/vendor/github.com/vishvananda/netlink/gtp_linux.go @@ -7,6 +7,7 @@ import ( "syscall" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) type PDP struct { @@ -82,9 +83,9 @@ func (h *Handle) GTPPDPList() ([]*PDP, error) { Command: nl.GENL_GTP_CMD_GETPDP, Version: nl.GENL_GTP_VERSION, } - req := h.newNetlinkRequest(int(f.ID), syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(int(f.ID), unix.NLM_F_DUMP) req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_GENERIC, 0) + msgs, err := req.Execute(unix.NETLINK_GENERIC, 0) if err != nil { return nil, err } @@ -96,7 +97,7 @@ func GTPPDPList() ([]*PDP, error) { } func gtpPDPGet(req *nl.NetlinkRequest) (*PDP, error) { - msgs, err := req.Execute(syscall.NETLINK_GENERIC, 0) + msgs, err := req.Execute(unix.NETLINK_GENERIC, 0) if err != nil { return nil, err } @@ -182,7 +183,7 @@ func (h *Handle) GTPPDPAdd(link Link, pdp *PDP) error { Command: nl.GENL_GTP_CMD_NEWPDP, Version: nl.GENL_GTP_VERSION, } - req := h.newNetlinkRequest(int(f.ID), syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(int(f.ID), unix.NLM_F_EXCL|unix.NLM_F_ACK) req.AddData(msg) req.AddData(nl.NewRtAttr(nl.GENL_GTP_ATTR_VERSION, nl.Uint32Attr(pdp.Version))) req.AddData(nl.NewRtAttr(nl.GENL_GTP_ATTR_LINK, nl.Uint32Attr(uint32(link.Attrs().Index)))) @@ -199,7 +200,7 @@ func (h *Handle) GTPPDPAdd(link Link, pdp *PDP) error { default: return fmt.Errorf("unsupported GTP version: %d", pdp.Version) } - _, err = req.Execute(syscall.NETLINK_GENERIC, 0) + _, err = req.Execute(unix.NETLINK_GENERIC, 0) return err } @@ -216,7 +217,7 @@ func (h *Handle) GTPPDPDel(link Link, pdp *PDP) error { Command: nl.GENL_GTP_CMD_DELPDP, Version: nl.GENL_GTP_VERSION, } - req := h.newNetlinkRequest(int(f.ID), syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(int(f.ID), unix.NLM_F_EXCL|unix.NLM_F_ACK) req.AddData(msg) req.AddData(nl.NewRtAttr(nl.GENL_GTP_ATTR_VERSION, nl.Uint32Attr(pdp.Version))) req.AddData(nl.NewRtAttr(nl.GENL_GTP_ATTR_LINK, nl.Uint32Attr(uint32(link.Attrs().Index)))) @@ -229,7 +230,7 @@ func (h *Handle) GTPPDPDel(link Link, pdp *PDP) error { default: return fmt.Errorf("unsupported GTP version: %d", pdp.Version) } - _, err = req.Execute(syscall.NETLINK_GENERIC, 0) + _, err = req.Execute(unix.NETLINK_GENERIC, 0) return err } diff --git a/vendor/github.com/vishvananda/netlink/handle_linux.go b/vendor/github.com/vishvananda/netlink/handle_linux.go index a04ceae6b..9f6d7fe0f 100644 --- a/vendor/github.com/vishvananda/netlink/handle_linux.go +++ b/vendor/github.com/vishvananda/netlink/handle_linux.go @@ -2,11 +2,11 @@ package netlink import ( "fmt" - "syscall" "time" "github.com/vishvananda/netlink/nl" "github.com/vishvananda/netns" + "golang.org/x/sys/unix" ) // Empty handle used by the netlink package methods @@ -43,14 +43,29 @@ func (h *Handle) SetSocketTimeout(to time.Duration) error { if to < time.Microsecond { return fmt.Errorf("invalid timeout, minimul value is %s", time.Microsecond) } - tv := syscall.NsecToTimeval(to.Nanoseconds()) + tv := unix.NsecToTimeval(to.Nanoseconds()) for _, sh := range h.sockets { - fd := sh.Socket.GetFd() - err := syscall.SetsockoptTimeval(fd, syscall.SOL_SOCKET, syscall.SO_RCVTIMEO, &tv) - if err != nil { + if err := sh.Socket.SetSendTimeout(&tv); err != nil { return err } - err = syscall.SetsockoptTimeval(fd, syscall.SOL_SOCKET, syscall.SO_SNDTIMEO, &tv) + if err := sh.Socket.SetReceiveTimeout(&tv); err != nil { + return err + } + } + return nil +} + +// SetSocketReceiveBufferSize sets the receive buffer size for each +// socket in the netlink handle. The maximum value is capped by +// /proc/sys/net/core/rmem_max. +func (h *Handle) SetSocketReceiveBufferSize(size int, force bool) error { + opt := unix.SO_RCVBUF + if force { + opt = unix.SO_RCVBUFFORCE + } + for _, sh := range h.sockets { + fd := sh.Socket.GetFd() + err := unix.SetsockoptInt(fd, unix.SOL_SOCKET, opt, size) if err != nil { return err } @@ -58,6 +73,24 @@ func (h *Handle) SetSocketTimeout(to time.Duration) error { return nil } +// GetSocketReceiveBufferSize gets the receiver buffer size for each +// socket in the netlink handle. The retrieved value should be the +// double to the one set for SetSocketReceiveBufferSize. +func (h *Handle) GetSocketReceiveBufferSize() ([]int, error) { + results := make([]int, len(h.sockets)) + i := 0 + for _, sh := range h.sockets { + fd := sh.Socket.GetFd() + size, err := unix.GetsockoptInt(fd, unix.SOL_SOCKET, unix.SO_RCVBUF) + if err != nil { + return nil, err + } + results[i] = size + i++ + } + return results, nil +} + // NewHandle returns a netlink handle on the network namespace // specified by ns. If ns=netns.None(), current network namespace // will be assumed @@ -101,10 +134,10 @@ func (h *Handle) newNetlinkRequest(proto, flags int) *nl.NetlinkRequest { return nl.NewNetlinkRequest(proto, flags) } return &nl.NetlinkRequest{ - NlMsghdr: syscall.NlMsghdr{ - Len: uint32(syscall.SizeofNlMsghdr), + NlMsghdr: unix.NlMsghdr{ + Len: uint32(unix.SizeofNlMsghdr), Type: uint16(proto), - Flags: syscall.NLM_F_REQUEST | uint16(flags), + Flags: unix.NLM_F_REQUEST | uint16(flags), }, Sockets: h.sockets, } diff --git a/vendor/github.com/vishvananda/netlink/handle_unspecified.go b/vendor/github.com/vishvananda/netlink/handle_unspecified.go index 32cf02273..915b765de 100644 --- a/vendor/github.com/vishvananda/netlink/handle_unspecified.go +++ b/vendor/github.com/vishvananda/netlink/handle_unspecified.go @@ -145,6 +145,10 @@ func (h *Handle) LinkSetFlood(link Link, mode bool) error { return ErrNotImplemented } +func (h *Handle) LinkSetTxQLen(link Link, qlen int) error { + return ErrNotImplemented +} + func (h *Handle) setProtinfoAttr(link Link, mode bool, attr int) error { return ErrNotImplemented } @@ -216,3 +220,39 @@ func (h *Handle) NeighList(linkIndex, family int) ([]Neigh, error) { func (h *Handle) NeighProxyList(linkIndex, family int) ([]Neigh, error) { return nil, ErrNotImplemented } + +func (h *Handle) RouteAdd(route *Route) error { + return ErrNotImplemented +} + +func (h *Handle) RouteDel(route *Route) error { + return ErrNotImplemented +} + +func (h *Handle) RouteGet(destination net.IP) ([]Route, error) { + return nil, ErrNotImplemented +} + +func (h *Handle) RouteList(link Link, family int) ([]Route, error) { + return nil, ErrNotImplemented +} + +func (h *Handle) RouteListFiltered(family int, filter *Route, filterMask uint64) ([]Route, error) { + return nil, ErrNotImplemented +} + +func (h *Handle) RouteReplace(route *Route) error { + return ErrNotImplemented +} + +func (h *Handle) RuleAdd(rule *Rule) error { + return ErrNotImplemented +} + +func (h *Handle) RuleDel(rule *Rule) error { + return ErrNotImplemented +} + +func (h *Handle) RuleList(family int) ([]Rule, error) { + return nil, ErrNotImplemented +} diff --git a/vendor/github.com/vishvananda/netlink/ioctl_linux.go b/vendor/github.com/vishvananda/netlink/ioctl_linux.go new file mode 100644 index 000000000..a8503126d --- /dev/null +++ b/vendor/github.com/vishvananda/netlink/ioctl_linux.go @@ -0,0 +1,98 @@ +package netlink + +import ( + "syscall" + "unsafe" + + "golang.org/x/sys/unix" +) + +// ioctl for statistics. +const ( + // ETHTOOL_GSSET_INFO gets string set info + ETHTOOL_GSSET_INFO = 0x00000037 + // SIOCETHTOOL is Ethtool interface + SIOCETHTOOL = 0x8946 + // ETHTOOL_GSTRINGS gets specified string set + ETHTOOL_GSTRINGS = 0x0000001b + // ETHTOOL_GSTATS gets NIC-specific statistics + ETHTOOL_GSTATS = 0x0000001d +) + +// string set id. +const ( + // ETH_SS_TEST is self-test result names, for use with %ETHTOOL_TEST + ETH_SS_TEST = iota + // ETH_SS_STATS statistic names, for use with %ETHTOOL_GSTATS + ETH_SS_STATS + // ETH_SS_PRIV_FLAGS are driver private flag names + ETH_SS_PRIV_FLAGS + // _ETH_SS_NTUPLE_FILTERS is deprecated + _ETH_SS_NTUPLE_FILTERS + // ETH_SS_FEATURES are device feature names + ETH_SS_FEATURES + // ETH_SS_RSS_HASH_FUNCS is RSS hush function names + ETH_SS_RSS_HASH_FUNCS +) + +// IfreqSlave is a struct for ioctl bond manipulation syscalls. +// It is used to assign slave to bond interface with Name. +type IfreqSlave struct { + Name [unix.IFNAMSIZ]byte + Slave [unix.IFNAMSIZ]byte +} + +// Ifreq is a struct for ioctl ethernet manipulation syscalls. +type Ifreq struct { + Name [unix.IFNAMSIZ]byte + Data uintptr +} + +// ethtoolSset is a string set information +type ethtoolSset struct { + cmd uint32 + reserved uint32 + mask uint64 + 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 +} + +// newIocltSlaveReq returns filled IfreqSlave with proper interface names +// It is used by ioctl to assign slave to bond master +func newIocltSlaveReq(slave, master string) *IfreqSlave { + ifreq := &IfreqSlave{} + copy(ifreq.Name[:unix.IFNAMSIZ-1], master) + copy(ifreq.Slave[:unix.IFNAMSIZ-1], slave) + return ifreq +} + +// newIocltStringSetReq creates request to get interface string set +func newIocltStringSetReq(linkName string) (*Ifreq, *ethtoolSset) { + e := ðtoolSset{ + cmd: ETHTOOL_GSSET_INFO, + mask: 1 << ETH_SS_STATS, + } + + ifreq := &Ifreq{Data: uintptr(unsafe.Pointer(e))} + copy(ifreq.Name[:unix.IFNAMSIZ-1], linkName) + return ifreq, e +} + +// getSocketUDP returns file descriptor to new UDP socket +// It is used for communication with ioctl interface. +func getSocketUDP() (int, error) { + return syscall.Socket(unix.AF_INET, unix.SOCK_DGRAM, 0) +} diff --git a/vendor/github.com/vishvananda/netlink/link.go b/vendor/github.com/vishvananda/netlink/link.go index 4e77037b5..fe74ffab9 100644 --- a/vendor/github.com/vishvananda/netlink/link.go +++ b/vendor/github.com/vishvananda/netlink/link.go @@ -3,6 +3,7 @@ package netlink import ( "fmt" "net" + "os" ) // Link represents a link device from netlink. Shared link attributes @@ -37,6 +38,9 @@ type LinkAttrs struct { EncapType string Protinfo *Protinfo OperState LinkOperState + NetNsID int + NumTxQueues int + NumRxQueues int } // LinkOperState represents the values of the IFLA_OPERSTATE link @@ -258,6 +262,9 @@ const ( type Macvlan struct { LinkAttrs Mode MacvlanMode + + // MACAddrs is only populated for Macvlan SOURCE links + MACAddrs []net.HardwareAddr } func (macvlan *Macvlan) Attrs() *LinkAttrs { @@ -283,8 +290,10 @@ type TuntapFlag uint16 // Tuntap links created via /dev/tun/tap, but can be destroyed via netlink type Tuntap struct { LinkAttrs - Mode TuntapMode - Flags TuntapFlag + Mode TuntapMode + Flags TuntapFlag + Queues int + Fds []*os.File } func (tuntap *Tuntap) Attrs() *LinkAttrs { @@ -326,26 +335,28 @@ func (generic *GenericLink) Type() string { type Vxlan struct { LinkAttrs - VxlanId int - VtepDevIndex int - SrcAddr net.IP - Group net.IP - TTL int - TOS int - Learning bool - Proxy bool - RSC bool - L2miss bool - L3miss bool - UDPCSum bool - NoAge bool - GBP bool - FlowBased bool - Age int - Limit int - Port int - PortLow int - PortHigh int + VxlanId int + VtepDevIndex int + SrcAddr net.IP + Group net.IP + TTL int + TOS int + Learning bool + Proxy bool + RSC bool + L2miss bool + L3miss bool + UDPCSum bool + UDP6ZeroCSumTx bool + UDP6ZeroCSumRx bool + NoAge bool + GBP bool + FlowBased bool + Age int + Limit int + Port int + PortLow int + PortHigh int } func (vxlan *Vxlan) Attrs() *LinkAttrs { @@ -694,17 +705,25 @@ func (gretap *Gretap) Attrs() *LinkAttrs { } func (gretap *Gretap) Type() string { + if gretap.Local.To4() == nil { + return "ip6gretap" + } return "gretap" } type Iptun struct { LinkAttrs - Ttl uint8 - Tos uint8 - PMtuDisc uint8 - Link uint32 - Local net.IP - Remote net.IP + Ttl uint8 + Tos uint8 + PMtuDisc uint8 + Link uint32 + Local net.IP + Remote net.IP + EncapSport uint16 + EncapDport uint16 + EncapType uint16 + EncapFlags uint16 + FlowBased bool } func (iptun *Iptun) Attrs() *LinkAttrs { @@ -715,6 +734,28 @@ func (iptun *Iptun) Type() string { return "ipip" } +type Sittun struct { + LinkAttrs + Link uint32 + Local net.IP + Remote net.IP + Ttl uint8 + Tos uint8 + PMtuDisc uint8 + EncapType uint16 + EncapFlags uint16 + EncapSport uint16 + EncapDport uint16 +} + +func (sittun *Sittun) Attrs() *LinkAttrs { + return &sittun.LinkAttrs +} + +func (sittun *Sittun) Type() string { + return "sit" +} + type Vti struct { LinkAttrs IKey uint32 @@ -732,6 +773,35 @@ func (iptun *Vti) Type() string { return "vti" } +type Gretun struct { + LinkAttrs + Link uint32 + IFlags uint16 + OFlags uint16 + IKey uint32 + OKey uint32 + Local net.IP + Remote net.IP + Ttl uint8 + Tos uint8 + PMtuDisc uint8 + EncapType uint16 + EncapFlags uint16 + EncapSport uint16 + EncapDport uint16 +} + +func (gretun *Gretun) Attrs() *LinkAttrs { + return &gretun.LinkAttrs +} + +func (gretun *Gretun) Type() string { + if gretun.Local.To4() == nil { + return "ip6gre" + } + return "gre" +} + type Vrf struct { LinkAttrs Table uint32 diff --git a/vendor/github.com/vishvananda/netlink/link_linux.go b/vendor/github.com/vishvananda/netlink/link_linux.go index 18a4d1a5f..540191ed8 100644 --- a/vendor/github.com/vishvananda/netlink/link_linux.go +++ b/vendor/github.com/vishvananda/netlink/link_linux.go @@ -11,22 +11,24 @@ import ( "github.com/vishvananda/netlink/nl" "github.com/vishvananda/netns" + "golang.org/x/sys/unix" ) const ( SizeofLinkStats32 = 0x5c SizeofLinkStats64 = 0xd8 - IFLA_STATS64 = 0x17 // syscall pkg does not contain this one ) const ( - TUNTAP_MODE_TUN TuntapMode = syscall.IFF_TUN - TUNTAP_MODE_TAP TuntapMode = syscall.IFF_TAP - TUNTAP_DEFAULTS TuntapFlag = syscall.IFF_TUN_EXCL | syscall.IFF_ONE_QUEUE - TUNTAP_VNET_HDR TuntapFlag = syscall.IFF_VNET_HDR - TUNTAP_TUN_EXCL TuntapFlag = syscall.IFF_TUN_EXCL - TUNTAP_NO_PI TuntapFlag = syscall.IFF_NO_PI - TUNTAP_ONE_QUEUE TuntapFlag = syscall.IFF_ONE_QUEUE + TUNTAP_MODE_TUN TuntapMode = unix.IFF_TUN + TUNTAP_MODE_TAP TuntapMode = unix.IFF_TAP + TUNTAP_DEFAULTS TuntapFlag = unix.IFF_TUN_EXCL | unix.IFF_ONE_QUEUE + TUNTAP_VNET_HDR TuntapFlag = unix.IFF_VNET_HDR + TUNTAP_TUN_EXCL TuntapFlag = unix.IFF_TUN_EXCL + TUNTAP_NO_PI TuntapFlag = unix.IFF_NO_PI + TUNTAP_ONE_QUEUE TuntapFlag = unix.IFF_ONE_QUEUE + TUNTAP_MULTI_QUEUE TuntapFlag = unix.IFF_MULTI_QUEUE + TUNTAP_MULTI_QUEUE_DEFAULTS TuntapFlag = TUNTAP_MULTI_QUEUE | TUNTAP_NO_PI ) var lookupByDump = false @@ -61,15 +63,15 @@ func (h *Handle) ensureIndex(link *LinkAttrs) { func (h *Handle) LinkSetARPOff(link Link) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) - msg.Change |= syscall.IFF_NOARP - msg.Flags |= syscall.IFF_NOARP + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Change |= unix.IFF_NOARP + msg.Flags |= unix.IFF_NOARP msg.Index = int32(base.Index) req.AddData(msg) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -80,15 +82,15 @@ func LinkSetARPOff(link Link) error { func (h *Handle) LinkSetARPOn(link Link) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) - msg.Change |= syscall.IFF_NOARP - msg.Flags &= ^uint32(syscall.IFF_NOARP) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Change |= unix.IFF_NOARP + msg.Flags &= ^uint32(unix.IFF_NOARP) msg.Index = int32(base.Index) req.AddData(msg) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -99,15 +101,84 @@ func LinkSetARPOn(link Link) error { func (h *Handle) SetPromiscOn(link Link) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) - msg.Change = syscall.IFF_PROMISC - msg.Flags = syscall.IFF_PROMISC + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Change = unix.IFF_PROMISC + msg.Flags = unix.IFF_PROMISC msg.Index = int32(base.Index) req.AddData(msg) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) + return err +} + +func MacvlanMACAddrAdd(link Link, addr net.HardwareAddr) error { + return pkgHandle.MacvlanMACAddrAdd(link, addr) +} + +func (h *Handle) MacvlanMACAddrAdd(link Link, addr net.HardwareAddr) error { + return h.macvlanMACAddrChange(link, []net.HardwareAddr{addr}, nl.MACVLAN_MACADDR_ADD) +} + +func MacvlanMACAddrDel(link Link, addr net.HardwareAddr) error { + return pkgHandle.MacvlanMACAddrDel(link, addr) +} + +func (h *Handle) MacvlanMACAddrDel(link Link, addr net.HardwareAddr) error { + return h.macvlanMACAddrChange(link, []net.HardwareAddr{addr}, nl.MACVLAN_MACADDR_DEL) +} + +func MacvlanMACAddrFlush(link Link) error { + return pkgHandle.MacvlanMACAddrFlush(link) +} + +func (h *Handle) MacvlanMACAddrFlush(link Link) error { + return h.macvlanMACAddrChange(link, nil, nl.MACVLAN_MACADDR_FLUSH) +} + +func MacvlanMACAddrSet(link Link, addrs []net.HardwareAddr) error { + return pkgHandle.MacvlanMACAddrSet(link, addrs) +} + +func (h *Handle) MacvlanMACAddrSet(link Link, addrs []net.HardwareAddr) error { + return h.macvlanMACAddrChange(link, addrs, nl.MACVLAN_MACADDR_SET) +} + +func (h *Handle) macvlanMACAddrChange(link Link, addrs []net.HardwareAddr, mode uint32) error { + base := link.Attrs() + h.ensureIndex(base) + req := h.newNetlinkRequest(unix.RTM_NEWLINK, unix.NLM_F_ACK) + + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Index = int32(base.Index) + req.AddData(msg) + + linkInfo := nl.NewRtAttr(unix.IFLA_LINKINFO, nil) + nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_KIND, nl.NonZeroTerminated(link.Type())) + inner := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil) + + // IFLA_MACVLAN_MACADDR_MODE = mode + b := make([]byte, 4) + native.PutUint32(b, mode) + nl.NewRtAttrChild(inner, nl.IFLA_MACVLAN_MACADDR_MODE, b) + + // populate message with MAC addrs, if necessary + switch mode { + case nl.MACVLAN_MACADDR_ADD, nl.MACVLAN_MACADDR_DEL: + if len(addrs) == 1 { + nl.NewRtAttrChild(inner, nl.IFLA_MACVLAN_MACADDR, []byte(addrs[0])) + } + case nl.MACVLAN_MACADDR_SET: + mad := nl.NewRtAttrChild(inner, nl.IFLA_MACVLAN_MACADDR_DATA, nil) + for _, addr := range addrs { + nl.NewRtAttrChild(mad, nl.IFLA_MACVLAN_MACADDR, []byte(addr)) + } + } + + req.AddData(linkInfo) + + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -118,7 +189,7 @@ func BridgeSetMcastSnoop(link Link, on bool) error { func (h *Handle) BridgeSetMcastSnoop(link Link, on bool) error { bridge := link.(*Bridge) bridge.MulticastSnooping = &on - return h.linkModify(bridge, syscall.NLM_F_ACK) + return h.linkModify(bridge, unix.NLM_F_ACK) } func SetPromiscOn(link Link) error { @@ -128,15 +199,15 @@ func SetPromiscOn(link Link) error { func (h *Handle) SetPromiscOff(link Link) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) - msg.Change = syscall.IFF_PROMISC - msg.Flags = 0 & ^syscall.IFF_PROMISC + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Change = unix.IFF_PROMISC + msg.Flags = 0 & ^unix.IFF_PROMISC msg.Index = int32(base.Index) req.AddData(msg) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -155,15 +226,15 @@ func LinkSetUp(link Link) error { func (h *Handle) LinkSetUp(link Link) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_NEWLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_NEWLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) - msg.Change = syscall.IFF_UP - msg.Flags = syscall.IFF_UP + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Change = unix.IFF_UP + msg.Flags = unix.IFF_UP msg.Index = int32(base.Index) req.AddData(msg) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -178,15 +249,15 @@ func LinkSetDown(link Link) error { func (h *Handle) LinkSetDown(link Link) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_NEWLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_NEWLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) - msg.Change = syscall.IFF_UP - msg.Flags = 0 & ^syscall.IFF_UP + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Change = unix.IFF_UP + msg.Flags = 0 & ^unix.IFF_UP msg.Index = int32(base.Index) req.AddData(msg) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -201,19 +272,19 @@ func LinkSetMTU(link Link, mtu int) error { func (h *Handle) LinkSetMTU(link Link, mtu int) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) b := make([]byte, 4) native.PutUint32(b, uint32(mtu)) - data := nl.NewRtAttr(syscall.IFLA_MTU, b) + data := nl.NewRtAttr(unix.IFLA_MTU, b) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -228,16 +299,16 @@ func LinkSetName(link Link, name string) error { func (h *Handle) LinkSetName(link Link, name string) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) - data := nl.NewRtAttr(syscall.IFLA_IFNAME, []byte(name)) + data := nl.NewRtAttr(unix.IFLA_IFNAME, []byte(name)) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -252,16 +323,16 @@ func LinkSetAlias(link Link, name string) error { func (h *Handle) LinkSetAlias(link Link, name string) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) - data := nl.NewRtAttr(syscall.IFLA_IFALIAS, []byte(name)) + data := nl.NewRtAttr(unix.IFLA_IFALIAS, []byte(name)) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -276,16 +347,16 @@ func LinkSetHardwareAddr(link Link, hwaddr net.HardwareAddr) error { func (h *Handle) LinkSetHardwareAddr(link Link, hwaddr net.HardwareAddr) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) - data := nl.NewRtAttr(syscall.IFLA_ADDRESS, []byte(hwaddr)) + data := nl.NewRtAttr(unix.IFLA_ADDRESS, []byte(hwaddr)) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -300,13 +371,13 @@ func LinkSetVfHardwareAddr(link Link, vf int, hwaddr net.HardwareAddr) error { func (h *Handle) LinkSetVfHardwareAddr(link Link, vf int, hwaddr net.HardwareAddr) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) - data := nl.NewRtAttr(nl.IFLA_VFINFO_LIST, nil) + data := nl.NewRtAttr(unix.IFLA_VFINFO_LIST, nil) info := nl.NewRtAttrChild(data, nl.IFLA_VF_INFO, nil) vfmsg := nl.VfMac{ Vf: uint32(vf), @@ -315,7 +386,7 @@ func (h *Handle) LinkSetVfHardwareAddr(link Link, vf int, hwaddr net.HardwareAdd nl.NewRtAttrChild(info, nl.IFLA_VF_MAC, vfmsg.Serialize()) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -330,13 +401,13 @@ func LinkSetVfVlan(link Link, vf, vlan int) error { func (h *Handle) LinkSetVfVlan(link Link, vf, vlan int) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) - data := nl.NewRtAttr(nl.IFLA_VFINFO_LIST, nil) + data := nl.NewRtAttr(unix.IFLA_VFINFO_LIST, nil) info := nl.NewRtAttrChild(data, nl.IFLA_VF_INFO, nil) vfmsg := nl.VfVlan{ Vf: uint32(vf), @@ -345,7 +416,7 @@ func (h *Handle) LinkSetVfVlan(link Link, vf, vlan int) error { nl.NewRtAttrChild(info, nl.IFLA_VF_VLAN, vfmsg.Serialize()) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -360,13 +431,13 @@ func LinkSetVfTxRate(link Link, vf, rate int) error { func (h *Handle) LinkSetVfTxRate(link Link, vf, rate int) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) - data := nl.NewRtAttr(nl.IFLA_VFINFO_LIST, nil) + data := nl.NewRtAttr(unix.IFLA_VFINFO_LIST, nil) info := nl.NewRtAttrChild(data, nl.IFLA_VF_INFO, nil) vfmsg := nl.VfTxRate{ Vf: uint32(vf), @@ -375,7 +446,75 @@ func (h *Handle) LinkSetVfTxRate(link Link, vf, rate int) error { nl.NewRtAttrChild(info, nl.IFLA_VF_TX_RATE, vfmsg.Serialize()) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) + return err +} + +// LinkSetVfSpoofchk enables/disables spoof check on a vf for the link. +// Equivalent to: `ip link set $link vf $vf spoofchk $check` +func LinkSetVfSpoofchk(link Link, vf int, check bool) error { + return pkgHandle.LinkSetVfSpoofchk(link, vf, check) +} + +// LinkSetVfSpookfchk enables/disables spoof check on a vf for the link. +// Equivalent to: `ip link set $link vf $vf spoofchk $check` +func (h *Handle) LinkSetVfSpoofchk(link Link, vf int, check bool) error { + var setting uint32 + base := link.Attrs() + h.ensureIndex(base) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) + + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Index = int32(base.Index) + req.AddData(msg) + + data := nl.NewRtAttr(unix.IFLA_VFINFO_LIST, nil) + info := nl.NewRtAttrChild(data, nl.IFLA_VF_INFO, nil) + if check { + setting = 1 + } + vfmsg := nl.VfSpoofchk{ + Vf: uint32(vf), + Setting: setting, + } + nl.NewRtAttrChild(info, nl.IFLA_VF_SPOOFCHK, vfmsg.Serialize()) + req.AddData(data) + + _, err := req.Execute(unix.NETLINK_ROUTE, 0) + return err +} + +// LinkSetVfTrust enables/disables trust state on a vf for the link. +// Equivalent to: `ip link set $link vf $vf trust $state` +func LinkSetVfTrust(link Link, vf int, state bool) error { + return pkgHandle.LinkSetVfTrust(link, vf, state) +} + +// LinkSetVfTrust enables/disables trust state on a vf for the link. +// Equivalent to: `ip link set $link vf $vf trust $state` +func (h *Handle) LinkSetVfTrust(link Link, vf int, state bool) error { + var setting uint32 + base := link.Attrs() + h.ensureIndex(base) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) + + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Index = int32(base.Index) + req.AddData(msg) + + data := nl.NewRtAttr(unix.IFLA_VFINFO_LIST, nil) + info := nl.NewRtAttrChild(data, nl.IFLA_VF_INFO, nil) + if state { + setting = 1 + } + vfmsg := nl.VfTrust{ + Vf: uint32(vf), + Setting: setting, + } + nl.NewRtAttrChild(info, nl.IFLA_VF_TRUST, vfmsg.Serialize()) + req.AddData(data) + + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -423,19 +562,19 @@ func LinkSetMasterByIndex(link Link, masterIndex int) error { func (h *Handle) LinkSetMasterByIndex(link Link, masterIndex int) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) b := make([]byte, 4) native.PutUint32(b, uint32(masterIndex)) - data := nl.NewRtAttr(syscall.IFLA_MASTER, b) + data := nl.NewRtAttr(unix.IFLA_MASTER, b) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -452,19 +591,19 @@ func LinkSetNsPid(link Link, nspid int) error { func (h *Handle) LinkSetNsPid(link Link, nspid int) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) b := make([]byte, 4) native.PutUint32(b, uint32(nspid)) - data := nl.NewRtAttr(syscall.IFLA_NET_NS_PID, b) + data := nl.NewRtAttr(unix.IFLA_NET_NS_PID, b) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -481,36 +620,42 @@ func LinkSetNsFd(link Link, fd int) error { func (h *Handle) LinkSetNsFd(link Link, fd int) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) b := make([]byte, 4) native.PutUint32(b, uint32(fd)) - data := nl.NewRtAttr(nl.IFLA_NET_NS_FD, b) + data := nl.NewRtAttr(unix.IFLA_NET_NS_FD, b) req.AddData(data) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } // LinkSetXdpFd adds a bpf function to the driver. The fd must be a bpf // program loaded with bpf(type=BPF_PROG_TYPE_XDP) func LinkSetXdpFd(link Link, fd int) error { + return LinkSetXdpFdWithFlags(link, fd, 0) +} + +// LinkSetXdpFdWithFlags adds a bpf function to the driver with the given +// options. The fd must be a bpf program loaded with bpf(type=BPF_PROG_TYPE_XDP) +func LinkSetXdpFdWithFlags(link Link, fd, flags int) error { base := link.Attrs() ensureIndex(base) - req := nl.NewNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := nl.NewNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) - addXdpAttrs(&LinkXdp{Fd: fd}, req) + addXdpAttrs(&LinkXdp{Fd: fd, Flags: uint32(flags)}, req) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -568,6 +713,8 @@ func addVxlanAttrs(vxlan *Vxlan, linkInfo *nl.RtAttr) { nl.NewRtAttrChild(data, nl.IFLA_VXLAN_RSC, boolAttr(vxlan.RSC)) nl.NewRtAttrChild(data, nl.IFLA_VXLAN_L2MISS, boolAttr(vxlan.L2miss)) nl.NewRtAttrChild(data, nl.IFLA_VXLAN_L3MISS, boolAttr(vxlan.L3miss)) + nl.NewRtAttrChild(data, nl.IFLA_VXLAN_UDP_ZERO_CSUM6_TX, boolAttr(vxlan.UDP6ZeroCSumTx)) + nl.NewRtAttrChild(data, nl.IFLA_VXLAN_UDP_ZERO_CSUM6_RX, boolAttr(vxlan.UDP6ZeroCSumRx)) if vxlan.UDPCSum { nl.NewRtAttrChild(data, nl.IFLA_VXLAN_UDP_CSUM, boolAttr(vxlan.UDPCSum)) @@ -692,6 +839,12 @@ func addBondAttrs(bond *Bond, linkInfo *nl.RtAttr) { } } +func cleanupFds(fds []*os.File) { + for _, f := range fds { + f.Close() + } +} + // LinkAdd adds a new link device. The type and features of the device // are taken from the parameters in the link object. // Equivalent to: `ip link add $link` @@ -703,7 +856,7 @@ func LinkAdd(link Link) error { // are taken fromt the parameters in the link object. // Equivalent to: `ip link add $link` func (h *Handle) LinkAdd(link Link) error { - return h.linkModify(link, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + return h.linkModify(link, unix.NLM_F_CREATE|unix.NLM_F_EXCL|unix.NLM_F_ACK) } func (h *Handle) linkModify(link Link, flags int) error { @@ -717,104 +870,155 @@ func (h *Handle) linkModify(link Link, flags int) error { if tuntap, ok := link.(*Tuntap); ok { // TODO: support user // TODO: support group - // TODO: multi_queue // TODO: support non- persistent - if tuntap.Mode < syscall.IFF_TUN || tuntap.Mode > syscall.IFF_TAP { + if tuntap.Mode < unix.IFF_TUN || tuntap.Mode > unix.IFF_TAP { return fmt.Errorf("Tuntap.Mode %v unknown!", tuntap.Mode) } - file, err := os.OpenFile("/dev/net/tun", os.O_RDWR, 0) - if err != nil { - return err - } - defer file.Close() + + queues := tuntap.Queues + + var fds []*os.File var req ifReq - if tuntap.Flags == 0 { - req.Flags = uint16(TUNTAP_DEFAULTS) + copy(req.Name[:15], base.Name) + + req.Flags = uint16(tuntap.Flags) + + if queues == 0 { //Legacy compatibility + queues = 1 + if tuntap.Flags == 0 { + req.Flags = uint16(TUNTAP_DEFAULTS) + } } else { - req.Flags = uint16(tuntap.Flags) + // For best peformance set Flags to TUNTAP_MULTI_QUEUE_DEFAULTS | TUNTAP_VNET_HDR + // when a) KVM has support for this ABI and + // b) the value of the flag is queryable using the TUNGETIFF ioctl + if tuntap.Flags == 0 { + req.Flags = uint16(TUNTAP_MULTI_QUEUE_DEFAULTS) + } } + req.Flags |= uint16(tuntap.Mode) - copy(req.Name[:15], base.Name) - _, _, errno := syscall.Syscall(syscall.SYS_IOCTL, file.Fd(), uintptr(syscall.TUNSETIFF), uintptr(unsafe.Pointer(&req))) - if errno != 0 { - return fmt.Errorf("Tuntap IOCTL TUNSETIFF failed, errno %v", errno) + + for i := 0; i < queues; i++ { + localReq := req + file, err := os.OpenFile("/dev/net/tun", os.O_RDWR, 0) + if err != nil { + cleanupFds(fds) + return err + } + + fds = append(fds, file) + _, _, errno := unix.Syscall(unix.SYS_IOCTL, file.Fd(), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&localReq))) + if errno != 0 { + cleanupFds(fds) + return fmt.Errorf("Tuntap IOCTL TUNSETIFF failed [%d], errno %v", i, errno) + } } - _, _, errno = syscall.Syscall(syscall.SYS_IOCTL, file.Fd(), uintptr(syscall.TUNSETPERSIST), 1) + + _, _, errno := unix.Syscall(unix.SYS_IOCTL, fds[0].Fd(), uintptr(unix.TUNSETPERSIST), 1) if errno != 0 { + cleanupFds(fds) return fmt.Errorf("Tuntap IOCTL TUNSETPERSIST failed, errno %v", errno) } + h.ensureIndex(base) // can't set master during create, so set it afterwards if base.MasterIndex != 0 { // TODO: verify MasterIndex is actually a bridge? - return h.LinkSetMasterByIndex(link, base.MasterIndex) + err := h.LinkSetMasterByIndex(link, base.MasterIndex) + if err != nil { + _, _, _ = unix.Syscall(unix.SYS_IOCTL, fds[0].Fd(), uintptr(unix.TUNSETPERSIST), 0) + cleanupFds(fds) + return err + } + } + + if tuntap.Queues == 0 { + cleanupFds(fds) + } else { + tuntap.Fds = fds } + return nil } - req := h.newNetlinkRequest(syscall.RTM_NEWLINK, flags) + req := h.newNetlinkRequest(unix.RTM_NEWLINK, flags) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) // TODO: make it shorter if base.Flags&net.FlagUp != 0 { - msg.Change = syscall.IFF_UP - msg.Flags = syscall.IFF_UP + msg.Change = unix.IFF_UP + msg.Flags = unix.IFF_UP } if base.Flags&net.FlagBroadcast != 0 { - msg.Change |= syscall.IFF_BROADCAST - msg.Flags |= syscall.IFF_BROADCAST + msg.Change |= unix.IFF_BROADCAST + msg.Flags |= unix.IFF_BROADCAST } if base.Flags&net.FlagLoopback != 0 { - msg.Change |= syscall.IFF_LOOPBACK - msg.Flags |= syscall.IFF_LOOPBACK + msg.Change |= unix.IFF_LOOPBACK + msg.Flags |= unix.IFF_LOOPBACK } if base.Flags&net.FlagPointToPoint != 0 { - msg.Change |= syscall.IFF_POINTOPOINT - msg.Flags |= syscall.IFF_POINTOPOINT + msg.Change |= unix.IFF_POINTOPOINT + msg.Flags |= unix.IFF_POINTOPOINT } if base.Flags&net.FlagMulticast != 0 { - msg.Change |= syscall.IFF_MULTICAST - msg.Flags |= syscall.IFF_MULTICAST + msg.Change |= unix.IFF_MULTICAST + msg.Flags |= unix.IFF_MULTICAST + } + if base.Index != 0 { + msg.Index = int32(base.Index) } + req.AddData(msg) if base.ParentIndex != 0 { b := make([]byte, 4) native.PutUint32(b, uint32(base.ParentIndex)) - data := nl.NewRtAttr(syscall.IFLA_LINK, b) + data := nl.NewRtAttr(unix.IFLA_LINK, b) req.AddData(data) } else if link.Type() == "ipvlan" { return fmt.Errorf("Can't create ipvlan link without ParentIndex") } - nameData := nl.NewRtAttr(syscall.IFLA_IFNAME, nl.ZeroTerminated(base.Name)) + nameData := nl.NewRtAttr(unix.IFLA_IFNAME, nl.ZeroTerminated(base.Name)) req.AddData(nameData) if base.MTU > 0 { - mtu := nl.NewRtAttr(syscall.IFLA_MTU, nl.Uint32Attr(uint32(base.MTU))) + mtu := nl.NewRtAttr(unix.IFLA_MTU, nl.Uint32Attr(uint32(base.MTU))) req.AddData(mtu) } if base.TxQLen >= 0 { - qlen := nl.NewRtAttr(syscall.IFLA_TXQLEN, nl.Uint32Attr(uint32(base.TxQLen))) + qlen := nl.NewRtAttr(unix.IFLA_TXQLEN, nl.Uint32Attr(uint32(base.TxQLen))) req.AddData(qlen) } if base.HardwareAddr != nil { - hwaddr := nl.NewRtAttr(syscall.IFLA_ADDRESS, []byte(base.HardwareAddr)) + hwaddr := nl.NewRtAttr(unix.IFLA_ADDRESS, []byte(base.HardwareAddr)) req.AddData(hwaddr) } + if base.NumTxQueues > 0 { + txqueues := nl.NewRtAttr(unix.IFLA_NUM_TX_QUEUES, nl.Uint32Attr(uint32(base.NumTxQueues))) + req.AddData(txqueues) + } + + if base.NumRxQueues > 0 { + rxqueues := nl.NewRtAttr(unix.IFLA_NUM_RX_QUEUES, nl.Uint32Attr(uint32(base.NumRxQueues))) + req.AddData(rxqueues) + } + if base.Namespace != nil { var attr *nl.RtAttr switch base.Namespace.(type) { case NsPid: val := nl.Uint32Attr(uint32(base.Namespace.(NsPid))) - attr = nl.NewRtAttr(syscall.IFLA_NET_NS_PID, val) + attr = nl.NewRtAttr(unix.IFLA_NET_NS_PID, val) case NsFd: val := nl.Uint32Attr(uint32(base.Namespace.(NsFd))) - attr = nl.NewRtAttr(nl.IFLA_NET_NS_FD, val) + attr = nl.NewRtAttr(unix.IFLA_NET_NS_FD, val) } req.AddData(attr) @@ -824,60 +1028,65 @@ func (h *Handle) linkModify(link Link, flags int) error { addXdpAttrs(base.Xdp, req) } - linkInfo := nl.NewRtAttr(syscall.IFLA_LINKINFO, nil) + linkInfo := nl.NewRtAttr(unix.IFLA_LINKINFO, nil) nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_KIND, nl.NonZeroTerminated(link.Type())) - if vlan, ok := link.(*Vlan); ok { + switch link := link.(type) { + case *Vlan: b := make([]byte, 2) - native.PutUint16(b, uint16(vlan.VlanId)) + native.PutUint16(b, uint16(link.VlanId)) data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil) nl.NewRtAttrChild(data, nl.IFLA_VLAN_ID, b) - } else if veth, ok := link.(*Veth); ok { + case *Veth: data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil) peer := nl.NewRtAttrChild(data, nl.VETH_INFO_PEER, nil) - nl.NewIfInfomsgChild(peer, syscall.AF_UNSPEC) - nl.NewRtAttrChild(peer, syscall.IFLA_IFNAME, nl.ZeroTerminated(veth.PeerName)) + nl.NewIfInfomsgChild(peer, unix.AF_UNSPEC) + nl.NewRtAttrChild(peer, unix.IFLA_IFNAME, nl.ZeroTerminated(link.PeerName)) if base.TxQLen >= 0 { - nl.NewRtAttrChild(peer, syscall.IFLA_TXQLEN, nl.Uint32Attr(uint32(base.TxQLen))) + nl.NewRtAttrChild(peer, unix.IFLA_TXQLEN, nl.Uint32Attr(uint32(base.TxQLen))) } if base.MTU > 0 { - nl.NewRtAttrChild(peer, syscall.IFLA_MTU, nl.Uint32Attr(uint32(base.MTU))) + nl.NewRtAttrChild(peer, unix.IFLA_MTU, nl.Uint32Attr(uint32(base.MTU))) } - } else if vxlan, ok := link.(*Vxlan); ok { - addVxlanAttrs(vxlan, linkInfo) - } else if bond, ok := link.(*Bond); ok { - addBondAttrs(bond, linkInfo) - } else if ipv, ok := link.(*IPVlan); ok { + case *Vxlan: + addVxlanAttrs(link, linkInfo) + case *Bond: + addBondAttrs(link, linkInfo) + case *IPVlan: data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil) - nl.NewRtAttrChild(data, nl.IFLA_IPVLAN_MODE, nl.Uint16Attr(uint16(ipv.Mode))) - } else if macv, ok := link.(*Macvlan); ok { - if macv.Mode != MACVLAN_MODE_DEFAULT { + nl.NewRtAttrChild(data, nl.IFLA_IPVLAN_MODE, nl.Uint16Attr(uint16(link.Mode))) + case *Macvlan: + if link.Mode != MACVLAN_MODE_DEFAULT { data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil) - nl.NewRtAttrChild(data, nl.IFLA_MACVLAN_MODE, nl.Uint32Attr(macvlanModes[macv.Mode])) + nl.NewRtAttrChild(data, nl.IFLA_MACVLAN_MODE, nl.Uint32Attr(macvlanModes[link.Mode])) } - } else if macv, ok := link.(*Macvtap); ok { - if macv.Mode != MACVLAN_MODE_DEFAULT { + case *Macvtap: + if link.Mode != MACVLAN_MODE_DEFAULT { data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil) - nl.NewRtAttrChild(data, nl.IFLA_MACVLAN_MODE, nl.Uint32Attr(macvlanModes[macv.Mode])) + nl.NewRtAttrChild(data, nl.IFLA_MACVLAN_MODE, nl.Uint32Attr(macvlanModes[link.Mode])) } - } else if gretap, ok := link.(*Gretap); ok { - addGretapAttrs(gretap, linkInfo) - } else if iptun, ok := link.(*Iptun); ok { - addIptunAttrs(iptun, linkInfo) - } else if vti, ok := link.(*Vti); ok { - addVtiAttrs(vti, linkInfo) - } else if vrf, ok := link.(*Vrf); ok { - addVrfAttrs(vrf, linkInfo) - } else if bridge, ok := link.(*Bridge); ok { - addBridgeAttrs(bridge, linkInfo) - } else if gtp, ok := link.(*GTP); ok { - addGTPAttrs(gtp, linkInfo) + case *Gretap: + addGretapAttrs(link, linkInfo) + case *Iptun: + addIptunAttrs(link, linkInfo) + case *Sittun: + addSittunAttrs(link, linkInfo) + case *Gretun: + addGretunAttrs(link, linkInfo) + case *Vti: + addVtiAttrs(link, linkInfo) + case *Vrf: + addVrfAttrs(link, linkInfo) + case *Bridge: + addBridgeAttrs(link, linkInfo) + case *GTP: + addGTPAttrs(link, linkInfo) } req.AddData(linkInfo) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) if err != nil { return err } @@ -907,13 +1116,13 @@ func (h *Handle) LinkDel(link Link) error { h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_DELLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_DELLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(base.Index) req.AddData(msg) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -956,16 +1165,16 @@ func (h *Handle) LinkByName(name string) (Link, error) { return h.linkByNameDump(name) } - req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_GETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) req.AddData(msg) - nameData := nl.NewRtAttr(syscall.IFLA_IFNAME, nl.ZeroTerminated(name)) + nameData := nl.NewRtAttr(unix.IFLA_IFNAME, nl.ZeroTerminated(name)) req.AddData(nameData) link, err := execGetLink(req) - if err == syscall.EINVAL { + if err == unix.EINVAL { // older kernels don't support looking up via IFLA_IFNAME // so fall back to dumping all links h.lookupByDump = true @@ -988,16 +1197,16 @@ func (h *Handle) LinkByAlias(alias string) (Link, error) { return h.linkByAliasDump(alias) } - req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_GETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) req.AddData(msg) - nameData := nl.NewRtAttr(syscall.IFLA_IFALIAS, nl.ZeroTerminated(alias)) + nameData := nl.NewRtAttr(unix.IFLA_IFALIAS, nl.ZeroTerminated(alias)) req.AddData(nameData) link, err := execGetLink(req) - if err == syscall.EINVAL { + if err == unix.EINVAL { // older kernels don't support looking up via IFLA_IFALIAS // so fall back to dumping all links h.lookupByDump = true @@ -1014,9 +1223,9 @@ func LinkByIndex(index int) (Link, error) { // LinkByIndex finds a link by index and returns a pointer to the object. func (h *Handle) LinkByIndex(index int) (Link, error) { - req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_GETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) msg.Index = int32(index) req.AddData(msg) @@ -1024,10 +1233,10 @@ func (h *Handle) LinkByIndex(index int) (Link, error) { } func execGetLink(req *nl.NetlinkRequest) (Link, error) { - msgs, err := req.Execute(syscall.NETLINK_ROUTE, 0) + msgs, err := req.Execute(unix.NETLINK_ROUTE, 0) if err != nil { if errno, ok := err.(syscall.Errno); ok { - if errno == syscall.ENODEV { + if errno == unix.ENODEV { return nil, LinkNotFoundError{fmt.Errorf("Link not found")} } } @@ -1048,7 +1257,7 @@ func execGetLink(req *nl.NetlinkRequest) (Link, error) { // linkDeserialize deserializes a raw message received from netlink into // a link object. -func LinkDeserialize(hdr *syscall.NlMsghdr, m []byte) (Link, error) { +func LinkDeserialize(hdr *unix.NlMsghdr, m []byte) (Link, error) { msg := nl.DeserializeIfInfomsg(m) attrs, err := nl.ParseRouteAttr(m[msg.Len():]) @@ -1057,7 +1266,7 @@ func LinkDeserialize(hdr *syscall.NlMsghdr, m []byte) (Link, error) { } base := LinkAttrs{Index: int(msg.Index), RawFlags: msg.Flags, Flags: linkFlags(msg.Flags), EncapType: msg.EncapType()} - if msg.Flags&syscall.IFF_PROMISC != 0 { + if msg.Flags&unix.IFF_PROMISC != 0 { base.Promisc = 1 } var ( @@ -1068,7 +1277,7 @@ func LinkDeserialize(hdr *syscall.NlMsghdr, m []byte) (Link, error) { ) for _, attr := range attrs { switch attr.Attr.Type { - case syscall.IFLA_LINKINFO: + case unix.IFLA_LINKINFO: infos, err := nl.ParseRouteAttr(attr.Value) if err != nil { return nil, err @@ -1100,8 +1309,16 @@ func LinkDeserialize(hdr *syscall.NlMsghdr, m []byte) (Link, error) { link = &Macvtap{} case "gretap": link = &Gretap{} + case "ip6gretap": + link = &Gretap{} case "ipip": link = &Iptun{} + case "sit": + link = &Sittun{} + case "gre": + link = &Gretun{} + case "ip6gre": + link = &Gretun{} case "vti": link = &Vti{} case "vrf": @@ -1131,8 +1348,16 @@ func LinkDeserialize(hdr *syscall.NlMsghdr, m []byte) (Link, error) { parseMacvtapData(link, data) case "gretap": parseGretapData(link, data) + case "ip6gretap": + parseGretapData(link, data) case "ipip": parseIptunData(link, data) + case "sit": + parseSittunData(link, data) + case "gre": + parseGretunData(link, data) + case "ip6gre": + parseGretunData(link, data) case "vti": parseVtiData(link, data) case "vrf": @@ -1144,7 +1369,7 @@ func LinkDeserialize(hdr *syscall.NlMsghdr, m []byte) (Link, error) { } } } - case syscall.IFLA_ADDRESS: + case unix.IFLA_ADDRESS: var nonzero bool for _, b := range attr.Value { if b != 0 { @@ -1154,39 +1379,41 @@ func LinkDeserialize(hdr *syscall.NlMsghdr, m []byte) (Link, error) { if nonzero { base.HardwareAddr = attr.Value[:] } - case syscall.IFLA_IFNAME: + case unix.IFLA_IFNAME: base.Name = string(attr.Value[:len(attr.Value)-1]) - case syscall.IFLA_MTU: + case unix.IFLA_MTU: base.MTU = int(native.Uint32(attr.Value[0:4])) - case syscall.IFLA_LINK: + case unix.IFLA_LINK: base.ParentIndex = int(native.Uint32(attr.Value[0:4])) - case syscall.IFLA_MASTER: + case unix.IFLA_MASTER: base.MasterIndex = int(native.Uint32(attr.Value[0:4])) - case syscall.IFLA_TXQLEN: + case unix.IFLA_TXQLEN: base.TxQLen = int(native.Uint32(attr.Value[0:4])) - case syscall.IFLA_IFALIAS: + case unix.IFLA_IFALIAS: base.Alias = string(attr.Value[:len(attr.Value)-1]) - case syscall.IFLA_STATS: + case unix.IFLA_STATS: stats32 = attr.Value[:] - case IFLA_STATS64: + case unix.IFLA_STATS64: stats64 = attr.Value[:] - case nl.IFLA_XDP: + case unix.IFLA_XDP: xdp, err := parseLinkXdp(attr.Value[:]) if err != nil { return nil, err } base.Xdp = xdp - case syscall.IFLA_PROTINFO | syscall.NLA_F_NESTED: - if hdr != nil && hdr.Type == syscall.RTM_NEWLINK && - msg.Family == syscall.AF_BRIDGE { + case unix.IFLA_PROTINFO | unix.NLA_F_NESTED: + if hdr != nil && hdr.Type == unix.RTM_NEWLINK && + msg.Family == unix.AF_BRIDGE { attrs, err := nl.ParseRouteAttr(attr.Value[:]) if err != nil { return nil, err } base.Protinfo = parseProtinfo(attrs) } - case syscall.IFLA_OPERSTATE: + case unix.IFLA_OPERSTATE: base.OperState = LinkOperState(uint8(attr.Value[0])) + case unix.IFLA_LINK_NETNSID: + base.NetNsID = int(native.Uint32(attr.Value[0:4])) } } @@ -1216,12 +1443,12 @@ func LinkList() ([]Link, error) { func (h *Handle) LinkList() ([]Link, error) { // NOTE(vish): This duplicates functionality in net/iface_linux.go, but we need // to get the message ourselves to parse link type. - req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(unix.RTM_GETLINK, unix.NLM_F_DUMP) - msg := nl.NewIfInfomsg(syscall.AF_UNSPEC) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWLINK) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWLINK) if err != nil { return nil, err } @@ -1241,24 +1468,43 @@ func (h *Handle) LinkList() ([]Link, error) { // LinkUpdate is used to pass information back from LinkSubscribe() type LinkUpdate struct { nl.IfInfomsg - Header syscall.NlMsghdr + Header unix.NlMsghdr Link } // LinkSubscribe takes a chan down which notifications will be sent // when links change. Close the 'done' chan to stop subscription. func LinkSubscribe(ch chan<- LinkUpdate, done <-chan struct{}) error { - return linkSubscribe(netns.None(), netns.None(), ch, done) + return linkSubscribeAt(netns.None(), netns.None(), ch, done, nil, false) } // LinkSubscribeAt works like LinkSubscribe plus it allows the caller // to choose the network namespace in which to subscribe (ns). func LinkSubscribeAt(ns netns.NsHandle, ch chan<- LinkUpdate, done <-chan struct{}) error { - return linkSubscribe(ns, netns.None(), ch, done) + return linkSubscribeAt(ns, netns.None(), ch, done, nil, false) +} + +// LinkSubscribeOptions contains a set of options to use with +// LinkSubscribeWithOptions. +type LinkSubscribeOptions struct { + Namespace *netns.NsHandle + ErrorCallback func(error) + ListExisting bool +} + +// LinkSubscribeWithOptions work like LinkSubscribe but enable to +// provide additional options to modify the behavior. Currently, the +// namespace can be provided as well as an error callback. +func LinkSubscribeWithOptions(ch chan<- LinkUpdate, done <-chan struct{}, options LinkSubscribeOptions) error { + if options.Namespace == nil { + none := netns.None() + options.Namespace = &none + } + return linkSubscribeAt(*options.Namespace, netns.None(), ch, done, options.ErrorCallback, options.ListExisting) } -func linkSubscribe(newNs, curNs netns.NsHandle, ch chan<- LinkUpdate, done <-chan struct{}) error { - s, err := nl.SubscribeAt(newNs, curNs, syscall.NETLINK_ROUTE, syscall.RTNLGRP_LINK) +func linkSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- LinkUpdate, done <-chan struct{}, cberr func(error), listExisting bool) error { + s, err := nl.SubscribeAt(newNs, curNs, unix.NETLINK_ROUTE, unix.RTNLGRP_LINK) if err != nil { return err } @@ -1268,20 +1514,50 @@ func linkSubscribe(newNs, curNs netns.NsHandle, ch chan<- LinkUpdate, done <-cha s.Close() }() } + if listExisting { + req := pkgHandle.newNetlinkRequest(unix.RTM_GETLINK, + unix.NLM_F_DUMP) + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + req.AddData(msg) + if err := s.Send(req); err != nil { + return err + } + } go func() { defer close(ch) for { msgs, err := s.Receive() if err != nil { + if cberr != nil { + cberr(err) + } return } for _, m := range msgs { + if m.Header.Type == unix.NLMSG_DONE { + 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 + } ifmsg := nl.DeserializeIfInfomsg(m.Data) - link, err := LinkDeserialize(&m.Header, m.Data) + header := unix.NlMsghdr(m.Header) + link, err := LinkDeserialize(&header, m.Data) if err != nil { + if cberr != nil { + cberr(err) + } return } - ch <- LinkUpdate{IfInfomsg: *ifmsg, Header: m.Header, Link: link} + ch <- LinkUpdate{IfInfomsg: *ifmsg, Header: header, Link: link} } } }() @@ -1356,22 +1632,49 @@ func (h *Handle) LinkSetBrProxyArpWiFi(link Link, mode bool) error { func (h *Handle) setProtinfoAttr(link Link, mode bool, attr int) error { base := link.Attrs() h.ensureIndex(base) - req := h.newNetlinkRequest(syscall.RTM_SETLINK, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) - msg := nl.NewIfInfomsg(syscall.AF_BRIDGE) + msg := nl.NewIfInfomsg(unix.AF_BRIDGE) msg.Index = int32(base.Index) req.AddData(msg) - br := nl.NewRtAttr(syscall.IFLA_PROTINFO|syscall.NLA_F_NESTED, nil) + br := nl.NewRtAttr(unix.IFLA_PROTINFO|unix.NLA_F_NESTED, nil) nl.NewRtAttrChild(br, attr, boolToByte(mode)) req.AddData(br) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) if err != nil { return err } return nil } +// LinkSetTxQLen sets the transaction queue length for the link. +// Equivalent to: `ip link set $link txqlen $qlen` +func LinkSetTxQLen(link Link, qlen int) error { + return pkgHandle.LinkSetTxQLen(link, qlen) +} + +// LinkSetTxQLen sets the transaction queue length for the link. +// Equivalent to: `ip link set $link txqlen $qlen` +func (h *Handle) LinkSetTxQLen(link Link, qlen int) error { + base := link.Attrs() + h.ensureIndex(base) + req := h.newNetlinkRequest(unix.RTM_SETLINK, unix.NLM_F_ACK) + + msg := nl.NewIfInfomsg(unix.AF_UNSPEC) + msg.Index = int32(base.Index) + req.AddData(msg) + + b := make([]byte, 4) + native.PutUint32(b, uint32(qlen)) + + data := nl.NewRtAttr(unix.IFLA_TXQLEN, b) + req.AddData(data) + + _, err := req.Execute(unix.NETLINK_ROUTE, 0) + return err +} + func parseVlanData(link Link, data []syscall.NetlinkRouteAttr) { vlan := link.(*Vlan) for _, datum := range data { @@ -1414,6 +1717,10 @@ func parseVxlanData(link Link, data []syscall.NetlinkRouteAttr) { vxlan.L3miss = int8(datum.Value[0]) != 0 case nl.IFLA_VXLAN_UDP_CSUM: vxlan.UDPCSum = int8(datum.Value[0]) != 0 + case nl.IFLA_VXLAN_UDP_ZERO_CSUM6_TX: + vxlan.UDP6ZeroCSumTx = int8(datum.Value[0]) != 0 + case nl.IFLA_VXLAN_UDP_ZERO_CSUM6_RX: + vxlan.UDP6ZeroCSumRx = int8(datum.Value[0]) != 0 case nl.IFLA_VXLAN_GBP: vxlan.GBP = true case nl.IFLA_VXLAN_FLOWBASED: @@ -1516,7 +1823,8 @@ func parseMacvtapData(link Link, data []syscall.NetlinkRouteAttr) { func parseMacvlanData(link Link, data []syscall.NetlinkRouteAttr) { macv := link.(*Macvlan) for _, datum := range data { - if datum.Attr.Type == nl.IFLA_MACVLAN_MODE { + switch datum.Attr.Type { + case nl.IFLA_MACVLAN_MODE: switch native.Uint32(datum.Value[0:4]) { case nl.MACVLAN_MODE_PRIVATE: macv.Mode = MACVLAN_MODE_PRIVATE @@ -1529,7 +1837,16 @@ func parseMacvlanData(link Link, data []syscall.NetlinkRouteAttr) { case nl.MACVLAN_MODE_SOURCE: macv.Mode = MACVLAN_MODE_SOURCE } - return + case nl.IFLA_MACVLAN_MACADDR_COUNT: + macv.MACAddrs = make([]net.HardwareAddr, 0, int(native.Uint32(datum.Value[0:4]))) + case nl.IFLA_MACVLAN_MACADDR_DATA: + macs, err := nl.ParseRouteAttr(datum.Value[:]) + if err != nil { + panic(fmt.Sprintf("failed to ParseRouteAttr for IFLA_MACVLAN_MACADDR_DATA: %v", err)) + } + for _, macDatum := range macs { + macv.MACAddrs = append(macv.MACAddrs, net.HardwareAddr(macDatum.Value[0:6])) + } } } } @@ -1537,19 +1854,19 @@ func parseMacvlanData(link Link, data []syscall.NetlinkRouteAttr) { // copied from pkg/net_linux.go func linkFlags(rawFlags uint32) net.Flags { var f net.Flags - if rawFlags&syscall.IFF_UP != 0 { + if rawFlags&unix.IFF_UP != 0 { f |= net.FlagUp } - if rawFlags&syscall.IFF_BROADCAST != 0 { + if rawFlags&unix.IFF_BROADCAST != 0 { f |= net.FlagBroadcast } - if rawFlags&syscall.IFF_LOOPBACK != 0 { + if rawFlags&unix.IFF_LOOPBACK != 0 { f |= net.FlagLoopback } - if rawFlags&syscall.IFF_POINTOPOINT != 0 { + if rawFlags&unix.IFF_POINTOPOINT != 0 { f |= net.FlagPointToPoint } - if rawFlags&syscall.IFF_MULTICAST != 0 { + if rawFlags&unix.IFF_MULTICAST != 0 { f |= net.FlagMulticast } return f @@ -1564,12 +1881,17 @@ func addGretapAttrs(gretap *Gretap, linkInfo *nl.RtAttr) { return } - ip := gretap.Local.To4() - if ip != nil { + if ip := gretap.Local; ip != nil { + if ip.To4() != nil { + ip = ip.To4() + } nl.NewRtAttrChild(data, nl.IFLA_GRE_LOCAL, []byte(ip)) } - ip = gretap.Remote.To4() - if ip != nil { + + if ip := gretap.Remote; ip != nil { + if ip.To4() != nil { + ip = ip.To4() + } nl.NewRtAttrChild(data, nl.IFLA_GRE_REMOTE, []byte(ip)) } @@ -1608,9 +1930,9 @@ func parseGretapData(link Link, data []syscall.NetlinkRouteAttr) { case nl.IFLA_GRE_IKEY: gre.OKey = ntohl(datum.Value[0:4]) case nl.IFLA_GRE_LOCAL: - gre.Local = net.IP(datum.Value[0:4]) + gre.Local = net.IP(datum.Value[0:16]) case nl.IFLA_GRE_REMOTE: - gre.Remote = net.IP(datum.Value[0:4]) + gre.Remote = net.IP(datum.Value[0:16]) case nl.IFLA_GRE_ENCAP_SPORT: gre.EncapSport = ntohs(datum.Value[0:2]) case nl.IFLA_GRE_ENCAP_DPORT: @@ -1631,7 +1953,87 @@ func parseGretapData(link Link, data []syscall.NetlinkRouteAttr) { case nl.IFLA_GRE_ENCAP_FLAGS: gre.EncapFlags = native.Uint16(datum.Value[0:2]) case nl.IFLA_GRE_COLLECT_METADATA: - gre.FlowBased = int8(datum.Value[0]) != 0 + if len(datum.Value) > 0 { + gre.FlowBased = int8(datum.Value[0]) != 0 + } + } + } +} + +func addGretunAttrs(gre *Gretun, linkInfo *nl.RtAttr) { + data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil) + + if ip := gre.Local; ip != nil { + if ip.To4() != nil { + ip = ip.To4() + } + nl.NewRtAttrChild(data, nl.IFLA_GRE_LOCAL, []byte(ip)) + } + + if ip := gre.Remote; ip != nil { + if ip.To4() != nil { + ip = ip.To4() + } + nl.NewRtAttrChild(data, nl.IFLA_GRE_REMOTE, []byte(ip)) + } + + if gre.IKey != 0 { + nl.NewRtAttrChild(data, nl.IFLA_GRE_IKEY, htonl(gre.IKey)) + gre.IFlags |= uint16(nl.GRE_KEY) + } + + if gre.OKey != 0 { + nl.NewRtAttrChild(data, nl.IFLA_GRE_OKEY, htonl(gre.OKey)) + gre.OFlags |= uint16(nl.GRE_KEY) + } + + nl.NewRtAttrChild(data, nl.IFLA_GRE_IFLAGS, htons(gre.IFlags)) + nl.NewRtAttrChild(data, nl.IFLA_GRE_OFLAGS, htons(gre.OFlags)) + + if gre.Link != 0 { + nl.NewRtAttrChild(data, nl.IFLA_GRE_LINK, nl.Uint32Attr(gre.Link)) + } + + nl.NewRtAttrChild(data, nl.IFLA_GRE_PMTUDISC, nl.Uint8Attr(gre.PMtuDisc)) + nl.NewRtAttrChild(data, nl.IFLA_GRE_TTL, nl.Uint8Attr(gre.Ttl)) + nl.NewRtAttrChild(data, nl.IFLA_GRE_TOS, nl.Uint8Attr(gre.Tos)) + nl.NewRtAttrChild(data, nl.IFLA_GRE_ENCAP_TYPE, nl.Uint16Attr(gre.EncapType)) + nl.NewRtAttrChild(data, nl.IFLA_GRE_ENCAP_FLAGS, nl.Uint16Attr(gre.EncapFlags)) + nl.NewRtAttrChild(data, nl.IFLA_GRE_ENCAP_SPORT, htons(gre.EncapSport)) + nl.NewRtAttrChild(data, nl.IFLA_GRE_ENCAP_DPORT, htons(gre.EncapDport)) +} + +func parseGretunData(link Link, data []syscall.NetlinkRouteAttr) { + gre := link.(*Gretun) + for _, datum := range data { + switch datum.Attr.Type { + case nl.IFLA_GRE_OKEY: + gre.IKey = ntohl(datum.Value[0:4]) + case nl.IFLA_GRE_IKEY: + gre.OKey = ntohl(datum.Value[0:4]) + case nl.IFLA_GRE_LOCAL: + gre.Local = net.IP(datum.Value[0:16]) + case nl.IFLA_GRE_REMOTE: + gre.Remote = net.IP(datum.Value[0:16]) + case nl.IFLA_GRE_IFLAGS: + gre.IFlags = ntohs(datum.Value[0:2]) + case nl.IFLA_GRE_OFLAGS: + gre.OFlags = ntohs(datum.Value[0:2]) + + case nl.IFLA_GRE_TTL: + gre.Ttl = uint8(datum.Value[0]) + case nl.IFLA_GRE_TOS: + gre.Tos = uint8(datum.Value[0]) + case nl.IFLA_GRE_PMTUDISC: + gre.PMtuDisc = uint8(datum.Value[0]) + case nl.IFLA_GRE_ENCAP_TYPE: + gre.EncapType = native.Uint16(datum.Value[0:2]) + case nl.IFLA_GRE_ENCAP_FLAGS: + gre.EncapFlags = native.Uint16(datum.Value[0:2]) + case nl.IFLA_GRE_ENCAP_SPORT: + gre.EncapSport = ntohs(datum.Value[0:2]) + case nl.IFLA_GRE_ENCAP_DPORT: + gre.EncapDport = ntohs(datum.Value[0:2]) } } } @@ -1645,12 +2047,15 @@ func parseLinkStats64(data []byte) *LinkStatistics { } func addXdpAttrs(xdp *LinkXdp, req *nl.NetlinkRequest) { - attrs := nl.NewRtAttr(nl.IFLA_XDP|syscall.NLA_F_NESTED, nil) + attrs := nl.NewRtAttr(unix.IFLA_XDP|unix.NLA_F_NESTED, nil) b := make([]byte, 4) native.PutUint32(b, uint32(xdp.Fd)) nl.NewRtAttrChild(attrs, nl.IFLA_XDP_FD, b) - native.PutUint32(b, xdp.Flags) - nl.NewRtAttrChild(attrs, nl.IFLA_XDP_FLAGS, b) + if xdp.Flags != 0 { + b := make([]byte, 4) + native.PutUint32(b, xdp.Flags) + nl.NewRtAttrChild(attrs, nl.IFLA_XDP_FLAGS, b) + } req.AddData(attrs) } @@ -1676,6 +2081,12 @@ func parseLinkXdp(data []byte) (*LinkXdp, error) { } func addIptunAttrs(iptun *Iptun, linkInfo *nl.RtAttr) { + if iptun.FlowBased { + // In flow based mode, no other attributes need to be configured + nl.NewRtAttrChild(linkInfo, nl.IFLA_IPTUN_COLLECT_METADATA, boolAttr(iptun.FlowBased)) + return + } + data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil) ip := iptun.Local.To4() @@ -1694,6 +2105,10 @@ func addIptunAttrs(iptun *Iptun, linkInfo *nl.RtAttr) { nl.NewRtAttrChild(data, nl.IFLA_IPTUN_PMTUDISC, nl.Uint8Attr(iptun.PMtuDisc)) nl.NewRtAttrChild(data, nl.IFLA_IPTUN_TTL, nl.Uint8Attr(iptun.Ttl)) nl.NewRtAttrChild(data, nl.IFLA_IPTUN_TOS, nl.Uint8Attr(iptun.Tos)) + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_ENCAP_TYPE, nl.Uint16Attr(iptun.EncapType)) + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_ENCAP_FLAGS, nl.Uint16Attr(iptun.EncapFlags)) + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_ENCAP_SPORT, htons(iptun.EncapSport)) + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_ENCAP_DPORT, htons(iptun.EncapDport)) } func parseIptunData(link Link, data []syscall.NetlinkRouteAttr) { @@ -1710,6 +2125,72 @@ func parseIptunData(link Link, data []syscall.NetlinkRouteAttr) { iptun.Tos = uint8(datum.Value[0]) case nl.IFLA_IPTUN_PMTUDISC: iptun.PMtuDisc = uint8(datum.Value[0]) + case nl.IFLA_IPTUN_ENCAP_SPORT: + iptun.EncapSport = ntohs(datum.Value[0:2]) + case nl.IFLA_IPTUN_ENCAP_DPORT: + iptun.EncapDport = ntohs(datum.Value[0:2]) + case nl.IFLA_IPTUN_ENCAP_TYPE: + iptun.EncapType = native.Uint16(datum.Value[0:2]) + case nl.IFLA_IPTUN_ENCAP_FLAGS: + iptun.EncapFlags = native.Uint16(datum.Value[0:2]) + case nl.IFLA_IPTUN_COLLECT_METADATA: + iptun.FlowBased = int8(datum.Value[0]) != 0 + } + } +} + +func addSittunAttrs(sittun *Sittun, linkInfo *nl.RtAttr) { + data := nl.NewRtAttrChild(linkInfo, nl.IFLA_INFO_DATA, nil) + + if sittun.Link != 0 { + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_LINK, nl.Uint32Attr(sittun.Link)) + } + + ip := sittun.Local.To4() + if ip != nil { + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_LOCAL, []byte(ip)) + } + + ip = sittun.Remote.To4() + if ip != nil { + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_REMOTE, []byte(ip)) + } + + if sittun.Ttl > 0 { + // Would otherwise fail on 3.10 kernel + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_TTL, nl.Uint8Attr(sittun.Ttl)) + } + + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_TOS, nl.Uint8Attr(sittun.Tos)) + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_PMTUDISC, nl.Uint8Attr(sittun.PMtuDisc)) + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_ENCAP_TYPE, nl.Uint16Attr(sittun.EncapType)) + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_ENCAP_FLAGS, nl.Uint16Attr(sittun.EncapFlags)) + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_ENCAP_SPORT, htons(sittun.EncapSport)) + nl.NewRtAttrChild(data, nl.IFLA_IPTUN_ENCAP_DPORT, htons(sittun.EncapDport)) +} + +func parseSittunData(link Link, data []syscall.NetlinkRouteAttr) { + sittun := link.(*Sittun) + for _, datum := range data { + switch datum.Attr.Type { + case nl.IFLA_IPTUN_LOCAL: + sittun.Local = net.IP(datum.Value[0:4]) + case nl.IFLA_IPTUN_REMOTE: + sittun.Remote = net.IP(datum.Value[0:4]) + case nl.IFLA_IPTUN_TTL: + sittun.Ttl = uint8(datum.Value[0]) + case nl.IFLA_IPTUN_TOS: + sittun.Tos = uint8(datum.Value[0]) + case nl.IFLA_IPTUN_PMTUDISC: + sittun.PMtuDisc = uint8(datum.Value[0]) + case nl.IFLA_IPTUN_ENCAP_TYPE: + sittun.EncapType = native.Uint16(datum.Value[0:2]) + case nl.IFLA_IPTUN_ENCAP_FLAGS: + sittun.EncapFlags = native.Uint16(datum.Value[0:2]) + case nl.IFLA_IPTUN_ENCAP_SPORT: + sittun.EncapSport = ntohs(datum.Value[0:2]) + case nl.IFLA_IPTUN_ENCAP_DPORT: + sittun.EncapDport = ntohs(datum.Value[0:2]) } } } @@ -1817,3 +2298,57 @@ func parseGTPData(link Link, data []syscall.NetlinkRouteAttr) { } } } + +// LinkSetBondSlave add slave to bond link via ioctl interface. +func LinkSetBondSlave(link Link, master *Bond) error { + fd, err := getSocketUDP() + if err != nil { + return err + } + defer syscall.Close(fd) + + ifreq := newIocltSlaveReq(link.Attrs().Name, master.Attrs().Name) + + _, _, errno := syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), unix.SIOCBONDENSLAVE, uintptr(unsafe.Pointer(ifreq))) + if errno != 0 { + return fmt.Errorf("Failed to enslave %q to %q, errno=%v", link.Attrs().Name, master.Attrs().Name, errno) + } + return nil +} + +// VethPeerIndex get veth peer index. +func VethPeerIndex(link *Veth) (int, error) { + fd, err := getSocketUDP() + if err != nil { + return -1, err + } + defer syscall.Close(fd) + + ifreq, sSet := newIocltStringSetReq(link.Name) + _, _, errno := syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), SIOCETHTOOL, uintptr(unsafe.Pointer(ifreq))) + if errno != 0 { + return -1, fmt.Errorf("SIOCETHTOOL request for %q failed, errno=%v", link.Attrs().Name, errno) + } + + gstrings := ðtoolGstrings{ + cmd: ETHTOOL_GSTRINGS, + stringSet: ETH_SS_STATS, + length: sSet.data[0], + } + ifreq.Data = uintptr(unsafe.Pointer(gstrings)) + _, _, errno = syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), SIOCETHTOOL, uintptr(unsafe.Pointer(ifreq))) + if errno != 0 { + return -1, fmt.Errorf("SIOCETHTOOL request for %q failed, errno=%v", link.Attrs().Name, errno) + } + + stats := ðtoolStats{ + cmd: ETHTOOL_GSTATS, + nStats: gstrings.length, + } + ifreq.Data = uintptr(unsafe.Pointer(stats)) + _, _, errno = syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), SIOCETHTOOL, uintptr(unsafe.Pointer(ifreq))) + if errno != 0 { + return -1, fmt.Errorf("SIOCETHTOOL request for %q failed, errno=%v", link.Attrs().Name, errno) + } + return int(stats.data[0]), nil +} diff --git a/vendor/github.com/vishvananda/netlink/neigh.go b/vendor/github.com/vishvananda/netlink/neigh.go index 0e5eb90c9..3f5cd497a 100644 --- a/vendor/github.com/vishvananda/netlink/neigh.go +++ b/vendor/github.com/vishvananda/netlink/neigh.go @@ -14,6 +14,9 @@ type Neigh struct { Flags int IP net.IP HardwareAddr net.HardwareAddr + LLIPAddr net.IP //Used in the case of NHRP + Vlan int + VNI int } // String returns $ip/$hwaddr $label diff --git a/vendor/github.com/vishvananda/netlink/neigh_linux.go b/vendor/github.com/vishvananda/netlink/neigh_linux.go index f069db257..f75c22649 100644 --- a/vendor/github.com/vishvananda/netlink/neigh_linux.go +++ b/vendor/github.com/vishvananda/netlink/neigh_linux.go @@ -2,10 +2,10 @@ package netlink import ( "net" - "syscall" "unsafe" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) const ( @@ -73,7 +73,7 @@ func NeighAdd(neigh *Neigh) error { // NeighAdd will add an IP to MAC mapping to the ARP table // Equivalent to: `ip neigh add ....` func (h *Handle) NeighAdd(neigh *Neigh) error { - return h.neighAdd(neigh, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL) + return h.neighAdd(neigh, unix.NLM_F_CREATE|unix.NLM_F_EXCL) } // NeighSet will add or replace an IP to MAC mapping to the ARP table @@ -85,7 +85,7 @@ func NeighSet(neigh *Neigh) error { // NeighSet will add or replace an IP to MAC mapping to the ARP table // Equivalent to: `ip neigh replace....` func (h *Handle) NeighSet(neigh *Neigh) error { - return h.neighAdd(neigh, syscall.NLM_F_CREATE|syscall.NLM_F_REPLACE) + return h.neighAdd(neigh, unix.NLM_F_CREATE|unix.NLM_F_REPLACE) } // NeighAppend will append an entry to FDB @@ -97,7 +97,7 @@ func NeighAppend(neigh *Neigh) error { // NeighAppend will append an entry to FDB // Equivalent to: `bridge fdb append...` func (h *Handle) NeighAppend(neigh *Neigh) error { - return h.neighAdd(neigh, syscall.NLM_F_CREATE|syscall.NLM_F_APPEND) + return h.neighAdd(neigh, unix.NLM_F_CREATE|unix.NLM_F_APPEND) } // NeighAppend will append an entry to FDB @@ -109,7 +109,7 @@ func neighAdd(neigh *Neigh, mode int) error { // NeighAppend will append an entry to FDB // Equivalent to: `bridge fdb append...` func (h *Handle) neighAdd(neigh *Neigh, mode int) error { - req := h.newNetlinkRequest(syscall.RTM_NEWNEIGH, mode|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_NEWNEIGH, mode|unix.NLM_F_ACK) return neighHandle(neigh, req) } @@ -122,12 +122,13 @@ func NeighDel(neigh *Neigh) error { // NeighDel will delete an IP address from a link device. // Equivalent to: `ip addr del $addr dev $link` func (h *Handle) NeighDel(neigh *Neigh) error { - req := h.newNetlinkRequest(syscall.RTM_DELNEIGH, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_DELNEIGH, unix.NLM_F_ACK) return neighHandle(neigh, req) } func neighHandle(neigh *Neigh, req *nl.NetlinkRequest) error { var family int + if neigh.Family > 0 { family = neigh.Family } else { @@ -151,12 +152,25 @@ func neighHandle(neigh *Neigh, req *nl.NetlinkRequest) error { dstData := nl.NewRtAttr(NDA_DST, ipData) req.AddData(dstData) - if neigh.Flags != NTF_PROXY || neigh.HardwareAddr != nil { + if neigh.LLIPAddr != nil { + llIPData := nl.NewRtAttr(NDA_LLADDR, neigh.LLIPAddr.To4()) + req.AddData(llIPData) + } else if neigh.Flags != NTF_PROXY || neigh.HardwareAddr != nil { hwData := nl.NewRtAttr(NDA_LLADDR, []byte(neigh.HardwareAddr)) req.AddData(hwData) } - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + if neigh.Vlan != 0 { + vlanData := nl.NewRtAttr(NDA_VLAN, nl.Uint16Attr(uint16(neigh.Vlan))) + req.AddData(vlanData) + } + + if neigh.VNI != 0 { + vniData := nl.NewRtAttr(NDA_VNI, nl.Uint32Attr(uint32(neigh.VNI))) + req.AddData(vniData) + } + + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -189,7 +203,7 @@ func (h *Handle) NeighProxyList(linkIndex, family int) ([]Neigh, error) { } func (h *Handle) neighList(linkIndex, family, flags int) ([]Neigh, error) { - req := h.newNetlinkRequest(syscall.RTM_GETNEIGH, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(unix.RTM_GETNEIGH, unix.NLM_F_DUMP) msg := Ndmsg{ Family: uint8(family), Index: uint32(linkIndex), @@ -197,7 +211,7 @@ func (h *Handle) neighList(linkIndex, family, flags int) ([]Neigh, error) { } req.AddData(&msg) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWNEIGH) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWNEIGH) if err != nil { return nil, err } @@ -237,12 +251,37 @@ 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: neigh.IP = net.IP(attr.Value) case NDA_LLADDR: - neigh.HardwareAddr = net.HardwareAddr(attr.Value) + // BUG: Is this a bug in the netlink library? + // #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") { + neigh.LLIPAddr = net.IP(attr.Value) + } else if attrLen == 16 && + encapType == "tunnel6" { + neigh.IP = net.IP(attr.Value) + } else { + neigh.HardwareAddr = net.HardwareAddr(attr.Value) + } + case NDA_VLAN: + neigh.Vlan = int(native.Uint16(attr.Value[0:2])) + case NDA_VNI: + neigh.VNI = int(native.Uint32(attr.Value[0:4])) } } diff --git a/vendor/github.com/vishvananda/netlink/netlink_unspecified.go b/vendor/github.com/vishvananda/netlink/netlink_unspecified.go index 2d57c16d7..86111b92c 100644 --- a/vendor/github.com/vishvananda/netlink/netlink_unspecified.go +++ b/vendor/github.com/vishvananda/netlink/netlink_unspecified.go @@ -108,6 +108,10 @@ func LinkSetFlood(link Link, mode bool) error { return ErrNotImplemented } +func LinkSetTxQLen(link Link, qlen int) error { + return ErrNotImplemented +} + func LinkAdd(link Link) error { return ErrNotImplemented } diff --git a/vendor/github.com/vishvananda/netlink/nl/addr_linux.go b/vendor/github.com/vishvananda/netlink/nl/addr_linux.go index fe362e9fa..50db3b4cd 100644 --- a/vendor/github.com/vishvananda/netlink/nl/addr_linux.go +++ b/vendor/github.com/vishvananda/netlink/nl/addr_linux.go @@ -1,17 +1,18 @@ package nl import ( - "syscall" "unsafe" + + "golang.org/x/sys/unix" ) type IfAddrmsg struct { - syscall.IfAddrmsg + unix.IfAddrmsg } func NewIfAddrmsg(family int) *IfAddrmsg { return &IfAddrmsg{ - IfAddrmsg: syscall.IfAddrmsg{ + IfAddrmsg: unix.IfAddrmsg{ Family: uint8(family), }, } @@ -35,15 +36,15 @@ func NewIfAddrmsg(family int) *IfAddrmsg { // SizeofIfAddrmsg = 0x8 func DeserializeIfAddrmsg(b []byte) *IfAddrmsg { - return (*IfAddrmsg)(unsafe.Pointer(&b[0:syscall.SizeofIfAddrmsg][0])) + return (*IfAddrmsg)(unsafe.Pointer(&b[0:unix.SizeofIfAddrmsg][0])) } func (msg *IfAddrmsg) Serialize() []byte { - return (*(*[syscall.SizeofIfAddrmsg]byte)(unsafe.Pointer(msg)))[:] + return (*(*[unix.SizeofIfAddrmsg]byte)(unsafe.Pointer(msg)))[:] } func (msg *IfAddrmsg) Len() int { - return syscall.SizeofIfAddrmsg + return unix.SizeofIfAddrmsg } // struct ifa_cacheinfo { diff --git a/vendor/github.com/vishvananda/netlink/nl/link_linux.go b/vendor/github.com/vishvananda/netlink/nl/link_linux.go index a492246d8..84a3498dd 100644 --- a/vendor/github.com/vishvananda/netlink/nl/link_linux.go +++ b/vendor/github.com/vishvananda/netlink/nl/link_linux.go @@ -1,35 +1,11 @@ package nl import ( - "syscall" "unsafe" ) const ( DEFAULT_CHANGE = 0xFFFFFFFF - // doesn't exist in syscall - IFLA_VFINFO_LIST = syscall.IFLA_IFALIAS + 1 + iota - IFLA_STATS64 - IFLA_VF_PORTS - IFLA_PORT_SELF - IFLA_AF_SPEC - IFLA_GROUP - IFLA_NET_NS_FD - IFLA_EXT_MASK - IFLA_PROMISCUITY - IFLA_NUM_TX_QUEUES - IFLA_NUM_RX_QUEUES - IFLA_CARRIER - IFLA_PHYS_PORT_ID - IFLA_CARRIER_CHANGES - IFLA_PHYS_SWITCH_ID - IFLA_LINK_NETNSID - IFLA_PHYS_PORT_NAME - IFLA_PROTO_DOWN - IFLA_GSO_MAX_SEGS - IFLA_GSO_MAX_SIZE - IFLA_PAD - IFLA_XDP ) const ( @@ -118,6 +94,10 @@ const ( IFLA_MACVLAN_UNSPEC = iota IFLA_MACVLAN_MODE IFLA_MACVLAN_FLAGS + IFLA_MACVLAN_MACADDR_MODE + IFLA_MACVLAN_MACADDR + IFLA_MACVLAN_MACADDR_DATA + IFLA_MACVLAN_MACADDR_COUNT IFLA_MACVLAN_MAX = IFLA_MACVLAN_FLAGS ) @@ -130,6 +110,13 @@ const ( ) const ( + MACVLAN_MACADDR_ADD = iota + MACVLAN_MACADDR_DEL + MACVLAN_MACADDR_FLUSH + MACVLAN_MACADDR_SET +) + +const ( IFLA_BOND_UNSPEC = iota IFLA_BOND_MODE IFLA_BOND_ACTIVE_SLAVE @@ -231,7 +218,8 @@ const ( * on/off switch */ IFLA_VF_STATS /* network device statistics */ - IFLA_VF_MAX = IFLA_VF_STATS + IFLA_VF_TRUST /* Trust state of VF */ + IFLA_VF_MAX = IFLA_VF_TRUST ) const ( @@ -259,6 +247,7 @@ const ( SizeofVfSpoofchk = 0x08 SizeofVfLinkState = 0x08 SizeofVfRssQueryEn = 0x08 + SizeofVfTrust = 0x08 ) // struct ifla_vf_mac { @@ -419,6 +408,35 @@ func (msg *VfRssQueryEn) Serialize() []byte { return (*(*[SizeofVfRssQueryEn]byte)(unsafe.Pointer(msg)))[:] } +// struct ifla_vf_trust { +// __u32 vf; +// __u32 setting; +// }; + +type VfTrust struct { + Vf uint32 + Setting uint32 +} + +func (msg *VfTrust) Len() int { + return SizeofVfTrust +} + +func DeserializeVfTrust(b []byte) *VfTrust { + return (*VfTrust)(unsafe.Pointer(&b[0:SizeofVfTrust][0])) +} + +func (msg *VfTrust) Serialize() []byte { + return (*(*[SizeofVfTrust]byte)(unsafe.Pointer(msg)))[:] +} + +const ( + XDP_FLAGS_UPDATE_IF_NOEXIST = 1 << iota + XDP_FLAGS_SKB_MODE + XDP_FLAGS_DRV_MODE + XDP_FLAGS_MASK = XDP_FLAGS_UPDATE_IF_NOEXIST | XDP_FLAGS_SKB_MODE | XDP_FLAGS_DRV_MODE +) + const ( IFLA_XDP_UNSPEC = iota IFLA_XDP_FD /* fd of xdp program to attach, or -1 to remove */ @@ -444,7 +462,12 @@ const ( IFLA_IPTUN_6RD_RELAY_PREFIX IFLA_IPTUN_6RD_PREFIXLEN IFLA_IPTUN_6RD_RELAY_PREFIXLEN - IFLA_IPTUN_MAX = IFLA_IPTUN_6RD_RELAY_PREFIXLEN + IFLA_IPTUN_ENCAP_TYPE + IFLA_IPTUN_ENCAP_FLAGS + IFLA_IPTUN_ENCAP_SPORT + IFLA_IPTUN_ENCAP_DPORT + IFLA_IPTUN_COLLECT_METADATA + IFLA_IPTUN_MAX = IFLA_IPTUN_COLLECT_METADATA ) const ( diff --git a/vendor/github.com/vishvananda/netlink/nl/nl_linux.go b/vendor/github.com/vishvananda/netlink/nl/nl_linux.go index 1329acd86..bc8e82c2c 100644 --- a/vendor/github.com/vishvananda/netlink/nl/nl_linux.go +++ b/vendor/github.com/vishvananda/netlink/nl/nl_linux.go @@ -13,18 +13,19 @@ import ( "unsafe" "github.com/vishvananda/netns" + "golang.org/x/sys/unix" ) const ( // Family type definitions - FAMILY_ALL = syscall.AF_UNSPEC - FAMILY_V4 = syscall.AF_INET - FAMILY_V6 = syscall.AF_INET6 + FAMILY_ALL = unix.AF_UNSPEC + FAMILY_V4 = unix.AF_INET + FAMILY_V6 = unix.AF_INET6 FAMILY_MPLS = AF_MPLS ) // SupportedNlFamilies contains the list of netlink families this netlink package supports -var SupportedNlFamilies = []int{syscall.NETLINK_ROUTE, syscall.NETLINK_XFRM, syscall.NETLINK_NETFILTER} +var SupportedNlFamilies = []int{unix.NETLINK_ROUTE, unix.NETLINK_XFRM, unix.NETLINK_NETFILTER} var nextSeqNr uint32 @@ -77,161 +78,161 @@ type NetlinkRequestData interface { // IfInfomsg is related to links, but it is used for list requests as well type IfInfomsg struct { - syscall.IfInfomsg + unix.IfInfomsg } // Create an IfInfomsg with family specified func NewIfInfomsg(family int) *IfInfomsg { return &IfInfomsg{ - IfInfomsg: syscall.IfInfomsg{ + IfInfomsg: unix.IfInfomsg{ Family: uint8(family), }, } } func DeserializeIfInfomsg(b []byte) *IfInfomsg { - return (*IfInfomsg)(unsafe.Pointer(&b[0:syscall.SizeofIfInfomsg][0])) + return (*IfInfomsg)(unsafe.Pointer(&b[0:unix.SizeofIfInfomsg][0])) } func (msg *IfInfomsg) Serialize() []byte { - return (*(*[syscall.SizeofIfInfomsg]byte)(unsafe.Pointer(msg)))[:] + return (*(*[unix.SizeofIfInfomsg]byte)(unsafe.Pointer(msg)))[:] } func (msg *IfInfomsg) Len() int { - return syscall.SizeofIfInfomsg + return unix.SizeofIfInfomsg } func (msg *IfInfomsg) EncapType() string { switch msg.Type { case 0: return "generic" - case syscall.ARPHRD_ETHER: + case unix.ARPHRD_ETHER: return "ether" - case syscall.ARPHRD_EETHER: + case unix.ARPHRD_EETHER: return "eether" - case syscall.ARPHRD_AX25: + case unix.ARPHRD_AX25: return "ax25" - case syscall.ARPHRD_PRONET: + case unix.ARPHRD_PRONET: return "pronet" - case syscall.ARPHRD_CHAOS: + case unix.ARPHRD_CHAOS: return "chaos" - case syscall.ARPHRD_IEEE802: + case unix.ARPHRD_IEEE802: return "ieee802" - case syscall.ARPHRD_ARCNET: + case unix.ARPHRD_ARCNET: return "arcnet" - case syscall.ARPHRD_APPLETLK: + case unix.ARPHRD_APPLETLK: return "atalk" - case syscall.ARPHRD_DLCI: + case unix.ARPHRD_DLCI: return "dlci" - case syscall.ARPHRD_ATM: + case unix.ARPHRD_ATM: return "atm" - case syscall.ARPHRD_METRICOM: + case unix.ARPHRD_METRICOM: return "metricom" - case syscall.ARPHRD_IEEE1394: + case unix.ARPHRD_IEEE1394: return "ieee1394" - case syscall.ARPHRD_INFINIBAND: + case unix.ARPHRD_INFINIBAND: return "infiniband" - case syscall.ARPHRD_SLIP: + case unix.ARPHRD_SLIP: return "slip" - case syscall.ARPHRD_CSLIP: + case unix.ARPHRD_CSLIP: return "cslip" - case syscall.ARPHRD_SLIP6: + case unix.ARPHRD_SLIP6: return "slip6" - case syscall.ARPHRD_CSLIP6: + case unix.ARPHRD_CSLIP6: return "cslip6" - case syscall.ARPHRD_RSRVD: + case unix.ARPHRD_RSRVD: return "rsrvd" - case syscall.ARPHRD_ADAPT: + case unix.ARPHRD_ADAPT: return "adapt" - case syscall.ARPHRD_ROSE: + case unix.ARPHRD_ROSE: return "rose" - case syscall.ARPHRD_X25: + case unix.ARPHRD_X25: return "x25" - case syscall.ARPHRD_HWX25: + case unix.ARPHRD_HWX25: return "hwx25" - case syscall.ARPHRD_PPP: + case unix.ARPHRD_PPP: return "ppp" - case syscall.ARPHRD_HDLC: + case unix.ARPHRD_HDLC: return "hdlc" - case syscall.ARPHRD_LAPB: + case unix.ARPHRD_LAPB: return "lapb" - case syscall.ARPHRD_DDCMP: + case unix.ARPHRD_DDCMP: return "ddcmp" - case syscall.ARPHRD_RAWHDLC: + case unix.ARPHRD_RAWHDLC: return "rawhdlc" - case syscall.ARPHRD_TUNNEL: + case unix.ARPHRD_TUNNEL: return "ipip" - case syscall.ARPHRD_TUNNEL6: + case unix.ARPHRD_TUNNEL6: return "tunnel6" - case syscall.ARPHRD_FRAD: + case unix.ARPHRD_FRAD: return "frad" - case syscall.ARPHRD_SKIP: + case unix.ARPHRD_SKIP: return "skip" - case syscall.ARPHRD_LOOPBACK: + case unix.ARPHRD_LOOPBACK: return "loopback" - case syscall.ARPHRD_LOCALTLK: + case unix.ARPHRD_LOCALTLK: return "ltalk" - case syscall.ARPHRD_FDDI: + case unix.ARPHRD_FDDI: return "fddi" - case syscall.ARPHRD_BIF: + case unix.ARPHRD_BIF: return "bif" - case syscall.ARPHRD_SIT: + case unix.ARPHRD_SIT: return "sit" - case syscall.ARPHRD_IPDDP: + case unix.ARPHRD_IPDDP: return "ip/ddp" - case syscall.ARPHRD_IPGRE: + case unix.ARPHRD_IPGRE: return "gre" - case syscall.ARPHRD_PIMREG: + case unix.ARPHRD_PIMREG: return "pimreg" - case syscall.ARPHRD_HIPPI: + case unix.ARPHRD_HIPPI: return "hippi" - case syscall.ARPHRD_ASH: + case unix.ARPHRD_ASH: return "ash" - case syscall.ARPHRD_ECONET: + case unix.ARPHRD_ECONET: return "econet" - case syscall.ARPHRD_IRDA: + case unix.ARPHRD_IRDA: return "irda" - case syscall.ARPHRD_FCPP: + case unix.ARPHRD_FCPP: return "fcpp" - case syscall.ARPHRD_FCAL: + case unix.ARPHRD_FCAL: return "fcal" - case syscall.ARPHRD_FCPL: + case unix.ARPHRD_FCPL: return "fcpl" - case syscall.ARPHRD_FCFABRIC: + case unix.ARPHRD_FCFABRIC: return "fcfb0" - case syscall.ARPHRD_FCFABRIC + 1: + case unix.ARPHRD_FCFABRIC + 1: return "fcfb1" - case syscall.ARPHRD_FCFABRIC + 2: + case unix.ARPHRD_FCFABRIC + 2: return "fcfb2" - case syscall.ARPHRD_FCFABRIC + 3: + case unix.ARPHRD_FCFABRIC + 3: return "fcfb3" - case syscall.ARPHRD_FCFABRIC + 4: + case unix.ARPHRD_FCFABRIC + 4: return "fcfb4" - case syscall.ARPHRD_FCFABRIC + 5: + case unix.ARPHRD_FCFABRIC + 5: return "fcfb5" - case syscall.ARPHRD_FCFABRIC + 6: + case unix.ARPHRD_FCFABRIC + 6: return "fcfb6" - case syscall.ARPHRD_FCFABRIC + 7: + case unix.ARPHRD_FCFABRIC + 7: return "fcfb7" - case syscall.ARPHRD_FCFABRIC + 8: + case unix.ARPHRD_FCFABRIC + 8: return "fcfb8" - case syscall.ARPHRD_FCFABRIC + 9: + case unix.ARPHRD_FCFABRIC + 9: return "fcfb9" - case syscall.ARPHRD_FCFABRIC + 10: + case unix.ARPHRD_FCFABRIC + 10: return "fcfb10" - case syscall.ARPHRD_FCFABRIC + 11: + case unix.ARPHRD_FCFABRIC + 11: return "fcfb11" - case syscall.ARPHRD_FCFABRIC + 12: + case unix.ARPHRD_FCFABRIC + 12: return "fcfb12" - case syscall.ARPHRD_IEEE802_TR: + case unix.ARPHRD_IEEE802_TR: return "tr" - case syscall.ARPHRD_IEEE80211: + case unix.ARPHRD_IEEE80211: return "ieee802.11" - case syscall.ARPHRD_IEEE80211_PRISM: + case unix.ARPHRD_IEEE80211_PRISM: return "ieee802.11/prism" - case syscall.ARPHRD_IEEE80211_RADIOTAP: + case unix.ARPHRD_IEEE80211_RADIOTAP: return "ieee802.11/radiotap" - case syscall.ARPHRD_IEEE802154: + case unix.ARPHRD_IEEE802154: return "ieee802.15.4" case 65534: @@ -243,7 +244,7 @@ func (msg *IfInfomsg) EncapType() string { } func rtaAlignOf(attrlen int) int { - return (attrlen + syscall.RTA_ALIGNTO - 1) & ^(syscall.RTA_ALIGNTO - 1) + return (attrlen + unix.RTA_ALIGNTO - 1) & ^(unix.RTA_ALIGNTO - 1) } func NewIfInfomsgChild(parent *RtAttr, family int) *IfInfomsg { @@ -254,7 +255,7 @@ func NewIfInfomsgChild(parent *RtAttr, family int) *IfInfomsg { // Extend RtAttr to handle data and children type RtAttr struct { - syscall.RtAttr + unix.RtAttr Data []byte children []NetlinkRequestData } @@ -262,7 +263,7 @@ type RtAttr struct { // Create a new Extended RtAttr object func NewRtAttr(attrType int, data []byte) *RtAttr { return &RtAttr{ - RtAttr: syscall.RtAttr{ + RtAttr: unix.RtAttr{ Type: uint16(attrType), }, children: []NetlinkRequestData{}, @@ -277,16 +278,21 @@ func NewRtAttrChild(parent *RtAttr, attrType int, data []byte) *RtAttr { return attr } +// AddChild adds an existing RtAttr as a child. +func (a *RtAttr) AddChild(attr *RtAttr) { + a.children = append(a.children, attr) +} + func (a *RtAttr) Len() int { if len(a.children) == 0 { - return (syscall.SizeofRtAttr + len(a.Data)) + return (unix.SizeofRtAttr + len(a.Data)) } l := 0 for _, child := range a.children { l += rtaAlignOf(child.Len()) } - l += syscall.SizeofRtAttr + l += unix.SizeofRtAttr return rtaAlignOf(l + len(a.Data)) } @@ -319,7 +325,7 @@ func (a *RtAttr) Serialize() []byte { } type NetlinkRequest struct { - syscall.NlMsghdr + unix.NlMsghdr Data []NetlinkRequestData RawData []byte Sockets map[int]*SocketHandle @@ -327,7 +333,7 @@ type NetlinkRequest struct { // Serialize the Netlink Request into a byte array func (req *NetlinkRequest) Serialize() []byte { - length := syscall.SizeofNlMsghdr + length := unix.SizeofNlMsghdr dataBytes := make([][]byte, len(req.Data)) for i, data := range req.Data { dataBytes[i] = data.Serialize() @@ -337,8 +343,8 @@ func (req *NetlinkRequest) Serialize() []byte { req.Len = uint32(length) b := make([]byte, length) - hdr := (*(*[syscall.SizeofNlMsghdr]byte)(unsafe.Pointer(req)))[:] - next := syscall.SizeofNlMsghdr + hdr := (*(*[unix.SizeofNlMsghdr]byte)(unsafe.Pointer(req)))[:] + next := unix.SizeofNlMsghdr copy(b[0:next], hdr) for _, data := range dataBytes { for _, dataByte := range data { @@ -421,10 +427,10 @@ done: if m.Header.Pid != pid { return nil, fmt.Errorf("Wrong pid %d, expected %d", m.Header.Pid, pid) } - if m.Header.Type == syscall.NLMSG_DONE { + if m.Header.Type == unix.NLMSG_DONE { break done } - if m.Header.Type == syscall.NLMSG_ERROR { + if m.Header.Type == unix.NLMSG_ERROR { native := NativeEndian() error := int32(native.Uint32(m.Data[0:4])) if error == 0 { @@ -436,7 +442,7 @@ done: continue } res = append(res, m.Data) - if m.Header.Flags&syscall.NLM_F_MULTI == 0 { + if m.Header.Flags&unix.NLM_F_MULTI == 0 { break done } } @@ -449,10 +455,10 @@ done: // the message is serialized func NewNetlinkRequest(proto, flags int) *NetlinkRequest { return &NetlinkRequest{ - NlMsghdr: syscall.NlMsghdr{ - Len: uint32(syscall.SizeofNlMsghdr), + NlMsghdr: unix.NlMsghdr{ + Len: uint32(unix.SizeofNlMsghdr), Type: uint16(proto), - Flags: syscall.NLM_F_REQUEST | uint16(flags), + Flags: unix.NLM_F_REQUEST | uint16(flags), Seq: atomic.AddUint32(&nextSeqNr, 1), }, } @@ -460,21 +466,21 @@ func NewNetlinkRequest(proto, flags int) *NetlinkRequest { type NetlinkSocket struct { fd int32 - lsa syscall.SockaddrNetlink + lsa unix.SockaddrNetlink sync.Mutex } func getNetlinkSocket(protocol int) (*NetlinkSocket, error) { - fd, err := syscall.Socket(syscall.AF_NETLINK, syscall.SOCK_RAW|syscall.SOCK_CLOEXEC, protocol) + fd, err := unix.Socket(unix.AF_NETLINK, unix.SOCK_RAW|unix.SOCK_CLOEXEC, protocol) if err != nil { return nil, err } s := &NetlinkSocket{ fd: int32(fd), } - s.lsa.Family = syscall.AF_NETLINK - if err := syscall.Bind(fd, &s.lsa); err != nil { - syscall.Close(fd) + s.lsa.Family = unix.AF_NETLINK + if err := unix.Bind(fd, &s.lsa); err != nil { + unix.Close(fd) return nil, err } @@ -551,21 +557,21 @@ func executeInNetns(newNs, curNs netns.NsHandle) (func(), error) { // Returns the netlink socket on which Receive() method can be called // to retrieve the messages from the kernel. func Subscribe(protocol int, groups ...uint) (*NetlinkSocket, error) { - fd, err := syscall.Socket(syscall.AF_NETLINK, syscall.SOCK_RAW, protocol) + fd, err := unix.Socket(unix.AF_NETLINK, unix.SOCK_RAW, protocol) if err != nil { return nil, err } s := &NetlinkSocket{ fd: int32(fd), } - s.lsa.Family = syscall.AF_NETLINK + s.lsa.Family = unix.AF_NETLINK for _, g := range groups { s.lsa.Groups |= (1 << (g - 1)) } - if err := syscall.Bind(fd, &s.lsa); err != nil { - syscall.Close(fd) + if err := unix.Bind(fd, &s.lsa); err != nil { + unix.Close(fd) return nil, err } @@ -586,7 +592,7 @@ func SubscribeAt(newNs, curNs netns.NsHandle, protocol int, groups ...uint) (*Ne func (s *NetlinkSocket) Close() { fd := int(atomic.SwapInt32(&s.fd, -1)) - syscall.Close(fd) + unix.Close(fd) } func (s *NetlinkSocket) GetFd() int { @@ -598,7 +604,7 @@ func (s *NetlinkSocket) Send(request *NetlinkRequest) error { if fd < 0 { return fmt.Errorf("Send called on a closed socket") } - if err := syscall.Sendto(fd, request.Serialize(), 0, &s.lsa); err != nil { + if err := unix.Sendto(fd, request.Serialize(), 0, &s.lsa); err != nil { return err } return nil @@ -609,26 +615,40 @@ func (s *NetlinkSocket) Receive() ([]syscall.NetlinkMessage, error) { if fd < 0 { return nil, fmt.Errorf("Receive called on a closed socket") } - rb := make([]byte, syscall.Getpagesize()) - nr, _, err := syscall.Recvfrom(fd, rb, 0) + rb := make([]byte, unix.Getpagesize()) + nr, _, err := unix.Recvfrom(fd, rb, 0) if err != nil { return nil, err } - if nr < syscall.NLMSG_HDRLEN { + if nr < unix.NLMSG_HDRLEN { return nil, fmt.Errorf("Got short response from netlink") } rb = rb[:nr] return syscall.ParseNetlinkMessage(rb) } +// SetSendTimeout allows to set a send timeout on the socket +func (s *NetlinkSocket) SetSendTimeout(timeout *unix.Timeval) error { + // Set a send timeout of SOCKET_SEND_TIMEOUT, this will allow the Send to periodically unblock and avoid that a routine + // remains stuck on a send on a closed fd + return unix.SetsockoptTimeval(int(s.fd), unix.SOL_SOCKET, unix.SO_SNDTIMEO, timeout) +} + +// SetReceiveTimeout allows to set a receive timeout on the socket +func (s *NetlinkSocket) SetReceiveTimeout(timeout *unix.Timeval) error { + // Set a read timeout of SOCKET_READ_TIMEOUT, this will allow the Read to periodically unblock and avoid that a routine + // remains stuck on a recvmsg on a closed fd + return unix.SetsockoptTimeval(int(s.fd), unix.SOL_SOCKET, unix.SO_RCVTIMEO, timeout) +} + func (s *NetlinkSocket) GetPid() (uint32, error) { fd := int(atomic.LoadInt32(&s.fd)) - lsa, err := syscall.Getsockname(fd) + lsa, err := unix.Getsockname(fd) if err != nil { return 0, err } switch v := lsa.(type) { - case *syscall.SockaddrNetlink: + case *unix.SockaddrNetlink: return v.Pid, nil } return 0, fmt.Errorf("Wrong socket type") @@ -683,24 +703,24 @@ func Uint64Attr(v uint64) []byte { func ParseRouteAttr(b []byte) ([]syscall.NetlinkRouteAttr, error) { var attrs []syscall.NetlinkRouteAttr - for len(b) >= syscall.SizeofRtAttr { + for len(b) >= unix.SizeofRtAttr { a, vbuf, alen, err := netlinkRouteAttrAndValue(b) if err != nil { return nil, err } - ra := syscall.NetlinkRouteAttr{Attr: *a, Value: vbuf[:int(a.Len)-syscall.SizeofRtAttr]} + ra := syscall.NetlinkRouteAttr{Attr: syscall.RtAttr(*a), Value: vbuf[:int(a.Len)-unix.SizeofRtAttr]} attrs = append(attrs, ra) b = b[alen:] } return attrs, nil } -func netlinkRouteAttrAndValue(b []byte) (*syscall.RtAttr, []byte, int, error) { - a := (*syscall.RtAttr)(unsafe.Pointer(&b[0])) - if int(a.Len) < syscall.SizeofRtAttr || int(a.Len) > len(b) { - return nil, nil, 0, syscall.EINVAL +func netlinkRouteAttrAndValue(b []byte) (*unix.RtAttr, []byte, int, error) { + a := (*unix.RtAttr)(unsafe.Pointer(&b[0])) + if int(a.Len) < unix.SizeofRtAttr || int(a.Len) > len(b) { + return nil, nil, 0, unix.EINVAL } - return a, b[syscall.SizeofRtAttr:], rtaAlignOf(int(a.Len)), nil + return a, b[unix.SizeofRtAttr:], rtaAlignOf(int(a.Len)), nil } // SocketHandle contains the netlink socket and the associated diff --git a/vendor/github.com/vishvananda/netlink/nl/route_linux.go b/vendor/github.com/vishvananda/netlink/nl/route_linux.go index 1a064d65d..f6906fcaf 100644 --- a/vendor/github.com/vishvananda/netlink/nl/route_linux.go +++ b/vendor/github.com/vishvananda/netlink/nl/route_linux.go @@ -1,65 +1,66 @@ package nl import ( - "syscall" "unsafe" + + "golang.org/x/sys/unix" ) type RtMsg struct { - syscall.RtMsg + unix.RtMsg } func NewRtMsg() *RtMsg { return &RtMsg{ - RtMsg: syscall.RtMsg{ - Table: syscall.RT_TABLE_MAIN, - Scope: syscall.RT_SCOPE_UNIVERSE, - Protocol: syscall.RTPROT_BOOT, - Type: syscall.RTN_UNICAST, + RtMsg: unix.RtMsg{ + Table: unix.RT_TABLE_MAIN, + Scope: unix.RT_SCOPE_UNIVERSE, + Protocol: unix.RTPROT_BOOT, + Type: unix.RTN_UNICAST, }, } } func NewRtDelMsg() *RtMsg { return &RtMsg{ - RtMsg: syscall.RtMsg{ - Table: syscall.RT_TABLE_MAIN, - Scope: syscall.RT_SCOPE_NOWHERE, + RtMsg: unix.RtMsg{ + Table: unix.RT_TABLE_MAIN, + Scope: unix.RT_SCOPE_NOWHERE, }, } } func (msg *RtMsg) Len() int { - return syscall.SizeofRtMsg + return unix.SizeofRtMsg } func DeserializeRtMsg(b []byte) *RtMsg { - return (*RtMsg)(unsafe.Pointer(&b[0:syscall.SizeofRtMsg][0])) + return (*RtMsg)(unsafe.Pointer(&b[0:unix.SizeofRtMsg][0])) } func (msg *RtMsg) Serialize() []byte { - return (*(*[syscall.SizeofRtMsg]byte)(unsafe.Pointer(msg)))[:] + return (*(*[unix.SizeofRtMsg]byte)(unsafe.Pointer(msg)))[:] } type RtNexthop struct { - syscall.RtNexthop + unix.RtNexthop Children []NetlinkRequestData } func DeserializeRtNexthop(b []byte) *RtNexthop { - return (*RtNexthop)(unsafe.Pointer(&b[0:syscall.SizeofRtNexthop][0])) + return (*RtNexthop)(unsafe.Pointer(&b[0:unix.SizeofRtNexthop][0])) } func (msg *RtNexthop) Len() int { if len(msg.Children) == 0 { - return syscall.SizeofRtNexthop + return unix.SizeofRtNexthop } l := 0 for _, child := range msg.Children { l += rtaAlignOf(child.Len()) } - l += syscall.SizeofRtNexthop + l += unix.SizeofRtNexthop return rtaAlignOf(l) } @@ -67,8 +68,8 @@ func (msg *RtNexthop) Serialize() []byte { length := msg.Len() msg.RtNexthop.Len = uint16(length) buf := make([]byte, length) - copy(buf, (*(*[syscall.SizeofRtNexthop]byte)(unsafe.Pointer(msg)))[:]) - next := rtaAlignOf(syscall.SizeofRtNexthop) + copy(buf, (*(*[unix.SizeofRtNexthop]byte)(unsafe.Pointer(msg)))[:]) + next := rtaAlignOf(unix.SizeofRtNexthop) if len(msg.Children) > 0 { for _, child := range msg.Children { childBuf := child.Serialize() diff --git a/vendor/github.com/vishvananda/netlink/nl/seg6_linux.go b/vendor/github.com/vishvananda/netlink/nl/seg6_linux.go new file mode 100644 index 000000000..b3425f6b0 --- /dev/null +++ b/vendor/github.com/vishvananda/netlink/nl/seg6_linux.go @@ -0,0 +1,111 @@ +package nl + +import ( + "errors" + "fmt" + "net" +) + +type IPv6SrHdr struct { + nextHdr uint8 + hdrLen uint8 + routingType uint8 + segmentsLeft uint8 + firstSegment uint8 + flags uint8 + reserved uint16 + + Segments []net.IP +} + +func (s1 *IPv6SrHdr) Equal(s2 IPv6SrHdr) bool { + if len(s1.Segments) != len(s2.Segments) { + return false + } + for i := range s1.Segments { + if s1.Segments[i].Equal(s2.Segments[i]) != true { + return false + } + } + return s1.nextHdr == s2.nextHdr && + s1.hdrLen == s2.hdrLen && + s1.routingType == s2.routingType && + s1.segmentsLeft == s2.segmentsLeft && + s1.firstSegment == s2.firstSegment && + s1.flags == s2.flags + // reserved doesn't need to be identical. +} + +// seg6 encap mode +const ( + SEG6_IPTUN_MODE_INLINE = iota + SEG6_IPTUN_MODE_ENCAP +) + +// number of nested RTATTR +// from include/uapi/linux/seg6_iptunnel.h +const ( + SEG6_IPTUNNEL_UNSPEC = iota + SEG6_IPTUNNEL_SRH + __SEG6_IPTUNNEL_MAX +) +const ( + SEG6_IPTUNNEL_MAX = __SEG6_IPTUNNEL_MAX - 1 +) + +func EncodeSEG6Encap(mode int, segments []net.IP) ([]byte, error) { + nsegs := len(segments) // nsegs: number of segments + if nsegs == 0 { + return nil, errors.New("EncodeSEG6Encap: No Segment in srh") + } + b := make([]byte, 12, 12+len(segments)*16) + native := NativeEndian() + native.PutUint32(b, uint32(mode)) + b[4] = 0 // srh.nextHdr (0 when calling netlink) + b[5] = uint8(16 * nsegs >> 3) // srh.hdrLen (in 8-octets unit) + b[6] = IPV6_SRCRT_TYPE_4 // srh.routingType (assigned by IANA) + b[7] = uint8(nsegs - 1) // srh.segmentsLeft + b[8] = uint8(nsegs - 1) // srh.firstSegment + b[9] = 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[10:], 0) // srh.reserved + for _, netIP := range segments { + b = append(b, netIP...) // srh.Segments + } + return b, nil +} + +func DecodeSEG6Encap(buf []byte) (int, []net.IP, error) { + native := NativeEndian() + mode := int(native.Uint32(buf)) + srh := IPv6SrHdr{ + nextHdr: buf[4], + hdrLen: buf[5], + routingType: buf[6], + segmentsLeft: buf[7], + firstSegment: buf[8], + flags: buf[9], + reserved: native.Uint16(buf[10:12]), + } + buf = buf[12:] + if len(buf)%16 != 0 { + err := fmt.Errorf("DecodeSEG6Encap: error parsing Segment List (buf len: %d)\n", len(buf)) + return mode, nil, err + } + for len(buf) > 0 { + srh.Segments = append(srh.Segments, net.IP(buf[:16])) + buf = buf[16:] + } + return mode, srh.Segments, nil +} + +// Helper functions +func SEG6EncapModeString(mode int) string { + switch mode { + case SEG6_IPTUN_MODE_INLINE: + return "inline" + case SEG6_IPTUN_MODE_ENCAP: + return "encap" + } + return "unknown" +} diff --git a/vendor/github.com/vishvananda/netlink/nl/syscall.go b/vendor/github.com/vishvananda/netlink/nl/syscall.go index 3473e5363..fc631e0e5 100644 --- a/vendor/github.com/vishvananda/netlink/nl/syscall.go +++ b/vendor/github.com/vishvananda/netlink/nl/syscall.go @@ -65,4 +65,14 @@ const ( LWTUNNEL_ENCAP_IP LWTUNNEL_ENCAP_ILA LWTUNNEL_ENCAP_IP6 + LWTUNNEL_ENCAP_SEG6 + LWTUNNEL_ENCAP_BPF +) + +// routing header types +const ( + IPV6_SRCRT_STRICT = 0x01 // Deprecated; will be removed + IPV6_SRCRT_TYPE_0 = 0 // Deprecated; will be removed + IPV6_SRCRT_TYPE_2 = 2 // IPv6 type 2 Routing Header + IPV6_SRCRT_TYPE_4 = 4 // Segment Routing with IPv6 ) diff --git a/vendor/github.com/vishvananda/netlink/nl/tc_linux.go b/vendor/github.com/vishvananda/netlink/nl/tc_linux.go index e91fb21c5..94ebc290a 100644 --- a/vendor/github.com/vishvananda/netlink/nl/tc_linux.go +++ b/vendor/github.com/vishvananda/netlink/nl/tc_linux.go @@ -673,3 +673,38 @@ const ( TCA_FW_MASK TCA_FW_MAX = TCA_FW_MASK ) + +const ( + TCA_MATCHALL_UNSPEC = iota + TCA_MATCHALL_CLASSID + TCA_MATCHALL_ACT + TCA_MATCHALL_FLAGS +) + +const ( + TCA_FQ_UNSPEC = iota + TCA_FQ_PLIMIT // limit of total number of packets in queue + TCA_FQ_FLOW_PLIMIT // limit of packets per flow + TCA_FQ_QUANTUM // RR quantum + TCA_FQ_INITIAL_QUANTUM // RR quantum for new flow + TCA_FQ_RATE_ENABLE // enable/disable rate limiting + TCA_FQ_FLOW_DEFAULT_RATE // obsolete do not use + TCA_FQ_FLOW_MAX_RATE // per flow max rate + TCA_FQ_BUCKETS_LOG // log2(number of buckets) + TCA_FQ_FLOW_REFILL_DELAY // flow credit refill delay in usec + TCA_FQ_ORPHAN_MASK // mask applied to orphaned skb hashes + TCA_FQ_LOW_RATE_THRESHOLD // per packet delay under this rate +) + +const ( + TCA_FQ_CODEL_UNSPEC = iota + TCA_FQ_CODEL_TARGET + TCA_FQ_CODEL_LIMIT + TCA_FQ_CODEL_INTERVAL + TCA_FQ_CODEL_ECN + TCA_FQ_CODEL_FLOWS + TCA_FQ_CODEL_QUANTUM + TCA_FQ_CODEL_CE_THRESHOLD + TCA_FQ_CODEL_DROP_BATCH_SIZE + TCA_FQ_CODEL_MEMORY_LIMIT +) diff --git a/vendor/github.com/vishvananda/netlink/protinfo_linux.go b/vendor/github.com/vishvananda/netlink/protinfo_linux.go index 10dd0d533..43c465f05 100644 --- a/vendor/github.com/vishvananda/netlink/protinfo_linux.go +++ b/vendor/github.com/vishvananda/netlink/protinfo_linux.go @@ -5,6 +5,7 @@ import ( "syscall" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) func LinkGetProtinfo(link Link) (Protinfo, error) { @@ -15,10 +16,10 @@ func (h *Handle) LinkGetProtinfo(link Link) (Protinfo, error) { base := link.Attrs() h.ensureIndex(base) var pi Protinfo - req := h.newNetlinkRequest(syscall.RTM_GETLINK, syscall.NLM_F_DUMP) - msg := nl.NewIfInfomsg(syscall.AF_BRIDGE) + req := h.newNetlinkRequest(unix.RTM_GETLINK, unix.NLM_F_DUMP) + msg := nl.NewIfInfomsg(unix.AF_BRIDGE) req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, 0) + msgs, err := req.Execute(unix.NETLINK_ROUTE, 0) if err != nil { return pi, err } @@ -33,7 +34,7 @@ func (h *Handle) LinkGetProtinfo(link Link) (Protinfo, error) { return pi, err } for _, attr := range attrs { - if attr.Attr.Type != syscall.IFLA_PROTINFO|syscall.NLA_F_NESTED { + if attr.Attr.Type != unix.IFLA_PROTINFO|unix.NLA_F_NESTED { continue } infos, err := nl.ParseRouteAttr(attr.Value) diff --git a/vendor/github.com/vishvananda/netlink/qdisc.go b/vendor/github.com/vishvananda/netlink/qdisc.go index 0ca86ebe8..3df4b5c29 100644 --- a/vendor/github.com/vishvananda/netlink/qdisc.go +++ b/vendor/github.com/vishvananda/netlink/qdisc.go @@ -230,3 +230,63 @@ func (qdisc *GenericQdisc) Attrs() *QdiscAttrs { func (qdisc *GenericQdisc) Type() string { return qdisc.QdiscType } + +// Fq is a classless packet scheduler meant to be mostly used for locally generated traffic. +type Fq struct { + QdiscAttrs + PacketLimit uint32 + FlowPacketLimit uint32 + // In bytes + Quantum uint32 + InitialQuantum uint32 + // called RateEnable under the hood + Pacing uint32 + FlowDefaultRate uint32 + FlowMaxRate uint32 + // called BucketsLog under the hood + Buckets uint32 + FlowRefillDelay uint32 + LowRateThreshold uint32 +} + +func NewFq(attrs QdiscAttrs) *Fq { + return &Fq{ + QdiscAttrs: attrs, + Pacing: 1, + } +} + +func (qdisc *Fq) Attrs() *QdiscAttrs { + return &qdisc.QdiscAttrs +} + +func (qdisc *Fq) Type() string { + return "fq" +} + +// FQ_Codel (Fair Queuing Controlled Delay) is queuing discipline that combines Fair Queuing with the CoDel AQM scheme. +type FqCodel struct { + QdiscAttrs + Target uint32 + Limit uint32 + Interval uint32 + ECN uint32 + Flows uint32 + Quantum uint32 + // There are some more attributes here, but support for them seems not ubiquitous +} + +func NewFqCodel(attrs QdiscAttrs) *FqCodel { + return &FqCodel{ + QdiscAttrs: attrs, + ECN: 1, + } +} + +func (qdisc *FqCodel) Attrs() *QdiscAttrs { + return &qdisc.QdiscAttrs +} + +func (qdisc *FqCodel) Type() string { + return "fq_codel" +} diff --git a/vendor/github.com/vishvananda/netlink/qdisc_linux.go b/vendor/github.com/vishvananda/netlink/qdisc_linux.go index 2c0deddb3..3794ac18a 100644 --- a/vendor/github.com/vishvananda/netlink/qdisc_linux.go +++ b/vendor/github.com/vishvananda/netlink/qdisc_linux.go @@ -8,6 +8,7 @@ import ( "syscall" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) // NOTE function is here because it uses other linux functions @@ -84,7 +85,7 @@ func QdiscDel(qdisc Qdisc) error { // QdiscDel will delete a qdisc from the system. // Equivalent to: `tc qdisc del $qdisc` func (h *Handle) QdiscDel(qdisc Qdisc) error { - return h.qdiscModify(syscall.RTM_DELQDISC, 0, qdisc) + return h.qdiscModify(unix.RTM_DELQDISC, 0, qdisc) } // QdiscChange will change a qdisc in place @@ -98,7 +99,7 @@ func QdiscChange(qdisc Qdisc) error { // Equivalent to: `tc qdisc change $qdisc` // The parent and handle MUST NOT be changed. func (h *Handle) QdiscChange(qdisc Qdisc) error { - return h.qdiscModify(syscall.RTM_NEWQDISC, 0, qdisc) + return h.qdiscModify(unix.RTM_NEWQDISC, 0, qdisc) } // QdiscReplace will replace a qdisc to the system. @@ -113,8 +114,8 @@ func QdiscReplace(qdisc Qdisc) error { // The handle MUST change. func (h *Handle) QdiscReplace(qdisc Qdisc) error { return h.qdiscModify( - syscall.RTM_NEWQDISC, - syscall.NLM_F_CREATE|syscall.NLM_F_REPLACE, + unix.RTM_NEWQDISC, + unix.NLM_F_CREATE|unix.NLM_F_REPLACE, qdisc) } @@ -128,13 +129,13 @@ func QdiscAdd(qdisc Qdisc) error { // Equivalent to: `tc qdisc add $qdisc` func (h *Handle) QdiscAdd(qdisc Qdisc) error { return h.qdiscModify( - syscall.RTM_NEWQDISC, - syscall.NLM_F_CREATE|syscall.NLM_F_EXCL, + unix.RTM_NEWQDISC, + unix.NLM_F_CREATE|unix.NLM_F_EXCL, qdisc) } func (h *Handle) qdiscModify(cmd, flags int, qdisc Qdisc) error { - req := h.newNetlinkRequest(cmd, flags|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(cmd, flags|unix.NLM_F_ACK) base := qdisc.Attrs() msg := &nl.TcMsg{ Family: nl.FAMILY_ALL, @@ -145,13 +146,13 @@ func (h *Handle) qdiscModify(cmd, flags int, qdisc Qdisc) error { req.AddData(msg) // When deleting don't bother building the rest of the netlink payload - if cmd != syscall.RTM_DELQDISC { + if cmd != unix.RTM_DELQDISC { if err := qdiscPayload(req, qdisc); err != nil { return err } } - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -160,75 +161,119 @@ func qdiscPayload(req *nl.NetlinkRequest, qdisc Qdisc) error { req.AddData(nl.NewRtAttr(nl.TCA_KIND, nl.ZeroTerminated(qdisc.Type()))) options := nl.NewRtAttr(nl.TCA_OPTIONS, nil) - if prio, ok := qdisc.(*Prio); ok { + + switch qdisc := qdisc.(type) { + case *Prio: tcmap := nl.TcPrioMap{ - Bands: int32(prio.Bands), - Priomap: prio.PriorityMap, + Bands: int32(qdisc.Bands), + Priomap: qdisc.PriorityMap, } options = nl.NewRtAttr(nl.TCA_OPTIONS, tcmap.Serialize()) - } else if tbf, ok := qdisc.(*Tbf); ok { + case *Tbf: opt := nl.TcTbfQopt{} - opt.Rate.Rate = uint32(tbf.Rate) - opt.Peakrate.Rate = uint32(tbf.Peakrate) - opt.Limit = tbf.Limit - opt.Buffer = tbf.Buffer + opt.Rate.Rate = uint32(qdisc.Rate) + opt.Peakrate.Rate = uint32(qdisc.Peakrate) + opt.Limit = qdisc.Limit + opt.Buffer = qdisc.Buffer nl.NewRtAttrChild(options, nl.TCA_TBF_PARMS, opt.Serialize()) - if tbf.Rate >= uint64(1<<32) { - nl.NewRtAttrChild(options, nl.TCA_TBF_RATE64, nl.Uint64Attr(tbf.Rate)) + if qdisc.Rate >= uint64(1<<32) { + nl.NewRtAttrChild(options, nl.TCA_TBF_RATE64, nl.Uint64Attr(qdisc.Rate)) } - if tbf.Peakrate >= uint64(1<<32) { - nl.NewRtAttrChild(options, nl.TCA_TBF_PRATE64, nl.Uint64Attr(tbf.Peakrate)) + if qdisc.Peakrate >= uint64(1<<32) { + nl.NewRtAttrChild(options, nl.TCA_TBF_PRATE64, nl.Uint64Attr(qdisc.Peakrate)) } - if tbf.Peakrate > 0 { - nl.NewRtAttrChild(options, nl.TCA_TBF_PBURST, nl.Uint32Attr(tbf.Minburst)) + if qdisc.Peakrate > 0 { + nl.NewRtAttrChild(options, nl.TCA_TBF_PBURST, nl.Uint32Attr(qdisc.Minburst)) } - } else if htb, ok := qdisc.(*Htb); ok { + case *Htb: opt := nl.TcHtbGlob{} - opt.Version = htb.Version - opt.Rate2Quantum = htb.Rate2Quantum - opt.Defcls = htb.Defcls + opt.Version = qdisc.Version + opt.Rate2Quantum = qdisc.Rate2Quantum + opt.Defcls = qdisc.Defcls // TODO: Handle Debug properly. For now default to 0 - opt.Debug = htb.Debug - opt.DirectPkts = htb.DirectPkts + 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()) - } else if netem, ok := qdisc.(*Netem); ok { + case *Netem: opt := nl.TcNetemQopt{} - opt.Latency = netem.Latency - opt.Limit = netem.Limit - opt.Loss = netem.Loss - opt.Gap = netem.Gap - opt.Duplicate = netem.Duplicate - opt.Jitter = netem.Jitter + opt.Latency = qdisc.Latency + opt.Limit = qdisc.Limit + opt.Loss = qdisc.Loss + opt.Gap = qdisc.Gap + opt.Duplicate = qdisc.Duplicate + opt.Jitter = qdisc.Jitter options = nl.NewRtAttr(nl.TCA_OPTIONS, opt.Serialize()) // Correlation corr := nl.TcNetemCorr{} - corr.DelayCorr = netem.DelayCorr - corr.LossCorr = netem.LossCorr - corr.DupCorr = netem.DuplicateCorr + corr.DelayCorr = qdisc.DelayCorr + corr.LossCorr = qdisc.LossCorr + corr.DupCorr = qdisc.DuplicateCorr if corr.DelayCorr > 0 || corr.LossCorr > 0 || corr.DupCorr > 0 { nl.NewRtAttrChild(options, nl.TCA_NETEM_CORR, corr.Serialize()) } // Corruption corruption := nl.TcNetemCorrupt{} - corruption.Probability = netem.CorruptProb - corruption.Correlation = netem.CorruptCorr + corruption.Probability = qdisc.CorruptProb + corruption.Correlation = qdisc.CorruptCorr if corruption.Probability > 0 { nl.NewRtAttrChild(options, nl.TCA_NETEM_CORRUPT, corruption.Serialize()) } // Reorder reorder := nl.TcNetemReorder{} - reorder.Probability = netem.ReorderProb - reorder.Correlation = netem.ReorderCorr + reorder.Probability = qdisc.ReorderProb + reorder.Correlation = qdisc.ReorderCorr if reorder.Probability > 0 { nl.NewRtAttrChild(options, nl.TCA_NETEM_REORDER, reorder.Serialize()) } - } else if _, ok := qdisc.(*Ingress); ok { + case *Ingress: // ingress filters must use the proper handle if qdisc.Attrs().Parent != HANDLE_INGRESS { 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)))) + if qdisc.Limit > 0 { + nl.NewRtAttrChild(options, 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)))) + } + if qdisc.Flows > 0 { + nl.NewRtAttrChild(options, 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)))) + } + + case *Fq: + nl.NewRtAttrChild(options, 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)))) + } + if qdisc.LowRateThreshold > 0 { + nl.NewRtAttrChild(options, 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)))) + } + if qdisc.InitialQuantum > 0 { + nl.NewRtAttrChild(options, 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)))) + } + if qdisc.FlowPacketLimit > 0 { + nl.NewRtAttrChild(options, 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)))) + } + if qdisc.FlowDefaultRate > 0 { + nl.NewRtAttrChild(options, nl.TCA_FQ_FLOW_DEFAULT_RATE, nl.Uint32Attr((uint32(qdisc.FlowDefaultRate)))) + } } req.AddData(options) @@ -246,7 +291,7 @@ func QdiscList(link Link) ([]Qdisc, error) { // Equivalent to: `tc qdisc show`. // The list can be filtered by link. func (h *Handle) QdiscList(link Link) ([]Qdisc, error) { - req := h.newNetlinkRequest(syscall.RTM_GETQDISC, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(unix.RTM_GETQDISC, unix.NLM_F_DUMP) index := int32(0) if link != nil { base := link.Attrs() @@ -259,7 +304,7 @@ func (h *Handle) QdiscList(link Link) ([]Qdisc, error) { } req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWQDISC) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWQDISC) if err != nil { return nil, err } @@ -301,6 +346,10 @@ func (h *Handle) QdiscList(link Link) ([]Qdisc, error) { qdisc = &Ingress{} case "htb": qdisc = &Htb{} + case "fq": + qdisc = &Fq{} + case "fq_codel": + qdisc = &FqCodel{} case "netem": qdisc = &Netem{} default: @@ -334,6 +383,22 @@ func (h *Handle) QdiscList(link Link) ([]Qdisc, error) { if err := parseHtbData(qdisc, data); err != nil { return nil, err } + case "fq": + data, err := nl.ParseRouteAttr(attr.Value) + if err != nil { + return nil, err + } + if err := parseFqData(qdisc, data); err != nil { + return nil, err + } + case "fq_codel": + data, err := nl.ParseRouteAttr(attr.Value) + if err != nil { + return nil, err + } + if err := parseFqCodelData(qdisc, data); err != nil { + return nil, err + } case "netem": if err := parseNetemData(qdisc, attr.Value); err != nil { return nil, err @@ -386,6 +451,61 @@ func parseHtbData(qdisc Qdisc, data []syscall.NetlinkRouteAttr) error { return nil } +func parseFqCodelData(qdisc Qdisc, data []syscall.NetlinkRouteAttr) error { + native = nl.NativeEndian() + fqCodel := qdisc.(*FqCodel) + for _, datum := range data { + + switch datum.Attr.Type { + case nl.TCA_FQ_CODEL_TARGET: + fqCodel.Target = native.Uint32(datum.Value) + case nl.TCA_FQ_CODEL_LIMIT: + fqCodel.Limit = native.Uint32(datum.Value) + case nl.TCA_FQ_CODEL_INTERVAL: + fqCodel.Interval = native.Uint32(datum.Value) + case nl.TCA_FQ_CODEL_ECN: + fqCodel.ECN = native.Uint32(datum.Value) + case nl.TCA_FQ_CODEL_FLOWS: + fqCodel.Flows = native.Uint32(datum.Value) + case nl.TCA_FQ_CODEL_QUANTUM: + fqCodel.Quantum = native.Uint32(datum.Value) + } + } + return nil +} + +func parseFqData(qdisc Qdisc, data []syscall.NetlinkRouteAttr) error { + native = nl.NativeEndian() + fq := qdisc.(*Fq) + for _, datum := range data { + switch datum.Attr.Type { + case nl.TCA_FQ_BUCKETS_LOG: + fq.Buckets = native.Uint32(datum.Value) + case nl.TCA_FQ_LOW_RATE_THRESHOLD: + fq.LowRateThreshold = native.Uint32(datum.Value) + case nl.TCA_FQ_QUANTUM: + fq.Quantum = native.Uint32(datum.Value) + case nl.TCA_FQ_RATE_ENABLE: + fq.Pacing = native.Uint32(datum.Value) + case nl.TCA_FQ_INITIAL_QUANTUM: + fq.InitialQuantum = native.Uint32(datum.Value) + case nl.TCA_FQ_ORPHAN_MASK: + // TODO + case nl.TCA_FQ_FLOW_REFILL_DELAY: + fq.FlowRefillDelay = native.Uint32(datum.Value) + case nl.TCA_FQ_FLOW_PLIMIT: + fq.FlowPacketLimit = native.Uint32(datum.Value) + case nl.TCA_FQ_PLIMIT: + fq.PacketLimit = native.Uint32(datum.Value) + case nl.TCA_FQ_FLOW_MAX_RATE: + fq.FlowMaxRate = native.Uint32(datum.Value) + case nl.TCA_FQ_FLOW_DEFAULT_RATE: + fq.FlowDefaultRate = native.Uint32(datum.Value) + } + } + return nil +} + func parseNetemData(qdisc Qdisc, value []byte) error { netem := qdisc.(*Netem) opt := nl.DeserializeTcNetemQopt(value) diff --git a/vendor/github.com/vishvananda/netlink/route.go b/vendor/github.com/vishvananda/netlink/route.go index 03ac4b239..2cd58ee33 100644 --- a/vendor/github.com/vishvananda/netlink/route.go +++ b/vendor/github.com/vishvananda/netlink/route.go @@ -16,6 +16,7 @@ type Destination interface { Decode([]byte) error Encode() ([]byte, error) String() string + Equal(Destination) bool } type Encap interface { @@ -23,6 +24,7 @@ type Encap interface { Decode([]byte) error Encode() ([]byte, error) String() string + Equal(Encap) bool } // Route represents a netlink route. @@ -43,6 +45,8 @@ type Route struct { MPLSDst *int NewDst Destination Encap Encap + MTU int + AdvMSS int } func (r Route) String() string { @@ -72,6 +76,25 @@ func (r Route) String() string { return fmt.Sprintf("{%s}", strings.Join(elems, " ")) } +func (r Route) Equal(x Route) bool { + return r.LinkIndex == x.LinkIndex && + r.ILinkIndex == x.ILinkIndex && + r.Scope == x.Scope && + ipNetEqual(r.Dst, x.Dst) && + r.Src.Equal(x.Src) && + r.Gw.Equal(x.Gw) && + nexthopInfoSlice(r.MultiPath).Equal(x.MultiPath) && + r.Protocol == x.Protocol && + r.Priority == x.Priority && + r.Table == x.Table && + r.Type == x.Type && + r.Tos == x.Tos && + 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))) && + (r.Encap == x.Encap || (r.Encap != nil && r.Encap.Equal(x.Encap))) +} + func (r *Route) SetFlag(flag NextHopFlag) { r.Flags |= int(flag) } @@ -110,7 +133,46 @@ func (n *NexthopInfo) String() string { elems = append(elems, fmt.Sprintf("Encap: %s", n.Encap)) } elems = append(elems, fmt.Sprintf("Weight: %d", n.Hops+1)) - elems = append(elems, fmt.Sprintf("Gw: %d", n.Gw)) + elems = append(elems, fmt.Sprintf("Gw: %s", n.Gw)) elems = append(elems, fmt.Sprintf("Flags: %s", n.ListFlags())) return fmt.Sprintf("{%s}", strings.Join(elems, " ")) } + +func (n NexthopInfo) Equal(x NexthopInfo) bool { + return n.LinkIndex == x.LinkIndex && + n.Hops == x.Hops && + n.Gw.Equal(x.Gw) && + n.Flags == x.Flags && + (n.NewDst == x.NewDst || (n.NewDst != nil && n.NewDst.Equal(x.NewDst))) && + (n.Encap == x.Encap || (n.Encap != nil && n.Encap.Equal(x.Encap))) +} + +type nexthopInfoSlice []*NexthopInfo + +func (n nexthopInfoSlice) Equal(x []*NexthopInfo) bool { + if len(n) != len(x) { + return false + } + for i := range n { + if n[i] == nil || x[i] == nil { + return false + } + if !n[i].Equal(*x[i]) { + return false + } + } + return true +} + +// ipNetEqual returns true iff both IPNet are equal +func ipNetEqual(ipn1 *net.IPNet, ipn2 *net.IPNet) bool { + if ipn1 == ipn2 { + return true + } + if ipn1 == nil || ipn2 == nil { + return false + } + m1, _ := ipn1.Mask.Size() + m2, _ := ipn2.Mask.Size() + return m1 == m2 && ipn1.IP.Equal(ipn2.IP) +} diff --git a/vendor/github.com/vishvananda/netlink/route_linux.go b/vendor/github.com/vishvananda/netlink/route_linux.go index cd739e714..3f856711f 100644 --- a/vendor/github.com/vishvananda/netlink/route_linux.go +++ b/vendor/github.com/vishvananda/netlink/route_linux.go @@ -8,16 +8,17 @@ import ( "github.com/vishvananda/netlink/nl" "github.com/vishvananda/netns" + "golang.org/x/sys/unix" ) // RtAttr is shared so it is in netlink_linux.go const ( - SCOPE_UNIVERSE Scope = syscall.RT_SCOPE_UNIVERSE - SCOPE_SITE Scope = syscall.RT_SCOPE_SITE - SCOPE_LINK Scope = syscall.RT_SCOPE_LINK - SCOPE_HOST Scope = syscall.RT_SCOPE_HOST - SCOPE_NOWHERE Scope = syscall.RT_SCOPE_NOWHERE + SCOPE_UNIVERSE Scope = unix.RT_SCOPE_UNIVERSE + SCOPE_SITE Scope = unix.RT_SCOPE_SITE + SCOPE_LINK Scope = unix.RT_SCOPE_LINK + SCOPE_HOST Scope = unix.RT_SCOPE_HOST + SCOPE_NOWHERE Scope = unix.RT_SCOPE_NOWHERE ) const ( @@ -34,8 +35,8 @@ const ( ) const ( - FLAG_ONLINK NextHopFlag = syscall.RTNH_F_ONLINK - FLAG_PERVASIVE NextHopFlag = syscall.RTNH_F_PERVASIVE + FLAG_ONLINK NextHopFlag = unix.RTNH_F_ONLINK + FLAG_PERVASIVE NextHopFlag = unix.RTNH_F_PERVASIVE ) var testFlags = []flagString{ @@ -86,6 +87,34 @@ func (d *MPLSDestination) String() string { return strings.Join(s, "/") } +func (d *MPLSDestination) Equal(x Destination) bool { + o, ok := x.(*MPLSDestination) + if !ok { + return false + } + if d == nil && o == nil { + return true + } + if d == nil || o == nil { + return false + } + if d.Labels == nil && o.Labels == nil { + return true + } + if d.Labels == nil || o.Labels == nil { + return false + } + if len(d.Labels) != len(o.Labels) { + return false + } + for i := range d.Labels { + if d.Labels[i] != o.Labels[i] { + return false + } + } + return true +} + type MPLSEncap struct { Labels []int } @@ -96,17 +125,17 @@ func (e *MPLSEncap) Type() int { func (e *MPLSEncap) Decode(buf []byte) error { if len(buf) < 4 { - return fmt.Errorf("Lack of bytes") + return fmt.Errorf("lack of bytes") } native := nl.NativeEndian() l := native.Uint16(buf) if len(buf) < int(l) { - return fmt.Errorf("Lack of bytes") + return fmt.Errorf("lack of bytes") } buf = buf[:l] typ := native.Uint16(buf[2:]) if typ != nl.MPLS_IPTUNNEL_DST { - return fmt.Errorf("Unknown MPLS Encap Type: %d", typ) + return fmt.Errorf("unknown MPLS Encap Type: %d", typ) } e.Labels = nl.DecodeMPLSStack(buf[4:]) return nil @@ -129,6 +158,107 @@ func (e *MPLSEncap) String() string { return strings.Join(s, "/") } +func (e *MPLSEncap) Equal(x Encap) bool { + o, ok := x.(*MPLSEncap) + if !ok { + return false + } + if e == nil && o == nil { + return true + } + if e == nil || o == nil { + return false + } + if e.Labels == nil && o.Labels == nil { + return true + } + if e.Labels == nil || o.Labels == nil { + return false + } + if len(e.Labels) != len(o.Labels) { + return false + } + for i := range e.Labels { + if e.Labels[i] != o.Labels[i] { + return false + } + } + return true +} + +// SEG6 definitions +type SEG6Encap struct { + Mode int + Segments []net.IP +} + +func (e *SEG6Encap) Type() int { + return nl.LWTUNNEL_ENCAP_SEG6 +} +func (e *SEG6Encap) Decode(buf []byte) error { + if len(buf) < 4 { + return fmt.Errorf("lack of bytes") + } + native := nl.NativeEndian() + // Get Length(l) & Type(typ) : 2 + 2 bytes + l := native.Uint16(buf) + if len(buf) < int(l) { + return fmt.Errorf("lack of bytes") + } + buf = buf[:l] // make sure buf size upper limit is Length + typ := native.Uint16(buf[2:]) + if typ != nl.SEG6_IPTUNNEL_SRH { + return fmt.Errorf("unknown SEG6 Type: %d", typ) + } + + var err error + e.Mode, e.Segments, err = nl.DecodeSEG6Encap(buf[4:]) + + return err +} +func (e *SEG6Encap) Encode() ([]byte, error) { + s, err := nl.EncodeSEG6Encap(e.Mode, e.Segments) + native := nl.NativeEndian() + hdr := make([]byte, 4) + native.PutUint16(hdr, uint16(len(s)+4)) + native.PutUint16(hdr[2:], nl.SEG6_IPTUNNEL_SRH) + return append(hdr, s...), err +} +func (e *SEG6Encap) String() string { + 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])) + } + str := fmt.Sprintf("mode %s segs %d [ %s ]", nl.SEG6EncapModeString(e.Mode), + len(e.Segments), strings.Join(segs, " ")) + return str +} +func (e *SEG6Encap) Equal(x Encap) bool { + o, ok := x.(*SEG6Encap) + if !ok { + return false + } + if e == o { + return true + } + if e == nil || o == nil { + return false + } + if e.Mode != o.Mode { + 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 + } + } + return true +} + // RouteAdd will add a route to the system. // Equivalent to: `ip route add $route` func RouteAdd(route *Route) error { @@ -138,8 +268,8 @@ func RouteAdd(route *Route) error { // RouteAdd will add a route to the system. // Equivalent to: `ip route add $route` func (h *Handle) RouteAdd(route *Route) error { - flags := syscall.NLM_F_CREATE | syscall.NLM_F_EXCL | syscall.NLM_F_ACK - req := h.newNetlinkRequest(syscall.RTM_NEWROUTE, flags) + flags := unix.NLM_F_CREATE | unix.NLM_F_EXCL | unix.NLM_F_ACK + req := h.newNetlinkRequest(unix.RTM_NEWROUTE, flags) return h.routeHandle(route, req, nl.NewRtMsg()) } @@ -152,8 +282,8 @@ func RouteReplace(route *Route) error { // RouteReplace will add a route to the system. // Equivalent to: `ip route replace $route` func (h *Handle) RouteReplace(route *Route) error { - flags := syscall.NLM_F_CREATE | syscall.NLM_F_REPLACE | syscall.NLM_F_ACK - req := h.newNetlinkRequest(syscall.RTM_NEWROUTE, flags) + flags := unix.NLM_F_CREATE | unix.NLM_F_REPLACE | unix.NLM_F_ACK + req := h.newNetlinkRequest(unix.RTM_NEWROUTE, flags) return h.routeHandle(route, req, nl.NewRtMsg()) } @@ -166,7 +296,7 @@ func RouteDel(route *Route) error { // RouteDel will delete a route from the system. // Equivalent to: `ip route del $route` func (h *Handle) RouteDel(route *Route) error { - req := h.newNetlinkRequest(syscall.RTM_DELROUTE, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_DELROUTE, unix.NLM_F_ACK) return h.routeHandle(route, req, nl.NewRtDelMsg()) } @@ -189,12 +319,12 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg } else { dstData = route.Dst.IP.To16() } - rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_DST, dstData)) + rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_DST, dstData)) } else if route.MPLSDst != nil { family = nl.FAMILY_MPLS msg.Dst_len = uint8(20) - msg.Type = syscall.RTN_UNICAST - rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_DST, nl.EncodeMPLSStack(*route.MPLSDst))) + msg.Type = unix.RTN_UNICAST + rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_DST, nl.EncodeMPLSStack(*route.MPLSDst))) } if route.NewDst != nil { @@ -232,7 +362,7 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg srcData = route.Src.To16() } // The commonly used src ip for routes is actually PREFSRC - rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_PREFSRC, srcData)) + rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_PREFSRC, srcData)) } if route.Gw != nil { @@ -247,14 +377,14 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg } else { gwData = route.Gw.To16() } - rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_GATEWAY, gwData)) + rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_GATEWAY, gwData)) } if len(route.MultiPath) > 0 { buf := []byte{} for _, nh := range route.MultiPath { rtnh := &nl.RtNexthop{ - RtNexthop: syscall.RtNexthop{ + RtNexthop: unix.RtNexthop{ Hops: uint8(nh.Hops), Ifindex: int32(nh.LinkIndex), Flags: uint8(nh.Flags), @@ -267,9 +397,9 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg return fmt.Errorf("gateway, source, and destination ip are not the same IP family") } if gwFamily == FAMILY_V4 { - children = append(children, nl.NewRtAttr(syscall.RTA_GATEWAY, []byte(nh.Gw.To4()))) + children = append(children, nl.NewRtAttr(unix.RTA_GATEWAY, []byte(nh.Gw.To4()))) } else { - children = append(children, nl.NewRtAttr(syscall.RTA_GATEWAY, []byte(nh.Gw.To16()))) + children = append(children, nl.NewRtAttr(unix.RTA_GATEWAY, []byte(nh.Gw.To16()))) } } if nh.NewDst != nil { @@ -295,15 +425,15 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg rtnh.Children = children buf = append(buf, rtnh.Serialize()...) } - rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_MULTIPATH, buf)) + rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_MULTIPATH, buf)) } if route.Table > 0 { if route.Table >= 256 { - msg.Table = syscall.RT_TABLE_UNSPEC + msg.Table = unix.RT_TABLE_UNSPEC b := make([]byte, 4) native.PutUint32(b, uint32(route.Table)) - rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_TABLE, b)) + rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_TABLE, b)) } else { msg.Table = uint8(route.Table) } @@ -312,7 +442,7 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg if route.Priority > 0 { b := make([]byte, 4) native.PutUint32(b, uint32(route.Priority)) - rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_PRIORITY, b)) + rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_PRIORITY, b)) } if route.Tos > 0 { msg.Tos = uint8(route.Tos) @@ -324,6 +454,25 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg msg.Type = uint8(route.Type) } + var metrics []*nl.RtAttr + // TODO: support other rta_metric values + if route.MTU > 0 { + b := nl.Uint32Attr(uint32(route.MTU)) + metrics = append(metrics, nl.NewRtAttr(unix.RTAX_MTU, b)) + } + if route.AdvMSS > 0 { + b := nl.Uint32Attr(uint32(route.AdvMSS)) + metrics = append(metrics, nl.NewRtAttr(unix.RTAX_ADVMSS, b)) + } + + if metrics != nil { + attr := nl.NewRtAttr(unix.RTA_METRICS, nil) + for _, metric := range metrics { + attr.AddChild(metric) + } + rtAttrs = append(rtAttrs, attr) + } + msg.Flags = uint32(route.Flags) msg.Scope = uint8(route.Scope) msg.Family = uint8(family) @@ -338,9 +487,9 @@ func (h *Handle) routeHandle(route *Route, req *nl.NetlinkRequest, msg *nl.RtMsg ) native.PutUint32(b, uint32(route.LinkIndex)) - req.AddData(nl.NewRtAttr(syscall.RTA_OIF, b)) + req.AddData(nl.NewRtAttr(unix.RTA_OIF, b)) - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -373,11 +522,11 @@ func RouteListFiltered(family int, filter *Route, filterMask uint64) ([]Route, e // RouteListFiltered gets a list of routes in the system filtered with specified rules. // All rules must be defined in RouteFilter struct func (h *Handle) RouteListFiltered(family int, filter *Route, filterMask uint64) ([]Route, error) { - req := h.newNetlinkRequest(syscall.RTM_GETROUTE, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(unix.RTM_GETROUTE, unix.NLM_F_DUMP) infmsg := nl.NewIfInfomsg(family) req.AddData(infmsg) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWROUTE) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWROUTE) if err != nil { return nil, err } @@ -385,11 +534,11 @@ func (h *Handle) RouteListFiltered(family int, filter *Route, filterMask uint64) var res []Route for _, m := range msgs { msg := nl.DeserializeRtMsg(m) - if msg.Flags&syscall.RTM_F_CLONED != 0 { + if msg.Flags&unix.RTM_F_CLONED != 0 { // Ignore cloned routes continue } - if msg.Table != syscall.RT_TABLE_MAIN { + if msg.Table != unix.RT_TABLE_MAIN { if filter == nil || filter != nil && filterMask&RT_FILTER_TABLE == 0 { // Ignore non-main tables continue @@ -401,7 +550,7 @@ func (h *Handle) RouteListFiltered(family int, filter *Route, filterMask uint64) } if filter != nil { switch { - case filterMask&RT_FILTER_TABLE != 0 && filter.Table != syscall.RT_TABLE_UNSPEC && route.Table != filter.Table: + case filterMask&RT_FILTER_TABLE != 0 && filter.Table != unix.RT_TABLE_UNSPEC && route.Table != filter.Table: continue case filterMask&RT_FILTER_PROTOCOL != 0 && route.Protocol != filter.Protocol: continue @@ -421,19 +570,8 @@ func (h *Handle) RouteListFiltered(family int, filter *Route, filterMask uint64) continue case filterMask&RT_FILTER_DST != 0: if filter.MPLSDst == nil || route.MPLSDst == nil || (*filter.MPLSDst) != (*route.MPLSDst) { - if filter.Dst == nil { - if route.Dst != nil { - continue - } - } else { - if route.Dst == nil { - continue - } - aMaskLen, aMaskBits := route.Dst.Mask.Size() - bMaskLen, bMaskBits := filter.Dst.Mask.Size() - if !(route.Dst.IP.Equal(filter.Dst.IP) && aMaskLen == bMaskLen && aMaskBits == bMaskBits) { - continue - } + if !ipNetEqual(route.Dst, filter.Dst) { + continue } } } @@ -463,11 +601,11 @@ func deserializeRoute(m []byte) (Route, error) { var encap, encapType syscall.NetlinkRouteAttr for _, attr := range attrs { switch attr.Attr.Type { - case syscall.RTA_GATEWAY: + case unix.RTA_GATEWAY: route.Gw = net.IP(attr.Value) - case syscall.RTA_PREFSRC: + case unix.RTA_PREFSRC: route.Src = net.IP(attr.Value) - case syscall.RTA_DST: + case unix.RTA_DST: if msg.Family == nl.FAMILY_MPLS { stack := nl.DecodeMPLSStack(attr.Value) if len(stack) == 0 || len(stack) > 1 { @@ -480,36 +618,36 @@ func deserializeRoute(m []byte) (Route, error) { Mask: net.CIDRMask(int(msg.Dst_len), 8*len(attr.Value)), } } - case syscall.RTA_OIF: + case unix.RTA_OIF: route.LinkIndex = int(native.Uint32(attr.Value[0:4])) - case syscall.RTA_IIF: + case unix.RTA_IIF: route.ILinkIndex = int(native.Uint32(attr.Value[0:4])) - case syscall.RTA_PRIORITY: + case unix.RTA_PRIORITY: route.Priority = int(native.Uint32(attr.Value[0:4])) - case syscall.RTA_TABLE: + case unix.RTA_TABLE: route.Table = int(native.Uint32(attr.Value[0:4])) - case syscall.RTA_MULTIPATH: + case unix.RTA_MULTIPATH: parseRtNexthop := func(value []byte) (*NexthopInfo, []byte, error) { - if len(value) < syscall.SizeofRtNexthop { - return nil, nil, fmt.Errorf("Lack of bytes") + if len(value) < unix.SizeofRtNexthop { + return nil, nil, fmt.Errorf("lack of bytes") } nh := nl.DeserializeRtNexthop(value) if len(value) < int(nh.RtNexthop.Len) { - return nil, nil, fmt.Errorf("Lack of bytes") + return nil, nil, fmt.Errorf("lack of bytes") } info := &NexthopInfo{ LinkIndex: int(nh.RtNexthop.Ifindex), Hops: int(nh.RtNexthop.Hops), Flags: int(nh.RtNexthop.Flags), } - attrs, err := nl.ParseRouteAttr(value[syscall.SizeofRtNexthop:int(nh.RtNexthop.Len)]) + attrs, err := nl.ParseRouteAttr(value[unix.SizeofRtNexthop:int(nh.RtNexthop.Len)]) if err != nil { return nil, nil, err } var encap, encapType syscall.NetlinkRouteAttr for _, attr := range attrs { switch attr.Attr.Type { - case syscall.RTA_GATEWAY: + case unix.RTA_GATEWAY: info.Gw = net.IP(attr.Value) case nl.RTA_NEWDST: var d Destination @@ -566,6 +704,19 @@ func deserializeRoute(m []byte) (Route, error) { encapType = attr case nl.RTA_ENCAP: encap = attr + case unix.RTA_METRICS: + metrics, err := nl.ParseRouteAttr(attr.Value) + if err != nil { + return route, err + } + for _, metric := range metrics { + switch metric.Attr.Type { + case unix.RTAX_MTU: + route.MTU = int(native.Uint32(metric.Value[0:4])) + case unix.RTAX_ADVMSS: + route.AdvMSS = int(native.Uint32(metric.Value[0:4])) + } + } } } @@ -578,6 +729,11 @@ func deserializeRoute(m []byte) (Route, error) { if err := e.Decode(encap.Value); err != nil { return route, err } + case nl.LWTUNNEL_ENCAP_SEG6: + e = &SEG6Encap{} + if err := e.Decode(encap.Value); err != nil { + return route, err + } } route.Encap = e } @@ -594,7 +750,7 @@ func RouteGet(destination net.IP) ([]Route, error) { // RouteGet gets a route to a specific destination from the host system. // Equivalent to: 'ip route get'. func (h *Handle) RouteGet(destination net.IP) ([]Route, error) { - req := h.newNetlinkRequest(syscall.RTM_GETROUTE, syscall.NLM_F_REQUEST) + req := h.newNetlinkRequest(unix.RTM_GETROUTE, unix.NLM_F_REQUEST) family := nl.GetIPFamily(destination) var destinationData []byte var bitlen uint8 @@ -610,10 +766,10 @@ func (h *Handle) RouteGet(destination net.IP) ([]Route, error) { msg.Dst_len = bitlen req.AddData(msg) - rtaDst := nl.NewRtAttr(syscall.RTA_DST, destinationData) + rtaDst := nl.NewRtAttr(unix.RTA_DST, destinationData) req.AddData(rtaDst) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWROUTE) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWROUTE) if err != nil { return nil, err } @@ -633,17 +789,36 @@ func (h *Handle) RouteGet(destination net.IP) ([]Route, error) { // RouteSubscribe takes a chan down which notifications will be sent // when routes are added or deleted. Close the 'done' chan to stop subscription. func RouteSubscribe(ch chan<- RouteUpdate, done <-chan struct{}) error { - return routeSubscribeAt(netns.None(), netns.None(), ch, done) + return routeSubscribeAt(netns.None(), netns.None(), ch, done, nil, false) } // RouteSubscribeAt works like RouteSubscribe plus it allows the caller // to choose the network namespace in which to subscribe (ns). func RouteSubscribeAt(ns netns.NsHandle, ch chan<- RouteUpdate, done <-chan struct{}) error { - return routeSubscribeAt(ns, netns.None(), ch, done) + return routeSubscribeAt(ns, netns.None(), ch, done, nil, false) +} + +// RouteSubscribeOptions contains a set of options to use with +// RouteSubscribeWithOptions. +type RouteSubscribeOptions struct { + Namespace *netns.NsHandle + ErrorCallback func(error) + ListExisting bool +} + +// RouteSubscribeWithOptions work like RouteSubscribe but enable to +// provide additional options to modify the behavior. Currently, the +// namespace can be provided as well as an error callback. +func RouteSubscribeWithOptions(ch chan<- RouteUpdate, done <-chan struct{}, options RouteSubscribeOptions) error { + if options.Namespace == nil { + none := netns.None() + options.Namespace = &none + } + return routeSubscribeAt(*options.Namespace, netns.None(), ch, done, options.ErrorCallback, options.ListExisting) } -func routeSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- RouteUpdate, done <-chan struct{}) error { - s, err := nl.SubscribeAt(newNs, curNs, syscall.NETLINK_ROUTE, syscall.RTNLGRP_IPV4_ROUTE, syscall.RTNLGRP_IPV6_ROUTE) +func routeSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- RouteUpdate, done <-chan struct{}, cberr func(error), listExisting bool) error { + s, err := nl.SubscribeAt(newNs, curNs, unix.NETLINK_ROUTE, unix.RTNLGRP_IPV4_ROUTE, unix.RTNLGRP_IPV6_ROUTE) if err != nil { return err } @@ -653,16 +828,45 @@ func routeSubscribeAt(newNs, curNs netns.NsHandle, ch chan<- RouteUpdate, done < s.Close() }() } + if listExisting { + req := pkgHandle.newNetlinkRequest(unix.RTM_GETROUTE, + unix.NLM_F_DUMP) + infmsg := nl.NewIfInfomsg(unix.AF_UNSPEC) + req.AddData(infmsg) + if err := s.Send(req); err != nil { + return err + } + } go func() { defer close(ch) for { msgs, err := s.Receive() if err != nil { + if cberr != nil { + cberr(err) + } return } for _, m := range msgs { + if m.Header.Type == unix.NLMSG_DONE { + 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 + } route, err := deserializeRoute(m.Data) if err != nil { + if cberr != nil { + cberr(err) + } return } ch <- RouteUpdate{Type: m.Header.Type, Route: route} diff --git a/vendor/github.com/vishvananda/netlink/rule.go b/vendor/github.com/vishvananda/netlink/rule.go index f0243defd..7fc8ae5df 100644 --- a/vendor/github.com/vishvananda/netlink/rule.go +++ b/vendor/github.com/vishvananda/netlink/rule.go @@ -8,6 +8,7 @@ import ( // Rule represents a netlink rule. type Rule struct { Priority int + Family int Table int Mark int Mask int @@ -20,6 +21,7 @@ type Rule struct { OifName string SuppressIfgroup int SuppressPrefixlen int + Invert bool } func (r Rule) String() string { diff --git a/vendor/github.com/vishvananda/netlink/rule_linux.go b/vendor/github.com/vishvananda/netlink/rule_linux.go index f9cdc855f..6238ae458 100644 --- a/vendor/github.com/vishvananda/netlink/rule_linux.go +++ b/vendor/github.com/vishvananda/netlink/rule_linux.go @@ -3,11 +3,13 @@ package netlink import ( "fmt" "net" - "syscall" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) +const FibRuleInvert = 0x2 + // RuleAdd adds a rule to the system. // Equivalent to: ip rule add func RuleAdd(rule *Rule) error { @@ -17,7 +19,7 @@ func RuleAdd(rule *Rule) error { // RuleAdd adds a rule to the system. // Equivalent to: ip rule add func (h *Handle) RuleAdd(rule *Rule) error { - req := h.newNetlinkRequest(syscall.RTM_NEWRULE, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_NEWRULE, unix.NLM_F_CREATE|unix.NLM_F_EXCL|unix.NLM_F_ACK) return ruleHandle(rule, req) } @@ -30,15 +32,31 @@ func RuleDel(rule *Rule) error { // RuleDel deletes a rule from the system. // Equivalent to: ip rule del func (h *Handle) RuleDel(rule *Rule) error { - req := h.newNetlinkRequest(syscall.RTM_DELRULE, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(unix.RTM_DELRULE, unix.NLM_F_ACK) return ruleHandle(rule, req) } func ruleHandle(rule *Rule, req *nl.NetlinkRequest) error { msg := nl.NewRtMsg() - msg.Family = syscall.AF_INET - var dstFamily uint8 + msg.Family = unix.AF_INET + msg.Protocol = unix.RTPROT_BOOT + msg.Scope = unix.RT_SCOPE_UNIVERSE + msg.Table = unix.RT_TABLE_UNSPEC + msg.Type = unix.RTN_UNSPEC + if req.NlMsghdr.Flags&unix.NLM_F_CREATE > 0 { + msg.Type = unix.RTN_UNICAST + } + if rule.Invert { + msg.Flags |= FibRuleInvert + } + if rule.Family != 0 { + msg.Family = uint8(rule.Family) + } + if rule.Table >= 0 && rule.Table < 256 { + msg.Table = uint8(rule.Table) + } + var dstFamily uint8 var rtAttrs []*nl.RtAttr if rule.Dst != nil && rule.Dst.IP != nil { dstLen, _ := rule.Dst.Mask.Size() @@ -46,12 +64,12 @@ func ruleHandle(rule *Rule, req *nl.NetlinkRequest) error { msg.Family = uint8(nl.GetIPFamily(rule.Dst.IP)) dstFamily = msg.Family var dstData []byte - if msg.Family == syscall.AF_INET { + if msg.Family == unix.AF_INET { dstData = rule.Dst.IP.To4() } else { dstData = rule.Dst.IP.To16() } - rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_DST, dstData)) + rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_DST, dstData)) } if rule.Src != nil && rule.Src.IP != nil { @@ -62,19 +80,12 @@ func ruleHandle(rule *Rule, req *nl.NetlinkRequest) error { srcLen, _ := rule.Src.Mask.Size() msg.Src_len = uint8(srcLen) var srcData []byte - if msg.Family == syscall.AF_INET { + if msg.Family == unix.AF_INET { srcData = rule.Src.IP.To4() } else { srcData = rule.Src.IP.To16() } - rtAttrs = append(rtAttrs, nl.NewRtAttr(syscall.RTA_SRC, srcData)) - } - - if rule.Table >= 0 { - msg.Table = uint8(rule.Table) - if rule.Table >= 256 { - msg.Table = syscall.RT_TABLE_UNSPEC - } + rtAttrs = append(rtAttrs, nl.NewRtAttr(unix.RTA_SRC, srcData)) } req.AddData(msg) @@ -139,7 +150,7 @@ func ruleHandle(rule *Rule, req *nl.NetlinkRequest) error { req.AddData(nl.NewRtAttr(nl.FRA_GOTO, b)) } - _, err := req.Execute(syscall.NETLINK_ROUTE, 0) + _, err := req.Execute(unix.NETLINK_ROUTE, 0) return err } @@ -152,11 +163,11 @@ func RuleList(family int) ([]Rule, error) { // RuleList lists rules in the system. // Equivalent to: ip rule list func (h *Handle) RuleList(family int) ([]Rule, error) { - req := h.newNetlinkRequest(syscall.RTM_GETRULE, syscall.NLM_F_DUMP|syscall.NLM_F_REQUEST) + req := h.newNetlinkRequest(unix.RTM_GETRULE, unix.NLM_F_DUMP|unix.NLM_F_REQUEST) msg := nl.NewIfInfomsg(family) req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_ROUTE, syscall.RTM_NEWRULE) + msgs, err := req.Execute(unix.NETLINK_ROUTE, unix.RTM_NEWRULE) if err != nil { return nil, err } @@ -172,9 +183,11 @@ func (h *Handle) RuleList(family int) ([]Rule, error) { rule := NewRule() + rule.Invert = msg.Flags&FibRuleInvert > 0 + for j := range attrs { switch attrs[j].Attr.Type { - case syscall.RTA_TABLE: + case unix.RTA_TABLE: rule.Table = int(native.Uint32(attrs[j].Value[0:4])) case nl.FRA_SRC: rule.Src = &net.IPNet{ diff --git a/vendor/github.com/vishvananda/netlink/socket_linux.go b/vendor/github.com/vishvananda/netlink/socket_linux.go index b42b84f0c..99e9fb4d8 100644 --- a/vendor/github.com/vishvananda/netlink/socket_linux.go +++ b/vendor/github.com/vishvananda/netlink/socket_linux.go @@ -4,9 +4,9 @@ import ( "errors" "fmt" "net" - "syscall" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) const ( @@ -123,15 +123,15 @@ func SocketGet(local, remote net.Addr) (*Socket, error) { return nil, ErrNotImplemented } - s, err := nl.Subscribe(syscall.NETLINK_INET_DIAG) + s, err := nl.Subscribe(unix.NETLINK_INET_DIAG) if err != nil { return nil, err } defer s.Close() req := nl.NewNetlinkRequest(nl.SOCK_DIAG_BY_FAMILY, 0) req.AddData(&socketRequest{ - Family: syscall.AF_INET, - Protocol: syscall.IPPROTO_TCP, + Family: unix.AF_INET, + Protocol: unix.IPPROTO_TCP, ID: SocketID{ SourcePort: uint16(localTCP.Port), DestinationPort: uint16(remoteTCP.Port), diff --git a/vendor/github.com/vishvananda/netlink/xfrm.go b/vendor/github.com/vishvananda/netlink/xfrm.go index 9962dcf70..02b41842e 100644 --- a/vendor/github.com/vishvananda/netlink/xfrm.go +++ b/vendor/github.com/vishvananda/netlink/xfrm.go @@ -2,19 +2,20 @@ package netlink import ( "fmt" - "syscall" + + "golang.org/x/sys/unix" ) // Proto is an enum representing an ipsec protocol. type Proto uint8 const ( - XFRM_PROTO_ROUTE2 Proto = syscall.IPPROTO_ROUTING - XFRM_PROTO_ESP Proto = syscall.IPPROTO_ESP - XFRM_PROTO_AH Proto = syscall.IPPROTO_AH - XFRM_PROTO_HAO Proto = syscall.IPPROTO_DSTOPTS + XFRM_PROTO_ROUTE2 Proto = unix.IPPROTO_ROUTING + XFRM_PROTO_ESP Proto = unix.IPPROTO_ESP + XFRM_PROTO_AH Proto = unix.IPPROTO_AH + XFRM_PROTO_HAO Proto = unix.IPPROTO_DSTOPTS XFRM_PROTO_COMP Proto = 0x6c // NOTE not defined on darwin - XFRM_PROTO_IPSEC_ANY Proto = syscall.IPPROTO_RAW + XFRM_PROTO_IPSEC_ANY Proto = unix.IPPROTO_RAW ) func (p Proto) String() string { diff --git a/vendor/github.com/vishvananda/netlink/xfrm_monitor_linux.go b/vendor/github.com/vishvananda/netlink/xfrm_monitor_linux.go index 7b98c9cb6..efe72ddf2 100644 --- a/vendor/github.com/vishvananda/netlink/xfrm_monitor_linux.go +++ b/vendor/github.com/vishvananda/netlink/xfrm_monitor_linux.go @@ -2,11 +2,10 @@ package netlink import ( "fmt" - "syscall" - - "github.com/vishvananda/netns" "github.com/vishvananda/netlink/nl" + "github.com/vishvananda/netns" + "golang.org/x/sys/unix" ) type XfrmMsg interface { @@ -39,7 +38,7 @@ func XfrmMonitor(ch chan<- XfrmMsg, done <-chan struct{}, errorChan chan<- error if err != nil { return nil } - s, err := nl.SubscribeAt(netns.None(), netns.None(), syscall.NETLINK_XFRM, groups...) + s, err := nl.SubscribeAt(netns.None(), netns.None(), unix.NETLINK_XFRM, groups...) if err != nil { return err } diff --git a/vendor/github.com/vishvananda/netlink/xfrm_policy_linux.go b/vendor/github.com/vishvananda/netlink/xfrm_policy_linux.go index c3d4e4222..fde0c2ca5 100644 --- a/vendor/github.com/vishvananda/netlink/xfrm_policy_linux.go +++ b/vendor/github.com/vishvananda/netlink/xfrm_policy_linux.go @@ -1,9 +1,8 @@ package netlink import ( - "syscall" - "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) func selFromPolicy(sel *nl.XfrmSelector, policy *XfrmPolicy) { @@ -55,7 +54,7 @@ func (h *Handle) XfrmPolicyUpdate(policy *XfrmPolicy) error { } func (h *Handle) xfrmPolicyAddOrUpdate(policy *XfrmPolicy, nlProto int) error { - req := h.newNetlinkRequest(nlProto, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(nlProto, unix.NLM_F_CREATE|unix.NLM_F_EXCL|unix.NLM_F_ACK) msg := &nl.XfrmUserpolicyInfo{} selFromPolicy(&msg.Sel, policy) @@ -91,7 +90,7 @@ func (h *Handle) xfrmPolicyAddOrUpdate(policy *XfrmPolicy, nlProto int) error { req.AddData(out) } - _, err := req.Execute(syscall.NETLINK_XFRM, 0) + _, err := req.Execute(unix.NETLINK_XFRM, 0) return err } @@ -121,12 +120,12 @@ func XfrmPolicyList(family int) ([]XfrmPolicy, error) { // Equivalent to: `ip xfrm policy show`. // The list can be filtered by ip family. func (h *Handle) XfrmPolicyList(family int) ([]XfrmPolicy, error) { - req := h.newNetlinkRequest(nl.XFRM_MSG_GETPOLICY, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(nl.XFRM_MSG_GETPOLICY, unix.NLM_F_DUMP) msg := nl.NewIfInfomsg(family) req.AddData(msg) - msgs, err := req.Execute(syscall.NETLINK_XFRM, nl.XFRM_MSG_NEWPOLICY) + msgs, err := req.Execute(unix.NETLINK_XFRM, nl.XFRM_MSG_NEWPOLICY) if err != nil { return nil, err } @@ -165,13 +164,13 @@ func XfrmPolicyFlush() error { // XfrmPolicyFlush will flush the policies on the system. // Equivalent to: `ip xfrm policy flush` func (h *Handle) XfrmPolicyFlush() error { - req := h.newNetlinkRequest(nl.XFRM_MSG_FLUSHPOLICY, syscall.NLM_F_ACK) - _, err := req.Execute(syscall.NETLINK_XFRM, 0) + req := h.newNetlinkRequest(nl.XFRM_MSG_FLUSHPOLICY, unix.NLM_F_ACK) + _, err := req.Execute(unix.NETLINK_XFRM, 0) return err } func (h *Handle) xfrmPolicyGetOrDelete(policy *XfrmPolicy, nlProto int) (*XfrmPolicy, error) { - req := h.newNetlinkRequest(nlProto, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(nlProto, unix.NLM_F_ACK) msg := &nl.XfrmUserpolicyId{} selFromPolicy(&msg.Sel, policy) @@ -189,7 +188,7 @@ func (h *Handle) xfrmPolicyGetOrDelete(policy *XfrmPolicy, nlProto int) (*XfrmPo resType = 0 } - msgs, err := req.Execute(syscall.NETLINK_XFRM, uint16(resType)) + msgs, err := req.Execute(unix.NETLINK_XFRM, uint16(resType)) if err != nil { return nil, err } diff --git a/vendor/github.com/vishvananda/netlink/xfrm_state.go b/vendor/github.com/vishvananda/netlink/xfrm_state.go index 368a9b986..d14740dc5 100644 --- a/vendor/github.com/vishvananda/netlink/xfrm_state.go +++ b/vendor/github.com/vishvananda/netlink/xfrm_state.go @@ -3,6 +3,7 @@ package netlink import ( "fmt" "net" + "time" ) // XfrmStateAlgo represents the algorithm to use for the ipsec encryption. @@ -67,6 +68,19 @@ type XfrmStateLimits struct { TimeUseHard uint64 } +// XfrmStateStats represents the current number of bytes/packets +// processed by this State, the State's installation and first use +// time and the replay window counters. +type XfrmStateStats struct { + ReplayWindow uint32 + Replay uint32 + Failed uint32 + Bytes uint64 + Packets uint64 + AddTime uint64 + UseTime uint64 +} + // XfrmState represents the state of an ipsec policy. It optionally // contains an XfrmStateAlgo for encryption and one for authentication. type XfrmState struct { @@ -78,6 +92,7 @@ type XfrmState struct { Reqid int ReplayWindow int Limits XfrmStateLimits + Statistics XfrmStateStats Mark *XfrmMark Auth *XfrmStateAlgo Crypt *XfrmStateAlgo @@ -94,10 +109,16 @@ func (sa XfrmState) Print(stats bool) string { if !stats { return sa.String() } - - return fmt.Sprintf("%s, ByteSoft: %s, ByteHard: %s, PacketSoft: %s, PacketHard: %s, TimeSoft: %d, TimeHard: %d, TimeUseSoft: %d, TimeUseHard: %d", + at := time.Unix(int64(sa.Statistics.AddTime), 0).Format(time.UnixDate) + ut := "-" + if sa.Statistics.UseTime > 0 { + ut = time.Unix(int64(sa.Statistics.UseTime), 0).Format(time.UnixDate) + } + return fmt.Sprintf("%s, ByteSoft: %s, ByteHard: %s, PacketSoft: %s, PacketHard: %s, TimeSoft: %d, TimeHard: %d, TimeUseSoft: %d, TimeUseHard: %d, Bytes: %d, Packets: %d, "+ + "AddTime: %s, UseTime: %s, ReplayWindow: %d, Replay: %d, Failed: %d", sa.String(), printLimit(sa.Limits.ByteSoft), printLimit(sa.Limits.ByteHard), printLimit(sa.Limits.PacketSoft), printLimit(sa.Limits.PacketHard), - sa.Limits.TimeSoft, sa.Limits.TimeHard, sa.Limits.TimeUseSoft, sa.Limits.TimeUseHard) + sa.Limits.TimeSoft, sa.Limits.TimeHard, sa.Limits.TimeUseSoft, sa.Limits.TimeUseHard, sa.Statistics.Bytes, sa.Statistics.Packets, at, ut, + sa.Statistics.ReplayWindow, sa.Statistics.Replay, sa.Statistics.Failed) } func printLimit(lmt uint64) string { diff --git a/vendor/github.com/vishvananda/netlink/xfrm_state_linux.go b/vendor/github.com/vishvananda/netlink/xfrm_state_linux.go index 6a7bc0dec..5dfdb33e4 100644 --- a/vendor/github.com/vishvananda/netlink/xfrm_state_linux.go +++ b/vendor/github.com/vishvananda/netlink/xfrm_state_linux.go @@ -2,10 +2,10 @@ package netlink import ( "fmt" - "syscall" "unsafe" "github.com/vishvananda/netlink/nl" + "golang.org/x/sys/unix" ) func writeStateAlgo(a *XfrmStateAlgo) []byte { @@ -69,8 +69,10 @@ func writeReplayEsn(replayWindow int) []byte { ReplayWindow: uint32(replayWindow), } - // taken from iproute2/ip/xfrm_state.c: - replayEsn.BmpLen = uint32((replayWindow + (4 * 8) - 1) / (4 * 8)) + // Linux stores the bitmap to identify the already received sequence packets in blocks of uint32 elements. + // Therefore bitmap length is the minimum number of uint32 elements needed. The following is a ceiling operation. + bytesPerElem := int(unsafe.Sizeof(replayEsn.BmpLen)) // Any uint32 variable is good for this + replayEsn.BmpLen = uint32((replayWindow + (bytesPerElem * 8) - 1) / (bytesPerElem * 8)) return replayEsn.Serialize() } @@ -111,7 +113,7 @@ func (h *Handle) xfrmStateAddOrUpdate(state *XfrmState, nlProto int) error { if state.Spi == 0 { return fmt.Errorf("Spi must be set when adding xfrm state.") } - req := h.newNetlinkRequest(nlProto, syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + req := h.newNetlinkRequest(nlProto, unix.NLM_F_CREATE|unix.NLM_F_EXCL|unix.NLM_F_ACK) msg := xfrmUsersaInfoFromXfrmState(state) @@ -157,13 +159,13 @@ func (h *Handle) xfrmStateAddOrUpdate(state *XfrmState, nlProto int) error { req.AddData(out) } - _, err := req.Execute(syscall.NETLINK_XFRM, 0) + _, err := req.Execute(unix.NETLINK_XFRM, 0) return err } func (h *Handle) xfrmStateAllocSpi(state *XfrmState) (*XfrmState, error) { req := h.newNetlinkRequest(nl.XFRM_MSG_ALLOCSPI, - syscall.NLM_F_CREATE|syscall.NLM_F_EXCL|syscall.NLM_F_ACK) + unix.NLM_F_CREATE|unix.NLM_F_EXCL|unix.NLM_F_ACK) msg := &nl.XfrmUserSpiInfo{} msg.XfrmUsersaInfo = *(xfrmUsersaInfoFromXfrmState(state)) @@ -177,7 +179,7 @@ func (h *Handle) xfrmStateAllocSpi(state *XfrmState) (*XfrmState, error) { req.AddData(out) } - msgs, err := req.Execute(syscall.NETLINK_XFRM, 0) + msgs, err := req.Execute(unix.NETLINK_XFRM, 0) if err != nil { return nil, err } @@ -216,9 +218,9 @@ func XfrmStateList(family int) ([]XfrmState, error) { // Equivalent to: `ip xfrm state show`. // The list can be filtered by ip family. func (h *Handle) XfrmStateList(family int) ([]XfrmState, error) { - req := h.newNetlinkRequest(nl.XFRM_MSG_GETSA, syscall.NLM_F_DUMP) + req := h.newNetlinkRequest(nl.XFRM_MSG_GETSA, unix.NLM_F_DUMP) - msgs, err := req.Execute(syscall.NETLINK_XFRM, nl.XFRM_MSG_NEWSA) + msgs, err := req.Execute(unix.NETLINK_XFRM, nl.XFRM_MSG_NEWSA) if err != nil { return nil, err } @@ -255,7 +257,7 @@ func (h *Handle) XfrmStateGet(state *XfrmState) (*XfrmState, error) { } func (h *Handle) xfrmStateGetOrDelete(state *XfrmState, nlProto int) (*XfrmState, error) { - req := h.newNetlinkRequest(nlProto, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(nlProto, unix.NLM_F_ACK) msg := &nl.XfrmUsersaId{} msg.Family = uint16(nl.GetIPFamily(state.Dst)) @@ -278,7 +280,7 @@ func (h *Handle) xfrmStateGetOrDelete(state *XfrmState, nlProto int) (*XfrmState resType = 0 } - msgs, err := req.Execute(syscall.NETLINK_XFRM, uint16(resType)) + msgs, err := req.Execute(unix.NETLINK_XFRM, uint16(resType)) if err != nil { return nil, err } @@ -308,6 +310,7 @@ func xfrmStateFromXfrmUsersaInfo(msg *nl.XfrmUsersaInfo) *XfrmState { state.Reqid = int(msg.Reqid) state.ReplayWindow = int(msg.ReplayWindow) lftToLimits(&msg.Lft, &state.Limits) + curToStats(&msg.Curlft, &msg.Stats, &state.Statistics) return &state } @@ -386,11 +389,11 @@ func XfrmStateFlush(proto Proto) error { // proto = 0 means any transformation protocols // Equivalent to: `ip xfrm state flush [ proto XFRM-PROTO ]` func (h *Handle) XfrmStateFlush(proto Proto) error { - req := h.newNetlinkRequest(nl.XFRM_MSG_FLUSHSA, syscall.NLM_F_ACK) + req := h.newNetlinkRequest(nl.XFRM_MSG_FLUSHSA, unix.NLM_F_ACK) req.AddData(&nl.XfrmUsersaFlush{Proto: uint8(proto)}) - _, err := req.Execute(syscall.NETLINK_XFRM, 0) + _, err := req.Execute(unix.NETLINK_XFRM, 0) if err != nil { return err } @@ -429,6 +432,16 @@ func lftToLimits(lft *nl.XfrmLifetimeCfg, lmts *XfrmStateLimits) { *lmts = *(*XfrmStateLimits)(unsafe.Pointer(lft)) } +func curToStats(cur *nl.XfrmLifetimeCur, wstats *nl.XfrmStats, stats *XfrmStateStats) { + stats.Bytes = cur.Bytes + stats.Packets = cur.Packets + stats.AddTime = cur.AddTime + stats.UseTime = cur.UseTime + stats.ReplayWindow = wstats.ReplayWindow + stats.Replay = wstats.Replay + stats.Failed = wstats.IntegrityFailed +} + func xfrmUsersaInfoFromXfrmState(state *XfrmState) *nl.XfrmUsersaInfo { msg := &nl.XfrmUsersaInfo{} msg.Family = uint16(nl.GetIPFamily(state.Dst)) diff --git a/vendor/github.com/vishvananda/netns/netns_linux.go b/vendor/github.com/vishvananda/netns/netns_linux.go index a267c7105..e665ef449 100644 --- a/vendor/github.com/vishvananda/netns/netns_linux.go +++ b/vendor/github.com/vishvananda/netns/netns_linux.go @@ -188,6 +188,12 @@ func getPidForContainer(id string) (int, error) { filepath.Join(cgroupRoot, "system.slice", "docker-"+id+".scope", "tasks"), // Even more recent docker versions under cgroup/systemd/docker/<id>/ filepath.Join(cgroupRoot, "..", "systemd", "docker", id, "tasks"), + // Kubernetes with docker and CNI is even more different + filepath.Join(cgroupRoot, "..", "systemd", "kubepods", "*", "pod*", id, "tasks"), + // Another flavor of containers location in recent kubernetes 1.11+ + filepath.Join(cgroupRoot, cgroupThis, "kubepods.slice", "kubepods-besteffort.slice", "*", "docker-"+id+".scope", "tasks"), + // When runs inside of a container with recent kubernetes 1.11+ + filepath.Join(cgroupRoot, "kubepods.slice", "kubepods-besteffort.slice", "*", "docker-"+id+".scope", "tasks"), } var filename string |