From 004274e828081d6a85654d2ee320286042987342 Mon Sep 17 00:00:00 2001 From: tc Date: Fri, 21 Jul 2023 12:06:13 +0800 Subject: [PATCH] add tc flower pedit action add pedit action in tc flower to support installing netlink hardware offload rules authored-by: tangchen --- filter.go | 27 ++++ filter_linux.go | 25 +++ filter_test.go | 160 ++++++++++++++++++++ nl/tc_linux.go | 393 ++++++++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 605 insertions(+) diff --git a/filter.go b/filter.go index fe41813..08e66ae 100644 --- a/filter.go +++ b/filter.go @@ -418,3 +418,30 @@ func (filter *GenericFilter) Attrs() *FilterAttrs { func (filter *GenericFilter) Type() string { return filter.FilterType } + +type PeditAction struct { + ActionAttrs + Proto uint8 + SrcMacAddr net.HardwareAddr + DstMacAddr net.HardwareAddr + SrcIP net.IP + DstIP net.IP + SrcPort uint16 + DstPort uint16 +} + +func (p *PeditAction) Attrs() *ActionAttrs { + return &p.ActionAttrs +} + +func (p *PeditAction) Type() string { + return "pedit" +} + +func NewPeditAction() *PeditAction { + return &PeditAction{ + ActionAttrs: ActionAttrs{ + Action: TC_ACT_PIPE, + }, + } +} diff --git a/filter_linux.go b/filter_linux.go index fbd434a..ae978b3 100644 --- a/filter_linux.go +++ b/filter_linux.go @@ -692,6 +692,29 @@ func EncodeActions(attr *nl.RtAttr, actions []Action) error { gen := nl.TcGen{} toTcGen(action.Attrs(), &gen) aopts.AddRtAttr(nl.TCA_GACT_PARMS, gen.Serialize()) + case *PeditAction: + table := attr.AddRtAttr(tabIndex, nil) + tabIndex++ + pedit := nl.TcPedit{} + if action.SrcMacAddr != nil { + pedit.SetEthSrc(action.SrcMacAddr) + } + if action.DstMacAddr != nil { + pedit.SetEthDst(action.DstMacAddr) + } + if action.SrcIP != nil { + pedit.SetSrcIP(action.SrcIP) + } + if action.DstIP != nil { + pedit.SetDstIP(action.DstIP) + } + if action.SrcPort != 0 { + pedit.SetSrcPort(action.SrcPort, action.Proto) + } + if action.DstPort != 0 { + pedit.SetDstPort(action.DstPort, action.Proto) + } + pedit.Encode(table) } } return nil @@ -752,6 +775,8 @@ func parseActions(tables []syscall.NetlinkRouteAttr) ([]Action, error) { action = &SkbEditAction{} case "police": action = &PoliceAction{} + case "pedit": + action = &PeditAction{} default: break nextattr } diff --git a/filter_test.go b/filter_test.go index 76ce866..5996512 100644 --- a/filter_test.go +++ b/filter_test.go @@ -1921,6 +1921,166 @@ func TestFilterFlowerAddDel(t *testing.T) { } } +func TestFilterIPv6FlowerPedit(t *testing.T) { + tearDown := setUpNetlinkTest(t) + defer tearDown() + if err := LinkAdd(&Ifb{LinkAttrs{Name: "foo"}}); err != nil { + t.Fatal(err) + } + if err := LinkAdd(&Ifb{LinkAttrs{Name: "bar"}}); err != nil { + t.Fatal(err) + } + link, err := LinkByName("foo") + if err != nil { + t.Fatal(err) + } + if err := LinkSetUp(link); err != nil { + t.Fatal(err) + } + redir, err := LinkByName("bar") + if err != nil { + t.Fatal(err) + } + if err := LinkSetUp(redir); err != nil { + t.Fatal(err) + } + + qdisc := &Ingress{ + QdiscAttrs: QdiscAttrs{ + LinkIndex: link.Attrs().Index, + Handle: MakeHandle(0xffff, 0), + Parent: HANDLE_INGRESS, + }, + } + if err := QdiscAdd(qdisc); err != nil { + t.Fatal(err) + } + qdiscs, err := SafeQdiscList(link) + if err != nil { + t.Fatal(err) + } + + found := false + for _, v := range qdiscs { + if _, ok := v.(*Ingress); ok { + found = true + break + } + } + if !found { + t.Fatal("Qdisc is the wrong type") + } + + testMask := net.CIDRMask(64, 128) + + ipproto := new(nl.IPProto) + *ipproto = nl.IPPROTO_TCP + + filter := &Flower{ + FilterAttrs: FilterAttrs{ + LinkIndex: link.Attrs().Index, + Parent: MakeHandle(0xffff, 0), + Priority: 1, + Protocol: unix.ETH_P_ALL, + }, + DestIP: net.ParseIP("ffff::fff1"), + DestIPMask: testMask, + EthType: unix.ETH_P_IPV6, + IPProto: ipproto, + DestPort: 6666, + Actions: []Action{}, + } + + peditAction := NewPeditAction() + peditAction.Proto = uint8(nl.IPPROTO_TCP) + peditAction.SrcPort = 7777 + peditAction.SrcIP = net.ParseIP("ffff::fff2") + filter.Actions = append(filter.Actions, peditAction) + + miaAction := &MirredAction{ + ActionAttrs: ActionAttrs{ + Action: TC_ACT_REDIRECT, + }, + MirredAction: TCA_EGRESS_REDIR, + Ifindex: redir.Attrs().Index, + } + filter.Actions = append(filter.Actions, miaAction) + + if err := FilterAdd(filter); err != nil { + t.Fatal(err) + } + + filters, err := FilterList(link, MakeHandle(0xffff, 0)) + if err != nil { + t.Fatal(err) + } + if len(filters) != 1 { + t.Fatal("Failed to add filter") + } + flower, ok := filters[0].(*Flower) + if !ok { + t.Fatal("Filter is the wrong type") + } + + if filter.EthType != flower.EthType { + t.Fatalf("Flower EthType doesn't match") + } + if !filter.DestIP.Equal(flower.DestIP) { + t.Fatalf("Flower DestIP doesn't match") + } + + if !reflect.DeepEqual(filter.DestIPMask, testMask) { + t.Fatalf("Flower DestIPMask doesn't match") + } + + if flower.IPProto == nil || *filter.IPProto != *flower.IPProto { + t.Fatalf("Flower IPProto doesn't match") + } + if filter.DestPort != flower.DestPort { + t.Fatalf("Flower DestPort doesn't match") + } + + _, ok = flower.Actions[0].(*PeditAction) + if !ok { + t.Fatal("Unable to find pedit action") + } + + _, ok = flower.Actions[1].(*MirredAction) + if !ok { + t.Fatal("Unable to find mirred action") + } + + if err := FilterDel(filter); err != nil { + t.Fatal(err) + } + filters, err = FilterList(link, MakeHandle(0xffff, 0)) + if err != nil { + t.Fatal(err) + } + if len(filters) != 0 { + t.Fatal("Failed to remove filter") + } + + if err := QdiscDel(qdisc); err != nil { + t.Fatal(err) + } + qdiscs, err = SafeQdiscList(link) + if err != nil { + t.Fatal(err) + } + + found = false + for _, v := range qdiscs { + if _, ok := v.(*Ingress); ok { + found = true + break + } + } + if found { + t.Fatal("Failed to remove qdisc") + } +} + func TestFilterU32PoliceAddDel(t *testing.T) { tearDown := setUpNetlinkTest(t) defer tearDown() diff --git a/nl/tc_linux.go b/nl/tc_linux.go index d8b60c2..c98173e 100644 --- a/nl/tc_linux.go +++ b/nl/tc_linux.go @@ -1,8 +1,10 @@ package nl import ( + "bytes" "encoding/binary" "fmt" + "net" "unsafe" "golang.org/x/sys/unix" @@ -1175,3 +1177,394 @@ func (i IPProto) String() string { } return fmt.Sprintf("%d", i) } + +const ( + MaxOffs = 128 + SizeOfPeditSel = 24 + SizeOfPeditKey = 24 + + TCA_PEDIT_KEY_EX_HTYPE = 1 + TCA_PEDIT_KEY_EX_CMD = 2 +) + +const ( + TCA_PEDIT_UNSPEC = iota + TCA_PEDIT_TM + TCA_PEDIT_PARMS + TCA_PEDIT_PAD + TCA_PEDIT_PARMS_EX + TCA_PEDIT_KEYS_EX + TCA_PEDIT_KEY_EX +) + +// /* TCA_PEDIT_KEY_EX_HDR_TYPE_NETWROK is a special case for legacy users. It +// * means no specific header type - offset is relative to the network layer +// */ +type PeditHeaderType uint16 + +const ( + TCA_PEDIT_KEY_EX_HDR_TYPE_NETWORK = iota + TCA_PEDIT_KEY_EX_HDR_TYPE_ETH + TCA_PEDIT_KEY_EX_HDR_TYPE_IP4 + TCA_PEDIT_KEY_EX_HDR_TYPE_IP6 + TCA_PEDIT_KEY_EX_HDR_TYPE_TCP + TCA_PEDIT_KEY_EX_HDR_TYPE_UDP + __PEDIT_HDR_TYPE_MAX +) + +type PeditCmd uint16 + +const ( + TCA_PEDIT_KEY_EX_CMD_SET = 0 + TCA_PEDIT_KEY_EX_CMD_ADD = 1 +) + +type TcPeditSel struct { + TcGen + NKeys uint8 + Flags uint8 +} + +func DeserializeTcPeditKey(b []byte) *TcPeditKey { + return (*TcPeditKey)(unsafe.Pointer(&b[0:SizeOfPeditKey][0])) +} + +func DeserializeTcPedit(b []byte) (*TcPeditSel, []TcPeditKey) { + x := &TcPeditSel{} + copy((*(*[SizeOfPeditSel]byte)(unsafe.Pointer(x)))[:SizeOfPeditSel], b) + + var keys []TcPeditKey + + next := SizeOfPeditKey + var i uint8 + for i = 0; i < x.NKeys; i++ { + keys = append(keys, *DeserializeTcPeditKey(b[next:])) + next += SizeOfPeditKey + } + + return x, keys +} + +type TcPeditKey struct { + Mask uint32 + Val uint32 + Off uint32 + At uint32 + OffMask uint32 + Shift uint32 +} + +type TcPeditKeyEx struct { + HeaderType PeditHeaderType + Cmd PeditCmd +} + +type TcPedit struct { + Sel TcPeditSel + Keys []TcPeditKey + KeysEx []TcPeditKeyEx + Extend uint8 +} + +func (p *TcPedit) Encode(parent *RtAttr) { + parent.AddRtAttr(TCA_ACT_KIND, ZeroTerminated("pedit")) + actOpts := parent.AddRtAttr(TCA_ACT_OPTIONS, nil) + + bbuf := bytes.NewBuffer(make([]byte, 0, int(unsafe.Sizeof(p.Sel)+unsafe.Sizeof(p.Keys)))) + + bbuf.Write((*(*[SizeOfPeditSel]byte)(unsafe.Pointer(&p.Sel)))[:]) + + for i := uint8(0); i < p.Sel.NKeys; i++ { + bbuf.Write((*(*[SizeOfPeditKey]byte)(unsafe.Pointer(&p.Keys[i])))[:]) + } + actOpts.AddRtAttr(TCA_PEDIT_PARMS_EX, bbuf.Bytes()) + + exAttrs := actOpts.AddRtAttr(int(TCA_PEDIT_KEYS_EX|NLA_F_NESTED), nil) + for i := uint8(0); i < p.Sel.NKeys; i++ { + keyAttr := exAttrs.AddRtAttr(int(TCA_PEDIT_KEY_EX|NLA_F_NESTED), nil) + + htypeBuf := make([]byte, 2) + cmdBuf := make([]byte, 2) + + NativeEndian().PutUint16(htypeBuf, uint16(p.KeysEx[i].HeaderType)) + NativeEndian().PutUint16(cmdBuf, uint16(p.KeysEx[i].Cmd)) + + keyAttr.AddRtAttr(TCA_PEDIT_KEY_EX_HTYPE, htypeBuf) + keyAttr.AddRtAttr(TCA_PEDIT_KEY_EX_CMD, cmdBuf) + } +} + +func (p *TcPedit) SetEthDst(mac net.HardwareAddr) { + u32 := NativeEndian().Uint32(mac) + u16 := NativeEndian().Uint16(mac[4:]) + + tKey := TcPeditKey{} + tKeyEx := TcPeditKeyEx{} + + tKey.Val = u32 + + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_ETH + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + p.Sel.NKeys++ + + tKey = TcPeditKey{} + tKeyEx = TcPeditKeyEx{} + + tKey.Val = uint32(u16) + tKey.Mask = 0xffff0000 + tKey.Off = 4 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_ETH + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + + p.Sel.NKeys++ +} + +func (p *TcPedit) SetEthSrc(mac net.HardwareAddr) { + u16 := NativeEndian().Uint16(mac) + u32 := NativeEndian().Uint32(mac[2:]) + + tKey := TcPeditKey{} + tKeyEx := TcPeditKeyEx{} + + tKey.Val = uint32(u16) << 16 + tKey.Mask = 0x0000ffff + tKey.Off = 4 + + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_ETH + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + p.Sel.NKeys++ + + tKey = TcPeditKey{} + tKeyEx = TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Mask = 0 + tKey.Off = 8 + + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_ETH + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + + p.Sel.NKeys++ +} + +func (p *TcPedit) SetIPv6Src(ip6 net.IP) { + u32 := NativeEndian().Uint32(ip6[:4]) + + tKey := TcPeditKey{} + tKeyEx := TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 8 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP6 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + p.Sel.NKeys++ + + u32 = NativeEndian().Uint32(ip6[4:8]) + tKey = TcPeditKey{} + tKeyEx = TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 12 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP6 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + + p.Sel.NKeys++ + + u32 = NativeEndian().Uint32(ip6[8:12]) + tKey = TcPeditKey{} + tKeyEx = TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 16 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP6 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + + p.Sel.NKeys++ + + u32 = NativeEndian().Uint32(ip6[12:16]) + tKey = TcPeditKey{} + tKeyEx = TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 20 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP6 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + + p.Sel.NKeys++ +} + +func (p *TcPedit) SetDstIP(ip net.IP) { + if ip.To4() != nil { + p.SetIPv4Dst(ip) + } else { + p.SetIPv6Dst(ip) + } +} + +func (p *TcPedit) SetSrcIP(ip net.IP) { + if ip.To4() != nil { + p.SetIPv4Src(ip) + } else { + p.SetIPv6Src(ip) + } +} + +func (p *TcPedit) SetIPv6Dst(ip6 net.IP) { + u32 := NativeEndian().Uint32(ip6[:4]) + + tKey := TcPeditKey{} + tKeyEx := TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 24 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP6 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + p.Sel.NKeys++ + + u32 = NativeEndian().Uint32(ip6[4:8]) + tKey = TcPeditKey{} + tKeyEx = TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 28 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP6 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + + p.Sel.NKeys++ + + u32 = NativeEndian().Uint32(ip6[8:12]) + tKey = TcPeditKey{} + tKeyEx = TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 32 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP6 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + + p.Sel.NKeys++ + + u32 = NativeEndian().Uint32(ip6[12:16]) + tKey = TcPeditKey{} + tKeyEx = TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 36 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP6 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + + p.Sel.NKeys++ +} + +func (p *TcPedit) SetIPv4Src(ip net.IP) { + u32 := NativeEndian().Uint32(ip[:4]) + + tKey := TcPeditKey{} + tKeyEx := TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 12 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP4 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + p.Sel.NKeys++ +} + +func (p *TcPedit) SetIPv4Dst(ip net.IP) { + u32 := NativeEndian().Uint32(ip[:4]) + + tKey := TcPeditKey{} + tKeyEx := TcPeditKeyEx{} + + tKey.Val = u32 + tKey.Off = 16 + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_IP4 + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + p.Sel.NKeys++ +} + +// SetDstPort only tcp and udp are supported to set port +func (p *TcPedit) SetDstPort(dstPort uint16, protocol uint8) { + tKey := TcPeditKey{} + tKeyEx := TcPeditKeyEx{} + + switch protocol { + case unix.IPPROTO_TCP: + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_TCP + case unix.IPPROTO_UDP: + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_UDP + default: + return + } + + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + tKey.Val = uint32(Swap16(dstPort)) << 16 + tKey.Mask = 0x0000ffff + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + p.Sel.NKeys++ +} + +// SetSrcPort only tcp and udp are supported to set port +func (p *TcPedit) SetSrcPort(srcPort uint16, protocol uint8) { + tKey := TcPeditKey{} + tKeyEx := TcPeditKeyEx{} + + switch protocol { + case unix.IPPROTO_TCP: + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_TCP + case unix.IPPROTO_UDP: + tKeyEx.HeaderType = TCA_PEDIT_KEY_EX_HDR_TYPE_UDP + default: + return + } + + tKeyEx.Cmd = TCA_PEDIT_KEY_EX_CMD_SET + + tKey.Val = uint32(Swap16(srcPort)) + tKey.Mask = 0xffff0000 + p.Keys = append(p.Keys, tKey) + p.KeysEx = append(p.KeysEx, tKeyEx) + p.Sel.NKeys++ +}