diff options
Diffstat (limited to 'pkg/tcpip/network')
-rw-r--r-- | pkg/tcpip/network/arp/arp.go | 36 | ||||
-rw-r--r-- | pkg/tcpip/network/arp/arp_test.go | 60 | ||||
-rw-r--r-- | pkg/tcpip/network/fragmentation/fragmentation.go | 98 | ||||
-rw-r--r-- | pkg/tcpip/network/fragmentation/fragmentation_test.go | 139 | ||||
-rw-r--r-- | pkg/tcpip/network/fragmentation/reassembler.go | 4 | ||||
-rw-r--r-- | pkg/tcpip/network/fragmentation/reassembler_test.go | 2 | ||||
-rw-r--r-- | pkg/tcpip/network/ip_test.go | 80 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv4/BUILD | 2 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv4/icmp.go | 18 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv4/ipv4.go | 153 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv4/ipv4_test.go | 231 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv6/BUILD | 3 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv6/icmp.go | 27 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv6/icmp_test.go | 109 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv6/ipv6.go | 141 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv6/ipv6_test.go | 365 | ||||
-rw-r--r-- | pkg/tcpip/network/ipv6/ndp_test.go | 43 |
17 files changed, 1145 insertions, 366 deletions
diff --git a/pkg/tcpip/network/arp/arp.go b/pkg/tcpip/network/arp/arp.go index 9d0797af7..31a242482 100644 --- a/pkg/tcpip/network/arp/arp.go +++ b/pkg/tcpip/network/arp/arp.go @@ -80,7 +80,7 @@ func (e *endpoint) MaxHeaderLength() uint16 { func (e *endpoint) Close() {} -func (e *endpoint) WritePacket(*stack.Route, *stack.GSO, stack.NetworkHeaderParams, stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WritePacket(*stack.Route, *stack.GSO, stack.NetworkHeaderParams, *stack.PacketBuffer) *tcpip.Error { return tcpip.ErrNotSupported } @@ -94,16 +94,12 @@ func (e *endpoint) WritePackets(*stack.Route, *stack.GSO, stack.PacketBufferList return 0, tcpip.ErrNotSupported } -func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuffer) *tcpip.Error { return tcpip.ErrNotSupported } -func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { - v, ok := pkt.Data.PullUp(header.ARPSize) - if !ok { - return - } - h := header.ARP(v) +func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { + h := header.ARP(pkt.NetworkHeader) if !h.IsValid() { return } @@ -122,7 +118,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { copy(packet.ProtocolAddressSender(), h.ProtocolAddressTarget()) copy(packet.HardwareAddressTarget(), h.HardwareAddressSender()) copy(packet.ProtocolAddressTarget(), h.ProtocolAddressSender()) - e.linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, stack.PacketBuffer{ + e.linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, &stack.PacketBuffer{ Header: hdr, }) fallthrough // also fill the cache from requests @@ -164,9 +160,12 @@ func (*protocol) LinkAddressProtocol() tcpip.NetworkProtocolNumber { } // LinkAddressRequest implements stack.LinkAddressResolver.LinkAddressRequest. -func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack.LinkEndpoint) *tcpip.Error { +func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, remoteLinkAddr tcpip.LinkAddress, linkEP stack.LinkEndpoint) *tcpip.Error { r := &stack.Route{ - RemoteLinkAddress: broadcastMAC, + RemoteLinkAddress: remoteLinkAddr, + } + if len(r.RemoteLinkAddress) == 0 { + r.RemoteLinkAddress = header.EthernetBroadcastAddress } hdr := buffer.NewPrependable(int(linkEP.MaxHeaderLength()) + header.ARPSize) @@ -177,7 +176,7 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack. copy(h.ProtocolAddressSender(), localAddr) copy(h.ProtocolAddressTarget(), addr) - return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, stack.PacketBuffer{ + return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, &stack.PacketBuffer{ Header: hdr, }) } @@ -185,7 +184,7 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack. // ResolveStaticAddress implements stack.LinkAddressResolver.ResolveStaticAddress. func (*protocol) ResolveStaticAddress(addr tcpip.Address) (tcpip.LinkAddress, bool) { if addr == header.IPv4Broadcast { - return broadcastMAC, true + return header.EthernetBroadcastAddress, true } if header.IsV4MulticastAddress(addr) { return header.EthernetAddressFromMulticastIPv4Address(addr), true @@ -209,7 +208,16 @@ func (*protocol) Close() {} // Wait implements stack.TransportProtocol.Wait. func (*protocol) Wait() {} -var broadcastMAC = tcpip.LinkAddress([]byte{0xff, 0xff, 0xff, 0xff, 0xff, 0xff}) +// Parse implements stack.NetworkProtocol.Parse. +func (*protocol) Parse(pkt *stack.PacketBuffer) (proto tcpip.TransportProtocolNumber, hasTransportHdr bool, ok bool) { + hdr, ok := pkt.Data.PullUp(header.ARPSize) + if !ok { + return 0, false, false + } + pkt.NetworkHeader = hdr + pkt.Data.TrimFront(header.ARPSize) + return 0, false, true +} // NewProtocol returns an ARP network protocol. func NewProtocol() stack.NetworkProtocol { diff --git a/pkg/tcpip/network/arp/arp_test.go b/pkg/tcpip/network/arp/arp_test.go index 1646d9cde..a35a64a0f 100644 --- a/pkg/tcpip/network/arp/arp_test.go +++ b/pkg/tcpip/network/arp/arp_test.go @@ -32,10 +32,14 @@ import ( ) const ( - stackLinkAddr = tcpip.LinkAddress("\x0a\x0a\x0b\x0b\x0c\x0c") - stackAddr1 = tcpip.Address("\x0a\x00\x00\x01") - stackAddr2 = tcpip.Address("\x0a\x00\x00\x02") - stackAddrBad = tcpip.Address("\x0a\x00\x00\x03") + stackLinkAddr1 = tcpip.LinkAddress("\x0a\x0a\x0b\x0b\x0c\x0c") + stackLinkAddr2 = tcpip.LinkAddress("\x0b\x0b\x0c\x0c\x0d\x0d") + stackAddr1 = tcpip.Address("\x0a\x00\x00\x01") + stackAddr2 = tcpip.Address("\x0a\x00\x00\x02") + stackAddrBad = tcpip.Address("\x0a\x00\x00\x03") + + defaultChannelSize = 1 + defaultMTU = 65536 ) type testContext struct { @@ -50,8 +54,7 @@ func newTestContext(t *testing.T) *testContext { TransportProtocols: []stack.TransportProtocol{icmp.NewProtocol4()}, }) - const defaultMTU = 65536 - ep := channel.New(256, defaultMTU, stackLinkAddr) + ep := channel.New(defaultChannelSize, defaultMTU, stackLinkAddr1) wep := stack.LinkEndpoint(ep) if testing.Verbose() { @@ -103,7 +106,7 @@ func TestDirectRequest(t *testing.T) { inject := func(addr tcpip.Address) { copy(h.ProtocolAddressTarget(), addr) - c.linkEP.InjectInbound(arp.ProtocolNumber, stack.PacketBuffer{ + c.linkEP.InjectInbound(arp.ProtocolNumber, &stack.PacketBuffer{ Data: v.ToVectorisedView(), }) } @@ -119,7 +122,7 @@ func TestDirectRequest(t *testing.T) { if !rep.IsValid() { t.Fatalf("invalid ARP response pi.Pkt.Header.UsedLength()=%d", pi.Pkt.Header.UsedLength()) } - if got, want := tcpip.LinkAddress(rep.HardwareAddressSender()), stackLinkAddr; got != want { + if got, want := tcpip.LinkAddress(rep.HardwareAddressSender()), stackLinkAddr1; got != want { t.Errorf("got HardwareAddressSender = %s, want = %s", got, want) } if got, want := tcpip.Address(rep.ProtocolAddressSender()), tcpip.Address(h.ProtocolAddressTarget()); got != want { @@ -144,3 +147,44 @@ func TestDirectRequest(t *testing.T) { t.Errorf("stackAddrBad: unexpected packet sent, Proto=%v", pkt.Proto) } } + +func TestLinkAddressRequest(t *testing.T) { + tests := []struct { + name string + remoteLinkAddr tcpip.LinkAddress + expectLinkAddr tcpip.LinkAddress + }{ + { + name: "Unicast", + remoteLinkAddr: stackLinkAddr2, + expectLinkAddr: stackLinkAddr2, + }, + { + name: "Multicast", + remoteLinkAddr: "", + expectLinkAddr: header.EthernetBroadcastAddress, + }, + } + + for _, test := range tests { + p := arp.NewProtocol() + linkRes, ok := p.(stack.LinkAddressResolver) + if !ok { + t.Fatal("expected ARP protocol to implement stack.LinkAddressResolver") + } + + linkEP := channel.New(defaultChannelSize, defaultMTU, stackLinkAddr1) + if err := linkRes.LinkAddressRequest(stackAddr1, stackAddr2, test.remoteLinkAddr, linkEP); err != nil { + t.Errorf("got p.LinkAddressRequest(%s, %s, %s, _) = %s", stackAddr1, stackAddr2, test.remoteLinkAddr, err) + } + + pkt, ok := linkEP.Read() + if !ok { + t.Fatal("expected to send a link address request") + } + + if got, want := pkt.Route.RemoteLinkAddress, test.expectLinkAddr; got != want { + t.Errorf("got pkt.Route.RemoteLinkAddress = %s, want = %s", got, want) + } + } +} diff --git a/pkg/tcpip/network/fragmentation/fragmentation.go b/pkg/tcpip/network/fragmentation/fragmentation.go index f42abc4bb..1827666c5 100644 --- a/pkg/tcpip/network/fragmentation/fragmentation.go +++ b/pkg/tcpip/network/fragmentation/fragmentation.go @@ -17,28 +17,58 @@ package fragmentation import ( + "errors" "fmt" "log" "time" "gvisor.dev/gvisor/pkg/sync" + "gvisor.dev/gvisor/pkg/tcpip" "gvisor.dev/gvisor/pkg/tcpip/buffer" ) -// DefaultReassembleTimeout is based on the linux stack: net.ipv4.ipfrag_time. -const DefaultReassembleTimeout = 30 * time.Second +const ( + // DefaultReassembleTimeout is based on the linux stack: net.ipv4.ipfrag_time. + DefaultReassembleTimeout = 30 * time.Second -// HighFragThreshold is the threshold at which we start trimming old -// fragmented packets. Linux uses a default value of 4 MB. See -// net.ipv4.ipfrag_high_thresh for more information. -const HighFragThreshold = 4 << 20 // 4MB + // HighFragThreshold is the threshold at which we start trimming old + // fragmented packets. Linux uses a default value of 4 MB. See + // net.ipv4.ipfrag_high_thresh for more information. + HighFragThreshold = 4 << 20 // 4MB -// LowFragThreshold is the threshold we reach to when we start dropping -// older fragmented packets. It's important that we keep enough room for newer -// packets to be re-assembled. Hence, this needs to be lower than -// HighFragThreshold enough. Linux uses a default value of 3 MB. See -// net.ipv4.ipfrag_low_thresh for more information. -const LowFragThreshold = 3 << 20 // 3MB + // LowFragThreshold is the threshold we reach to when we start dropping + // older fragmented packets. It's important that we keep enough room for newer + // packets to be re-assembled. Hence, this needs to be lower than + // HighFragThreshold enough. Linux uses a default value of 3 MB. See + // net.ipv4.ipfrag_low_thresh for more information. + LowFragThreshold = 3 << 20 // 3MB + + // minBlockSize is the minimum block size for fragments. + minBlockSize = 1 +) + +var ( + // ErrInvalidArgs indicates to the caller that that an invalid argument was + // provided. + ErrInvalidArgs = errors.New("invalid args") +) + +// FragmentID is the identifier for a fragment. +type FragmentID struct { + // Source is the source address of the fragment. + Source tcpip.Address + + // Destination is the destination address of the fragment. + Destination tcpip.Address + + // ID is the identification value of the fragment. + // + // This is a uint32 because IPv6 uses a 32-bit identification value. + ID uint32 + + // The protocol for the packet. + Protocol uint8 +} // Fragmentation is the main structure that other modules // of the stack should use to implement IP Fragmentation. @@ -46,14 +76,17 @@ type Fragmentation struct { mu sync.Mutex highLimit int lowLimit int - reassemblers map[uint32]*reassembler + reassemblers map[FragmentID]*reassembler rList reassemblerList size int timeout time.Duration + blockSize uint16 } // NewFragmentation creates a new Fragmentation. // +// blockSize specifies the fragment block size, in bytes. +// // highMemoryLimit specifies the limit on the memory consumed // by the fragments stored by Fragmentation (overhead of internal data-structures // is not accounted). Fragments are dropped when the limit is reached. @@ -64,7 +97,7 @@ type Fragmentation struct { // reassemblingTimeout specifies the maximum time allowed to reassemble a packet. // Fragments are lazily evicted only when a new a packet with an // already existing fragmentation-id arrives after the timeout. -func NewFragmentation(highMemoryLimit, lowMemoryLimit int, reassemblingTimeout time.Duration) *Fragmentation { +func NewFragmentation(blockSize uint16, highMemoryLimit, lowMemoryLimit int, reassemblingTimeout time.Duration) *Fragmentation { if lowMemoryLimit >= highMemoryLimit { lowMemoryLimit = highMemoryLimit } @@ -73,17 +106,46 @@ func NewFragmentation(highMemoryLimit, lowMemoryLimit int, reassemblingTimeout t lowMemoryLimit = 0 } + if blockSize < minBlockSize { + blockSize = minBlockSize + } + return &Fragmentation{ - reassemblers: make(map[uint32]*reassembler), + reassemblers: make(map[FragmentID]*reassembler), highLimit: highMemoryLimit, lowLimit: lowMemoryLimit, timeout: reassemblingTimeout, + blockSize: blockSize, } } -// Process processes an incoming fragment belonging to an ID -// and returns a complete packet when all the packets belonging to that ID have been received. -func (f *Fragmentation) Process(id uint32, first, last uint16, more bool, vv buffer.VectorisedView) (buffer.VectorisedView, bool, error) { +// Process processes an incoming fragment belonging to an ID and returns a +// complete packet when all the packets belonging to that ID have been received. +// +// [first, last] is the range of the fragment bytes. +// +// first must be a multiple of the block size f is configured with. The size +// of the fragment data must be a multiple of the block size, unless there are +// no fragments following this fragment (more set to false). +func (f *Fragmentation) Process(id FragmentID, first, last uint16, more bool, vv buffer.VectorisedView) (buffer.VectorisedView, bool, error) { + if first > last { + return buffer.VectorisedView{}, false, fmt.Errorf("first=%d is greater than last=%d: %w", first, last, ErrInvalidArgs) + } + + if first%f.blockSize != 0 { + return buffer.VectorisedView{}, false, fmt.Errorf("first=%d is not a multiple of block size=%d: %w", first, f.blockSize, ErrInvalidArgs) + } + + fragmentSize := last - first + 1 + if more && fragmentSize%f.blockSize != 0 { + return buffer.VectorisedView{}, false, fmt.Errorf("fragment size=%d bytes is not a multiple of block size=%d on non-final fragment: %w", fragmentSize, f.blockSize, ErrInvalidArgs) + } + + if l := vv.Size(); l < int(fragmentSize) { + return buffer.VectorisedView{}, false, fmt.Errorf("got fragment size=%d bytes less than the expected fragment size=%d bytes (first=%d last=%d): %w", l, fragmentSize, first, last, ErrInvalidArgs) + } + vv.CapLength(int(fragmentSize)) + f.mu.Lock() r, ok := f.reassemblers[id] if ok && r.tooOld(f.timeout) { diff --git a/pkg/tcpip/network/fragmentation/fragmentation_test.go b/pkg/tcpip/network/fragmentation/fragmentation_test.go index 72c0f53be..9eedd33c4 100644 --- a/pkg/tcpip/network/fragmentation/fragmentation_test.go +++ b/pkg/tcpip/network/fragmentation/fragmentation_test.go @@ -15,6 +15,7 @@ package fragmentation import ( + "errors" "reflect" "testing" "time" @@ -33,7 +34,7 @@ func vv(size int, pieces ...string) buffer.VectorisedView { } type processInput struct { - id uint32 + id FragmentID first uint16 last uint16 more bool @@ -53,8 +54,8 @@ var processTestCases = []struct { { comment: "One ID", in: []processInput{ - {id: 0, first: 0, last: 1, more: true, vv: vv(2, "01")}, - {id: 0, first: 2, last: 3, more: false, vv: vv(2, "23")}, + {id: FragmentID{ID: 0}, first: 0, last: 1, more: true, vv: vv(2, "01")}, + {id: FragmentID{ID: 0}, first: 2, last: 3, more: false, vv: vv(2, "23")}, }, out: []processOutput{ {vv: buffer.VectorisedView{}, done: false}, @@ -64,10 +65,10 @@ var processTestCases = []struct { { comment: "Two IDs", in: []processInput{ - {id: 0, first: 0, last: 1, more: true, vv: vv(2, "01")}, - {id: 1, first: 0, last: 1, more: true, vv: vv(2, "ab")}, - {id: 1, first: 2, last: 3, more: false, vv: vv(2, "cd")}, - {id: 0, first: 2, last: 3, more: false, vv: vv(2, "23")}, + {id: FragmentID{ID: 0}, first: 0, last: 1, more: true, vv: vv(2, "01")}, + {id: FragmentID{ID: 1}, first: 0, last: 1, more: true, vv: vv(2, "ab")}, + {id: FragmentID{ID: 1}, first: 2, last: 3, more: false, vv: vv(2, "cd")}, + {id: FragmentID{ID: 0}, first: 2, last: 3, more: false, vv: vv(2, "23")}, }, out: []processOutput{ {vv: buffer.VectorisedView{}, done: false}, @@ -81,7 +82,7 @@ var processTestCases = []struct { func TestFragmentationProcess(t *testing.T) { for _, c := range processTestCases { t.Run(c.comment, func(t *testing.T) { - f := NewFragmentation(1024, 512, DefaultReassembleTimeout) + f := NewFragmentation(minBlockSize, 1024, 512, DefaultReassembleTimeout) for i, in := range c.in { vv, done, err := f.Process(in.id, in.first, in.last, in.more, in.vv) if err != nil { @@ -110,14 +111,14 @@ func TestFragmentationProcess(t *testing.T) { func TestReassemblingTimeout(t *testing.T) { timeout := time.Millisecond - f := NewFragmentation(1024, 512, timeout) + f := NewFragmentation(minBlockSize, 1024, 512, timeout) // Send first fragment with id = 0, first = 0, last = 0, and more = true. - f.Process(0, 0, 0, true, vv(1, "0")) + f.Process(FragmentID{}, 0, 0, true, vv(1, "0")) // Sleep more than the timeout. time.Sleep(2 * timeout) // Send another fragment that completes a packet. // However, no packet should be reassembled because the fragment arrived after the timeout. - _, done, err := f.Process(0, 1, 1, false, vv(1, "1")) + _, done, err := f.Process(FragmentID{}, 1, 1, false, vv(1, "1")) if err != nil { t.Fatalf("f.Process(0, 1, 1, false, vv(1, \"1\")) failed: %v", err) } @@ -127,35 +128,35 @@ func TestReassemblingTimeout(t *testing.T) { } func TestMemoryLimits(t *testing.T) { - f := NewFragmentation(3, 1, DefaultReassembleTimeout) + f := NewFragmentation(minBlockSize, 3, 1, DefaultReassembleTimeout) // Send first fragment with id = 0. - f.Process(0, 0, 0, true, vv(1, "0")) + f.Process(FragmentID{ID: 0}, 0, 0, true, vv(1, "0")) // Send first fragment with id = 1. - f.Process(1, 0, 0, true, vv(1, "1")) + f.Process(FragmentID{ID: 1}, 0, 0, true, vv(1, "1")) // Send first fragment with id = 2. - f.Process(2, 0, 0, true, vv(1, "2")) + f.Process(FragmentID{ID: 2}, 0, 0, true, vv(1, "2")) // Send first fragment with id = 3. This should caused id = 0 and id = 1 to be // evicted. - f.Process(3, 0, 0, true, vv(1, "3")) + f.Process(FragmentID{ID: 3}, 0, 0, true, vv(1, "3")) - if _, ok := f.reassemblers[0]; ok { + if _, ok := f.reassemblers[FragmentID{ID: 0}]; ok { t.Errorf("Memory limits are not respected: id=0 has not been evicted.") } - if _, ok := f.reassemblers[1]; ok { + if _, ok := f.reassemblers[FragmentID{ID: 1}]; ok { t.Errorf("Memory limits are not respected: id=1 has not been evicted.") } - if _, ok := f.reassemblers[3]; !ok { + if _, ok := f.reassemblers[FragmentID{ID: 3}]; !ok { t.Errorf("Implementation of memory limits is wrong: id=3 is not present.") } } func TestMemoryLimitsIgnoresDuplicates(t *testing.T) { - f := NewFragmentation(1, 0, DefaultReassembleTimeout) + f := NewFragmentation(minBlockSize, 1, 0, DefaultReassembleTimeout) // Send first fragment with id = 0. - f.Process(0, 0, 0, true, vv(1, "0")) + f.Process(FragmentID{}, 0, 0, true, vv(1, "0")) // Send the same packet again. - f.Process(0, 0, 0, true, vv(1, "0")) + f.Process(FragmentID{}, 0, 0, true, vv(1, "0")) got := f.size want := 1 @@ -163,3 +164,97 @@ func TestMemoryLimitsIgnoresDuplicates(t *testing.T) { t.Errorf("Wrong size, duplicates are not handled correctly: got=%d, want=%d.", got, want) } } + +func TestErrors(t *testing.T) { + tests := []struct { + name string + blockSize uint16 + first uint16 + last uint16 + more bool + data string + err error + }{ + { + name: "exact block size without more", + blockSize: 2, + first: 2, + last: 3, + more: false, + data: "01", + }, + { + name: "exact block size with more", + blockSize: 2, + first: 2, + last: 3, + more: true, + data: "01", + }, + { + name: "exact block size with more and extra data", + blockSize: 2, + first: 2, + last: 3, + more: true, + data: "012", + }, + { + name: "exact block size with more and too little data", + blockSize: 2, + first: 2, + last: 3, + more: true, + data: "0", + err: ErrInvalidArgs, + }, + { + name: "not exact block size with more", + blockSize: 2, + first: 2, + last: 2, + more: true, + data: "0", + err: ErrInvalidArgs, + }, + { + name: "not exact block size without more", + blockSize: 2, + first: 2, + last: 2, + more: false, + data: "0", + }, + { + name: "first not a multiple of block size", + blockSize: 2, + first: 3, + last: 4, + more: true, + data: "01", + err: ErrInvalidArgs, + }, + { + name: "first more than last", + blockSize: 2, + first: 4, + last: 3, + more: true, + data: "01", + err: ErrInvalidArgs, + }, + } + + for _, test := range tests { + t.Run(test.name, func(t *testing.T) { + f := NewFragmentation(test.blockSize, HighFragThreshold, LowFragThreshold, DefaultReassembleTimeout) + _, done, err := f.Process(FragmentID{}, test.first, test.last, test.more, vv(len(test.data), test.data)) + if !errors.Is(err, test.err) { + t.Errorf("got Proceess(_, %d, %d, %t, %q) = (_, _, %v), want = (_, _, %v)", test.first, test.last, test.more, test.data, err, test.err) + } + if done { + t.Errorf("got Proceess(_, %d, %d, %t, %q) = (_, true, _), want = (_, false, _)", test.first, test.last, test.more, test.data) + } + }) + } +} diff --git a/pkg/tcpip/network/fragmentation/reassembler.go b/pkg/tcpip/network/fragmentation/reassembler.go index 0a83d81f2..50d30bbf0 100644 --- a/pkg/tcpip/network/fragmentation/reassembler.go +++ b/pkg/tcpip/network/fragmentation/reassembler.go @@ -32,7 +32,7 @@ type hole struct { type reassembler struct { reassemblerEntry - id uint32 + id FragmentID size int mu sync.Mutex holes []hole @@ -42,7 +42,7 @@ type reassembler struct { creationTime time.Time } -func newReassembler(id uint32) *reassembler { +func newReassembler(id FragmentID) *reassembler { r := &reassembler{ id: id, holes: make([]hole, 0, 16), diff --git a/pkg/tcpip/network/fragmentation/reassembler_test.go b/pkg/tcpip/network/fragmentation/reassembler_test.go index 7eee0710d..dff7c9dcb 100644 --- a/pkg/tcpip/network/fragmentation/reassembler_test.go +++ b/pkg/tcpip/network/fragmentation/reassembler_test.go @@ -94,7 +94,7 @@ var holesTestCases = []struct { func TestUpdateHoles(t *testing.T) { for _, c := range holesTestCases { - r := newReassembler(0) + r := newReassembler(FragmentID{}) for _, i := range c.in { r.updateHoles(i.first, i.last, i.more) } diff --git a/pkg/tcpip/network/ip_test.go b/pkg/tcpip/network/ip_test.go index 4c20301c6..615bae648 100644 --- a/pkg/tcpip/network/ip_test.go +++ b/pkg/tcpip/network/ip_test.go @@ -96,7 +96,7 @@ func (t *testObject) checkValues(protocol tcpip.TransportProtocolNumber, vv buff // DeliverTransportPacket is called by network endpoints after parsing incoming // packets. This is used by the test object to verify that the results of the // parsing are expected. -func (t *testObject) DeliverTransportPacket(r *stack.Route, protocol tcpip.TransportProtocolNumber, pkt stack.PacketBuffer) { +func (t *testObject) DeliverTransportPacket(r *stack.Route, protocol tcpip.TransportProtocolNumber, pkt *stack.PacketBuffer) { t.checkValues(protocol, pkt.Data, r.RemoteAddress, r.LocalAddress) t.dataCalls++ } @@ -104,7 +104,7 @@ func (t *testObject) DeliverTransportPacket(r *stack.Route, protocol tcpip.Trans // DeliverTransportControlPacket is called by network endpoints after parsing // incoming control (ICMP) packets. This is used by the test object to verify // that the results of the parsing are expected. -func (t *testObject) DeliverTransportControlPacket(local, remote tcpip.Address, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ stack.ControlType, extra uint32, pkt stack.PacketBuffer) { +func (t *testObject) DeliverTransportControlPacket(local, remote tcpip.Address, net tcpip.NetworkProtocolNumber, trans tcpip.TransportProtocolNumber, typ stack.ControlType, extra uint32, pkt *stack.PacketBuffer) { t.checkValues(trans, pkt.Data, remote, local) if typ != t.typ { t.t.Errorf("typ = %v, want %v", typ, t.typ) @@ -150,7 +150,7 @@ func (*testObject) Wait() {} // WritePacket is called by network endpoints after producing a packet and // writing it to the link endpoint. This is used by the test object to verify // that the produced packet is as expected. -func (t *testObject) WritePacket(_ *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error { +func (t *testObject) WritePacket(_ *stack.Route, _ *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error { var prot tcpip.TransportProtocolNumber var srcAddr tcpip.Address var dstAddr tcpip.Address @@ -172,14 +172,24 @@ func (t *testObject) WritePacket(_ *stack.Route, _ *stack.GSO, protocol tcpip.Ne } // WritePackets implements stack.LinkEndpoint.WritePackets. -func (t *testObject) WritePackets(_ *stack.Route, _ *stack.GSO, pkt stack.PacketBufferList, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) { +func (*testObject) WritePackets(_ *stack.Route, _ *stack.GSO, pkt stack.PacketBufferList, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) { panic("not implemented") } -func (t *testObject) WriteRawPacket(_ buffer.VectorisedView) *tcpip.Error { +func (*testObject) WriteRawPacket(_ buffer.VectorisedView) *tcpip.Error { return tcpip.ErrNotSupported } +// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType. +func (*testObject) ARPHardwareType() header.ARPHardwareType { + panic("not implemented") +} + +// AddHeader implements stack.LinkEndpoint.AddHeader. +func (*testObject) AddHeader(local, remote tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) { + panic("not implemented") +} + func buildIPv4Route(local, remote tcpip.Address) (stack.Route, *tcpip.Error) { s := stack.New(stack.Options{ NetworkProtocols: []stack.NetworkProtocol{ipv4.NewProtocol()}, @@ -246,7 +256,11 @@ func TestIPv4Send(t *testing.T) { if err != nil { t.Fatalf("could not find route: %v", err) } - if err := ep.WritePacket(&r, nil /* gso */, stack.NetworkHeaderParams{Protocol: 123, TTL: 123, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := ep.WritePacket(&r, nil /* gso */, stack.NetworkHeaderParams{ + Protocol: 123, + TTL: 123, + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: payload.ToVectorisedView(), }); err != nil { @@ -289,9 +303,9 @@ func TestIPv4Receive(t *testing.T) { if err != nil { t.Fatalf("could not find route: %v", err) } - ep.HandlePacket(&r, stack.PacketBuffer{ - Data: view.ToVectorisedView(), - }) + pkt := stack.PacketBuffer{Data: view.ToVectorisedView()} + proto.Parse(&pkt) + ep.HandlePacket(&r, &pkt) if o.dataCalls != 1 { t.Fatalf("Bad number of data calls: got %x, want 1", o.dataCalls) } @@ -378,10 +392,7 @@ func TestIPv4ReceiveControl(t *testing.T) { o.typ = c.expectedTyp o.extra = c.expectedExtra - vv := view[:len(view)-c.trunc].ToVectorisedView() - ep.HandlePacket(&r, stack.PacketBuffer{ - Data: vv, - }) + ep.HandlePacket(&r, truncatedPacket(view, c.trunc, header.IPv4MinimumSize)) if want := c.expectedCount; o.controlCalls != want { t.Fatalf("Bad number of control calls for %q case: got %v, want %v", c.name, o.controlCalls, want) } @@ -444,17 +455,17 @@ func TestIPv4FragmentationReceive(t *testing.T) { } // Send first segment. - ep.HandlePacket(&r, stack.PacketBuffer{ - Data: frag1.ToVectorisedView(), - }) + pkt := stack.PacketBuffer{Data: frag1.ToVectorisedView()} + proto.Parse(&pkt) + ep.HandlePacket(&r, &pkt) if o.dataCalls != 0 { t.Fatalf("Bad number of data calls: got %x, want 0", o.dataCalls) } // Send second segment. - ep.HandlePacket(&r, stack.PacketBuffer{ - Data: frag2.ToVectorisedView(), - }) + pkt = stack.PacketBuffer{Data: frag2.ToVectorisedView()} + proto.Parse(&pkt) + ep.HandlePacket(&r, &pkt) if o.dataCalls != 1 { t.Fatalf("Bad number of data calls: got %x, want 1", o.dataCalls) } @@ -487,7 +498,11 @@ func TestIPv6Send(t *testing.T) { if err != nil { t.Fatalf("could not find route: %v", err) } - if err := ep.WritePacket(&r, nil /* gso */, stack.NetworkHeaderParams{Protocol: 123, TTL: 123, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := ep.WritePacket(&r, nil /* gso */, stack.NetworkHeaderParams{ + Protocol: 123, + TTL: 123, + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: payload.ToVectorisedView(), }); err != nil { @@ -530,9 +545,9 @@ func TestIPv6Receive(t *testing.T) { t.Fatalf("could not find route: %v", err) } - ep.HandlePacket(&r, stack.PacketBuffer{ - Data: view.ToVectorisedView(), - }) + pkt := stack.PacketBuffer{Data: view.ToVectorisedView()} + proto.Parse(&pkt) + ep.HandlePacket(&r, &pkt) if o.dataCalls != 1 { t.Fatalf("Bad number of data calls: got %x, want 1", o.dataCalls) } @@ -644,12 +659,25 @@ func TestIPv6ReceiveControl(t *testing.T) { // Set ICMPv6 checksum. icmp.SetChecksum(header.ICMPv6Checksum(icmp, outerSrcAddr, localIpv6Addr, buffer.VectorisedView{})) - ep.HandlePacket(&r, stack.PacketBuffer{ - Data: view[:len(view)-c.trunc].ToVectorisedView(), - }) + ep.HandlePacket(&r, truncatedPacket(view, c.trunc, header.IPv6MinimumSize)) if want := c.expectedCount; o.controlCalls != want { t.Fatalf("Bad number of control calls for %q case: got %v, want %v", c.name, o.controlCalls, want) } }) } } + +// truncatedPacket returns a PacketBuffer based on a truncated view. If view, +// after truncation, is large enough to hold a network header, it makes part of +// view the packet's NetworkHeader and the rest its Data. Otherwise all of view +// becomes Data. +func truncatedPacket(view buffer.View, trunc, netHdrLen int) *stack.PacketBuffer { + v := view[:len(view)-trunc] + if len(v) < netHdrLen { + return &stack.PacketBuffer{Data: v.ToVectorisedView()} + } + return &stack.PacketBuffer{ + NetworkHeader: v[:netHdrLen], + Data: v[netHdrLen:].ToVectorisedView(), + } +} diff --git a/pkg/tcpip/network/ipv4/BUILD b/pkg/tcpip/network/ipv4/BUILD index 78420d6e6..d142b4ffa 100644 --- a/pkg/tcpip/network/ipv4/BUILD +++ b/pkg/tcpip/network/ipv4/BUILD @@ -34,6 +34,6 @@ go_test( "//pkg/tcpip/transport/tcp", "//pkg/tcpip/transport/udp", "//pkg/waiter", - "@com_github_google_go-cmp//cmp:go_default_library", + "@com_github_google_go_cmp//cmp:go_default_library", ], ) diff --git a/pkg/tcpip/network/ipv4/icmp.go b/pkg/tcpip/network/ipv4/icmp.go index 4cbefe5ab..83e71cb8c 100644 --- a/pkg/tcpip/network/ipv4/icmp.go +++ b/pkg/tcpip/network/ipv4/icmp.go @@ -24,7 +24,7 @@ import ( // the original packet that caused the ICMP one to be sent. This information is // used to find out which transport endpoint must be notified about the ICMP // packet. -func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt stack.PacketBuffer) { +func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt *stack.PacketBuffer) { h, ok := pkt.Data.PullUp(header.IPv4MinimumSize) if !ok { return @@ -56,9 +56,12 @@ func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt stack. e.dispatcher.DeliverTransportControlPacket(e.id.LocalAddress, hdr.DestinationAddress(), ProtocolNumber, p, typ, extra, pkt) } -func (e *endpoint) handleICMP(r *stack.Route, pkt stack.PacketBuffer) { +func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer) { stats := r.Stats() received := stats.ICMP.V4PacketsReceived + // TODO(gvisor.dev/issue/170): ICMP packets don't have their + // TransportHeader fields set. See icmp/protocol.go:protocol.Parse for a + // full explanation. v, ok := pkt.Data.PullUp(header.ICMPv4MinimumSize) if !ok { received.Invalid.Increment() @@ -88,7 +91,7 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt stack.PacketBuffer) { // It's possible that a raw socket expects to receive this. h.SetChecksum(wantChecksum) - e.dispatcher.DeliverTransportPacket(r, header.ICMPv4ProtocolNumber, stack.PacketBuffer{ + e.dispatcher.DeliverTransportPacket(r, header.ICMPv4ProtocolNumber, &stack.PacketBuffer{ Data: pkt.Data.Clone(nil), NetworkHeader: append(buffer.View(nil), pkt.NetworkHeader...), }) @@ -102,7 +105,11 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt stack.PacketBuffer) { pkt.SetChecksum(0) pkt.SetChecksum(^header.Checksum(pkt, header.ChecksumVV(vv, 0))) sent := stats.ICMP.V4PacketsSent - if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv4ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{ + Protocol: header.ICMPv4ProtocolNumber, + TTL: r.DefaultTTL(), + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: vv, TransportHeader: buffer.View(pkt), @@ -122,6 +129,9 @@ func (e *endpoint) handleICMP(r *stack.Route, pkt stack.PacketBuffer) { pkt.Data.TrimFront(header.ICMPv4MinimumSize) switch h.Code() { + case header.ICMPv4HostUnreachable: + e.handleControl(stack.ControlNoRoute, 0, pkt) + case header.ICMPv4PortUnreachable: e.handleControl(stack.ControlPortUnreachable, 0, pkt) diff --git a/pkg/tcpip/network/ipv4/ipv4.go b/pkg/tcpip/network/ipv4/ipv4.go index 64046cbbf..d5f5d38f7 100644 --- a/pkg/tcpip/network/ipv4/ipv4.go +++ b/pkg/tcpip/network/ipv4/ipv4.go @@ -21,6 +21,7 @@ package ipv4 import ( + "fmt" "sync/atomic" "gvisor.dev/gvisor/pkg/tcpip" @@ -44,6 +45,10 @@ const ( // buckets is the number of identifier buckets. buckets = 2048 + + // The size of a fragment block, in bytes, as per RFC 791 section 3.1, + // page 14. + fragmentblockSize = 8 ) type endpoint struct { @@ -65,7 +70,7 @@ func (p *protocol) NewEndpoint(nicID tcpip.NICID, addrWithPrefix tcpip.AddressWi prefixLen: addrWithPrefix.PrefixLen, linkEP: linkEP, dispatcher: dispatcher, - fragmentation: fragmentation.NewFragmentation(fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, fragmentation.DefaultReassembleTimeout), + fragmentation: fragmentation.NewFragmentation(fragmentblockSize, fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, fragmentation.DefaultReassembleTimeout), protocol: p, stack: st, } @@ -129,7 +134,7 @@ func (e *endpoint) NetworkProtocolNumber() tcpip.NetworkProtocolNumber { // packet's stated length matches the length of the header+payload. mtu // includes the IP header and options. This does not support the DontFragment // IP flag. -func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, pkt *stack.PacketBuffer) *tcpip.Error { // This packet is too big, it needs to be fragmented. ip := header.IPv4(pkt.Header.View()) flags := ip.Flags() @@ -169,7 +174,7 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, if i > 0 { newPayload := pkt.Data.Clone(nil) newPayload.CapLength(innerMTU) - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, stack.PacketBuffer{ + if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, &stack.PacketBuffer{ Header: pkt.Header, Data: newPayload, NetworkHeader: buffer.View(h), @@ -188,7 +193,7 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, newPayload := pkt.Data.Clone(nil) newPayloadLength := outerMTU - pkt.Header.UsedLength() newPayload.CapLength(newPayloadLength) - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, stack.PacketBuffer{ + if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, &stack.PacketBuffer{ Header: pkt.Header, Data: newPayload, NetworkHeader: buffer.View(h), @@ -202,7 +207,7 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, startOfHdr := pkt.Header startOfHdr.TrimBack(pkt.Header.UsedLength() - outerMTU) emptyVV := buffer.NewVectorisedView(0, []buffer.View{}) - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, stack.PacketBuffer{ + if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, &stack.PacketBuffer{ Header: startOfHdr, Data: emptyVV, NetworkHeader: buffer.View(h), @@ -224,12 +229,10 @@ func (e *endpoint) writePacketFragments(r *stack.Route, gso *stack.GSO, mtu int, func (e *endpoint) addIPHeader(r *stack.Route, hdr *buffer.Prependable, payloadSize int, params stack.NetworkHeaderParams) header.IPv4 { ip := header.IPv4(hdr.Prepend(header.IPv4MinimumSize)) length := uint16(hdr.UsedLength() + payloadSize) - id := uint32(0) - if length > header.IPv4MaximumHeaderSize+8 { - // Packets of 68 bytes or less are required by RFC 791 to not be - // fragmented, so we only assign ids to larger packets. - id = atomic.AddUint32(&e.protocol.ids[hashRoute(r, params.Protocol, e.protocol.hashIV)%buckets], 1) - } + // RFC 6864 section 4.3 mandates uniqueness of ID values for non-atomic + // datagrams. Since the DF bit is never being set here, all datagrams + // are non-atomic and need an ID. + id := atomic.AddUint32(&e.protocol.ids[hashRoute(r, params.Protocol, e.protocol.hashIV)%buckets], 1) ip.Encode(&header.IPv4Fields{ IHL: header.IPv4MinimumSize, TotalLength: length, @@ -245,7 +248,7 @@ func (e *endpoint) addIPHeader(r *stack.Route, hdr *buffer.Prependable, payloadS } // WritePacket writes a packet to the given destination address and protocol. -func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt *stack.PacketBuffer) *tcpip.Error { ip := e.addIPHeader(r, &pkt.Header, pkt.Data.Size(), params) pkt.NetworkHeader = buffer.View(ip) @@ -253,43 +256,29 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.Netw // iptables filtering. All packets that reach here are locally // generated. ipt := e.stack.IPTables() - if ok := ipt.Check(stack.Output, &pkt, gso, r, "", nicName); !ok { + if ok := ipt.Check(stack.Output, pkt, gso, r, "", nicName); !ok { // iptables is telling us to drop the packet. return nil } + // If the packet is manipulated as per NAT Ouput rules, handle packet + // based on destination address and do not send the packet to link layer. + // TODO(gvisor.dev/issue/170): We should do this for every packet, rather than + // only NATted packets, but removing this check short circuits broadcasts + // before they are sent out to other hosts. if pkt.NatDone { - // If the packet is manipulated as per NAT Ouput rules, handle packet - // based on destination address and do not send the packet to link layer. netHeader := header.IPv4(pkt.NetworkHeader) ep, err := e.stack.FindNetworkEndpoint(header.IPv4ProtocolNumber, netHeader.DestinationAddress()) if err == nil { - src := netHeader.SourceAddress() - dst := netHeader.DestinationAddress() - route := r.ReverseRoute(src, dst) - - views := make([]buffer.View, 1, 1+len(pkt.Data.Views())) - views[0] = pkt.Header.View() - views = append(views, pkt.Data.Views()...) - packet := stack.PacketBuffer{ - Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views)} - ep.HandlePacket(&route, packet) + route := r.ReverseRoute(netHeader.SourceAddress(), netHeader.DestinationAddress()) + ep.HandlePacket(&route, pkt) return nil } } if r.Loop&stack.PacketLoop != 0 { - // The inbound path expects the network header to still be in - // the PacketBuffer's Data field. - views := make([]buffer.View, 1, 1+len(pkt.Data.Views())) - views[0] = pkt.Header.View() - views = append(views, pkt.Data.Views()...) loopedR := r.MakeLoopedRoute() - - e.HandlePacket(&loopedR, stack.PacketBuffer{ - Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views), - }) - + e.HandlePacket(&loopedR, pkt) loopedR.Release() } if r.Loop&stack.PacketOut == 0 { @@ -342,23 +331,16 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe } if _, ok := natPkts[pkt]; ok { netHeader := header.IPv4(pkt.NetworkHeader) - ep, err := e.stack.FindNetworkEndpoint(header.IPv4ProtocolNumber, netHeader.DestinationAddress()) - if err == nil { + if ep, err := e.stack.FindNetworkEndpoint(header.IPv4ProtocolNumber, netHeader.DestinationAddress()); err == nil { src := netHeader.SourceAddress() dst := netHeader.DestinationAddress() route := r.ReverseRoute(src, dst) - - views := make([]buffer.View, 1, 1+len(pkt.Data.Views())) - views[0] = pkt.Header.View() - views = append(views, pkt.Data.Views()...) - packet := stack.PacketBuffer{ - Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views)} - ep.HandlePacket(&route, packet) + ep.HandlePacket(&route, pkt) n++ continue } } - if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, *pkt); err != nil { + if err := e.linkEP.WritePacket(r, gso, ProtocolNumber, pkt); err != nil { r.Stats().IP.PacketsSent.IncrementBy(uint64(n)) return n, err } @@ -370,7 +352,7 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe // WriteHeaderIncludedPacket writes a packet already containing a network // header through the given route. -func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuffer) *tcpip.Error { // The packet already has an IP header, but there are a few required // checks. h, ok := pkt.Data.PullUp(header.IPv4MinimumSize) @@ -396,13 +378,12 @@ func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuf // Set the packet ID when zero. if ip.ID() == 0 { - id := uint32(0) - if pkt.Data.Size() > header.IPv4MaximumHeaderSize+8 { - // Packets of 68 bytes or less are required by RFC 791 to not be - // fragmented, so we only assign ids to larger packets. - id = atomic.AddUint32(&e.protocol.ids[hashRoute(r, 0 /* protocol */, e.protocol.hashIV)%buckets], 1) + // RFC 6864 section 4.3 mandates uniqueness of ID values for + // non-atomic datagrams, so assign an ID to all such datagrams + // according to the definition given in RFC 6864 section 4. + if ip.Flags()&header.IPv4FlagDontFragment == 0 || ip.Flags()&header.IPv4FlagMoreFragments != 0 || ip.FragmentOffset() > 0 { + ip.SetID(uint16(atomic.AddUint32(&e.protocol.ids[hashRoute(r, 0 /* protocol */, e.protocol.hashIV)%buckets], 1))) } - ip.SetID(uint16(id)) } // Always set the checksum. @@ -426,35 +407,23 @@ func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuf // HandlePacket is called by the link layer when new ipv4 packets arrive for // this endpoint. -func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { - headerView, ok := pkt.Data.PullUp(header.IPv4MinimumSize) - if !ok { +func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { + h := header.IPv4(pkt.NetworkHeader) + if !h.IsValid(pkt.Data.Size() + len(pkt.NetworkHeader) + len(pkt.TransportHeader)) { r.Stats().IP.MalformedPacketsReceived.Increment() return } - h := header.IPv4(headerView) - if !h.IsValid(pkt.Data.Size()) { - r.Stats().IP.MalformedPacketsReceived.Increment() - return - } - pkt.NetworkHeader = headerView[:h.HeaderLength()] - - hlen := int(h.HeaderLength()) - tlen := int(h.TotalLength()) - pkt.Data.TrimFront(hlen) - pkt.Data.CapLength(tlen - hlen) // iptables filtering. All packets that reach here are intended for // this machine and will not be forwarded. ipt := e.stack.IPTables() - if ok := ipt.Check(stack.Input, &pkt, nil, nil, "", ""); !ok { + if ok := ipt.Check(stack.Input, pkt, nil, nil, "", ""); !ok { // iptables is telling us to drop the packet. return } - more := (h.Flags() & header.IPv4FlagMoreFragments) != 0 - if more || h.FragmentOffset() != 0 { - if pkt.Data.Size() == 0 { + if h.More() || h.FragmentOffset() != 0 { + if pkt.Data.Size()+len(pkt.TransportHeader) == 0 { // Drop the packet as it's marked as a fragment but has // no payload. r.Stats().IP.MalformedPacketsReceived.Increment() @@ -473,7 +442,18 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { } var ready bool var err error - pkt.Data, ready, err = e.fragmentation.Process(hash.IPv4FragmentHash(h), h.FragmentOffset(), last, more, pkt.Data) + pkt.Data, ready, err = e.fragmentation.Process( + fragmentation.FragmentID{ + Source: h.SourceAddress(), + Destination: h.DestinationAddress(), + ID: uint32(h.ID()), + Protocol: h.Protocol(), + }, + h.FragmentOffset(), + last, + h.More(), + pkt.Data, + ) if err != nil { r.Stats().IP.MalformedPacketsReceived.Increment() r.Stats().IP.MalformedFragmentsReceived.Increment() @@ -485,7 +465,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { } p := h.TransportProtocol() if p == header.ICMPv4ProtocolNumber { - headerView.CapLength(hlen) + pkt.NetworkHeader.CapLength(int(h.HeaderLength())) e.handleICMP(r, pkt) return } @@ -565,6 +545,35 @@ func (*protocol) Close() {} // Wait implements stack.TransportProtocol.Wait. func (*protocol) Wait() {} +// Parse implements stack.TransportProtocol.Parse. +func (*protocol) Parse(pkt *stack.PacketBuffer) (proto tcpip.TransportProtocolNumber, hasTransportHdr bool, ok bool) { + hdr, ok := pkt.Data.PullUp(header.IPv4MinimumSize) + if !ok { + return 0, false, false + } + ipHdr := header.IPv4(hdr) + + // If there are options, pull those into hdr as well. + if headerLen := int(ipHdr.HeaderLength()); headerLen > header.IPv4MinimumSize && headerLen <= pkt.Data.Size() { + hdr, ok = pkt.Data.PullUp(headerLen) + if !ok { + panic(fmt.Sprintf("There are only %d bytes in pkt.Data, but there should be at least %d", pkt.Data.Size(), headerLen)) + } + ipHdr = header.IPv4(hdr) + } + + // If this is a fragment, don't bother parsing the transport header. + parseTransportHeader := true + if ipHdr.More() || ipHdr.FragmentOffset() != 0 { + parseTransportHeader = false + } + + pkt.NetworkHeader = hdr + pkt.Data.TrimFront(len(hdr)) + pkt.Data.CapLength(int(ipHdr.TotalLength()) - len(hdr)) + return ipHdr.TransportProtocol(), parseTransportHeader, true +} + // calculateMTU calculates the network-layer payload MTU based on the link-layer // payload mtu. func calculateMTU(mtu uint32) uint32 { diff --git a/pkg/tcpip/network/ipv4/ipv4_test.go b/pkg/tcpip/network/ipv4/ipv4_test.go index 36035c820..ded97ac64 100644 --- a/pkg/tcpip/network/ipv4/ipv4_test.go +++ b/pkg/tcpip/network/ipv4/ipv4_test.go @@ -114,7 +114,7 @@ func makeHdrAndPayload(hdrLength int, extraLength int, viewSizes []int) (buffer. // comparePayloads compared the contents of all the packets against the contents // of the source packet. -func compareFragments(t *testing.T, packets []stack.PacketBuffer, sourcePacketInfo stack.PacketBuffer, mtu uint32) { +func compareFragments(t *testing.T, packets []*stack.PacketBuffer, sourcePacketInfo *stack.PacketBuffer, mtu uint32) { t.Helper() // Make a complete array of the sourcePacketInfo packet. source := header.IPv4(packets[0].Header.View()[:header.IPv4MinimumSize]) @@ -174,7 +174,7 @@ func compareFragments(t *testing.T, packets []stack.PacketBuffer, sourcePacketIn type errorChannel struct { *channel.Endpoint - Ch chan stack.PacketBuffer + Ch chan *stack.PacketBuffer packetCollectorErrors []*tcpip.Error } @@ -184,7 +184,7 @@ type errorChannel struct { func newErrorChannel(size int, mtu uint32, linkAddr tcpip.LinkAddress, packetCollectorErrors []*tcpip.Error) *errorChannel { return &errorChannel{ Endpoint: channel.New(size, mtu, linkAddr), - Ch: make(chan stack.PacketBuffer, size), + Ch: make(chan *stack.PacketBuffer, size), packetCollectorErrors: packetCollectorErrors, } } @@ -203,7 +203,7 @@ func (e *errorChannel) Drain() int { } // WritePacket stores outbound packets into the channel. -func (e *errorChannel) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt stack.PacketBuffer) *tcpip.Error { +func (e *errorChannel) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.NetworkProtocolNumber, pkt *stack.PacketBuffer) *tcpip.Error { select { case e.Ch <- pkt: default: @@ -282,13 +282,17 @@ func TestFragmentation(t *testing.T) { for _, ft := range fragTests { t.Run(ft.description, func(t *testing.T) { hdr, payload := makeHdrAndPayload(ft.hdrLength, ft.extraLength, ft.payloadViewsSizes) - source := stack.PacketBuffer{ + source := &stack.PacketBuffer{ Header: hdr, // Save the source payload because WritePacket will modify it. Data: payload.Clone(nil), } c := buildContext(t, nil, ft.mtu) - err := c.Route.WritePacket(ft.gso, stack.NetworkHeaderParams{Protocol: tcp.ProtocolNumber, TTL: 42, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + err := c.Route.WritePacket(ft.gso, stack.NetworkHeaderParams{ + Protocol: tcp.ProtocolNumber, + TTL: 42, + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: payload, }) @@ -296,7 +300,7 @@ func TestFragmentation(t *testing.T) { t.Errorf("err got %v, want %v", err, nil) } - var results []stack.PacketBuffer + var results []*stack.PacketBuffer L: for { select { @@ -338,7 +342,11 @@ func TestFragmentationErrors(t *testing.T) { t.Run(ft.description, func(t *testing.T) { hdr, payload := makeHdrAndPayload(ft.hdrLength, header.IPv4MinimumSize, ft.payloadViewsSizes) c := buildContext(t, ft.packetCollectorErrors, ft.mtu) - err := c.Route.WritePacket(&stack.GSO{}, stack.NetworkHeaderParams{Protocol: tcp.ProtocolNumber, TTL: 42, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + err := c.Route.WritePacket(&stack.GSO{}, stack.NetworkHeaderParams{ + Protocol: tcp.ProtocolNumber, + TTL: 42, + TOS: stack.DefaultTOS, + }, &stack.PacketBuffer{ Header: hdr, Data: payload, }) @@ -460,7 +468,7 @@ func TestInvalidFragments(t *testing.T) { s.CreateNIC(nicID, sniffer.New(ep)) for _, pkt := range tc.packets { - ep.InjectLinkAddr(header.IPv4ProtocolNumber, remoteLinkAddr, stack.PacketBuffer{ + ep.InjectLinkAddr(header.IPv4ProtocolNumber, remoteLinkAddr, &stack.PacketBuffer{ Data: buffer.NewVectorisedView(len(pkt), []buffer.View{pkt}), }) } @@ -478,12 +486,16 @@ func TestInvalidFragments(t *testing.T) { // TestReceiveFragments feeds fragments in through the incoming packet path to // test reassembly func TestReceiveFragments(t *testing.T) { - const addr1 = "\x0c\xa8\x00\x01" // 192.168.0.1 - const addr2 = "\x0c\xa8\x00\x02" // 192.168.0.2 - const nicID = 1 + const ( + nicID = 1 + + addr1 = "\x0c\xa8\x00\x01" // 192.168.0.1 + addr2 = "\x0c\xa8\x00\x02" // 192.168.0.2 + addr3 = "\x0c\xa8\x00\x03" // 192.168.0.3 + ) // Build and return a UDP header containing payload. - udpGen := func(payloadLen int, multiplier uint8) buffer.View { + udpGen := func(payloadLen int, multiplier uint8, src, dst tcpip.Address) buffer.View { payload := buffer.NewView(payloadLen) for i := 0; i < len(payload); i++ { payload[i] = uint8(i) * multiplier @@ -499,20 +511,29 @@ func TestReceiveFragments(t *testing.T) { Length: uint16(udpLength), }) copy(u.Payload(), payload) - sum := header.PseudoHeaderChecksum(udp.ProtocolNumber, addr1, addr2, uint16(udpLength)) + sum := header.PseudoHeaderChecksum(udp.ProtocolNumber, src, dst, uint16(udpLength)) sum = header.Checksum(payload, sum) u.SetChecksum(^u.CalculateChecksum(sum)) return hdr.View() } // UDP header plus a payload of 0..256 - ipv4Payload1 := udpGen(256, 1) - udpPayload1 := ipv4Payload1[header.UDPMinimumSize:] + ipv4Payload1Addr1ToAddr2 := udpGen(256, 1, addr1, addr2) + udpPayload1Addr1ToAddr2 := ipv4Payload1Addr1ToAddr2[header.UDPMinimumSize:] + ipv4Payload1Addr3ToAddr2 := udpGen(256, 1, addr3, addr2) + udpPayload1Addr3ToAddr2 := ipv4Payload1Addr3ToAddr2[header.UDPMinimumSize:] // UDP header plus a payload of 0..256 in increments of 2. - ipv4Payload2 := udpGen(128, 2) - udpPayload2 := ipv4Payload2[header.UDPMinimumSize:] + ipv4Payload2Addr1ToAddr2 := udpGen(128, 2, addr1, addr2) + udpPayload2Addr1ToAddr2 := ipv4Payload2Addr1ToAddr2[header.UDPMinimumSize:] + // UDP header plus a payload of 0..256 in increments of 3. + // Used to test cases where the fragment blocks are not a multiple of + // the fragment block size of 8 (RFC 791 section 3.1 page 14). + ipv4Payload3Addr1ToAddr2 := udpGen(127, 3, addr1, addr2) + udpPayload3Addr1ToAddr2 := ipv4Payload3Addr1ToAddr2[header.UDPMinimumSize:] type fragmentData struct { + srcAddr tcpip.Address + dstAddr tcpip.Address id uint16 flags uint8 fragmentOffset uint16 @@ -528,22 +549,40 @@ func TestReceiveFragments(t *testing.T) { name: "No fragmentation", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: 0, fragmentOffset: 0, - payload: ipv4Payload1, + payload: ipv4Payload1Addr1ToAddr2, }, }, - expectedPayloads: [][]byte{udpPayload1}, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2}, + }, + { + name: "No fragmentation with size not a multiple of fragment block size", + fragments: []fragmentData{ + { + srcAddr: addr1, + dstAddr: addr2, + id: 1, + flags: 0, + fragmentOffset: 0, + payload: ipv4Payload3Addr1ToAddr2, + }, + }, + expectedPayloads: [][]byte{udpPayload3Addr1ToAddr2}, }, { name: "More fragments without payload", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: header.IPv4FlagMoreFragments, fragmentOffset: 0, - payload: ipv4Payload1, + payload: ipv4Payload1Addr1ToAddr2, }, }, expectedPayloads: nil, @@ -552,10 +591,12 @@ func TestReceiveFragments(t *testing.T) { name: "Non-zero fragment offset without payload", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: 0, fragmentOffset: 8, - payload: ipv4Payload1, + payload: ipv4Payload1Addr1ToAddr2, }, }, expectedPayloads: nil, @@ -564,34 +605,86 @@ func TestReceiveFragments(t *testing.T) { name: "Two fragments", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: header.IPv4FlagMoreFragments, fragmentOffset: 0, - payload: ipv4Payload1[:64], + payload: ipv4Payload1Addr1ToAddr2[:64], }, { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: 0, fragmentOffset: 64, - payload: ipv4Payload1[64:], + payload: ipv4Payload1Addr1ToAddr2[64:], }, }, - expectedPayloads: [][]byte{udpPayload1}, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2}, + }, + { + name: "Two fragments with last fragment size not a multiple of fragment block size", + fragments: []fragmentData{ + { + srcAddr: addr1, + dstAddr: addr2, + id: 1, + flags: header.IPv4FlagMoreFragments, + fragmentOffset: 0, + payload: ipv4Payload3Addr1ToAddr2[:64], + }, + { + srcAddr: addr1, + dstAddr: addr2, + id: 1, + flags: 0, + fragmentOffset: 64, + payload: ipv4Payload3Addr1ToAddr2[64:], + }, + }, + expectedPayloads: [][]byte{udpPayload3Addr1ToAddr2}, + }, + { + name: "Two fragments with first fragment size not a multiple of fragment block size", + fragments: []fragmentData{ + { + srcAddr: addr1, + dstAddr: addr2, + id: 1, + flags: header.IPv4FlagMoreFragments, + fragmentOffset: 0, + payload: ipv4Payload3Addr1ToAddr2[:63], + }, + { + srcAddr: addr1, + dstAddr: addr2, + id: 1, + flags: 0, + fragmentOffset: 63, + payload: ipv4Payload3Addr1ToAddr2[63:], + }, + }, + expectedPayloads: nil, }, { name: "Second fragment has MoreFlags set", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: header.IPv4FlagMoreFragments, fragmentOffset: 0, - payload: ipv4Payload1[:64], + payload: ipv4Payload1Addr1ToAddr2[:64], }, { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: header.IPv4FlagMoreFragments, fragmentOffset: 64, - payload: ipv4Payload1[64:], + payload: ipv4Payload1Addr1ToAddr2[64:], }, }, expectedPayloads: nil, @@ -600,16 +693,20 @@ func TestReceiveFragments(t *testing.T) { name: "Two fragments with different IDs", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: header.IPv4FlagMoreFragments, fragmentOffset: 0, - payload: ipv4Payload1[:64], + payload: ipv4Payload1Addr1ToAddr2[:64], }, { + srcAddr: addr1, + dstAddr: addr2, id: 2, flags: 0, fragmentOffset: 64, - payload: ipv4Payload1[64:], + payload: ipv4Payload1Addr1ToAddr2[64:], }, }, expectedPayloads: nil, @@ -618,31 +715,91 @@ func TestReceiveFragments(t *testing.T) { name: "Two interleaved fragmented packets", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: header.IPv4FlagMoreFragments, fragmentOffset: 0, - payload: ipv4Payload1[:64], + payload: ipv4Payload1Addr1ToAddr2[:64], }, { + srcAddr: addr1, + dstAddr: addr2, id: 2, flags: header.IPv4FlagMoreFragments, fragmentOffset: 0, - payload: ipv4Payload2[:64], + payload: ipv4Payload2Addr1ToAddr2[:64], }, { + srcAddr: addr1, + dstAddr: addr2, id: 1, flags: 0, fragmentOffset: 64, - payload: ipv4Payload1[64:], + payload: ipv4Payload1Addr1ToAddr2[64:], }, { + srcAddr: addr1, + dstAddr: addr2, id: 2, flags: 0, fragmentOffset: 64, - payload: ipv4Payload2[64:], + payload: ipv4Payload2Addr1ToAddr2[64:], + }, + }, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2, udpPayload2Addr1ToAddr2}, + }, + { + name: "Two interleaved fragmented packets from different sources but with same ID", + fragments: []fragmentData{ + { + srcAddr: addr1, + dstAddr: addr2, + id: 1, + flags: header.IPv4FlagMoreFragments, + fragmentOffset: 0, + payload: ipv4Payload1Addr1ToAddr2[:64], + }, + { + srcAddr: addr3, + dstAddr: addr2, + id: 1, + flags: header.IPv4FlagMoreFragments, + fragmentOffset: 0, + payload: ipv4Payload1Addr3ToAddr2[:32], + }, + { + srcAddr: addr1, + dstAddr: addr2, + id: 1, + flags: 0, + fragmentOffset: 64, + payload: ipv4Payload1Addr1ToAddr2[64:], + }, + { + srcAddr: addr3, + dstAddr: addr2, + id: 1, + flags: 0, + fragmentOffset: 32, + payload: ipv4Payload1Addr3ToAddr2[32:], + }, + }, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2, udpPayload1Addr3ToAddr2}, + }, + { + name: "Fragment without followup", + fragments: []fragmentData{ + { + srcAddr: addr1, + dstAddr: addr2, + id: 1, + flags: header.IPv4FlagMoreFragments, + fragmentOffset: 0, + payload: ipv4Payload1Addr1ToAddr2[:64], }, }, - expectedPayloads: [][]byte{udpPayload1, udpPayload2}, + expectedPayloads: nil, }, } @@ -691,14 +848,14 @@ func TestReceiveFragments(t *testing.T) { FragmentOffset: frag.fragmentOffset, TTL: 64, Protocol: uint8(header.UDPProtocolNumber), - SrcAddr: addr1, - DstAddr: addr2, + SrcAddr: frag.srcAddr, + DstAddr: frag.dstAddr, }) vv := hdr.View().ToVectorisedView() vv.AppendView(frag.payload) - e.InjectInbound(header.IPv4ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(header.IPv4ProtocolNumber, &stack.PacketBuffer{ Data: vv, }) } diff --git a/pkg/tcpip/network/ipv6/BUILD b/pkg/tcpip/network/ipv6/BUILD index 3f71fc520..bcc64994e 100644 --- a/pkg/tcpip/network/ipv6/BUILD +++ b/pkg/tcpip/network/ipv6/BUILD @@ -14,7 +14,6 @@ go_library( "//pkg/tcpip/buffer", "//pkg/tcpip/header", "//pkg/tcpip/network/fragmentation", - "//pkg/tcpip/network/hash", "//pkg/tcpip/stack", ], ) @@ -39,6 +38,6 @@ go_test( "//pkg/tcpip/transport/icmp", "//pkg/tcpip/transport/udp", "//pkg/waiter", - "@com_github_google_go-cmp//cmp:go_default_library", + "@com_github_google_go_cmp//cmp:go_default_library", ], ) diff --git a/pkg/tcpip/network/ipv6/icmp.go b/pkg/tcpip/network/ipv6/icmp.go index bdf3a0d25..24600d877 100644 --- a/pkg/tcpip/network/ipv6/icmp.go +++ b/pkg/tcpip/network/ipv6/icmp.go @@ -27,7 +27,7 @@ import ( // the original packet that caused the ICMP one to be sent. This information is // used to find out which transport endpoint must be notified about the ICMP // packet. -func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt stack.PacketBuffer) { +func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt *stack.PacketBuffer) { h, ok := pkt.Data.PullUp(header.IPv6MinimumSize) if !ok { return @@ -70,17 +70,20 @@ func (e *endpoint) handleControl(typ stack.ControlType, extra uint32, pkt stack. e.dispatcher.DeliverTransportControlPacket(e.id.LocalAddress, hdr.DestinationAddress(), ProtocolNumber, p, typ, extra, pkt) } -func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.PacketBuffer, hasFragmentHeader bool) { +func (e *endpoint) handleICMP(r *stack.Route, pkt *stack.PacketBuffer, hasFragmentHeader bool) { stats := r.Stats().ICMP sent := stats.V6PacketsSent received := stats.V6PacketsReceived + // TODO(gvisor.dev/issue/170): ICMP packets don't have their + // TransportHeader fields set. See icmp/protocol.go:protocol.Parse for a + // full explanation. v, ok := pkt.Data.PullUp(header.ICMPv6HeaderSize) if !ok { received.Invalid.Increment() return } h := header.ICMPv6(v) - iph := header.IPv6(netHeader) + iph := header.IPv6(pkt.NetworkHeader) // Validate ICMPv6 checksum before processing the packet. // @@ -125,6 +128,8 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.P } pkt.Data.TrimFront(header.ICMPv6DstUnreachableMinimumSize) switch header.ICMPv6(hdr).Code() { + case header.ICMPv6NetworkUnreachable: + e.handleControl(stack.ControlNetworkUnreachable, 0, pkt) case header.ICMPv6PortUnreachable: e.handleControl(stack.ControlPortUnreachable, 0, pkt) } @@ -288,7 +293,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.P // // The IP Hop Limit field has a value of 255, i.e., the packet // could not possibly have been forwarded by a router. - if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: header.NDPHopLimit, TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: header.NDPHopLimit, TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, }); err != nil { sent.Dropped.Increment() @@ -390,7 +395,7 @@ func (e *endpoint) handleICMP(r *stack.Route, netHeader buffer.View, pkt stack.P copy(packet, icmpHdr) packet.SetType(header.ICMPv6EchoReply) packet.SetChecksum(header.ICMPv6Checksum(packet, r.LocalAddress, r.RemoteAddress, pkt.Data)) - if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, stack.PacketBuffer{ + if err := r.WritePacket(nil /* gso */, stack.NetworkHeaderParams{Protocol: header.ICMPv6ProtocolNumber, TTL: r.DefaultTTL(), TOS: stack.DefaultTOS}, &stack.PacketBuffer{ Header: hdr, Data: pkt.Data, }); err != nil { @@ -491,8 +496,6 @@ const ( icmpV6LengthOffset = 25 ) -var broadcastMAC = tcpip.LinkAddress([]byte{0xff, 0xff, 0xff, 0xff, 0xff, 0xff}) - var _ stack.LinkAddressResolver = (*protocol)(nil) // LinkAddressProtocol implements stack.LinkAddressResolver. @@ -501,7 +504,7 @@ func (*protocol) LinkAddressProtocol() tcpip.NetworkProtocolNumber { } // LinkAddressRequest implements stack.LinkAddressResolver. -func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack.LinkEndpoint) *tcpip.Error { +func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, remoteLinkAddr tcpip.LinkAddress, linkEP stack.LinkEndpoint) *tcpip.Error { snaddr := header.SolicitedNodeAddr(addr) // TODO(b/148672031): Use stack.FindRoute instead of manually creating the @@ -510,8 +513,12 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack. r := &stack.Route{ LocalAddress: localAddr, RemoteAddress: snaddr, - RemoteLinkAddress: header.EthernetAddressFromMulticastIPv6Address(snaddr), + RemoteLinkAddress: remoteLinkAddr, + } + if len(r.RemoteLinkAddress) == 0 { + r.RemoteLinkAddress = header.EthernetAddressFromMulticastIPv6Address(snaddr) } + hdr := buffer.NewPrependable(int(linkEP.MaxHeaderLength()) + header.IPv6MinimumSize + header.ICMPv6NeighborAdvertSize) pkt := header.ICMPv6(hdr.Prepend(header.ICMPv6NeighborAdvertSize)) pkt.SetType(header.ICMPv6NeighborSolicit) @@ -532,7 +539,7 @@ func (*protocol) LinkAddressRequest(addr, localAddr tcpip.Address, linkEP stack. }) // TODO(stijlist): count this in ICMP stats. - return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, stack.PacketBuffer{ + return linkEP.WritePacket(r, nil /* gso */, ProtocolNumber, &stack.PacketBuffer{ Header: hdr, }) } diff --git a/pkg/tcpip/network/ipv6/icmp_test.go b/pkg/tcpip/network/ipv6/icmp_test.go index d412ff688..f86aaed1d 100644 --- a/pkg/tcpip/network/ipv6/icmp_test.go +++ b/pkg/tcpip/network/ipv6/icmp_test.go @@ -34,6 +34,9 @@ const ( linkAddr0 = tcpip.LinkAddress("\x02\x02\x03\x04\x05\x06") linkAddr1 = tcpip.LinkAddress("\x0a\x0b\x0c\x0d\x0e\x0e") linkAddr2 = tcpip.LinkAddress("\x0a\x0b\x0c\x0d\x0e\x0f") + + defaultChannelSize = 1 + defaultMTU = 65536 ) var ( @@ -57,7 +60,7 @@ func (*stubLinkEndpoint) LinkAddress() tcpip.LinkAddress { return "" } -func (*stubLinkEndpoint) WritePacket(*stack.Route, *stack.GSO, tcpip.NetworkProtocolNumber, stack.PacketBuffer) *tcpip.Error { +func (*stubLinkEndpoint) WritePacket(*stack.Route, *stack.GSO, tcpip.NetworkProtocolNumber, *stack.PacketBuffer) *tcpip.Error { return nil } @@ -67,7 +70,7 @@ type stubDispatcher struct { stack.TransportDispatcher } -func (*stubDispatcher) DeliverTransportPacket(*stack.Route, tcpip.TransportProtocolNumber, stack.PacketBuffer) { +func (*stubDispatcher) DeliverTransportPacket(*stack.Route, tcpip.TransportProtocolNumber, *stack.PacketBuffer) { } type stubLinkAddressCache struct { @@ -179,36 +182,32 @@ func TestICMPCounts(t *testing.T) { }, } - handleIPv6Payload := func(hdr buffer.Prependable) { - payloadLength := hdr.UsedLength() - ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize)) + handleIPv6Payload := func(icmp header.ICMPv6) { + ip := header.IPv6(buffer.NewView(header.IPv6MinimumSize)) ip.Encode(&header.IPv6Fields{ - PayloadLength: uint16(payloadLength), + PayloadLength: uint16(len(icmp)), NextHeader: uint8(header.ICMPv6ProtocolNumber), HopLimit: header.NDPHopLimit, SrcAddr: r.LocalAddress, DstAddr: r.RemoteAddress, }) - ep.HandlePacket(&r, stack.PacketBuffer{ - Data: hdr.View().ToVectorisedView(), + ep.HandlePacket(&r, &stack.PacketBuffer{ + NetworkHeader: buffer.View(ip), + Data: buffer.View(icmp).ToVectorisedView(), }) } for _, typ := range types { - extraDataLen := len(typ.extraData) - hdr := buffer.NewPrependable(header.IPv6MinimumSize + typ.size + extraDataLen) - extraData := buffer.View(hdr.Prepend(extraDataLen)) - copy(extraData, typ.extraData) - pkt := header.ICMPv6(hdr.Prepend(typ.size)) - pkt.SetType(typ.typ) - pkt.SetChecksum(header.ICMPv6Checksum(pkt, r.LocalAddress, r.RemoteAddress, extraData.ToVectorisedView())) - - handleIPv6Payload(hdr) + icmp := header.ICMPv6(buffer.NewView(typ.size + len(typ.extraData))) + copy(icmp[typ.size:], typ.extraData) + icmp.SetType(typ.typ) + icmp.SetChecksum(header.ICMPv6Checksum(icmp[:typ.size], r.LocalAddress, r.RemoteAddress, buffer.View(typ.extraData).ToVectorisedView())) + handleIPv6Payload(icmp) } // Construct an empty ICMP packet so that // Stats().ICMP.ICMPv6ReceivedPacketStats.Invalid is incremented. - handleIPv6Payload(buffer.NewPrependable(header.IPv6MinimumSize)) + handleIPv6Payload(header.ICMPv6(buffer.NewView(header.IPv6MinimumSize))) icmpv6Stats := s.Stats().ICMP.V6PacketsReceived visitStats(reflect.ValueOf(&icmpv6Stats).Elem(), func(name string, s *tcpip.StatCounter) { @@ -261,8 +260,7 @@ func newTestContext(t *testing.T) *testContext { }), } - const defaultMTU = 65536 - c.linkEP0 = channel.New(256, defaultMTU, linkAddr0) + c.linkEP0 = channel.New(defaultChannelSize, defaultMTU, linkAddr0) wrappedEP0 := stack.LinkEndpoint(endpointWithResolutionCapability{LinkEndpoint: c.linkEP0}) if testing.Verbose() { @@ -275,7 +273,7 @@ func newTestContext(t *testing.T) *testContext { t.Fatalf("AddAddress lladdr0: %v", err) } - c.linkEP1 = channel.New(256, defaultMTU, linkAddr1) + c.linkEP1 = channel.New(defaultChannelSize, defaultMTU, linkAddr1) wrappedEP1 := stack.LinkEndpoint(endpointWithResolutionCapability{LinkEndpoint: c.linkEP1}) if err := c.s1.CreateNIC(1, wrappedEP1); err != nil { t.Fatalf("CreateNIC failed: %v", err) @@ -328,7 +326,7 @@ func routeICMPv6Packet(t *testing.T, args routeArgs, fn func(*testing.T, header. views := []buffer.View{pi.Pkt.Header.View(), pi.Pkt.Data.ToView()} size := pi.Pkt.Header.UsedLength() + pi.Pkt.Data.Size() vv := buffer.NewVectorisedView(size, views) - args.dst.InjectLinkAddr(pi.Proto, args.dst.LinkAddress(), stack.PacketBuffer{ + args.dst.InjectLinkAddr(pi.Proto, args.dst.LinkAddress(), &stack.PacketBuffer{ Data: vv, }) } @@ -546,25 +544,22 @@ func TestICMPChecksumValidationSimple(t *testing.T) { } handleIPv6Payload := func(checksum bool) { - extraDataLen := len(typ.extraData) - hdr := buffer.NewPrependable(header.IPv6MinimumSize + typ.size + extraDataLen) - extraData := buffer.View(hdr.Prepend(extraDataLen)) - copy(extraData, typ.extraData) - pkt := header.ICMPv6(hdr.Prepend(typ.size)) - pkt.SetType(typ.typ) + icmp := header.ICMPv6(buffer.NewView(typ.size + len(typ.extraData))) + copy(icmp[typ.size:], typ.extraData) + icmp.SetType(typ.typ) if checksum { - pkt.SetChecksum(header.ICMPv6Checksum(pkt, lladdr1, lladdr0, extraData.ToVectorisedView())) + icmp.SetChecksum(header.ICMPv6Checksum(icmp, lladdr1, lladdr0, buffer.View{}.ToVectorisedView())) } - ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize)) + ip := header.IPv6(buffer.NewView(header.IPv6MinimumSize)) ip.Encode(&header.IPv6Fields{ - PayloadLength: uint16(typ.size + extraDataLen), + PayloadLength: uint16(len(icmp)), NextHeader: uint8(header.ICMPv6ProtocolNumber), HopLimit: header.NDPHopLimit, SrcAddr: lladdr1, DstAddr: lladdr0, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ - Data: hdr.View().ToVectorisedView(), + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ + Data: buffer.NewVectorisedView(len(ip)+len(icmp), []buffer.View{buffer.View(ip), buffer.View(icmp)}), }) } @@ -740,7 +735,7 @@ func TestICMPChecksumValidationWithPayload(t *testing.T) { SrcAddr: lladdr1, DstAddr: lladdr0, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) } @@ -918,7 +913,7 @@ func TestICMPChecksumValidationWithPayloadMultipleViews(t *testing.T) { SrcAddr: lladdr1, DstAddr: lladdr0, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: buffer.NewVectorisedView(header.IPv6MinimumSize+size+payloadSize, []buffer.View{hdr.View(), payload}), }) } @@ -958,3 +953,47 @@ func TestICMPChecksumValidationWithPayloadMultipleViews(t *testing.T) { }) } } + +func TestLinkAddressRequest(t *testing.T) { + snaddr := header.SolicitedNodeAddr(lladdr0) + mcaddr := header.EthernetAddressFromMulticastIPv6Address(snaddr) + + tests := []struct { + name string + remoteLinkAddr tcpip.LinkAddress + expectLinkAddr tcpip.LinkAddress + }{ + { + name: "Unicast", + remoteLinkAddr: linkAddr1, + expectLinkAddr: linkAddr1, + }, + { + name: "Multicast", + remoteLinkAddr: "", + expectLinkAddr: mcaddr, + }, + } + + for _, test := range tests { + p := NewProtocol() + linkRes, ok := p.(stack.LinkAddressResolver) + if !ok { + t.Fatalf("expected IPv6 protocol to implement stack.LinkAddressResolver") + } + + linkEP := channel.New(defaultChannelSize, defaultMTU, linkAddr0) + if err := linkRes.LinkAddressRequest(lladdr0, lladdr1, test.remoteLinkAddr, linkEP); err != nil { + t.Errorf("got p.LinkAddressRequest(%s, %s, %s, _) = %s", lladdr0, lladdr1, test.remoteLinkAddr, err) + } + + pkt, ok := linkEP.Read() + if !ok { + t.Fatal("expected to send a link address request") + } + + if got, want := pkt.Route.RemoteLinkAddress, test.expectLinkAddr; got != want { + t.Errorf("got pkt.Route.RemoteLinkAddress = %s, want = %s", got, want) + } + } +} diff --git a/pkg/tcpip/network/ipv6/ipv6.go b/pkg/tcpip/network/ipv6/ipv6.go index daf1fcbc6..a0a5c9c01 100644 --- a/pkg/tcpip/network/ipv6/ipv6.go +++ b/pkg/tcpip/network/ipv6/ipv6.go @@ -28,7 +28,6 @@ import ( "gvisor.dev/gvisor/pkg/tcpip/buffer" "gvisor.dev/gvisor/pkg/tcpip/header" "gvisor.dev/gvisor/pkg/tcpip/network/fragmentation" - "gvisor.dev/gvisor/pkg/tcpip/network/hash" "gvisor.dev/gvisor/pkg/tcpip/stack" ) @@ -116,7 +115,7 @@ func (e *endpoint) addIPHeader(r *stack.Route, hdr *buffer.Prependable, payloadS } // WritePacket writes a packet to the given destination address and protocol. -func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt stack.PacketBuffer) *tcpip.Error { +func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt *stack.PacketBuffer) *tcpip.Error { ip := e.addIPHeader(r, &pkt.Header, pkt.Data.Size(), params) pkt.NetworkHeader = buffer.View(ip) @@ -128,7 +127,7 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.Netw views = append(views, pkt.Data.Views()...) loopedR := r.MakeLoopedRoute() - e.HandlePacket(&loopedR, stack.PacketBuffer{ + e.HandlePacket(&loopedR, &stack.PacketBuffer{ Data: buffer.NewVectorisedView(len(views[0])+pkt.Data.Size(), views), }) @@ -163,30 +162,28 @@ func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.Packe // WriteHeaderIncludedPacker implements stack.NetworkEndpoint. It is not yet // supported by IPv6. -func (*endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt stack.PacketBuffer) *tcpip.Error { +func (*endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuffer) *tcpip.Error { // TODO(b/146666412): Support IPv6 header-included packets. return tcpip.ErrNotSupported } // HandlePacket is called by the link layer when new ipv6 packets arrive for // this endpoint. -func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { - headerView, ok := pkt.Data.PullUp(header.IPv6MinimumSize) - if !ok { +func (e *endpoint) HandlePacket(r *stack.Route, pkt *stack.PacketBuffer) { + h := header.IPv6(pkt.NetworkHeader) + if !h.IsValid(pkt.Data.Size() + len(pkt.NetworkHeader) + len(pkt.TransportHeader)) { r.Stats().IP.MalformedPacketsReceived.Increment() return } - h := header.IPv6(headerView) - if !h.IsValid(pkt.Data.Size()) { - r.Stats().IP.MalformedPacketsReceived.Increment() - return - } - - pkt.NetworkHeader = headerView[:header.IPv6MinimumSize] - pkt.Data.TrimFront(header.IPv6MinimumSize) - pkt.Data.CapLength(int(h.PayloadLength())) - it := header.MakeIPv6PayloadIterator(header.IPv6ExtensionHeaderIdentifier(h.NextHeader()), pkt.Data) + // vv consists of: + // - Any IPv6 header bytes after the first 40 (i.e. extensions). + // - The transport header, if present. + // - Any other payload data. + vv := pkt.NetworkHeader[header.IPv6MinimumSize:].ToVectorisedView() + vv.AppendView(pkt.TransportHeader) + vv.Append(pkt.Data) + it := header.MakeIPv6PayloadIterator(header.IPv6ExtensionHeaderIdentifier(h.NextHeader()), vv) hasFragmentHeader := false for firstHeader := true; ; firstHeader = false { @@ -262,9 +259,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { case header.IPv6FragmentExtHdr: hasFragmentHeader = true - fragmentOffset := extHdr.FragmentOffset() - more := extHdr.More() - if !more && fragmentOffset == 0 { + if extHdr.IsAtomic() { // This fragment extension header indicates that this packet is an // atomic fragment. An atomic fragment is a fragment that contains // all the data required to reassemble a full packet. As per RFC 6946, @@ -277,9 +272,9 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { // Don't consume the iterator if we have the first fragment because we // will use it to validate that the first fragment holds the upper layer // header. - rawPayload := it.AsRawHeader(fragmentOffset != 0 /* consume */) + rawPayload := it.AsRawHeader(extHdr.FragmentOffset() != 0 /* consume */) - if fragmentOffset == 0 { + if extHdr.FragmentOffset() == 0 { // Check that the iterator ends with a raw payload as the first fragment // should include all headers up to and including any upper layer // headers, as per RFC 8200 section 4.5; only upper layer data @@ -332,7 +327,7 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { } // The packet is a fragment, let's try to reassemble it. - start := fragmentOffset * header.IPv6FragmentExtHdrFragmentOffsetBytesPerUnit + start := extHdr.FragmentOffset() * header.IPv6FragmentExtHdrFragmentOffsetBytesPerUnit last := start + uint16(fragmentPayloadLen) - 1 // Drop the packet if the fragmentOffset is incorrect. i.e the @@ -345,7 +340,21 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { } var ready bool - pkt.Data, ready, err = e.fragmentation.Process(hash.IPv6FragmentHash(h, extHdr.ID()), start, last, more, rawPayload.Buf) + // Note that pkt doesn't have its transport header set after reassembly, + // and won't until DeliverNetworkPacket sets it. + pkt.Data, ready, err = e.fragmentation.Process( + // IPv6 ignores the Protocol field since the ID only needs to be unique + // across source-destination pairs, as per RFC 8200 section 4.5. + fragmentation.FragmentID{ + Source: h.SourceAddress(), + Destination: h.DestinationAddress(), + ID: extHdr.ID(), + }, + start, + last, + extHdr.More(), + rawPayload.Buf, + ) if err != nil { r.Stats().IP.MalformedPacketsReceived.Increment() r.Stats().IP.MalformedFragmentsReceived.Increment() @@ -394,10 +403,17 @@ func (e *endpoint) HandlePacket(r *stack.Route, pkt stack.PacketBuffer) { case header.IPv6RawPayloadHeader: // If the last header in the payload isn't a known IPv6 extension header, // handle it as if it is transport layer data. + + // For unfragmented packets, extHdr still contains the transport header. + // Get rid of it. + // + // For reassembled fragments, pkt.TransportHeader is unset, so this is a + // no-op and pkt.Data begins with the transport header. + extHdr.Buf.TrimFront(len(pkt.TransportHeader)) pkt.Data = extHdr.Buf if p := tcpip.TransportProtocolNumber(extHdr.Identifier); p == header.ICMPv6ProtocolNumber { - e.handleICMP(r, headerView, pkt, hasFragmentHeader) + e.handleICMP(r, pkt, hasFragmentHeader) } else { r.Stats().IP.PacketsDelivered.Increment() // TODO(b/152019344): Send an ICMPv6 Parameter Problem, Code 1 error @@ -462,7 +478,7 @@ func (p *protocol) NewEndpoint(nicID tcpip.NICID, addrWithPrefix tcpip.AddressWi linkEP: linkEP, linkAddrCache: linkAddrCache, dispatcher: dispatcher, - fragmentation: fragmentation.NewFragmentation(fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, fragmentation.DefaultReassembleTimeout), + fragmentation: fragmentation.NewFragmentation(header.IPv6FragmentExtHdrFragmentOffsetBytesPerUnit, fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, fragmentation.DefaultReassembleTimeout), protocol: p, }, nil } @@ -505,6 +521,79 @@ func (*protocol) Close() {} // Wait implements stack.TransportProtocol.Wait. func (*protocol) Wait() {} +// Parse implements stack.TransportProtocol.Parse. +func (*protocol) Parse(pkt *stack.PacketBuffer) (proto tcpip.TransportProtocolNumber, hasTransportHdr bool, ok bool) { + hdr, ok := pkt.Data.PullUp(header.IPv6MinimumSize) + if !ok { + return 0, false, false + } + ipHdr := header.IPv6(hdr) + + // dataClone consists of: + // - Any IPv6 header bytes after the first 40 (i.e. extensions). + // - The transport header, if present. + // - Any other payload data. + views := [8]buffer.View{} + dataClone := pkt.Data.Clone(views[:]) + dataClone.TrimFront(header.IPv6MinimumSize) + it := header.MakeIPv6PayloadIterator(header.IPv6ExtensionHeaderIdentifier(ipHdr.NextHeader()), dataClone) + + // Iterate over the IPv6 extensions to find their length. + // + // Parsing occurs again in HandlePacket because we don't track the + // extensions in PacketBuffer. Unfortunately, that means HandlePacket + // has to do the parsing work again. + var nextHdr tcpip.TransportProtocolNumber + foundNext := true + extensionsSize := 0 +traverseExtensions: + for extHdr, done, err := it.Next(); ; extHdr, done, err = it.Next() { + if err != nil { + break + } + // If we exhaust the extension list, the entire packet is the IPv6 header + // and (possibly) extensions. + if done { + extensionsSize = dataClone.Size() + foundNext = false + break + } + + switch extHdr := extHdr.(type) { + case header.IPv6FragmentExtHdr: + // If this is an atomic fragment, we don't have to treat it specially. + if !extHdr.More() && extHdr.FragmentOffset() == 0 { + continue + } + // This is a non-atomic fragment and has to be re-assembled before we can + // examine the payload for a transport header. + foundNext = false + + case header.IPv6RawPayloadHeader: + // We've found the payload after any extensions. + extensionsSize = dataClone.Size() - extHdr.Buf.Size() + nextHdr = tcpip.TransportProtocolNumber(extHdr.Identifier) + break traverseExtensions + + default: + // Any other extension is a no-op, keep looping until we find the payload. + } + } + + // Put the IPv6 header with extensions in pkt.NetworkHeader. + hdr, ok = pkt.Data.PullUp(header.IPv6MinimumSize + extensionsSize) + if !ok { + panic(fmt.Sprintf("pkt.Data should have at least %d bytes, but only has %d.", header.IPv6MinimumSize+extensionsSize, pkt.Data.Size())) + } + ipHdr = header.IPv6(hdr) + + pkt.NetworkHeader = hdr + pkt.Data.TrimFront(len(hdr)) + pkt.Data.CapLength(int(ipHdr.PayloadLength())) + + return nextHdr, foundNext, true +} + // calculateMTU calculates the network-layer payload MTU based on the link-layer // payload mtu. func calculateMTU(mtu uint32) uint32 { diff --git a/pkg/tcpip/network/ipv6/ipv6_test.go b/pkg/tcpip/network/ipv6/ipv6_test.go index 841a0cb7a..3d65814de 100644 --- a/pkg/tcpip/network/ipv6/ipv6_test.go +++ b/pkg/tcpip/network/ipv6/ipv6_test.go @@ -65,7 +65,7 @@ func testReceiveICMP(t *testing.T, s *stack.Stack, e *channel.Endpoint, src, dst DstAddr: dst, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -123,7 +123,7 @@ func testReceiveUDP(t *testing.T, s *stack.Stack, e *channel.Endpoint, src, dst DstAddr: dst, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -637,7 +637,7 @@ func TestReceiveIPv6ExtHdrs(t *testing.T) { DstAddr: addr2, }) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -673,20 +673,27 @@ func TestReceiveIPv6ExtHdrs(t *testing.T) { // fragmentData holds the IPv6 payload for a fragmented IPv6 packet. type fragmentData struct { + srcAddr tcpip.Address + dstAddr tcpip.Address nextHdr uint8 data buffer.VectorisedView } func TestReceiveIPv6Fragments(t *testing.T) { - const nicID = 1 - const udpPayload1Length = 256 - const udpPayload2Length = 128 - const fragmentExtHdrLen = 8 - // Note, not all routing extension headers will be 8 bytes but this test - // uses 8 byte routing extension headers for most sub tests. - const routingExtHdrLen = 8 - - udpGen := func(payload []byte, multiplier uint8) buffer.View { + const ( + nicID = 1 + udpPayload1Length = 256 + udpPayload2Length = 128 + // Used to test cases where the fragment blocks are not a multiple of + // the fragment block size of 8 (RFC 8200 section 4.5). + udpPayload3Length = 127 + fragmentExtHdrLen = 8 + // Note, not all routing extension headers will be 8 bytes but this test + // uses 8 byte routing extension headers for most sub tests. + routingExtHdrLen = 8 + ) + + udpGen := func(payload []byte, multiplier uint8, src, dst tcpip.Address) buffer.View { payloadLen := len(payload) for i := 0; i < payloadLen; i++ { payload[i] = uint8(i) * multiplier @@ -702,19 +709,27 @@ func TestReceiveIPv6Fragments(t *testing.T) { Length: uint16(udpLength), }) copy(u.Payload(), payload) - sum := header.PseudoHeaderChecksum(udp.ProtocolNumber, addr1, addr2, uint16(udpLength)) + sum := header.PseudoHeaderChecksum(udp.ProtocolNumber, src, dst, uint16(udpLength)) sum = header.Checksum(payload, sum) u.SetChecksum(^u.CalculateChecksum(sum)) return hdr.View() } - var udpPayload1Buf [udpPayload1Length]byte - udpPayload1 := udpPayload1Buf[:] - ipv6Payload1 := udpGen(udpPayload1, 1) + var udpPayload1Addr1ToAddr2Buf [udpPayload1Length]byte + udpPayload1Addr1ToAddr2 := udpPayload1Addr1ToAddr2Buf[:] + ipv6Payload1Addr1ToAddr2 := udpGen(udpPayload1Addr1ToAddr2, 1, addr1, addr2) + + var udpPayload1Addr3ToAddr2Buf [udpPayload1Length]byte + udpPayload1Addr3ToAddr2 := udpPayload1Addr3ToAddr2Buf[:] + ipv6Payload1Addr3ToAddr2 := udpGen(udpPayload1Addr3ToAddr2, 4, addr3, addr2) - var udpPayload2Buf [udpPayload2Length]byte - udpPayload2 := udpPayload2Buf[:] - ipv6Payload2 := udpGen(udpPayload2, 2) + var udpPayload2Addr1ToAddr2Buf [udpPayload2Length]byte + udpPayload2Addr1ToAddr2 := udpPayload2Addr1ToAddr2Buf[:] + ipv6Payload2Addr1ToAddr2 := udpGen(udpPayload2Addr1ToAddr2, 2, addr1, addr2) + + var udpPayload3Addr1ToAddr2Buf [udpPayload3Length]byte + udpPayload3Addr1ToAddr2 := udpPayload3Addr1ToAddr2Buf[:] + ipv6Payload3Addr1ToAddr2 := udpGen(udpPayload3Addr1ToAddr2, 3, addr1, addr2) tests := []struct { name string @@ -726,34 +741,98 @@ func TestReceiveIPv6Fragments(t *testing.T) { name: "No fragmentation", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: uint8(header.UDPProtocolNumber), - data: ipv6Payload1.ToVectorisedView(), + data: ipv6Payload1Addr1ToAddr2.ToVectorisedView(), }, }, - expectedPayloads: [][]byte{udpPayload1}, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2}, }, { name: "Atomic fragment", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, + nextHdr: fragmentExtHdrID, + data: buffer.NewVectorisedView( + fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2), + []buffer.View{ + // Fragment extension header. + buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 0, 0, 0, 0, 0}), + + ipv6Payload1Addr1ToAddr2, + }, + ), + }, + }, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2}, + }, + { + name: "Atomic fragment with size not a multiple of fragment block size", + fragments: []fragmentData{ + { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( - fragmentExtHdrLen+len(ipv6Payload1), + fragmentExtHdrLen+len(ipv6Payload3Addr1ToAddr2), []buffer.View{ // Fragment extension header. buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 0, 0, 0, 0, 0}), - ipv6Payload1, + ipv6Payload3Addr1ToAddr2, }, ), }, }, - expectedPayloads: [][]byte{udpPayload1}, + expectedPayloads: [][]byte{udpPayload3Addr1ToAddr2}, }, { name: "Two fragments", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, + nextHdr: fragmentExtHdrID, + data: buffer.NewVectorisedView( + fragmentExtHdrLen+64, + []buffer.View{ + // Fragment extension header. + // + // Fragment offset = 0, More = true, ID = 1 + buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), + + ipv6Payload1Addr1ToAddr2[:64], + }, + ), + }, + { + srcAddr: addr1, + dstAddr: addr2, + nextHdr: fragmentExtHdrID, + data: buffer.NewVectorisedView( + fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64, + []buffer.View{ + // Fragment extension header. + // + // Fragment offset = 8, More = false, ID = 1 + buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}), + + ipv6Payload1Addr1ToAddr2[64:], + }, + ), + }, + }, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2}, + }, + { + name: "Two fragments with last fragment size not a multiple of fragment block size", + fragments: []fragmentData{ + { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( fragmentExtHdrLen+64, @@ -763,31 +842,73 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Fragment offset = 0, More = true, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), - ipv6Payload1[:64], + ipv6Payload3Addr1ToAddr2[:64], }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( - fragmentExtHdrLen+len(ipv6Payload1)-64, + fragmentExtHdrLen+len(ipv6Payload3Addr1ToAddr2)-64, []buffer.View{ // Fragment extension header. // // Fragment offset = 8, More = false, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}), - ipv6Payload1[64:], + ipv6Payload3Addr1ToAddr2[64:], }, ), }, }, - expectedPayloads: [][]byte{udpPayload1}, + expectedPayloads: [][]byte{udpPayload3Addr1ToAddr2}, + }, + { + name: "Two fragments with first fragment size not a multiple of fragment block size", + fragments: []fragmentData{ + { + srcAddr: addr1, + dstAddr: addr2, + nextHdr: fragmentExtHdrID, + data: buffer.NewVectorisedView( + fragmentExtHdrLen+63, + []buffer.View{ + // Fragment extension header. + // + // Fragment offset = 0, More = true, ID = 1 + buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), + + ipv6Payload3Addr1ToAddr2[:63], + }, + ), + }, + { + srcAddr: addr1, + dstAddr: addr2, + nextHdr: fragmentExtHdrID, + data: buffer.NewVectorisedView( + fragmentExtHdrLen+len(ipv6Payload3Addr1ToAddr2)-63, + []buffer.View{ + // Fragment extension header. + // + // Fragment offset = 8, More = false, ID = 1 + buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}), + + ipv6Payload3Addr1ToAddr2[63:], + }, + ), + }, + }, + expectedPayloads: nil, }, { name: "Two fragments with different IDs", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( fragmentExtHdrLen+64, @@ -797,21 +918,23 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Fragment offset = 0, More = true, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), - ipv6Payload1[:64], + ipv6Payload1Addr1ToAddr2[:64], }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( - fragmentExtHdrLen+len(ipv6Payload1)-64, + fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64, []buffer.View{ // Fragment extension header. // // Fragment offset = 8, More = false, ID = 2 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 2}), - ipv6Payload1[64:], + ipv6Payload1Addr1ToAddr2[64:], }, ), }, @@ -822,6 +945,8 @@ func TestReceiveIPv6Fragments(t *testing.T) { name: "Two fragments with per-fragment routing header with zero segments left", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: routingExtHdrID, data: buffer.NewVectorisedView( routingExtHdrLen+fragmentExtHdrLen+64, @@ -836,14 +961,16 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Fragment offset = 0, More = true, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), - ipv6Payload1[:64], + ipv6Payload1Addr1ToAddr2[:64], }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: routingExtHdrID, data: buffer.NewVectorisedView( - routingExtHdrLen+fragmentExtHdrLen+len(ipv6Payload1)-64, + routingExtHdrLen+fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64, []buffer.View{ // Routing extension header. // @@ -855,17 +982,19 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Fragment offset = 8, More = false, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}), - ipv6Payload1[64:], + ipv6Payload1Addr1ToAddr2[64:], }, ), }, }, - expectedPayloads: [][]byte{udpPayload1}, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2}, }, { name: "Two fragments with per-fragment routing header with non-zero segments left", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: routingExtHdrID, data: buffer.NewVectorisedView( routingExtHdrLen+fragmentExtHdrLen+64, @@ -880,14 +1009,16 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Fragment offset = 0, More = true, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), - ipv6Payload1[:64], + ipv6Payload1Addr1ToAddr2[:64], }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: routingExtHdrID, data: buffer.NewVectorisedView( - routingExtHdrLen+fragmentExtHdrLen+len(ipv6Payload1)-64, + routingExtHdrLen+fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64, []buffer.View{ // Routing extension header. // @@ -899,7 +1030,7 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Fragment offset = 9, More = false, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 72, 0, 0, 0, 1}), - ipv6Payload1[64:], + ipv6Payload1Addr1ToAddr2[64:], }, ), }, @@ -910,6 +1041,8 @@ func TestReceiveIPv6Fragments(t *testing.T) { name: "Two fragments with routing header with zero segments left", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( routingExtHdrLen+fragmentExtHdrLen+64, @@ -924,31 +1057,35 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Segments left = 0. buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 1, 0, 2, 3, 4, 5}), - ipv6Payload1[:64], + ipv6Payload1Addr1ToAddr2[:64], }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( - fragmentExtHdrLen+len(ipv6Payload1)-64, + fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64, []buffer.View{ // Fragment extension header. // // Fragment offset = 9, More = false, ID = 1 buffer.View([]byte{routingExtHdrID, 0, 0, 72, 0, 0, 0, 1}), - ipv6Payload1[64:], + ipv6Payload1Addr1ToAddr2[64:], }, ), }, }, - expectedPayloads: [][]byte{udpPayload1}, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2}, }, { name: "Two fragments with routing header with non-zero segments left", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( routingExtHdrLen+fragmentExtHdrLen+64, @@ -963,21 +1100,23 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Segments left = 1. buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 1, 1, 2, 3, 4, 5}), - ipv6Payload1[:64], + ipv6Payload1Addr1ToAddr2[:64], }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( - fragmentExtHdrLen+len(ipv6Payload1)-64, + fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64, []buffer.View{ // Fragment extension header. // // Fragment offset = 9, More = false, ID = 1 buffer.View([]byte{routingExtHdrID, 0, 0, 72, 0, 0, 0, 1}), - ipv6Payload1[64:], + ipv6Payload1Addr1ToAddr2[64:], }, ), }, @@ -988,6 +1127,8 @@ func TestReceiveIPv6Fragments(t *testing.T) { name: "Two fragments with routing header with zero segments left across fragments", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( // The length of this payload is fragmentExtHdrLen+8 because the @@ -1008,12 +1149,14 @@ func TestReceiveIPv6Fragments(t *testing.T) { ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( // The length of this payload is - // fragmentExtHdrLen+8+len(ipv6Payload1) because the last 8 bytes of + // fragmentExtHdrLen+8+len(ipv6Payload1Addr1ToAddr2) because the last 8 bytes of // the 16 byte routing extension header is in this fagment. - fragmentExtHdrLen+8+len(ipv6Payload1), + fragmentExtHdrLen+8+len(ipv6Payload1Addr1ToAddr2), []buffer.View{ // Fragment extension header. // @@ -1023,7 +1166,7 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Routing extension header (part 2) buffer.View([]byte{6, 7, 8, 9, 10, 11, 12, 13}), - ipv6Payload1, + ipv6Payload1Addr1ToAddr2, }, ), }, @@ -1034,6 +1177,8 @@ func TestReceiveIPv6Fragments(t *testing.T) { name: "Two fragments with routing header with non-zero segments left across fragments", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( // The length of this payload is fragmentExtHdrLen+8 because the @@ -1054,12 +1199,14 @@ func TestReceiveIPv6Fragments(t *testing.T) { ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( // The length of this payload is - // fragmentExtHdrLen+8+len(ipv6Payload1) because the last 8 bytes of + // fragmentExtHdrLen+8+len(ipv6Payload1Addr1ToAddr2) because the last 8 bytes of // the 16 byte routing extension header is in this fagment. - fragmentExtHdrLen+8+len(ipv6Payload1), + fragmentExtHdrLen+8+len(ipv6Payload1Addr1ToAddr2), []buffer.View{ // Fragment extension header. // @@ -1069,7 +1216,7 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Routing extension header (part 2) buffer.View([]byte{6, 7, 8, 9, 10, 11, 12, 13}), - ipv6Payload1, + ipv6Payload1Addr1ToAddr2, }, ), }, @@ -1082,6 +1229,8 @@ func TestReceiveIPv6Fragments(t *testing.T) { name: "Two fragments with atomic", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( fragmentExtHdrLen+64, @@ -1091,47 +1240,53 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Fragment offset = 0, More = true, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), - ipv6Payload1[:64], + ipv6Payload1Addr1ToAddr2[:64], }, ), }, // This fragment has the same ID as the other fragments but is an atomic // fragment. It should not interfere with the other fragments. { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( - fragmentExtHdrLen+len(ipv6Payload2), + fragmentExtHdrLen+len(ipv6Payload2Addr1ToAddr2), []buffer.View{ // Fragment extension header. // // Fragment offset = 0, More = false, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 0, 0, 0, 0, 1}), - ipv6Payload2, + ipv6Payload2Addr1ToAddr2, }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( - fragmentExtHdrLen+len(ipv6Payload1)-64, + fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64, []buffer.View{ // Fragment extension header. // // Fragment offset = 8, More = false, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}), - ipv6Payload1[64:], + ipv6Payload1Addr1ToAddr2[64:], }, ), }, }, - expectedPayloads: [][]byte{udpPayload2, udpPayload1}, + expectedPayloads: [][]byte{udpPayload2Addr1ToAddr2, udpPayload1Addr1ToAddr2}, }, { name: "Two interleaved fragmented packets", fragments: []fragmentData{ { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( fragmentExtHdrLen+64, @@ -1141,11 +1296,13 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Fragment offset = 0, More = true, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), - ipv6Payload1[:64], + ipv6Payload1Addr1ToAddr2[:64], }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( fragmentExtHdrLen+32, @@ -1155,40 +1312,114 @@ func TestReceiveIPv6Fragments(t *testing.T) { // Fragment offset = 0, More = true, ID = 2 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 2}), - ipv6Payload2[:32], + ipv6Payload2Addr1ToAddr2[:32], }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( - fragmentExtHdrLen+len(ipv6Payload1)-64, + fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64, []buffer.View{ // Fragment extension header. // // Fragment offset = 8, More = false, ID = 1 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}), - ipv6Payload1[64:], + ipv6Payload1Addr1ToAddr2[64:], }, ), }, { + srcAddr: addr1, + dstAddr: addr2, nextHdr: fragmentExtHdrID, data: buffer.NewVectorisedView( - fragmentExtHdrLen+len(ipv6Payload2)-32, + fragmentExtHdrLen+len(ipv6Payload2Addr1ToAddr2)-32, []buffer.View{ // Fragment extension header. // // Fragment offset = 4, More = false, ID = 2 buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 32, 0, 0, 0, 2}), - ipv6Payload2[32:], + ipv6Payload2Addr1ToAddr2[32:], + }, + ), + }, + }, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2, udpPayload2Addr1ToAddr2}, + }, + { + name: "Two interleaved fragmented packets from different sources but with same ID", + fragments: []fragmentData{ + { + srcAddr: addr1, + dstAddr: addr2, + nextHdr: fragmentExtHdrID, + data: buffer.NewVectorisedView( + fragmentExtHdrLen+64, + []buffer.View{ + // Fragment extension header. + // + // Fragment offset = 0, More = true, ID = 1 + buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), + + ipv6Payload1Addr1ToAddr2[:64], + }, + ), + }, + { + srcAddr: addr3, + dstAddr: addr2, + nextHdr: fragmentExtHdrID, + data: buffer.NewVectorisedView( + fragmentExtHdrLen+32, + []buffer.View{ + // Fragment extension header. + // + // Fragment offset = 0, More = true, ID = 1 + buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 1, 0, 0, 0, 1}), + + ipv6Payload1Addr3ToAddr2[:32], + }, + ), + }, + { + srcAddr: addr1, + dstAddr: addr2, + nextHdr: fragmentExtHdrID, + data: buffer.NewVectorisedView( + fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-64, + []buffer.View{ + // Fragment extension header. + // + // Fragment offset = 8, More = false, ID = 1 + buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 64, 0, 0, 0, 1}), + + ipv6Payload1Addr1ToAddr2[64:], + }, + ), + }, + { + srcAddr: addr3, + dstAddr: addr2, + nextHdr: fragmentExtHdrID, + data: buffer.NewVectorisedView( + fragmentExtHdrLen+len(ipv6Payload1Addr1ToAddr2)-32, + []buffer.View{ + // Fragment extension header. + // + // Fragment offset = 4, More = false, ID = 1 + buffer.View([]byte{uint8(header.UDPProtocolNumber), 0, 0, 32, 0, 0, 0, 1}), + + ipv6Payload1Addr3ToAddr2[32:], }, ), }, }, - expectedPayloads: [][]byte{udpPayload1, udpPayload2}, + expectedPayloads: [][]byte{udpPayload1Addr1ToAddr2, udpPayload1Addr3ToAddr2}, }, } @@ -1231,14 +1462,14 @@ func TestReceiveIPv6Fragments(t *testing.T) { PayloadLength: uint16(f.data.Size()), NextHeader: f.nextHdr, HopLimit: 255, - SrcAddr: addr1, - DstAddr: addr2, + SrcAddr: f.srcAddr, + DstAddr: f.dstAddr, }) vv := hdr.View().ToVectorisedView() vv.Append(f.data) - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: vv, }) } diff --git a/pkg/tcpip/network/ipv6/ndp_test.go b/pkg/tcpip/network/ipv6/ndp_test.go index 12b70f7e9..64239ce9a 100644 --- a/pkg/tcpip/network/ipv6/ndp_test.go +++ b/pkg/tcpip/network/ipv6/ndp_test.go @@ -136,7 +136,7 @@ func TestNeighorSolicitationWithSourceLinkLayerOption(t *testing.T) { t.Fatalf("got invalid = %d, want = 0", got) } - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -380,7 +380,7 @@ func TestNeighorSolicitationResponse(t *testing.T) { t.Fatalf("got invalid = %d, want = 0", got) } - e.InjectLinkAddr(ProtocolNumber, test.nsSrcLinkAddr, stack.PacketBuffer{ + e.InjectLinkAddr(ProtocolNumber, test.nsSrcLinkAddr, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -497,7 +497,7 @@ func TestNeighorAdvertisementWithTargetLinkLayerOption(t *testing.T) { t.Fatalf("got invalid = %d, want = 0", got) } - e.InjectInbound(ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) @@ -551,25 +551,29 @@ func TestNDPValidation(t *testing.T) { return s, ep, r } - handleIPv6Payload := func(hdr buffer.Prependable, hopLimit uint8, atomicFragment bool, ep stack.NetworkEndpoint, r *stack.Route) { + handleIPv6Payload := func(payload buffer.View, hopLimit uint8, atomicFragment bool, ep stack.NetworkEndpoint, r *stack.Route) { nextHdr := uint8(header.ICMPv6ProtocolNumber) + var extensions buffer.View if atomicFragment { - bytes := hdr.Prepend(header.IPv6FragmentExtHdrLength) - bytes[0] = nextHdr + extensions = buffer.NewView(header.IPv6FragmentExtHdrLength) + extensions[0] = nextHdr nextHdr = uint8(header.IPv6FragmentExtHdrIdentifier) } - payloadLength := hdr.UsedLength() - ip := header.IPv6(hdr.Prepend(header.IPv6MinimumSize)) + ip := header.IPv6(buffer.NewView(header.IPv6MinimumSize + len(extensions))) ip.Encode(&header.IPv6Fields{ - PayloadLength: uint16(payloadLength), + PayloadLength: uint16(len(payload) + len(extensions)), NextHeader: nextHdr, HopLimit: hopLimit, SrcAddr: r.LocalAddress, DstAddr: r.RemoteAddress, }) - ep.HandlePacket(r, stack.PacketBuffer{ - Data: hdr.View().ToVectorisedView(), + if n := copy(ip[header.IPv6MinimumSize:], extensions); n != len(extensions) { + t.Fatalf("expected to write %d bytes of extensions, but wrote %d", len(extensions), n) + } + ep.HandlePacket(r, &stack.PacketBuffer{ + NetworkHeader: buffer.View(ip), + Data: payload.ToVectorisedView(), }) } @@ -676,14 +680,11 @@ func TestNDPValidation(t *testing.T) { invalid := stats.Invalid typStat := typ.statCounter(stats) - extraDataLen := len(typ.extraData) - hdr := buffer.NewPrependable(header.IPv6MinimumSize + typ.size + extraDataLen + header.IPv6FragmentExtHdrLength) - extraData := buffer.View(hdr.Prepend(extraDataLen)) - copy(extraData, typ.extraData) - pkt := header.ICMPv6(hdr.Prepend(typ.size)) - pkt.SetType(typ.typ) - pkt.SetCode(test.code) - pkt.SetChecksum(header.ICMPv6Checksum(pkt, r.LocalAddress, r.RemoteAddress, extraData.ToVectorisedView())) + icmp := header.ICMPv6(buffer.NewView(typ.size + len(typ.extraData))) + copy(icmp[typ.size:], typ.extraData) + icmp.SetType(typ.typ) + icmp.SetCode(test.code) + icmp.SetChecksum(header.ICMPv6Checksum(icmp[:typ.size], r.LocalAddress, r.RemoteAddress, buffer.View(typ.extraData).ToVectorisedView())) // Rx count of the NDP message should initially be 0. if got := typStat.Value(); got != 0 { @@ -699,7 +700,7 @@ func TestNDPValidation(t *testing.T) { t.FailNow() } - handleIPv6Payload(hdr, test.hopLimit, test.atomicFragment, ep, &r) + handleIPv6Payload(buffer.View(icmp), test.hopLimit, test.atomicFragment, ep, &r) // Rx count of the NDP packet should have increased. if got := typStat.Value(); got != 1 { @@ -884,7 +885,7 @@ func TestRouterAdvertValidation(t *testing.T) { t.Fatalf("got rxRA = %d, want = 0", got) } - e.InjectInbound(header.IPv6ProtocolNumber, stack.PacketBuffer{ + e.InjectInbound(header.IPv6ProtocolNumber, &stack.PacketBuffer{ Data: hdr.View().ToVectorisedView(), }) |