summaryrefslogtreecommitdiffhomepage
path: root/pkg/tcpip
diff options
context:
space:
mode:
Diffstat (limited to 'pkg/tcpip')
-rw-r--r--pkg/tcpip/header/arp.go77
-rw-r--r--pkg/tcpip/link/channel/BUILD1
-rw-r--r--pkg/tcpip/link/channel/channel.go6
-rw-r--r--pkg/tcpip/link/fdbased/BUILD1
-rw-r--r--pkg/tcpip/link/fdbased/endpoint.go77
-rw-r--r--pkg/tcpip/link/loopback/loopback.go5
-rw-r--r--pkg/tcpip/link/muxed/BUILD1
-rw-r--r--pkg/tcpip/link/muxed/injectable.go6
-rw-r--r--pkg/tcpip/link/nested/BUILD1
-rw-r--r--pkg/tcpip/link/nested/nested.go6
-rw-r--r--pkg/tcpip/link/qdisc/fifo/BUILD1
-rw-r--r--pkg/tcpip/link/qdisc/fifo/endpoint.go6
-rw-r--r--pkg/tcpip/link/rawfile/rawfile_unsafe.go32
-rw-r--r--pkg/tcpip/link/sharedmem/sharedmem.go5
-rw-r--r--pkg/tcpip/link/tun/device.go10
-rw-r--r--pkg/tcpip/link/waitable/BUILD2
-rw-r--r--pkg/tcpip/link/waitable/waitable.go6
-rw-r--r--pkg/tcpip/link/waitable/waitable_test.go6
-rw-r--r--pkg/tcpip/network/ip_test.go9
-rw-r--r--pkg/tcpip/stack/forwarder_test.go6
-rw-r--r--pkg/tcpip/stack/iptables.go30
-rw-r--r--pkg/tcpip/stack/nic_test.go5
-rw-r--r--pkg/tcpip/stack/registration.go7
-rw-r--r--pkg/tcpip/stack/stack.go6
-rw-r--r--pkg/tcpip/transport/packet/endpoint.go25
-rw-r--r--pkg/tcpip/transport/packet/endpoint_state.go19
26 files changed, 230 insertions, 126 deletions
diff --git a/pkg/tcpip/header/arp.go b/pkg/tcpip/header/arp.go
index 718a4720a..83189676e 100644
--- a/pkg/tcpip/header/arp.go
+++ b/pkg/tcpip/header/arp.go
@@ -14,14 +14,33 @@
package header
-import "gvisor.dev/gvisor/pkg/tcpip"
+import (
+ "encoding/binary"
+
+ "gvisor.dev/gvisor/pkg/tcpip"
+)
const (
// ARPProtocolNumber is the ARP network protocol number.
ARPProtocolNumber tcpip.NetworkProtocolNumber = 0x0806
// ARPSize is the size of an IPv4-over-Ethernet ARP packet.
- ARPSize = 2 + 2 + 1 + 1 + 2 + 2*6 + 2*4
+ ARPSize = 28
+)
+
+// ARPHardwareType is the hardware type for LinkEndpoint in an ARP header.
+type ARPHardwareType uint16
+
+// Typical ARP HardwareType values. Some of the constants have to be specific
+// values as they are egressed on the wire in the HTYPE field of an ARP header.
+const (
+ ARPHardwareNone ARPHardwareType = 0
+ // ARPHardwareEther specifically is the HTYPE for Ethernet as specified
+ // in the IANA list here:
+ //
+ // https://www.iana.org/assignments/arp-parameters/arp-parameters.xhtml#arp-parameters-2
+ ARPHardwareEther ARPHardwareType = 1
+ ARPHardwareLoopback ARPHardwareType = 2
)
// ARPOp is an ARP opcode.
@@ -36,54 +55,64 @@ const (
// ARP is an ARP packet stored in a byte array as described in RFC 826.
type ARP []byte
-func (a ARP) hardwareAddressSpace() uint16 { return uint16(a[0])<<8 | uint16(a[1]) }
-func (a ARP) protocolAddressSpace() uint16 { return uint16(a[2])<<8 | uint16(a[3]) }
-func (a ARP) hardwareAddressSize() int { return int(a[4]) }
-func (a ARP) protocolAddressSize() int { return int(a[5]) }
+const (
+ hTypeOffset = 0
+ protocolOffset = 2
+ haAddressSizeOffset = 4
+ protoAddressSizeOffset = 5
+ opCodeOffset = 6
+ senderHAAddressOffset = 8
+ senderProtocolAddressOffset = senderHAAddressOffset + EthernetAddressSize
+ targetHAAddressOffset = senderProtocolAddressOffset + IPv4AddressSize
+ targetProtocolAddressOffset = targetHAAddressOffset + EthernetAddressSize
+)
+
+func (a ARP) hardwareAddressType() ARPHardwareType {
+ return ARPHardwareType(binary.BigEndian.Uint16(a[hTypeOffset:]))
+}
+
+func (a ARP) protocolAddressSpace() uint16 { return binary.BigEndian.Uint16(a[protocolOffset:]) }
+func (a ARP) hardwareAddressSize() int { return int(a[haAddressSizeOffset]) }
+func (a ARP) protocolAddressSize() int { return int(a[protoAddressSizeOffset]) }
// Op is the ARP opcode.
-func (a ARP) Op() ARPOp { return ARPOp(a[6])<<8 | ARPOp(a[7]) }
+func (a ARP) Op() ARPOp { return ARPOp(binary.BigEndian.Uint16(a[opCodeOffset:])) }
// SetOp sets the ARP opcode.
func (a ARP) SetOp(op ARPOp) {
- a[6] = uint8(op >> 8)
- a[7] = uint8(op)
+ binary.BigEndian.PutUint16(a[opCodeOffset:], uint16(op))
}
// SetIPv4OverEthernet configures the ARP packet for IPv4-over-Ethernet.
func (a ARP) SetIPv4OverEthernet() {
- a[0], a[1] = 0, 1 // htypeEthernet
- a[2], a[3] = 0x08, 0x00 // IPv4ProtocolNumber
- a[4] = 6 // macSize
- a[5] = uint8(IPv4AddressSize)
+ binary.BigEndian.PutUint16(a[hTypeOffset:], uint16(ARPHardwareEther))
+ binary.BigEndian.PutUint16(a[protocolOffset:], uint16(IPv4ProtocolNumber))
+ a[haAddressSizeOffset] = EthernetAddressSize
+ a[protoAddressSizeOffset] = uint8(IPv4AddressSize)
}
// HardwareAddressSender is the link address of the sender.
// It is a view on to the ARP packet so it can be used to set the value.
func (a ARP) HardwareAddressSender() []byte {
- const s = 8
- return a[s : s+6]
+ return a[senderHAAddressOffset : senderHAAddressOffset+EthernetAddressSize]
}
// ProtocolAddressSender is the protocol address of the sender.
// It is a view on to the ARP packet so it can be used to set the value.
func (a ARP) ProtocolAddressSender() []byte {
- const s = 8 + 6
- return a[s : s+4]
+ return a[senderProtocolAddressOffset : senderProtocolAddressOffset+IPv4AddressSize]
}
// HardwareAddressTarget is the link address of the target.
// It is a view on to the ARP packet so it can be used to set the value.
func (a ARP) HardwareAddressTarget() []byte {
- const s = 8 + 6 + 4
- return a[s : s+6]
+ return a[targetHAAddressOffset : targetHAAddressOffset+EthernetAddressSize]
}
// ProtocolAddressTarget is the protocol address of the target.
// It is a view on to the ARP packet so it can be used to set the value.
func (a ARP) ProtocolAddressTarget() []byte {
- const s = 8 + 6 + 4 + 6
- return a[s : s+4]
+ return a[targetProtocolAddressOffset : targetProtocolAddressOffset+IPv4AddressSize]
}
// IsValid reports whether this is an ARP packet for IPv4 over Ethernet.
@@ -91,10 +120,8 @@ func (a ARP) IsValid() bool {
if len(a) < ARPSize {
return false
}
- const htypeEthernet = 1
- const macSize = 6
- return a.hardwareAddressSpace() == htypeEthernet &&
+ return a.hardwareAddressType() == ARPHardwareEther &&
a.protocolAddressSpace() == uint16(IPv4ProtocolNumber) &&
- a.hardwareAddressSize() == macSize &&
+ a.hardwareAddressSize() == EthernetAddressSize &&
a.protocolAddressSize() == IPv4AddressSize
}
diff --git a/pkg/tcpip/link/channel/BUILD b/pkg/tcpip/link/channel/BUILD
index b8b93e78e..39ca774ef 100644
--- a/pkg/tcpip/link/channel/BUILD
+++ b/pkg/tcpip/link/channel/BUILD
@@ -10,6 +10,7 @@ go_library(
"//pkg/sync",
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/channel/channel.go b/pkg/tcpip/link/channel/channel.go
index 20b183da0..a2bb773d4 100644
--- a/pkg/tcpip/link/channel/channel.go
+++ b/pkg/tcpip/link/channel/channel.go
@@ -23,6 +23,7 @@ import (
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -296,3 +297,8 @@ func (e *Endpoint) AddNotify(notify Notification) *NotificationHandle {
func (e *Endpoint) RemoveNotify(handle *NotificationHandle) {
e.q.RemoveNotify(handle)
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*Endpoint) ARPHardwareType() header.ARPHardwareType {
+ return header.ARPHardwareNone
+}
diff --git a/pkg/tcpip/link/fdbased/BUILD b/pkg/tcpip/link/fdbased/BUILD
index aa6db9aea..507b44abc 100644
--- a/pkg/tcpip/link/fdbased/BUILD
+++ b/pkg/tcpip/link/fdbased/BUILD
@@ -15,6 +15,7 @@ go_library(
visibility = ["//visibility:public"],
deps = [
"//pkg/binary",
+ "//pkg/iovec",
"//pkg/sync",
"//pkg/tcpip",
"//pkg/tcpip/buffer",
diff --git a/pkg/tcpip/link/fdbased/endpoint.go b/pkg/tcpip/link/fdbased/endpoint.go
index f34082e1a..6aa1badc7 100644
--- a/pkg/tcpip/link/fdbased/endpoint.go
+++ b/pkg/tcpip/link/fdbased/endpoint.go
@@ -45,6 +45,7 @@ import (
"golang.org/x/sys/unix"
"gvisor.dev/gvisor/pkg/binary"
+ "gvisor.dev/gvisor/pkg/iovec"
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
@@ -406,6 +407,8 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.Ne
eth.Encode(ethHdr)
}
+ var builder iovec.Builder
+
fd := e.fds[pkt.Hash%uint32(len(e.fds))]
if e.Capabilities()&stack.CapabilityHardwareGSO != 0 {
vnetHdr := virtioNetHdr{}
@@ -430,29 +433,25 @@ func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, protocol tcpip.Ne
}
vnetHdrBuf := binary.Marshal(make([]byte, 0, virtioNetHdrSize), binary.LittleEndian, vnetHdr)
- return rawfile.NonBlockingWrite3(fd, vnetHdrBuf, pkt.Header.View(), pkt.Data.ToView())
+ builder.Add(vnetHdrBuf)
}
- if pkt.Data.Size() == 0 {
- return rawfile.NonBlockingWrite(fd, pkt.Header.View())
- }
- if pkt.Header.UsedLength() == 0 {
- return rawfile.NonBlockingWrite(fd, pkt.Data.ToView())
+ builder.Add(pkt.Header.View())
+ for _, v := range pkt.Data.Views() {
+ builder.Add(v)
}
- return rawfile.NonBlockingWrite3(fd, pkt.Header.View(), pkt.Data.ToView(), nil)
+ return rawfile.NonBlockingWriteIovec(fd, builder.Build())
}
func (e *endpoint) sendBatch(batchFD int, batch []*stack.PacketBuffer) (int, *tcpip.Error) {
// Send a batch of packets through batchFD.
mmsgHdrs := make([]rawfile.MMsgHdr, 0, len(batch))
for _, pkt := range batch {
- var ethHdrBuf []byte
- iovLen := 0
+ var eth header.Ethernet
if e.hdrSize > 0 {
// Add ethernet header if needed.
- ethHdrBuf = make([]byte, header.EthernetMinimumSize)
- eth := header.Ethernet(ethHdrBuf)
+ eth = make(header.Ethernet, header.EthernetMinimumSize)
ethHdr := &header.EthernetFields{
DstAddr: pkt.EgressRoute.RemoteLinkAddress,
Type: pkt.NetworkProtocolNumber,
@@ -465,12 +464,11 @@ func (e *endpoint) sendBatch(batchFD int, batch []*stack.PacketBuffer) (int, *tc
ethHdr.SrcAddr = e.addr
}
eth.Encode(ethHdr)
- iovLen++
}
- vnetHdr := virtioNetHdr{}
var vnetHdrBuf []byte
if e.Capabilities()&stack.CapabilityHardwareGSO != 0 {
+ vnetHdr := virtioNetHdr{}
if pkt.GSOOptions != nil {
vnetHdr.hdrLen = uint16(pkt.Header.UsedLength())
if pkt.GSOOptions.NeedsCsum {
@@ -491,45 +489,20 @@ func (e *endpoint) sendBatch(batchFD int, batch []*stack.PacketBuffer) (int, *tc
}
}
vnetHdrBuf = binary.Marshal(make([]byte, 0, virtioNetHdrSize), binary.LittleEndian, vnetHdr)
- iovLen++
}
- iovecs := make([]syscall.Iovec, iovLen+1+len(pkt.Data.Views()))
+ var builder iovec.Builder
+ builder.Add(vnetHdrBuf)
+ builder.Add(eth)
+ builder.Add(pkt.Header.View())
+ for _, v := range pkt.Data.Views() {
+ builder.Add(v)
+ }
+ iovecs := builder.Build()
+
var mmsgHdr rawfile.MMsgHdr
mmsgHdr.Msg.Iov = &iovecs[0]
- iovecIdx := 0
- if vnetHdrBuf != nil {
- v := &iovecs[iovecIdx]
- v.Base = &vnetHdrBuf[0]
- v.Len = uint64(len(vnetHdrBuf))
- iovecIdx++
- }
- if ethHdrBuf != nil {
- v := &iovecs[iovecIdx]
- v.Base = &ethHdrBuf[0]
- v.Len = uint64(len(ethHdrBuf))
- iovecIdx++
- }
- pktSize := uint64(0)
- // Encode L3 Header
- v := &iovecs[iovecIdx]
- hdr := &pkt.Header
- hdrView := hdr.View()
- v.Base = &hdrView[0]
- v.Len = uint64(len(hdrView))
- pktSize += v.Len
- iovecIdx++
-
- // Now encode the Transport Payload.
- pktViews := pkt.Data.Views()
- for i := range pktViews {
- vec := &iovecs[iovecIdx]
- iovecIdx++
- vec.Base = &pktViews[i][0]
- vec.Len = uint64(len(pktViews[i]))
- pktSize += vec.Len
- }
- mmsgHdr.Msg.Iovlen = uint64(iovecIdx)
+ mmsgHdr.Msg.Iovlen = uint64(len(iovecs))
mmsgHdrs = append(mmsgHdrs, mmsgHdr)
}
@@ -626,6 +599,14 @@ func (e *endpoint) GSOMaxSize() uint32 {
return e.gsoMaxSize
}
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (e *endpoint) ARPHardwareType() header.ARPHardwareType {
+ if e.hdrSize > 0 {
+ return header.ARPHardwareEther
+ }
+ return header.ARPHardwareNone
+}
+
// InjectableEndpoint is an injectable fd-based endpoint. The endpoint writes
// to the FD, but does not read from it. All reads come from injected packets.
type InjectableEndpoint struct {
diff --git a/pkg/tcpip/link/loopback/loopback.go b/pkg/tcpip/link/loopback/loopback.go
index 568c6874f..3b17d8c28 100644
--- a/pkg/tcpip/link/loopback/loopback.go
+++ b/pkg/tcpip/link/loopback/loopback.go
@@ -113,3 +113,8 @@ func (e *endpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error {
return nil
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*endpoint) ARPHardwareType() header.ARPHardwareType {
+ return header.ARPHardwareLoopback
+}
diff --git a/pkg/tcpip/link/muxed/BUILD b/pkg/tcpip/link/muxed/BUILD
index 82b441b79..e7493e5c5 100644
--- a/pkg/tcpip/link/muxed/BUILD
+++ b/pkg/tcpip/link/muxed/BUILD
@@ -9,6 +9,7 @@ go_library(
deps = [
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/muxed/injectable.go b/pkg/tcpip/link/muxed/injectable.go
index c69d6b7e9..c305d9e86 100644
--- a/pkg/tcpip/link/muxed/injectable.go
+++ b/pkg/tcpip/link/muxed/injectable.go
@@ -18,6 +18,7 @@ package muxed
import (
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -129,6 +130,11 @@ func (m *InjectableEndpoint) Wait() {
}
}
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*InjectableEndpoint) ARPHardwareType() header.ARPHardwareType {
+ panic("unsupported operation")
+}
+
// NewInjectableEndpoint creates a new multi-endpoint injectable endpoint.
func NewInjectableEndpoint(routes map[tcpip.Address]stack.InjectableLinkEndpoint) *InjectableEndpoint {
return &InjectableEndpoint{
diff --git a/pkg/tcpip/link/nested/BUILD b/pkg/tcpip/link/nested/BUILD
index bdd5276ad..2cdb23475 100644
--- a/pkg/tcpip/link/nested/BUILD
+++ b/pkg/tcpip/link/nested/BUILD
@@ -12,6 +12,7 @@ go_library(
"//pkg/sync",
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/nested/nested.go b/pkg/tcpip/link/nested/nested.go
index 2998f9c4f..328bd048e 100644
--- a/pkg/tcpip/link/nested/nested.go
+++ b/pkg/tcpip/link/nested/nested.go
@@ -20,6 +20,7 @@ import (
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -129,3 +130,8 @@ func (e *Endpoint) GSOMaxSize() uint32 {
}
return 0
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType
+func (e *Endpoint) ARPHardwareType() header.ARPHardwareType {
+ return e.child.ARPHardwareType()
+}
diff --git a/pkg/tcpip/link/qdisc/fifo/BUILD b/pkg/tcpip/link/qdisc/fifo/BUILD
index 054c213bc..1d0079bd6 100644
--- a/pkg/tcpip/link/qdisc/fifo/BUILD
+++ b/pkg/tcpip/link/qdisc/fifo/BUILD
@@ -14,6 +14,7 @@ go_library(
"//pkg/sync",
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/qdisc/fifo/endpoint.go b/pkg/tcpip/link/qdisc/fifo/endpoint.go
index b5dfb7850..c84fe1bb9 100644
--- a/pkg/tcpip/link/qdisc/fifo/endpoint.go
+++ b/pkg/tcpip/link/qdisc/fifo/endpoint.go
@@ -22,6 +22,7 @@ import (
"gvisor.dev/gvisor/pkg/sync"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -207,3 +208,8 @@ func (e *endpoint) Wait() {
e.wg.Wait()
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType
+func (e *endpoint) ARPHardwareType() header.ARPHardwareType {
+ return e.lower.ARPHardwareType()
+}
diff --git a/pkg/tcpip/link/rawfile/rawfile_unsafe.go b/pkg/tcpip/link/rawfile/rawfile_unsafe.go
index 69de6eb3e..f4c32c2da 100644
--- a/pkg/tcpip/link/rawfile/rawfile_unsafe.go
+++ b/pkg/tcpip/link/rawfile/rawfile_unsafe.go
@@ -66,38 +66,14 @@ func NonBlockingWrite(fd int, buf []byte) *tcpip.Error {
return nil
}
-// NonBlockingWrite3 writes up to three byte slices to a file descriptor in a
-// single syscall. It fails if partial data is written.
-func NonBlockingWrite3(fd int, b1, b2, b3 []byte) *tcpip.Error {
- // If there is no second and third buffer, issue a regular write.
- if len(b2) == 0 && len(b3) == 0 {
- return NonBlockingWrite(fd, b1)
- }
-
- // Build the iovec that represents them and issue a writev syscall.
- iovec := [3]syscall.Iovec{
- {
- Base: &b1[0],
- Len: uint64(len(b1)),
- },
- {
- Base: &b2[0],
- Len: uint64(len(b2)),
- },
- }
- iovecLen := uintptr(2)
-
- if len(b3) > 0 {
- iovecLen++
- iovec[2].Base = &b3[0]
- iovec[2].Len = uint64(len(b3))
- }
-
+// NonBlockingWriteIovec writes iovec to a file descriptor in a single syscall.
+// It fails if partial data is written.
+func NonBlockingWriteIovec(fd int, iovec []syscall.Iovec) *tcpip.Error {
+ iovecLen := uintptr(len(iovec))
_, _, e := syscall.RawSyscall(syscall.SYS_WRITEV, uintptr(fd), uintptr(unsafe.Pointer(&iovec[0])), iovecLen)
if e != 0 {
return TranslateErrno(e)
}
-
return nil
}
diff --git a/pkg/tcpip/link/sharedmem/sharedmem.go b/pkg/tcpip/link/sharedmem/sharedmem.go
index 0374a2441..a36862c67 100644
--- a/pkg/tcpip/link/sharedmem/sharedmem.go
+++ b/pkg/tcpip/link/sharedmem/sharedmem.go
@@ -287,3 +287,8 @@ func (e *endpoint) dispatchLoop(d stack.NetworkDispatcher) {
e.completed.Done()
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType
+func (*endpoint) ARPHardwareType() header.ARPHardwareType {
+ return header.ARPHardwareEther
+}
diff --git a/pkg/tcpip/link/tun/device.go b/pkg/tcpip/link/tun/device.go
index 6bc9033d0..47446efec 100644
--- a/pkg/tcpip/link/tun/device.go
+++ b/pkg/tcpip/link/tun/device.go
@@ -139,6 +139,7 @@ func attachOrCreateNIC(s *stack.Stack, name, prefix string, linkCaps stack.LinkE
stack: s,
nicID: id,
name: name,
+ isTap: prefix == "tap",
}
endpoint.Endpoint.LinkEPCapabilities = linkCaps
if endpoint.name == "" {
@@ -348,6 +349,7 @@ type tunEndpoint struct {
stack *stack.Stack
nicID tcpip.NICID
name string
+ isTap bool
}
// DecRef decrements refcount of e, removes NIC if refcount goes to 0.
@@ -356,3 +358,11 @@ func (e *tunEndpoint) DecRef() {
e.stack.RemoveNIC(e.nicID)
})
}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (e *tunEndpoint) ARPHardwareType() header.ARPHardwareType {
+ if e.isTap {
+ return header.ARPHardwareEther
+ }
+ return header.ARPHardwareNone
+}
diff --git a/pkg/tcpip/link/waitable/BUILD b/pkg/tcpip/link/waitable/BUILD
index 0956d2c65..ee84c3d96 100644
--- a/pkg/tcpip/link/waitable/BUILD
+++ b/pkg/tcpip/link/waitable/BUILD
@@ -12,6 +12,7 @@ go_library(
"//pkg/gate",
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
@@ -25,6 +26,7 @@ go_test(
deps = [
"//pkg/tcpip",
"//pkg/tcpip/buffer",
+ "//pkg/tcpip/header",
"//pkg/tcpip/stack",
],
)
diff --git a/pkg/tcpip/link/waitable/waitable.go b/pkg/tcpip/link/waitable/waitable.go
index 949b3f2b2..24a8dc2eb 100644
--- a/pkg/tcpip/link/waitable/waitable.go
+++ b/pkg/tcpip/link/waitable/waitable.go
@@ -25,6 +25,7 @@ import (
"gvisor.dev/gvisor/pkg/gate"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -147,3 +148,8 @@ func (e *Endpoint) WaitDispatch() {
// Wait implements stack.LinkEndpoint.Wait.
func (e *Endpoint) Wait() {}
+
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (e *Endpoint) ARPHardwareType() header.ARPHardwareType {
+ return e.lower.ARPHardwareType()
+}
diff --git a/pkg/tcpip/link/waitable/waitable_test.go b/pkg/tcpip/link/waitable/waitable_test.go
index 63bf40562..ffb2354be 100644
--- a/pkg/tcpip/link/waitable/waitable_test.go
+++ b/pkg/tcpip/link/waitable/waitable_test.go
@@ -19,6 +19,7 @@ import (
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -81,6 +82,11 @@ func (e *countedEndpoint) WriteRawPacket(buffer.VectorisedView) *tcpip.Error {
return nil
}
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*countedEndpoint) ARPHardwareType() header.ARPHardwareType {
+ panic("unimplemented")
+}
+
// Wait implements stack.LinkEndpoint.Wait.
func (*countedEndpoint) Wait() {}
diff --git a/pkg/tcpip/network/ip_test.go b/pkg/tcpip/network/ip_test.go
index 7c8fb3e0a..a5b780ca2 100644
--- a/pkg/tcpip/network/ip_test.go
+++ b/pkg/tcpip/network/ip_test.go
@@ -172,14 +172,19 @@ 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")
+}
+
func buildIPv4Route(local, remote tcpip.Address) (stack.Route, *tcpip.Error) {
s := stack.New(stack.Options{
NetworkProtocols: []stack.NetworkProtocol{ipv4.NewProtocol()},
diff --git a/pkg/tcpip/stack/forwarder_test.go b/pkg/tcpip/stack/forwarder_test.go
index a6546cef0..eefb4b07f 100644
--- a/pkg/tcpip/stack/forwarder_test.go
+++ b/pkg/tcpip/stack/forwarder_test.go
@@ -22,6 +22,7 @@ import (
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
)
const (
@@ -301,6 +302,11 @@ func (e *fwdTestLinkEndpoint) WriteRawPacket(vv buffer.VectorisedView) *tcpip.Er
// Wait implements stack.LinkEndpoint.Wait.
func (*fwdTestLinkEndpoint) Wait() {}
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*fwdTestLinkEndpoint) ARPHardwareType() header.ARPHardwareType {
+ panic("not implemented")
+}
+
func fwdTestNetFactory(t *testing.T, proto *fwdTestNetworkProtocol) (ep1, ep2 *fwdTestLinkEndpoint) {
// Create a stack with the network protocol and two NICs.
s := New(Options{
diff --git a/pkg/tcpip/stack/iptables.go b/pkg/tcpip/stack/iptables.go
index f846ea2e5..bbf3b60e8 100644
--- a/pkg/tcpip/stack/iptables.go
+++ b/pkg/tcpip/stack/iptables.go
@@ -292,10 +292,9 @@ func (it *IPTables) startReaper(interval time.Duration) {
// CheckPackets runs pkts through the rules for hook and returns a map of packets that
// should not go forward.
//
-// Precondition: pkt is a IPv4 packet of at least length header.IPv4MinimumSize.
-//
-// TODO(gvisor.dev/issue/170): pk.NetworkHeader will always be set as a
-// precondition.
+// Preconditions:
+// - pkt is a IPv4 packet of at least length header.IPv4MinimumSize.
+// - pkt.NetworkHeader is not nil.
//
// NOTE: unlike the Check API the returned map contains packets that should be
// dropped.
@@ -319,9 +318,9 @@ func (it *IPTables) CheckPackets(hook Hook, pkts PacketBufferList, gso *GSO, r *
return drop, natPkts
}
-// Precondition: pkt is a IPv4 packet of at least length header.IPv4MinimumSize.
-// TODO(gvisor.dev/issue/170): pkt.NetworkHeader will always be set as a
-// precondition.
+// Preconditions:
+// - pkt is a IPv4 packet of at least length header.IPv4MinimumSize.
+// - pkt.NetworkHeader is not nil.
func (it *IPTables) checkChain(hook Hook, pkt *PacketBuffer, table Table, ruleIdx int, gso *GSO, r *Route, address tcpip.Address, nicName string) chainVerdict {
// Start from ruleIdx and walk the list of rules until a rule gives us
// a verdict.
@@ -366,23 +365,12 @@ func (it *IPTables) checkChain(hook Hook, pkt *PacketBuffer, table Table, ruleId
return chainDrop
}
-// Precondition: pkt is a IPv4 packet of at least length header.IPv4MinimumSize.
-// TODO(gvisor.dev/issue/170): pkt.NetworkHeader will always be set as a
-// precondition.
+// Preconditions:
+// - pkt is a IPv4 packet of at least length header.IPv4MinimumSize.
+// - pkt.NetworkHeader is not nil.
func (it *IPTables) checkRule(hook Hook, pkt *PacketBuffer, table Table, ruleIdx int, gso *GSO, r *Route, address tcpip.Address, nicName string) (RuleVerdict, int) {
rule := table.Rules[ruleIdx]
- // If pkt.NetworkHeader hasn't been set yet, it will be contained in
- // pkt.Data.
- if pkt.NetworkHeader == nil {
- var ok bool
- pkt.NetworkHeader, ok = pkt.Data.PullUp(header.IPv4MinimumSize)
- if !ok {
- // Precondition has been violated.
- panic(fmt.Sprintf("iptables checks require IPv4 headers of at least %d bytes", header.IPv4MinimumSize))
- }
- }
-
// Check whether the packet matches the IP header filter.
if !rule.Filter.match(header.IPv4(pkt.NetworkHeader), hook, nicName) {
// Continue on to the next rule.
diff --git a/pkg/tcpip/stack/nic_test.go b/pkg/tcpip/stack/nic_test.go
index 31f865260..3bc9fd831 100644
--- a/pkg/tcpip/stack/nic_test.go
+++ b/pkg/tcpip/stack/nic_test.go
@@ -84,6 +84,11 @@ func (e *testLinkEndpoint) WriteRawPacket(buffer.VectorisedView) *tcpip.Error {
return tcpip.ErrNotSupported
}
+// ARPHardwareType implements stack.LinkEndpoint.ARPHardwareType.
+func (*testLinkEndpoint) ARPHardwareType() header.ARPHardwareType {
+ panic("not implemented")
+}
+
var _ NetworkEndpoint = (*testIPv6Endpoint)(nil)
// An IPv6 NetworkEndpoint that throws away outgoing packets.
diff --git a/pkg/tcpip/stack/registration.go b/pkg/tcpip/stack/registration.go
index 5cbc946b6..f260eeb7f 100644
--- a/pkg/tcpip/stack/registration.go
+++ b/pkg/tcpip/stack/registration.go
@@ -18,6 +18,7 @@ import (
"gvisor.dev/gvisor/pkg/sleep"
"gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
+ "gvisor.dev/gvisor/pkg/tcpip/header"
"gvisor.dev/gvisor/pkg/waiter"
)
@@ -436,6 +437,12 @@ type LinkEndpoint interface {
// Wait will not block if the endpoint hasn't started any goroutines
// yet, even if it might later.
Wait()
+
+ // ARPHardwareType returns the ARPHRD_TYPE of the link endpoint.
+ //
+ // See:
+ // https://github.com/torvalds/linux/blob/aa0c9086b40c17a7ad94425b3b70dd1fdd7497bf/include/uapi/linux/if_arp.h#L30
+ ARPHardwareType() header.ARPHardwareType
}
// InjectableLinkEndpoint is a LinkEndpoint where inbound packets are
diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go
index 0aa815447..2b7ece851 100644
--- a/pkg/tcpip/stack/stack.go
+++ b/pkg/tcpip/stack/stack.go
@@ -1095,6 +1095,11 @@ type NICInfo struct {
// Context is user-supplied data optionally supplied in CreateNICWithOptions.
// See type NICOptions for more details.
Context NICContext
+
+ // ARPHardwareType holds the ARP Hardware type of the NIC. This is the
+ // value sent in haType field of an ARP Request sent by this NIC and the
+ // value expected in the haType field of an ARP response.
+ ARPHardwareType header.ARPHardwareType
}
// HasNIC returns true if the NICID is defined in the stack.
@@ -1126,6 +1131,7 @@ func (s *Stack) NICInfo() map[tcpip.NICID]NICInfo {
MTU: nic.linkEP.MTU(),
Stats: nic.stats,
Context: nic.context,
+ ARPHardwareType: nic.linkEP.ARPHardwareType(),
}
}
return nics
diff --git a/pkg/tcpip/transport/packet/endpoint.go b/pkg/tcpip/transport/packet/endpoint.go
index 57b7f5c19..92b487381 100644
--- a/pkg/tcpip/transport/packet/endpoint.go
+++ b/pkg/tcpip/transport/packet/endpoint.go
@@ -79,6 +79,11 @@ type endpoint struct {
closed bool
stats tcpip.TransportEndpointStats `state:"nosave"`
bound bool
+ boundNIC tcpip.NICID
+
+ // lastErrorMu protects lastError.
+ lastErrorMu sync.Mutex `state:"nosave"`
+ lastError *tcpip.Error `state:".(string)"`
}
// NewEndpoint returns a new packet endpoint.
@@ -229,12 +234,14 @@ func (ep *endpoint) Bind(addr tcpip.FullAddress) *tcpip.Error {
ep.mu.Lock()
defer ep.mu.Unlock()
- if ep.bound {
- return tcpip.ErrAlreadyBound
+ if ep.bound && ep.boundNIC == addr.NIC {
+ // If the NIC being bound is the same then just return success.
+ return nil
}
// Unregister endpoint with all the nics.
ep.stack.UnregisterPacketEndpoint(0, ep.netProto, ep)
+ ep.bound = false
// Bind endpoint to receive packets from specific interface.
if err := ep.stack.RegisterPacketEndpoint(addr.NIC, ep.netProto, ep); err != nil {
@@ -242,6 +249,7 @@ func (ep *endpoint) Bind(addr tcpip.FullAddress) *tcpip.Error {
}
ep.bound = true
+ ep.boundNIC = addr.NIC
return nil
}
@@ -336,8 +344,21 @@ func (ep *endpoint) SetSockOptInt(opt tcpip.SockOptInt, v int) *tcpip.Error {
}
}
+func (ep *endpoint) takeLastError() *tcpip.Error {
+ ep.lastErrorMu.Lock()
+ defer ep.lastErrorMu.Unlock()
+
+ err := ep.lastError
+ ep.lastError = nil
+ return err
+}
+
// GetSockOpt implements tcpip.Endpoint.GetSockOpt.
func (ep *endpoint) GetSockOpt(opt interface{}) *tcpip.Error {
+ switch opt.(type) {
+ case tcpip.ErrorOption:
+ return ep.takeLastError()
+ }
return tcpip.ErrNotSupported
}
diff --git a/pkg/tcpip/transport/packet/endpoint_state.go b/pkg/tcpip/transport/packet/endpoint_state.go
index 9b88f17e4..e2fa96d17 100644
--- a/pkg/tcpip/transport/packet/endpoint_state.go
+++ b/pkg/tcpip/transport/packet/endpoint_state.go
@@ -15,6 +15,7 @@
package packet
import (
+ "gvisor.dev/gvisor/pkg/tcpip"
"gvisor.dev/gvisor/pkg/tcpip/buffer"
"gvisor.dev/gvisor/pkg/tcpip/stack"
)
@@ -70,3 +71,21 @@ func (ep *endpoint) afterLoad() {
panic(*err)
}
}
+
+// saveLastError is invoked by stateify.
+func (ep *endpoint) saveLastError() string {
+ if ep.lastError == nil {
+ return ""
+ }
+
+ return ep.lastError.String()
+}
+
+// loadLastError is invoked by stateify.
+func (ep *endpoint) loadLastError(s string) {
+ if s == "" {
+ return
+ }
+
+ ep.lastError = tcpip.StringToError(s)
+}
generated by cgit v1.2.3 (git 2.39.1) at 2025-01-12 07:09:13 +0000