Files
goreplay/capture/sock_linux.go
Leonid BugaevandGitHub 58ff8865df Fix windows packet capture (#943)
Issues is that Go built-in net.Interfaces function in newer Windows versions return wrong interface names, which libpcap can't consume.
Now we use pcap.FindDevices instead of net.Interfaces.
See this Article for deep understanding of the issue https://haydz.github.io/2020/07/06/Go-Windows-NIC.html

Additionally, found a bug causing big memory allocations, for large requests, when we perform check if messages finished or not.
Because of this bug chunked body encoding check was not working properly.
Was not caught in tests, because test was working on packet array level, and this issue happens when dealing with TCP message object.

Additionally, added a small fix for windows Makefile task, it now generates proper file name.
2021-06-10 14:46:17 +03:00

277 lines
7.1 KiB
Go

package capture
import (
"fmt"
"net"
"sync"
"time"
"unsafe"
"golang.org/x/sys/unix"
"github.com/google/gopacket"
"github.com/google/gopacket/layers"
"github.com/google/gopacket/pcap"
)
const (
// ETHALL htons(ETH_P_ALL)
ETHALL uint16 = unix.ETH_P_ALL<<8 | unix.ETH_P_ALL>>8
// BLOCKSIZE ring buffer block_size
BLOCKSIZE = 64 << 10
// BLOCKNR ring buffer block_nr
BLOCKNR = (2 << 20) / BLOCKSIZE // 2mb / 64kb
// FRAMESIZE ring buffer frame_size
FRAMESIZE = BLOCKSIZE
// FRAMENR ring buffer frame_nr
FRAMENR = BLOCKNR * BLOCKSIZE / FRAMESIZE
// MAPHUGE2MB 2mb huge map
MAPHUGE2MB = 21 << unix.MAP_HUGE_SHIFT
)
var tpacket2hdrlen = tpAlign(int(unsafe.Sizeof(unix.Tpacket2Hdr{})))
// SockRaw is a linux M'maped af_packet socket
type SockRaw struct {
mu sync.Mutex
fd int
ifindex int
snaplen int
pollTimeout uintptr
frame uint32 // current frame
buf []byte // points to the memory space of the ring buffer shared with the kernel.
loopIndex int32 // this field must filled to avoid reading packet twice on a loopback device
}
// NewSocket returns new M'maped sock_raw on packet version 2.
func NewSocket(pifi pcap.Interface) (*SockRaw, error) {
var ifi net.Interface
infs, _ := net.Interfaces()
found := false
for _, i := range infs {
if i.Name == pifi.Name {
ifi = i
found = true
break
}
}
if !found {
return nil, fmt.Errorf("Can't find matching interface")
}
// sock create
fd, err := unix.Socket(unix.AF_PACKET, unix.SOCK_RAW, int(ETHALL))
if err != nil {
return nil, err
}
sock := &SockRaw{
fd: fd,
ifindex: ifi.Index,
snaplen: FRAMESIZE,
pollTimeout: ^uintptr(0),
}
// set packet version
err = unix.SetsockoptInt(fd, unix.SOL_PACKET, unix.PACKET_VERSION, unix.TPACKET_V2)
if err != nil {
unix.Close(fd)
return nil, fmt.Errorf("setsockopt packet_version: %v", err)
}
// bind to interface
addr := unix.RawSockaddrLinklayer{
Family: unix.AF_PACKET,
Protocol: ETHALL,
Ifindex: int32(ifi.Index),
}
_, _, e := unix.Syscall(
unix.SYS_BIND,
uintptr(fd),
uintptr(unsafe.Pointer(&addr)),
uintptr(unix.SizeofSockaddrLinklayer),
)
if e != 0 {
unix.Close(fd)
return nil, e
}
// create shared-memory ring buffer
tp := &unix.TpacketReq{
Block_size: BLOCKSIZE,
Block_nr: BLOCKNR,
Frame_size: FRAMESIZE,
Frame_nr: FRAMENR,
}
err = unix.SetsockoptTpacketReq(sock.fd, unix.SOL_PACKET, unix.PACKET_RX_RING, tp)
if err != nil {
unix.Close(fd)
return nil, fmt.Errorf("setsockopt packet_rx_ring: %v", err)
}
sock.buf, err = unix.Mmap(
sock.fd,
0,
BLOCKSIZE*BLOCKNR,
unix.PROT_READ|unix.PROT_WRITE,
unix.MAP_SHARED|MAPHUGE2MB,
)
if err != nil {
unix.Close(fd)
return nil, fmt.Errorf("socket mmap error: %v", err)
}
return sock, nil
}
// ReadPacketData implements gopacket.PacketDataSource.
func (sock *SockRaw) ZeroCopyReadPacketData() (buf []byte, ci gopacket.CaptureInfo, err error) {
sock.mu.Lock()
defer sock.mu.Unlock()
var tpHdr *unix.Tpacket2Hdr
poll := &unix.PollFd{
Fd: int32(sock.fd),
Events: unix.POLLIN,
}
var i int
read:
i = int(sock.frame * FRAMESIZE)
tpHdr = (*unix.Tpacket2Hdr)(unsafe.Pointer(&sock.buf[i]))
sock.frame = (sock.frame + 1) % FRAMENR
if tpHdr.Status&unix.TP_STATUS_USER == 0 {
_, _, e := unix.Syscall(unix.SYS_POLL, uintptr(unsafe.Pointer(poll)), 1, sock.pollTimeout)
if e != 0 && e != unix.EINTR {
return buf, ci, e
}
// it might be some other frame with data!
if tpHdr.Status&unix.TP_STATUS_USER == 0 {
goto read
}
}
tpHdr.Status = unix.TP_STATUS_KERNEL
sockAddr := (*unix.RawSockaddrLinklayer)(unsafe.Pointer(&sock.buf[i+tpacket2hdrlen]))
// parse out repeating packets on loopback
if sockAddr.Ifindex == sock.loopIndex && sock.frame%2 != 0 {
goto read
}
ci.Length = int(tpHdr.Len)
ci.Timestamp = time.Unix(int64(tpHdr.Sec), int64(tpHdr.Nsec))
ci.InterfaceIndex = int(sockAddr.Ifindex)
buf = make([]byte, tpHdr.Snaplen)
ci.CaptureLength = copy(buf, sock.buf[i+int(tpHdr.Mac):])
return
}
// Close closes the underlying socket
func (sock *SockRaw) Close() (err error) {
sock.mu.Lock()
defer sock.mu.Unlock()
if sock.fd != -1 {
unix.Munmap(sock.buf)
sock.buf = nil
err = unix.Close(sock.fd)
sock.fd = -1
}
return
}
// SetSnapLen sets the maximum capture length to the given value.
// for this to take effects on the kernel level SetBPFilter should be called too.
func (sock *SockRaw) SetSnapLen(snap int) error {
sock.mu.Lock()
defer sock.mu.Unlock()
if snap < 0 {
return fmt.Errorf("expected %d snap length to be at least 0", snap)
}
if snap > FRAMESIZE {
snap = FRAMESIZE
}
sock.snaplen = snap
return nil
}
// SetTimeout sets poll wait timeout for the socket.
// negative value will block forever
func (sock *SockRaw) SetTimeout(t time.Duration) error {
sock.mu.Lock()
defer sock.mu.Unlock()
sock.pollTimeout = uintptr(t)
return nil
}
// GetSnapLen returns the maximum capture length
func (sock *SockRaw) GetSnapLen() int {
sock.mu.Lock()
defer sock.mu.Unlock()
return sock.snaplen
}
// SetBPFFilter compiles and sets a BPF filter for the socket handle.
func (sock *SockRaw) SetBPFFilter(expr string) error {
sock.mu.Lock()
defer sock.mu.Unlock()
if expr == "" {
return unix.SetsockoptInt(sock.fd, unix.SOL_SOCKET, unix.SO_DETACH_FILTER, 0)
}
filter, err := pcap.CompileBPFFilter(layers.LinkTypeEthernet, sock.snaplen, expr)
if err != nil {
return err
}
if len(filter) > int(^uint16(0)) {
return fmt.Errorf("filters out of range 0-%d", ^uint16(0))
}
if len(filter) == 0 {
return unix.SetsockoptInt(sock.fd, unix.SOL_SOCKET, unix.SO_DETACH_FILTER, 0)
}
fprog := &unix.SockFprog{
Len: uint16(len(filter)),
Filter: &(*(*[]unix.SockFilter)(unsafe.Pointer(&filter)))[0],
}
return unix.SetsockoptSockFprog(sock.fd, unix.SOL_SOCKET, unix.SO_ATTACH_FILTER, fprog)
}
// SetPromiscuous sets promiscuous mode to the required value. for better result capture on all interfaces instead.
// If it is enabled, traffic not destined for the interface will also be captured.
func (sock *SockRaw) SetPromiscuous(b bool) error {
sock.mu.Lock()
defer sock.mu.Unlock()
mreq := unix.PacketMreq{
Ifindex: int32(sock.ifindex),
Type: unix.PACKET_MR_PROMISC,
}
opt := unix.PACKET_ADD_MEMBERSHIP
if !b {
opt = unix.PACKET_DROP_MEMBERSHIP
}
return unix.SetsockoptPacketMreq(sock.fd, unix.SOL_PACKET, opt, &mreq)
}
// Stats returns number of packets and dropped packets. This will be the number of packets/dropped packets since the last call to stats (not the cummulative sum!).
func (sock *SockRaw) Stats() (*unix.TpacketStats, error) {
sock.mu.Lock()
defer sock.mu.Unlock()
return unix.GetsockoptTpacketStats(sock.fd, unix.SOL_PACKET, unix.PACKET_STATISTICS)
}
// SetLoopbackIndex necessary to avoid reading packet twice on a loopback device
func (sock *SockRaw) SetLoopbackIndex(i int32) {
sock.mu.Lock()
defer sock.mu.Unlock()
sock.loopIndex = i
}
// WritePacketData transmits a raw packet.
func (sock *SockRaw) WritePacketData(pkt []byte) error {
_, err := unix.Write(sock.fd, pkt)
return err
}
func tpAlign(x int) int {
return int((uint(x) + unix.TPACKET_ALIGNMENT - 1) &^ (unix.TPACKET_ALIGNMENT - 1))
}