/*
* Reset algorithm is not stated directly in RFC 793,
* but we can conclude it according to all reset generation.
* NOTE: maybe @tsk is NULL
*/
void tcp_send_reset(struct tcp_sock *tsk, struct tcp_segment *seg)
{
struct tcp *otcp, *tcphdr = seg->tcphdr;
struct pkbuf *opkb;
if (tcphdr->rst)
return;
opkb = alloc_pkb(ETH_HRD_SZ + IP_HRD_SZ + TCP_HRD_SZ);
/* fill tcp head */
otcp = (struct tcp *)pkb2ip(opkb)->ip_data;
otcp->src = tcphdr->dst;
otcp->dst = tcphdr->src;
if (tcphdr->ack) {
/*
* Should we set ack?
* -Yes for xinu, it always set ack to seq+len
* +No for Linux
* +No for tapip
*/
otcp->seq = tcphdr->ackn;
} else {
otcp->ackn = _htonl(seg->seq + seg->len);
otcp->ack = 1;
}
otcp->doff = TCP_HRD_DOFF;
otcp->rst = 1;
tcpdbg("send RESET from "IPFMT":%d to "IPFMT":%d",
ipfmt(seg->iphdr->ip_dst), _ntohs(otcp->src),
ipfmt(seg->iphdr->ip_src), _ntohs(otcp->dst));
tcp_send_out(NULL, opkb, seg);
}
static int udp_init_pkb(struct sock *sk, struct pkbuf *pkb,
void *buf, int size, struct sock_addr *skaddr)
{
struct ip *iphdr = pkb2ip(pkb);
struct udp *udphdr = (struct udp *)iphdr->ip_data;
/* fill ip head */
iphdr->ip_hlen = IP_HRD_SZ >> 2;
iphdr->ip_ver = IP_VERSION_4;
iphdr->ip_tos = 0;
iphdr->ip_len = _htons(pkb->pk_len - ETH_HRD_SZ);
iphdr->ip_id = _htons(udp_id);
iphdr->ip_fragoff = 0;
iphdr->ip_ttl = UDP_DEFAULT_TTL;
iphdr->ip_pro = sk->protocol; /* IP_P_UDP */
iphdr->ip_dst = skaddr->dst_addr;
/* FIXME:use the sk->rt_dst */
if (rt_output(pkb) < 0) /* fill ip src */
return -1;
/* fill udp */
udphdr->src = sk->sk_sport; /* bound local address */
udphdr->dst = skaddr->dst_port;
udphdr->length = _htons(size + UDP_HRD_SZ);
memcpy(udphdr->data, buf, size);
udpdbg(IPFMT":%d" "->" IPFMT":%d(proto %d)",
ipfmt(iphdr->ip_src), _ntohs(udphdr->src),
ipfmt(iphdr->ip_dst), _ntohs(udphdr->dst),
iphdr->ip_pro);
udp_set_checksum(iphdr, udphdr);
return 0;
}
static void icmp_redirect(struct icmp_desc *icmp_desc, struct pkbuf *pkb)
{
struct ip *iphdr = pkb2ip(pkb);
struct icmp *icmphdr = ip2icmp(iphdr);
if (icmphdr->icmp_code > 4)
icmpdbg("Redirect code %d is error", icmphdr->icmp_code);
else
icmpdbg("from " IPFMT " %s(new nexthop "IPFMT")",
ipfmt(iphdr->ip_src),
redirectstr[icmphdr->icmp_code],
ipfmt(icmphdr->icmp_gw));
free_pkb(pkb);
}
void tcp_send_synack(struct tcp_sock *tsk, struct tcp_segment *seg)
{
/*
* LISTEN :
* SYN-SENT:
* SEG: SYN, no ACK, no RST
* <SEQ=ISS><ACK=RCV.NXT><CTL=SYN,ACK>
* (ISS == SND.NXT)
*/
struct tcp *otcp, *tcphdr = seg->tcphdr;
struct pkbuf *opkb;
if (tcphdr->rst)
return;
opkb = alloc_pkb(ETH_HRD_SZ + IP_HRD_SZ + TCP_HRD_SZ);
/* fill tcp head */
otcp = (struct tcp *)pkb2ip(opkb)->ip_data;
otcp->src = tcphdr->dst;
otcp->dst = tcphdr->src;
otcp->doff = TCP_HRD_DOFF;
otcp->seq = _htonl(tsk->iss);
otcp->ackn = _htonl(tsk->rcv_nxt);
otcp->syn = 1;
otcp->ack = 1;
otcp->window = _htons(tsk->rcv_wnd);
tcpdbg("send SYN(%u)/ACK(%u) [WIN %d] to "IPFMT":%d",
_ntohl(otcp->seq), _ntohs(otcp->window),
_ntohl(otcp->ackn), ipfmt(seg->iphdr->ip_dst),
_ntohs(otcp->dst));
tcp_send_out(tsk, opkb, seg);
}
void tcp_send_fin(struct tcp_sock *tsk)
{
struct tcp *otcp;
struct pkbuf *opkb;
opkb = alloc_pkb(ETH_HRD_SZ + IP_HRD_SZ + TCP_HRD_SZ);
/* fill tcp head */
otcp = (struct tcp *)pkb2ip(opkb)->ip_data;
otcp->src = tsk->sk.sk_sport;
otcp->dst = tsk->sk.sk_dport;
otcp->doff = TCP_HRD_DOFF;
otcp->seq = _htonl(tsk->snd_nxt);
otcp->window = _htons(tsk->rcv_wnd);
otcp->fin = 1;
/*
* Should we send an ACK?
* Yes, tcp stack will drop packet if it has no ACK bit
* according to RFC 793 #SEGMENT RECEIVE
*/
otcp->ackn = _htonl(tsk->rcv_nxt);
otcp->ack = 1;
tcpdbg("send FIN(%u)/ACK(%u) [WIN %d] to "IPFMT":%d",
_ntohl(otcp->seq), _ntohl(otcp->ackn),
_ntohs(otcp->window), ipfmt(tsk->sk.sk_daddr),
_ntohs(otcp->dst));
tcp_send_out(tsk, opkb, NULL);
}
static void send_packet(void)
{
if (!buf)
buf = xmalloc(size + ICMP_HRD_SZ);
struct icmp *icmphdr = (struct icmp *)buf;
static int first = 1;
if (first) {
printf("PING "IPFMT" %d(%d) bytes of data\n",
ipfmt(ipaddr),
size,
size + ICMP_HRD_SZ + IP_HRD_SZ);
first = 0;
}
/* fill icmp data */
memset(icmphdr->icmp_data, 'x', size);
icmphdr->icmp_type = ICMP_T_ECHOREQ;
icmphdr->icmp_code = 0;
icmphdr->icmp_id = _htons(id);
icmphdr->icmp_seq = _htons(seq);
icmphdr->icmp_cksum = 0;
icmphdr->icmp_cksum = icmp_chksum((unsigned short *)icmphdr,
ICMP_HRD_SZ + size);
seq++;
/* socket apis */
_send(sock, buf, ICMP_HRD_SZ + size, &skaddr);
psend++;
}
static void ping_stat(void)
{
printf(
"\n"
"--- " IPFMT " ping statistics ---\n"
"%d packets transmitted, %d received, %d%% packet loss\n",
ipfmt(ipaddr), psend, precv, (psend - precv) * 100 / psend);
}
static void icmp_echo_reply(struct icmp_desc *icmp_desc, struct pkbuf *pkb)
{
struct ip *iphdr = pkb2ip(pkb);
struct icmp *icmphdr = ip2icmp(iphdr);
icmpdbg("from "IPFMT" id %d seq %d ttl %d",
ipfmt(iphdr->ip_src),
_ntohs(icmphdr->icmp_id),
_ntohs(icmphdr->icmp_seq),
iphdr->ip_ttl);
free_pkb(pkb);
}
void getipaddr_tap(unsigned char *name, unsigned int *ipaddr)
{
struct ifreq ifr = {};
struct sockaddr_in *saddr;
strcpy(ifr.ifr_name, (char *)name);
if (ioctl(skfd, SIOCGIFADDR, (void *)&ifr) < 0) {
close(skfd);
perrx("socket SIOCGIFADDR");
}
saddr = (struct sockaddr_in *)&ifr.ifr_addr;
*ipaddr = saddr->sin_addr.s_addr;
dbg("get IPaddr: "IPFMT, ipfmt(*ipaddr));
}
/* NOTE: icmp dont drop @ipkb */
void icmp_send(unsigned char type, unsigned char code,
unsigned int data, struct pkbuf *pkb_in)
{
struct pkbuf *pkb;
struct ip *iphdr = pkb2ip(pkb_in);
struct icmp *icmphdr;
int paylen = _ntohs(iphdr->ip_len); /* icmp payload length */
if (paylen < iphlen(iphdr) + 8)
return;
/*
* RFC 1812 Section 4.3.2.7 for sanity check
* An ICMP error message MUST NOT be sent as the result of receiving:
* 1. A packet sent as a Link Layer broadcast or multicast
* 2. A packet destined to an IP broadcast or IP multicast address
*[3] A packet whose source address has a network prefix of zero or is an
* invalid source address (as defined in Section [5.3.7])
* 4. Any fragment of a datagram other then the first fragment (i.e., a
* packet for which the fragment offset in the IP header is nonzero).
* 5. An ICMP error message
*/
if (pkb_in->pk_type != PKT_LOCALHOST)
return;
if (MULTICAST(iphdr->ip_dst) || BROADCAST(iphdr->ip_dst))
return;
if (iphdr->ip_fragoff & _htons(IP_FRAG_OFF))
return;
if (icmp_type_error(type) && iphdr->ip_pro == IP_P_ICMP) {
icmphdr = ip2icmp(iphdr);
if (icmphdr->icmp_type > ICMP_T_MAXNUM || icmp_error(icmphdr))
return;
}
/* build icmp packet and send */
/* ip packet size must be smaller than 576 bytes */
if (IP_HRD_SZ + ICMP_HRD_SZ + paylen > 576)
paylen = 576 - IP_HRD_SZ - ICMP_HRD_SZ;
pkb = alloc_pkb(ETH_HRD_SZ + IP_HRD_SZ + ICMP_HRD_SZ + paylen);
icmphdr = (struct icmp *)(pkb2ip(pkb)->ip_data);
icmphdr->icmp_type = type;
icmphdr->icmp_code = code;
icmphdr->icmp_cksum = 0;
icmphdr->icmp_undata = data;
memcpy(icmphdr->icmp_data, (unsigned char *)iphdr, paylen);
icmphdr->icmp_cksum =
icmp_chksum((unsigned short *)icmphdr, ICMP_HRD_SZ + paylen);
icmpdbg("to "IPFMT"(payload %d) [type %d code %d]\n",
ipfmt(iphdr->ip_src), paylen, type, code);
ip_send_info(pkb, 0, IP_HRD_SZ + ICMP_HRD_SZ + paylen,
0, IP_P_ICMP, iphdr->ip_src);
}
void setipaddr_tap(unsigned char *name, unsigned int ipaddr)
{
struct ifreq ifr = {};
struct sockaddr_in *saddr;
strcpy(ifr.ifr_name, (char *)name);
saddr = (struct sockaddr_in *)&ifr.ifr_addr;
saddr->sin_family = AF_INET;
saddr->sin_addr.s_addr = ipaddr;
if (ioctl(skfd, SIOCSIFADDR, (void *)&ifr) < 0) {
close(skfd);
perrx("socket SIOCSIFADDR");
}
dbg("set IPaddr: "IPFMT, ipfmt(ipaddr));
}
void setnetmask_tap(unsigned char *name, unsigned int netmask)
{
struct ifreq ifr = {};
struct sockaddr_in *saddr;
strcpy(ifr.ifr_name, (char *)name);
saddr = (struct sockaddr_in *)&ifr.ifr_netmask;
saddr->sin_family = AF_INET;
saddr->sin_addr.s_addr = netmask;
if (ioctl(skfd, SIOCSIFNETMASK, (void *)&ifr) < 0) {
close(skfd);
perrx("socket SIOCSIFNETMASK");
}
dbg("set Netmask: "IPFMT, ipfmt(netmask));
}
struct arpentry *arp_lookup(unsigned short pro, unsigned int ipaddr)
{
struct arpentry *ae, *ret = NULL;
arp_cache_lock();
arpdbg("pro:%d "IPFMT, pro, ipfmt(ipaddr));
for (ae = arp_cache_head; ae < arp_cache_end; ae++) {
if (ae->ae_state == ARP_FREE)
continue;
if (ae->ae_pro == pro && ae->ae_ipaddr == ipaddr) {
ret = ae;
break;
}
}
arp_cache_unlock();
return ret;
}
static int veth_dev_init(struct netdev *dev)
{
/* init tap: out network nic */
if (tap_dev_init() < 0)
perrx("Cannot init tap device");
/* init veth: information for our netstack */
dev->net_mtu = tap->dev.net_mtu;
dev->net_ipaddr = FAKE_IPADDR;
dev->net_mask = FAKE_NETMASK;
hwacpy(dev->net_hwaddr, FAKE_HWADDR);
dbg("%s ip address: " IPFMT, dev->net_name, ipfmt(dev->net_ipaddr));
dbg("%s hw address: " MACFMT, dev->net_name, macfmt(dev->net_hwaddr));
/* net stats have been zero */
return 0;
}
static void tcp_init_text(struct tcp_sock *tsk, struct pkbuf *pkb,
void *buf, int size)
{
struct tcp *tcphdr = pkb2tcp(pkb);
tcphdr->src = tsk->sk.sk_sport;
tcphdr->dst = tsk->sk.sk_dport;
tcphdr->doff = TCP_HRD_DOFF;
tcphdr->seq = _htonl(tsk->snd_nxt);
tcphdr->ackn = _htonl(tsk->rcv_nxt);
tcphdr->ack = 1;
tcphdr->window = _htons(tsk->rcv_wnd);
memcpy(tcphdr->data, buf, size);
tsk->snd_nxt += size;
tsk->snd_wnd -= size;
tcpsdbg("send TEXT(%u:%d) [WIN %d] to "IPFMT":%d",
_ntohl(tcphdr->seq), size, _ntohs(tcphdr->window),
ipfmt(tsk->sk.sk_daddr), _ntohs(tcphdr->dst));
}
void ifinfo(struct netdev *dev)
{
printf("%-10sHWaddr "MACFMT"\n"
" IPaddr "IPFMT"\n"
" mtu %d\n"
" RX packet:%u bytes:%u errors:%u\n"
" TX packet:%u bytes:%u errors:%u\n",
dev->net_name,
macfmt(dev->net_hwaddr),
ipfmt(dev->net_ipaddr),
dev->net_mtu,
dev->net_stats.rx_packets,
dev->net_stats.rx_bytes,
dev->net_stats.rx_errors,
dev->net_stats.tx_packets,
dev->net_stats.tx_bytes,
dev->net_stats.tx_errors);
}
void tcp_send_syn(struct tcp_sock *tsk, struct tcp_segment *seg)
{
/*
* SYN-SENT:
*/
struct tcp *otcp;
struct pkbuf *opkb;
opkb = alloc_pkb(ETH_HRD_SZ + IP_HRD_SZ + TCP_HRD_SZ);
/* fill tcp head */
otcp = (struct tcp *)pkb2ip(opkb)->ip_data;
otcp->src = tsk->sk.sk_sport;
otcp->dst = tsk->sk.sk_dport;
otcp->doff = TCP_HRD_DOFF;
otcp->seq = _htonl(tsk->iss);
otcp->syn = 1;
otcp->window = _htons(tsk->rcv_wnd);
tcpdbg("send SYN(%u) [WIN %d] to "IPFMT":%d",
_ntohl(otcp->seq), _ntohs(otcp->window),
ipfmt(tsk->sk.sk_daddr), _ntohs(otcp->dst));
tcp_send_out(tsk, opkb, seg);
}
static void recv_packet(void)
{
struct pkbuf *pkb;
struct ip *iphdr;
struct icmp *icmphdr;
while (!finite || recv > 0) {
pkb = _recv(sock);
if (!pkb)
break;
iphdr = pkb2ip(pkb);
icmphdr = ip2icmp(iphdr);
if (iphdr->ip_pro == IP_P_ICMP &&
_ntohs(icmphdr->icmp_id) == id &&
icmphdr->icmp_type == ICMP_T_ECHORLY) {
recv--;
printf("%d bytes from " IPFMT ": icmp_seq=%d ttl=%d\n",
ipdlen(iphdr), ipfmt(iphdr->ip_src),
_ntohs(icmphdr->icmp_seq), iphdr->ip_ttl);
precv++;
}
free_pkb(pkb);
}
}
/*
* Acknowledgment algorithm is not stated directly in RFC 793,
* but we can conclude it from all acknowledgment situation.
*/
void tcp_send_ack(struct tcp_sock *tsk, struct tcp_segment *seg)
{
/*
* SYN-SENT :
* SEG: SYN, acceptable ACK, no RST (SND.NXT = SEG.SEQ+1)
* <SEQ=SND.NXT><ACK=RCV.NXT><CTL=ACK>
* SYN-RECEIVED / ESTABLISHED / FIN-WAIT-1 / FIN-WAIT-2 /
* CLOSE-WAIT / CLOSING / LAST-ACK / TIME-WAIT :
* SEG: no RST, ??ACK, ??SYN (segment is not acceptable)
* <SEQ=SND.NXT><ACK=RCV.NXT><CTL=ACK>
* ESTABLISHED / FIN-WAIT-1 / FIN-WAIT-2 / process the segment text:
* SEG: ACK, no RST
* <SEQ=SND.NXT><ACK=RCV.NXT><CTL=ACK>
* (This acknowledgment should be piggybacked on a segment being
* transmitted if possible without incurring undue delay.)
*/
struct tcp *otcp, *tcphdr = seg->tcphdr;
struct pkbuf *opkb;
if (tcphdr->rst)
return;
opkb = alloc_pkb(ETH_HRD_SZ + IP_HRD_SZ + TCP_HRD_SZ);
/* fill tcp head */
otcp = (struct tcp *)pkb2ip(opkb)->ip_data;
otcp->src = tcphdr->dst;
otcp->dst = tcphdr->src;
otcp->doff = TCP_HRD_DOFF;
otcp->seq = _htonl(tsk->snd_nxt);
otcp->ackn = _htonl(tsk->rcv_nxt);
otcp->ack = 1;
otcp->window = _htons(tsk->rcv_wnd);
tcpdbg("send ACK(%u) [WIN %d] to "IPFMT":%d",
_ntohl(otcp->ackn), _ntohs(otcp->window),
ipfmt(seg->iphdr->ip_src), _ntohs(otcp->dst));
tcp_send_out(tsk, opkb, seg);
}
char *ipnfmt(unsigned int ipaddr)
{
static char ipbuf[16];
snprintf(ipbuf, 16, IPFMT, ipfmt(ipaddr));
return ipbuf;
}
int cl_setup_promisc(struct netdev *nd)
{
char buf[1024]={0};
struct ifconf ifc;
struct ifreq *ifr;
int if_cnt;
int yes;
struct sockaddr_ll sll;
/* set netlink socket */
nd->nd_fd = socket(PF_PACKET,SOCK_RAW,htons(ETH_P_ALL));
if(nd->nd_fd == -1) {
perrx("socket:PF_PACKET");
return -1;
}
memset(&ifc,0,sizeof(ifc));
memset(&ifr,0,sizeof(ifr));
ifc.ifc_len = sizeof(buf);
ifc.ifc_buf = buf;
if(ioctl(nd->nd_fd,SIOCGIFCONF,&ifc) < 0) {
perrx("ioctl:SIOCGIFCONF");
return -1;
}
/* list all network interfaces
then look for the desired one */
ifr = ifc.ifc_req;
if_cnt =ifc.ifc_len / sizeof(struct ifreq);
ifr = lookup_dev(ifr,&if_cnt,nd);
if(ifr == NULL) {
perrx("Couldn't get device name \n");
return -1;
}
#ifdef IFCONF
printfd(2,"interface index : %d\n",ifr->ifr_ifindex);
printfd(2,"interface address : "IPFMT"\n",
ipfmt((int)(((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr.s_addr)));
printfd(2,"interface netmask : "IPFMT"\n",
ipfmt((int)(((struct sockaddr_in *)&ifr->ifr_netmask)->sin_addr.s_addr)));
#endif
/* make sure that the interface is UP & RUNNING */
ifr->ifr_flags |= IFF_PROMISC;
ifr->ifr_flags |= IFF_UP |IFF_RUNNING;
if(ioctl(nd->nd_fd, SIOCSIFFLAGS, ifr) < 0) {
perrx("ioctl:SIOCGIFCONF");
return -1;
}
/* get interface index */
if (ioctl(nd->nd_fd,SIOCGIFINDEX,ifr)==-1) {
perrx("SIOCGIFINDEX");
}
if (setsockopt(nd->nd_fd, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(yes)) == -1) {
perrx("setsockopt:SO_REUSEADDR");
return -1;
}
/* No need for this
if (setsockopt(nd->nd_fd, SOL_SOCKET, SO_BINDTODEVICE, ifr->ifr_name, IFNAMSIZ-1) == -1) {
perrx("SO_BINDTODEVICE");
return -1;
}
*/
memset(&sll,0,sizeof(sll));
sll.sll_family = AF_PACKET;
sll.sll_ifindex = ifr->ifr_ifindex;
sll.sll_protocol = htons(ETH_P_ALL);
if(bind(nd->nd_fd,(struct sockaddr*)&sll,sizeof(sll)) == -1) {
perrx("bind");
return -1;
}
nd->nd_ops->xmit(nd,NULL);
return 0;
}