u_int16_t
inet_cksum(struct mbuf *m, unsigned int nxt, unsigned int skip,
unsigned int len)
{
u_int32_t sum = 0;
/* sanity check */
if ((m->m_flags & M_PKTHDR) && m->m_pkthdr.len < skip + len) {
panic("inet_cksum: mbuf len (%d) < off+len (%d+%d)\n",
m->m_pkthdr.len, skip, len);
}
/* include pseudo header checksum? */
if (nxt != 0) {
struct ip *iph;
if (m->m_len < sizeof (struct ip))
panic("inet_cksum: bad mbuf chain");
iph = mtod(m, struct ip *);
sum = in_pseudo(iph->ip_src.s_addr, iph->ip_dst.s_addr,
htonl(len + nxt));
}
return (cpu_in_cksum(m, len, skip, sum));
}
static void
udp_send(netmsg_t msg)
{
struct socket *so = msg->send.base.nm_so;
struct mbuf *m = msg->send.nm_m;
struct sockaddr *dstaddr = msg->send.nm_addr;
int pru_flags = msg->send.nm_flags;
struct inpcb *inp = so->so_pcb;
struct thread *td = msg->send.nm_td;
int flags;
struct udpiphdr *ui;
int len = m->m_pkthdr.len;
struct sockaddr_in *sin; /* really is initialized before use */
int error = 0, cpu;
KKASSERT(msg->send.nm_control == NULL);
logudp(send_beg, inp);
if (inp == NULL) {
error = EINVAL;
goto release;
}
if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
error = EMSGSIZE;
goto release;
}
if (inp->inp_lport == 0) { /* unbound socket */
boolean_t forwarded;
error = in_pcbbind(inp, NULL, td);
if (error)
goto release;
/*
* Need to call udp_send again, after this inpcb is
* inserted into wildcard hash table.
*/
msg->send.base.lmsg.ms_flags |= MSGF_UDP_SEND;
forwarded = udp_inswildcardhash(inp, &msg->send.base, 0);
if (forwarded) {
/*
* The message is further forwarded, so we are
* done here.
*/
logudp(send_inswildcard, inp);
return;
}
}
if (dstaddr != NULL) { /* destination address specified */
if (inp->inp_faddr.s_addr != INADDR_ANY) {
/* already connected */
error = EISCONN;
goto release;
}
sin = (struct sockaddr_in *)dstaddr;
if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
error = EAFNOSUPPORT; /* IPv6 only jail */
goto release;
}
} else {
if (inp->inp_faddr.s_addr == INADDR_ANY) {
/* no destination specified and not already connected */
error = ENOTCONN;
goto release;
}
sin = NULL;
}
/*
* Calculate data length and get a mbuf
* for UDP and IP headers.
*/
M_PREPEND(m, sizeof(struct udpiphdr), M_NOWAIT);
if (m == NULL) {
error = ENOBUFS;
goto release;
}
/*
* Fill in mbuf with extended UDP header
* and addresses and length put into network format.
*/
ui = mtod(m, struct udpiphdr *);
bzero(ui->ui_x1, sizeof ui->ui_x1); /* XXX still needed? */
ui->ui_pr = IPPROTO_UDP;
/*
* Set destination address.
*/
if (dstaddr != NULL) { /* use specified destination */
ui->ui_dst = sin->sin_addr;
ui->ui_dport = sin->sin_port;
} else { /* use connected destination */
ui->ui_dst = inp->inp_faddr;
ui->ui_dport = inp->inp_fport;
}
/*
* Set source address.
*/
if (inp->inp_laddr.s_addr == INADDR_ANY ||
IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
struct sockaddr_in *if_sin;
if (dstaddr == NULL) {
/*
* connect() had (or should have) failed because
* the interface had no IP address, but the
* application proceeded to call send() anyways.
*/
error = ENOTCONN;
goto release;
}
/* Look up outgoing interface. */
error = in_pcbladdr_find(inp, dstaddr, &if_sin, td, 1);
if (error)
goto release;
ui->ui_src = if_sin->sin_addr; /* use address of interface */
} else {
ui->ui_src = inp->inp_laddr; /* use non-null bound address */
}
ui->ui_sport = inp->inp_lport;
KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
/*
* Release the original thread, since it is no longer used
*/
if (pru_flags & PRUS_HELDTD) {
lwkt_rele(td);
pru_flags &= ~PRUS_HELDTD;
}
/*
* Free the dest address, since it is no longer needed
*/
if (pru_flags & PRUS_FREEADDR) {
kfree(dstaddr, M_SONAME);
pru_flags &= ~PRUS_FREEADDR;
}
ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
/*
* Set up checksum and output datagram.
*/
if (udpcksum) {
ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
m->m_pkthdr.csum_flags = CSUM_UDP;
m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
m->m_pkthdr.csum_thlen = sizeof(struct udphdr);
} else {
int
udp_input(struct mbuf **mp, int *offp, int proto)
{
struct sockaddr_in udp_in = { sizeof udp_in, AF_INET };
int iphlen;
struct ip *ip;
struct udphdr *uh;
struct inpcb *inp;
struct mbuf *m;
struct mbuf *opts = NULL;
int len, off;
struct ip save_ip;
struct inpcbinfo *pcbinfo = &udbinfo[mycpuid];
off = *offp;
m = *mp;
*mp = NULL;
iphlen = off;
udp_stat.udps_ipackets++;
/*
* Strip IP options, if any; should skip this,
* make available to user, and use on returned packets,
* but we don't yet have a way to check the checksum
* with options still present.
*/
if (iphlen > sizeof(struct ip)) {
ip_stripoptions(m);
iphlen = sizeof(struct ip);
}
/*
* IP and UDP headers are together in first mbuf.
* Already checked and pulled up in ip_demux().
*/
KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
("UDP header not in one mbuf"));
ip = mtod(m, struct ip *);
uh = (struct udphdr *)((caddr_t)ip + iphlen);
/* destination port of 0 is illegal, based on RFC768. */
if (uh->uh_dport == 0)
goto bad;
/*
* Make mbuf data length reflect UDP length.
* If not enough data to reflect UDP length, drop.
*/
len = ntohs((u_short)uh->uh_ulen);
if (ip->ip_len != len) {
if (len > ip->ip_len || len < sizeof(struct udphdr)) {
udp_stat.udps_badlen++;
goto bad;
}
m_adj(m, len - ip->ip_len);
/* ip->ip_len = len; */
}
/*
* Save a copy of the IP header in case we want restore it
* for sending an ICMP error message in response.
*/
save_ip = *ip;
/*
* Checksum extended UDP header and data.
*/
if (uh->uh_sum) {
if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
uh->uh_sum = m->m_pkthdr.csum_data;
else
uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
ip->ip_dst.s_addr, htonl((u_short)len +
m->m_pkthdr.csum_data + IPPROTO_UDP));
uh->uh_sum ^= 0xffff;
} else {
char b[9];
bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
bzero(((struct ipovly *)ip)->ih_x1, 9);
((struct ipovly *)ip)->ih_len = uh->uh_ulen;
uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
}
if (uh->uh_sum) {
udp_stat.udps_badsum++;
m_freem(m);
return(IPPROTO_DONE);
}
} else
udp_stat.udps_nosum++;
if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
struct inpcbhead *connhead;
struct inpcontainer *ic, *ic_marker;
struct inpcontainerhead *ichead;
struct udp_mcast_arg arg;
struct inpcb *last;
int error;
/*
* Deliver a multicast or broadcast datagram to *all* sockets
* for which the local and remote addresses and ports match
* those of the incoming datagram. This allows more than
* one process to receive multi/broadcasts on the same port.
* (This really ought to be done for unicast datagrams as
* well, but that would cause problems with existing
* applications that open both address-specific sockets and
* a wildcard socket listening to the same port -- they would
* end up receiving duplicates of every unicast datagram.
* Those applications open the multiple sockets to overcome an
* inadequacy of the UDP socket interface, but for backwards
* compatibility we avoid the problem here rather than
* fixing the interface. Maybe 4.5BSD will remedy this?)
*/
/*
* Construct sockaddr format source address.
*/
udp_in.sin_port = uh->uh_sport;
udp_in.sin_addr = ip->ip_src;
arg.udp_in = &udp_in;
/*
* Locate pcb(s) for datagram.
* (Algorithm copied from raw_intr().)
*/
last = NULL;
arg.iphlen = iphlen;
connhead = &pcbinfo->hashbase[
INP_PCBCONNHASH(ip->ip_src.s_addr, uh->uh_sport,
ip->ip_dst.s_addr, uh->uh_dport, pcbinfo->hashmask)];
LIST_FOREACH(inp, connhead, inp_hash) {
#ifdef INET6
if (!INP_ISIPV4(inp))
continue;
#endif
if (!in_hosteq(inp->inp_faddr, ip->ip_src) ||
!in_hosteq(inp->inp_laddr, ip->ip_dst) ||
inp->inp_fport != uh->uh_sport ||
inp->inp_lport != uh->uh_dport)
continue;
arg.inp = inp;
arg.last = last;
arg.ip = ip;
arg.m = m;
error = udp_mcast_input(&arg);
if (error == ERESTART)
continue;
last = arg.last;
if (error == EJUSTRETURN)
goto done;
}
ichead = &pcbinfo->wildcardhashbase[
INP_PCBWILDCARDHASH(uh->uh_dport,
pcbinfo->wildcardhashmask)];
ic_marker = in_pcbcontainer_marker(mycpuid);
GET_PCBINFO_TOKEN(pcbinfo);
LIST_INSERT_HEAD(ichead, ic_marker, ic_list);
while ((ic = LIST_NEXT(ic_marker, ic_list)) != NULL) {
LIST_REMOVE(ic_marker, ic_list);
LIST_INSERT_AFTER(ic, ic_marker, ic_list);
inp = ic->ic_inp;
if (inp->inp_flags & INP_PLACEMARKER)
continue;
#ifdef INET6
if (!INP_ISIPV4(inp))
continue;
#endif
if (inp->inp_lport != uh->uh_dport)
continue;
if (inp->inp_laddr.s_addr != INADDR_ANY &&
inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
continue;
arg.inp = inp;
arg.last = last;
arg.ip = ip;
arg.m = m;
error = udp_mcast_input(&arg);
if (error == ERESTART)
continue;
last = arg.last;
if (error == EJUSTRETURN)
break;
}
LIST_REMOVE(ic_marker, ic_list);
REL_PCBINFO_TOKEN(pcbinfo);
done:
if (last == NULL) {
/*
* No matching pcb found; discard datagram.
* (No need to send an ICMP Port Unreachable
* for a broadcast or multicast datgram.)
*/
udp_stat.udps_noportbcast++;
goto bad;
}
#ifdef IPSEC
/* check AH/ESP integrity. */
if (ipsec4_in_reject_so(m, last->inp_socket)) {
ipsecstat.in_polvio++;
goto bad;
}
#endif /*IPSEC*/
#ifdef FAST_IPSEC
/* check AH/ESP integrity. */
if (ipsec4_in_reject(m, last))
goto bad;
#endif /*FAST_IPSEC*/
udp_append(last, ip, m, iphlen + sizeof(struct udphdr),
&udp_in);
return(IPPROTO_DONE);
}
/*
* Locate pcb for datagram.
*/
inp = in_pcblookup_pkthash(pcbinfo, ip->ip_src, uh->uh_sport,
ip->ip_dst, uh->uh_dport, TRUE, m->m_pkthdr.rcvif,
udp_reuseport_ext ? m : NULL);
if (inp == NULL) {
if (log_in_vain) {
char buf[sizeof "aaa.bbb.ccc.ddd"];
strcpy(buf, inet_ntoa(ip->ip_dst));
log(LOG_INFO,
"Connection attempt to UDP %s:%d from %s:%d\n",
buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
ntohs(uh->uh_sport));
}
udp_stat.udps_noport++;
if (m->m_flags & (M_BCAST | M_MCAST)) {
udp_stat.udps_noportbcast++;
goto bad;
}
if (blackhole)
goto bad;
#ifdef ICMP_BANDLIM
if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
goto bad;
#endif
*ip = save_ip;
ip->ip_len += iphlen;
icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
return(IPPROTO_DONE);
}
KASSERT(INP_ISIPV4(inp), ("not inet inpcb"));
#ifdef IPSEC
if (ipsec4_in_reject_so(m, inp->inp_socket)) {
ipsecstat.in_polvio++;
goto bad;
}
#endif /*IPSEC*/
#ifdef FAST_IPSEC
if (ipsec4_in_reject(m, inp))
goto bad;
#endif /*FAST_IPSEC*/
/*
* Check the minimum TTL for socket.
*/
if (ip->ip_ttl < inp->inp_ip_minttl)
goto bad;
/*
* Construct sockaddr format source address.
* Stuff source address and datagram in user buffer.
*/
udp_in.sin_port = uh->uh_sport;
udp_in.sin_addr = ip->ip_src;
if ((inp->inp_flags & INP_CONTROLOPTS) ||
(inp->inp_socket->so_options & SO_TIMESTAMP))
ip_savecontrol(inp, &opts, ip, m);
m_adj(m, iphlen + sizeof(struct udphdr));
lwkt_gettoken(&inp->inp_socket->so_rcv.ssb_token);
if (ssb_appendaddr(&inp->inp_socket->so_rcv,
(struct sockaddr *)&udp_in, m, opts) == 0) {
lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
udp_stat.udps_fullsock++;
goto bad;
}
lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
sorwakeup(inp->inp_socket);
return(IPPROTO_DONE);
bad:
m_freem(m);
if (opts)
m_freem(opts);
return(IPPROTO_DONE);
}
void
udp_input(struct mbuf *m, int off)
{
int iphlen = off;
struct ip *ip;
struct udphdr *uh;
struct ifnet *ifp;
struct inpcb *inp;
uint16_t len, ip_len;
struct inpcbinfo *pcbinfo;
struct ip save_ip;
struct sockaddr_in udp_in;
struct m_tag *fwd_tag;
int cscov_partial;
uint8_t pr;
ifp = m->m_pkthdr.rcvif;
UDPSTAT_INC(udps_ipackets);
/*
* Strip IP options, if any; should skip this, make available to
* user, and use on returned packets, but we don't yet have a way to
* check the checksum with options still present.
*/
if (iphlen > sizeof (struct ip)) {
ip_stripoptions(m);
iphlen = sizeof(struct ip);
}
/*
* Get IP and UDP header together in first mbuf.
*/
ip = mtod(m, struct ip *);
if (m->m_len < iphlen + sizeof(struct udphdr)) {
if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == NULL) {
UDPSTAT_INC(udps_hdrops);
return;
}
ip = mtod(m, struct ip *);
}
uh = (struct udphdr *)((caddr_t)ip + iphlen);
pr = ip->ip_p;
cscov_partial = (pr == IPPROTO_UDPLITE) ? 1 : 0;
/*
* Destination port of 0 is illegal, based on RFC768.
*/
if (uh->uh_dport == 0)
goto badunlocked;
/*
* Construct sockaddr format source address. Stuff source address
* and datagram in user buffer.
*/
bzero(&udp_in, sizeof(udp_in));
udp_in.sin_len = sizeof(udp_in);
udp_in.sin_family = AF_INET;
udp_in.sin_port = uh->uh_sport;
udp_in.sin_addr = ip->ip_src;
/*
* Make mbuf data length reflect UDP length. If not enough data to
* reflect UDP length, drop.
*/
len = ntohs((u_short)uh->uh_ulen);
ip_len = ntohs(ip->ip_len) - iphlen;
if (pr == IPPROTO_UDPLITE && len == 0) {
/* Zero means checksum over the complete packet. */
len = ip_len;
cscov_partial = 0;
}
if (ip_len != len) {
if (len > ip_len || len < sizeof(struct udphdr)) {
UDPSTAT_INC(udps_badlen);
goto badunlocked;
}
if (pr == IPPROTO_UDP)
m_adj(m, len - ip_len);
}
/*
* Save a copy of the IP header in case we want restore it for
* sending an ICMP error message in response.
*/
if (!V_udp_blackhole)
save_ip = *ip;
else
memset(&save_ip, 0, sizeof(save_ip));
/*
* Checksum extended UDP header and data.
*/
if (uh->uh_sum) {
u_short uh_sum;
if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID) &&
!cscov_partial) {
if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
uh_sum = m->m_pkthdr.csum_data;
else
uh_sum = in_pseudo(ip->ip_src.s_addr,
ip->ip_dst.s_addr, htonl((u_short)len +
m->m_pkthdr.csum_data + pr));
uh_sum ^= 0xffff;
} else {
char b[9];
bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
bzero(((struct ipovly *)ip)->ih_x1, 9);
((struct ipovly *)ip)->ih_len = (pr == IPPROTO_UDP) ?
uh->uh_ulen : htons(ip_len);
uh_sum = in_cksum(m, len + sizeof (struct ip));
bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
}
if (uh_sum) {
UDPSTAT_INC(udps_badsum);
m_freem(m);
return;
}
} else
UDPSTAT_INC(udps_nosum);
pcbinfo = get_inpcbinfo(pr);
if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
in_broadcast(ip->ip_dst, ifp)) {
struct inpcb *last;
struct inpcbhead *pcblist;
struct ip_moptions *imo;
INP_INFO_RLOCK(pcbinfo);
pcblist = get_pcblist(pr);
last = NULL;
LIST_FOREACH(inp, pcblist, inp_list) {
if (inp->inp_lport != uh->uh_dport)
continue;
#ifdef INET6
if ((inp->inp_vflag & INP_IPV4) == 0)
continue;
#endif
if (inp->inp_laddr.s_addr != INADDR_ANY &&
inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
continue;
if (inp->inp_faddr.s_addr != INADDR_ANY &&
inp->inp_faddr.s_addr != ip->ip_src.s_addr)
continue;
if (inp->inp_fport != 0 &&
inp->inp_fport != uh->uh_sport)
continue;
INP_RLOCK(inp);
/*
* XXXRW: Because we weren't holding either the inpcb
* or the hash lock when we checked for a match
* before, we should probably recheck now that the
* inpcb lock is held.
*/
/*
* Handle socket delivery policy for any-source
* and source-specific multicast. [RFC3678]
*/
imo = inp->inp_moptions;
if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
struct sockaddr_in group;
int blocked;
if (imo == NULL) {
INP_RUNLOCK(inp);
continue;
}
bzero(&group, sizeof(struct sockaddr_in));
group.sin_len = sizeof(struct sockaddr_in);
group.sin_family = AF_INET;
group.sin_addr = ip->ip_dst;
blocked = imo_multi_filter(imo, ifp,
(struct sockaddr *)&group,
(struct sockaddr *)&udp_in);
if (blocked != MCAST_PASS) {
if (blocked == MCAST_NOTGMEMBER)
IPSTAT_INC(ips_notmember);
if (blocked == MCAST_NOTSMEMBER ||
blocked == MCAST_MUTED)
UDPSTAT_INC(udps_filtermcast);
INP_RUNLOCK(inp);
continue;
}
}
if (last != NULL) {
struct mbuf *n;
n = m_copy(m, 0, M_COPYALL);
udp_append(last, ip, n, iphlen, &udp_in);
INP_RUNLOCK(last);
}
last = inp;
/*
* Don't look for additional matches if this one does
* not have either the SO_REUSEPORT or SO_REUSEADDR
* socket options set. This heuristic avoids
* searching through all pcbs in the common case of a
* non-shared port. It assumes that an application
* will never clear these options after setting them.
*/
if ((last->inp_socket->so_options &
(SO_REUSEPORT|SO_REUSEADDR)) == 0)
break;
}
if (last == NULL) {
/*
* No matching pcb found; discard datagram. (No need
* to send an ICMP Port Unreachable for a broadcast
* or multicast datgram.)
*/
UDPSTAT_INC(udps_noportbcast);
if (inp)
INP_RUNLOCK(inp);
INP_INFO_RUNLOCK(pcbinfo);
goto badunlocked;
}
udp_append(last, ip, m, iphlen, &udp_in);
INP_RUNLOCK(last);
INP_INFO_RUNLOCK(pcbinfo);
return;
}
/*
* Locate pcb for datagram.
*/
/*
* Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
*/
if ((m->m_flags & M_IP_NEXTHOP) &&
(fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
struct sockaddr_in *next_hop;
next_hop = (struct sockaddr_in *)(fwd_tag + 1);
/*
* Transparently forwarded. Pretend to be the destination.
* Already got one like this?
*/
inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport,
ip->ip_dst, uh->uh_dport, INPLOOKUP_RLOCKPCB, ifp, m);
if (!inp) {
/*
* It's new. Try to find the ambushing socket.
* Because we've rewritten the destination address,
* any hardware-generated hash is ignored.
*/
inp = in_pcblookup(pcbinfo, ip->ip_src,
uh->uh_sport, next_hop->sin_addr,
next_hop->sin_port ? htons(next_hop->sin_port) :
uh->uh_dport, INPLOOKUP_WILDCARD |
INPLOOKUP_RLOCKPCB, ifp);
}
/* Remove the tag from the packet. We don't need it anymore. */
m_tag_delete(m, fwd_tag);
m->m_flags &= ~M_IP_NEXTHOP;
} else
inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport,
ip->ip_dst, uh->uh_dport, INPLOOKUP_WILDCARD |
INPLOOKUP_RLOCKPCB, ifp, m);
if (inp == NULL) {
if (udp_log_in_vain) {
char buf[4*sizeof "123"];
strcpy(buf, inet_ntoa(ip->ip_dst));
log(LOG_INFO,
"Connection attempt to UDP %s:%d from %s:%d\n",
buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
ntohs(uh->uh_sport));
}
UDPSTAT_INC(udps_noport);
if (m->m_flags & (M_BCAST | M_MCAST)) {
UDPSTAT_INC(udps_noportbcast);
goto badunlocked;
}
if (V_udp_blackhole)
goto badunlocked;
if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
goto badunlocked;
*ip = save_ip;
icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
return;
}
/*
* Check the minimum TTL for socket.
*/
INP_RLOCK_ASSERT(inp);
if (inp->inp_ip_minttl && inp->inp_ip_minttl > ip->ip_ttl) {
INP_RUNLOCK(inp);
m_freem(m);
return;
}
if (cscov_partial) {
struct udpcb *up;
up = intoudpcb(inp);
if (up->u_rxcslen > len) {
INP_RUNLOCK(inp);
m_freem(m);
return;
}
}
UDP_PROBE(receive, NULL, inp, ip, inp, uh);
udp_append(inp, ip, m, iphlen, &udp_in);
INP_RUNLOCK(inp);
return;
badunlocked:
m_freem(m);
}
static int
udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr,
struct mbuf *control, struct thread *td)
{
struct udpiphdr *ui;
int len = m->m_pkthdr.len;
struct in_addr faddr, laddr;
struct cmsghdr *cm;
struct inpcbinfo *pcbinfo;
struct sockaddr_in *sin, src;
int cscov_partial = 0;
int error = 0;
int ipflags;
u_short fport, lport;
int unlock_udbinfo;
u_char tos;
uint8_t pr;
uint16_t cscov = 0;
/*
* udp_output() may need to temporarily bind or connect the current
* inpcb. As such, we don't know up front whether we will need the
* pcbinfo lock or not. Do any work to decide what is needed up
* front before acquiring any locks.
*/
if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
if (control)
m_freem(control);
m_freem(m);
return (EMSGSIZE);
}
src.sin_family = 0;
INP_RLOCK(inp);
tos = inp->inp_ip_tos;
if (control != NULL) {
/*
* XXX: Currently, we assume all the optional information is
* stored in a single mbuf.
*/
if (control->m_next) {
INP_RUNLOCK(inp);
m_freem(control);
m_freem(m);
return (EINVAL);
}
for (; control->m_len > 0;
control->m_data += CMSG_ALIGN(cm->cmsg_len),
control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
cm = mtod(control, struct cmsghdr *);
if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0
|| cm->cmsg_len > control->m_len) {
error = EINVAL;
break;
}
if (cm->cmsg_level != IPPROTO_IP)
continue;
switch (cm->cmsg_type) {
case IP_SENDSRCADDR:
if (cm->cmsg_len !=
CMSG_LEN(sizeof(struct in_addr))) {
error = EINVAL;
break;
}
bzero(&src, sizeof(src));
src.sin_family = AF_INET;
src.sin_len = sizeof(src);
src.sin_port = inp->inp_lport;
src.sin_addr =
*(struct in_addr *)CMSG_DATA(cm);
break;
case IP_TOS:
if (cm->cmsg_len != CMSG_LEN(sizeof(u_char))) {
error = EINVAL;
break;
}
tos = *(u_char *)CMSG_DATA(cm);
break;
default:
error = ENOPROTOOPT;
break;
}
if (error)
break;
}
m_freem(control);
}
if (error) {
INP_RUNLOCK(inp);
m_freem(m);
return (error);
}
/*
* Depending on whether or not the application has bound or connected
* the socket, we may have to do varying levels of work. The optimal
* case is for a connected UDP socket, as a global lock isn't
* required at all.
*
* In order to decide which we need, we require stability of the
* inpcb binding, which we ensure by acquiring a read lock on the
* inpcb. This doesn't strictly follow the lock order, so we play
* the trylock and retry game; note that we may end up with more
* conservative locks than required the second time around, so later
* assertions have to accept that. Further analysis of the number of
* misses under contention is required.
*
* XXXRW: Check that hash locking update here is correct.
*/
pr = inp->inp_socket->so_proto->pr_protocol;
pcbinfo = get_inpcbinfo(pr);
sin = (struct sockaddr_in *)addr;
if (sin != NULL &&
(inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0)) {
INP_RUNLOCK(inp);
INP_WLOCK(inp);
INP_HASH_WLOCK(pcbinfo);
unlock_udbinfo = UH_WLOCKED;
} else if ((sin != NULL && (
(sin->sin_addr.s_addr == INADDR_ANY) ||
(sin->sin_addr.s_addr == INADDR_BROADCAST) ||
(inp->inp_laddr.s_addr == INADDR_ANY) ||
(inp->inp_lport == 0))) ||
(src.sin_family == AF_INET)) {
INP_HASH_RLOCK(pcbinfo);
unlock_udbinfo = UH_RLOCKED;
} else
unlock_udbinfo = UH_UNLOCKED;
/*
* If the IP_SENDSRCADDR control message was specified, override the
* source address for this datagram. Its use is invalidated if the
* address thus specified is incomplete or clobbers other inpcbs.
*/
laddr = inp->inp_laddr;
lport = inp->inp_lport;
if (src.sin_family == AF_INET) {
INP_HASH_LOCK_ASSERT(pcbinfo);
if ((lport == 0) ||
(laddr.s_addr == INADDR_ANY &&
src.sin_addr.s_addr == INADDR_ANY)) {
error = EINVAL;
goto release;
}
error = in_pcbbind_setup(inp, (struct sockaddr *)&src,
&laddr.s_addr, &lport, td->td_ucred);
if (error)
goto release;
}
/*
* If a UDP socket has been connected, then a local address/port will
* have been selected and bound.
*
* If a UDP socket has not been connected to, then an explicit
* destination address must be used, in which case a local
* address/port may not have been selected and bound.
*/
if (sin != NULL) {
INP_LOCK_ASSERT(inp);
if (inp->inp_faddr.s_addr != INADDR_ANY) {
error = EISCONN;
goto release;
}
/*
* Jail may rewrite the destination address, so let it do
* that before we use it.
*/
error = prison_remote_ip4(td->td_ucred, &sin->sin_addr);
if (error)
goto release;
/*
* If a local address or port hasn't yet been selected, or if
* the destination address needs to be rewritten due to using
* a special INADDR_ constant, invoke in_pcbconnect_setup()
* to do the heavy lifting. Once a port is selected, we
* commit the binding back to the socket; we also commit the
* binding of the address if in jail.
*
* If we already have a valid binding and we're not
* requesting a destination address rewrite, use a fast path.
*/
if (inp->inp_laddr.s_addr == INADDR_ANY ||
inp->inp_lport == 0 ||
sin->sin_addr.s_addr == INADDR_ANY ||
sin->sin_addr.s_addr == INADDR_BROADCAST) {
INP_HASH_LOCK_ASSERT(pcbinfo);
error = in_pcbconnect_setup(inp, addr, &laddr.s_addr,
&lport, &faddr.s_addr, &fport, NULL,
td->td_ucred);
if (error)
goto release;
/*
* XXXRW: Why not commit the port if the address is
* !INADDR_ANY?
*/
/* Commit the local port if newly assigned. */
if (inp->inp_laddr.s_addr == INADDR_ANY &&
inp->inp_lport == 0) {
INP_WLOCK_ASSERT(inp);
INP_HASH_WLOCK_ASSERT(pcbinfo);
/*
* Remember addr if jailed, to prevent
* rebinding.
*/
if (prison_flag(td->td_ucred, PR_IP4))
inp->inp_laddr = laddr;
inp->inp_lport = lport;
if (in_pcbinshash(inp) != 0) {
inp->inp_lport = 0;
error = EAGAIN;
goto release;
}
inp->inp_flags |= INP_ANONPORT;
}
} else {
faddr = sin->sin_addr;
fport = sin->sin_port;
}
} else {
INP_LOCK_ASSERT(inp);
faddr = inp->inp_faddr;
fport = inp->inp_fport;
if (faddr.s_addr == INADDR_ANY) {
error = ENOTCONN;
goto release;
}
}
/*
* Calculate data length and get a mbuf for UDP, IP, and possible
* link-layer headers. Immediate slide the data pointer back forward
* since we won't use that space at this layer.
*/
M_PREPEND(m, sizeof(struct udpiphdr) + max_linkhdr, M_NOWAIT);
if (m == NULL) {
error = ENOBUFS;
goto release;
}
m->m_data += max_linkhdr;
m->m_len -= max_linkhdr;
m->m_pkthdr.len -= max_linkhdr;
/*
* Fill in mbuf with extended UDP header and addresses and length put
* into network format.
*/
ui = mtod(m, struct udpiphdr *);
bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */
ui->ui_pr = pr;
ui->ui_src = laddr;
ui->ui_dst = faddr;
ui->ui_sport = lport;
ui->ui_dport = fport;
ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
if (pr == IPPROTO_UDPLITE) {
struct udpcb *up;
uint16_t plen;
up = intoudpcb(inp);
cscov = up->u_txcslen;
plen = (u_short)len + sizeof(struct udphdr);
if (cscov >= plen)
cscov = 0;
ui->ui_len = htons(plen);
ui->ui_ulen = htons(cscov);
/*
* For UDP-Lite, checksum coverage length of zero means
* the entire UDPLite packet is covered by the checksum.
*/
cscov_partial = (cscov == 0) ? 0 : 1;
} else
ui->ui_v = IPVERSION << 4;
/*
* Set the Don't Fragment bit in the IP header.
*/
if (inp->inp_flags & INP_DONTFRAG) {
struct ip *ip;
ip = (struct ip *)&ui->ui_i;
ip->ip_off |= htons(IP_DF);
}
ipflags = 0;
if (inp->inp_socket->so_options & SO_DONTROUTE)
ipflags |= IP_ROUTETOIF;
if (inp->inp_socket->so_options & SO_BROADCAST)
ipflags |= IP_ALLOWBROADCAST;
if (inp->inp_flags & INP_ONESBCAST)
ipflags |= IP_SENDONES;
#ifdef MAC
mac_inpcb_create_mbuf(inp, m);
#endif
/*
* Set up checksum and output datagram.
*/
ui->ui_sum = 0;
if (cscov_partial) {
if (inp->inp_flags & INP_ONESBCAST)
faddr.s_addr = INADDR_BROADCAST;
if ((ui->ui_sum = in_cksum(m, sizeof(struct ip) + cscov)) == 0)
ui->ui_sum = 0xffff;
} else if (V_udp_cksum || !cscov_partial) {
if (inp->inp_flags & INP_ONESBCAST)
faddr.s_addr = INADDR_BROADCAST;
ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr,
htons((u_short)len + sizeof(struct udphdr) + pr));
m->m_pkthdr.csum_flags = CSUM_UDP;
m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
}
((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len);
((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
((struct ip *)ui)->ip_tos = tos; /* XXX */
UDPSTAT_INC(udps_opackets);
if (unlock_udbinfo == UH_WLOCKED)
INP_HASH_WUNLOCK(pcbinfo);
else if (unlock_udbinfo == UH_RLOCKED)
INP_HASH_RUNLOCK(pcbinfo);
UDP_PROBE(send, NULL, inp, &ui->ui_i, inp, &ui->ui_u);
error = ip_output(m, inp->inp_options, NULL, ipflags,
inp->inp_moptions, inp);
if (unlock_udbinfo == UH_WLOCKED)
INP_WUNLOCK(inp);
else
INP_RUNLOCK(inp);
return (error);
release:
if (unlock_udbinfo == UH_WLOCKED) {
INP_HASH_WUNLOCK(pcbinfo);
INP_WUNLOCK(inp);
} else if (unlock_udbinfo == UH_RLOCKED) {
INP_HASH_RUNLOCK(pcbinfo);
INP_RUNLOCK(inp);
} else
INP_RUNLOCK(inp);
m_freem(m);
return (error);
}
/*
* [udp_validate]
*
* Validates UDP length and checksum
* Almost verbatim from udp_input
* returns: 1 -> good, 0 -> bad
*/
static int
udp_validate(register struct mbuf *m)
{
register struct ip *ip = mtod(m, struct ip *);
register struct udphdr *uh;
int len;
// int iphlen = IP_VHL_HL(ip->ip_vhl) << 2;
int iphlen = ip->ip_hl << 2;
/*
* Can't check cksum with options present
*/
if (iphlen > sizeof (struct ip)) {
printf("udp_cksum_validate: ip options present, skipping checks\n");
#ifdef NOTYET
ip_stripoptions(m, (struct mbuf *)0);
iphlen = sizeof(struct ip);
#endif
return(1);
}
/*
* Get IP and UDP header together in first mbuf.
*/
if (m->m_len < iphlen + sizeof(struct udphdr)) {
if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
return(0);
}
ip = mtod(m, struct ip *);
}
uh = (struct udphdr *)((caddr_t)ip + iphlen);
/*
* Make mbuf data length reflect UDP length.
* If not enough data to reflect UDP length, drop.
*/
len = ntohs((u_short)uh->uh_ulen);
if (ip->ip_len != len) {
if (len > ip->ip_len || len < sizeof(struct udphdr)) {
printf("udp_validate: bad udp length\n");
m_freem(m);
return(0);
}
}
/*
* Checksum extended UDP header and data.
*/
if (uh->uh_sum) {
if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
uh->uh_sum = m->m_pkthdr.csum_data;
else
uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
ip->ip_dst.s_addr, htonl((u_short)len +
m->m_pkthdr.csum_data + IPPROTO_UDP));
uh->uh_sum ^= 0xffff;
} else {
char b[9];
bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
bzero(((struct ipovly *)ip)->ih_x1, 9);
((struct ipovly *)ip)->ih_len = uh->uh_ulen;
uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
}
if (uh->uh_sum) {
printf("udp_validate: failed, bad checksum\n");
m_freem(m);
return(0);
}
}
return(1);
}
/**********************************************************************
*
* Setup work for hardware segmentation offload (TSO) on
* adapters using advanced tx descriptors
*
**********************************************************************/
static int
ixgbe_tso_setup(struct tx_ring *txr, int ctxd, struct mbuf *mp,
u32 *cmd_type_len, u32 *olinfo_status)
{
struct ixgbe_adv_tx_context_desc *TXD;
u32 vlan_macip_lens = 0, type_tucmd_mlhl = 0;
u32 mss_l4len_idx = 0, paylen;
u16 vtag = 0, eh_type;
int ehdrlen, ip_hlen, tcp_hlen;
struct ether_vlan_header *eh;
#ifdef INET6
struct ip6_hdr *ip6;
#endif
#ifdef INET
struct ip *ip;
#endif
struct tcphdr *th;
/*
* Determine where frame payload starts.
* Jump over vlan headers if already present
*/
eh = mtod(mp, struct ether_vlan_header *);
if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
eh_type = eh->evl_proto;
} else {
ehdrlen = ETHER_HDR_LEN;
eh_type = eh->evl_encap_proto;
}
switch (ntohs(eh_type)) {
#ifdef INET6
case ETHERTYPE_IPV6:
ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen);
/* XXX-BZ For now we do not pretend to support ext. hdrs. */
if (ip6->ip6_nxt != IPPROTO_TCP)
return (ENXIO);
ip_hlen = sizeof(struct ip6_hdr);
ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen);
th = (struct tcphdr *)((caddr_t)ip6 + ip_hlen);
th->th_sum = in6_cksum_pseudo(ip6, 0, IPPROTO_TCP, 0);
type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV6;
break;
#endif
#ifdef INET
case ETHERTYPE_IP:
ip = (struct ip *)(mp->m_data + ehdrlen);
if (ip->ip_p != IPPROTO_TCP)
return (ENXIO);
ip->ip_sum = 0;
ip_hlen = ip->ip_hl << 2;
th = (struct tcphdr *)((caddr_t)ip + ip_hlen);
th->th_sum = in_pseudo(ip->ip_src.s_addr,
ip->ip_dst.s_addr, htons(IPPROTO_TCP));
type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_IPV4;
/* Tell transmit desc to also do IPv4 checksum. */
*olinfo_status |= IXGBE_TXD_POPTS_IXSM << 8;
break;
#endif
default:
panic("%s: CSUM_TSO but no supported IP version (0x%04x)",
__func__, ntohs(eh_type));
break;
}
TXD = (struct ixgbe_adv_tx_context_desc *) &txr->tx_base[ctxd];
tcp_hlen = th->th_off << 2;
/* This is used in the transmit desc in encap */
paylen = mp->m_pkthdr.len - ehdrlen - ip_hlen - tcp_hlen;
/* VLAN MACLEN IPLEN */
if (mp->m_flags & M_VLANTAG) {
vtag = htole16(mp->m_pkthdr.ether_vtag);
vlan_macip_lens |= (vtag << IXGBE_ADVTXD_VLAN_SHIFT);
}
vlan_macip_lens |= ehdrlen << IXGBE_ADVTXD_MACLEN_SHIFT;
vlan_macip_lens |= ip_hlen;
TXD->vlan_macip_lens = htole32(vlan_macip_lens);
/* ADV DTYPE TUCMD */
type_tucmd_mlhl |= IXGBE_ADVTXD_DCMD_DEXT | IXGBE_ADVTXD_DTYP_CTXT;
type_tucmd_mlhl |= IXGBE_ADVTXD_TUCMD_L4T_TCP;
TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl);
/* MSS L4LEN IDX */
mss_l4len_idx |= (mp->m_pkthdr.tso_segsz << IXGBE_ADVTXD_MSS_SHIFT);
mss_l4len_idx |= (tcp_hlen << IXGBE_ADVTXD_L4LEN_SHIFT);
TXD->mss_l4len_idx = htole32(mss_l4len_idx);
TXD->seqnum_seed = htole32(0);
*cmd_type_len |= IXGBE_ADVTXD_DCMD_TSE;
*olinfo_status |= IXGBE_TXD_POPTS_TXSM << 8;
*olinfo_status |= paylen << IXGBE_ADVTXD_PAYLEN_SHIFT;
++txr->tso_tx;
return (0);
}
u_int16_t
inet_cksum(struct mbuf *m, unsigned int nxt, unsigned int skip,
unsigned int len)
{
u_short *w;
u_int32_t sum = 0;
int mlen = 0;
int byte_swapped = 0;
union s_util s_util;
union l_util l_util;
KERNEL_DEBUG(DBG_FNC_IN_CKSUM | DBG_FUNC_START, len,0,0,0,0);
/* sanity check */
if ((m->m_flags & M_PKTHDR) && m->m_pkthdr.len < skip + len) {
panic("inet_cksum: mbuf len (%d) < off+len (%d+%d)\n",
m->m_pkthdr.len, skip, len);
}
/* include pseudo header checksum? */
if (nxt != 0) {
struct ip *iph;
if (m->m_len < sizeof (struct ip))
panic("inet_cksum: bad mbuf chain");
iph = mtod(m, struct ip *);
sum = in_pseudo(iph->ip_src.s_addr, iph->ip_dst.s_addr,
htonl(len + nxt));
}
if (skip != 0) {
for (; skip && m; m = m->m_next) {
if (m->m_len > skip) {
mlen = m->m_len - skip;
w = (u_short *)(m->m_data+skip);
goto skip_start;
} else {
skip -= m->m_len;
}
}
}
for (;m && len; m = m->m_next) {
if (m->m_len == 0)
continue;
w = mtod(m, u_short *);
if (mlen == -1) {
/*
* The first byte of this mbuf is the continuation
* of a word spanning between this mbuf and the
* last mbuf.
*
* s_util.c[0] is already saved when scanning previous
* mbuf.
*/
s_util.c[1] = *(char *)w;
sum += s_util.s;
w = (u_short *)((char *)w + 1);
mlen = m->m_len - 1;
len--;
} else {
mlen = m->m_len;
}
skip_start:
if (len < mlen)
mlen = len;
len -= mlen;
/*
* Force to even boundary.
*/
if ((1 & (uintptr_t) w) && (mlen > 0)) {
REDUCE;
sum <<= 8;
s_util.c[0] = *(u_char *)w;
w = (u_short *)((char *)w + 1);
mlen--;
byte_swapped = 1;
}
/*
* Unroll the loop to make overhead from
* branches &c small.
*/
while ((mlen -= 32) >= 0) {
sum += w[0]; sum += w[1]; sum += w[2]; sum += w[3];
sum += w[4]; sum += w[5]; sum += w[6]; sum += w[7];
sum += w[8]; sum += w[9]; sum += w[10]; sum += w[11];
sum += w[12]; sum += w[13]; sum += w[14]; sum += w[15];
w += 16;
}
mlen += 32;
while ((mlen -= 8) >= 0) {
sum += w[0]; sum += w[1]; sum += w[2]; sum += w[3];
w += 4;
}
mlen += 8;
if (mlen == 0 && byte_swapped == 0)
continue;
REDUCE;
while ((mlen -= 2) >= 0) {
sum += *w++;
}
if (byte_swapped) {
REDUCE;
sum <<= 8;
byte_swapped = 0;
if (mlen == -1) {
s_util.c[1] = *(char *)w;
sum += s_util.s;
mlen = 0;
} else
mlen = -1;
} else if (mlen == -1)
s_util.c[0] = *(char *)w;
}
if (len)
printf("cksum: out of data by %d\n", len);
if (mlen == -1) {
/* The last mbuf has odd # of bytes. Follow the
standard (the odd byte may be shifted left by 8 bits
or not as determined by endian-ness of the machine) */
s_util.c[1] = 0;
sum += s_util.s;
}
REDUCE;
KERNEL_DEBUG(DBG_FNC_IN_CKSUM | DBG_FUNC_END, 0,0,0,0,0);
return (~sum & 0xffff);
}
