int
in6_pcbbind(
struct inpcb *inp,
struct sockaddr *nam,
struct proc *p)
{
struct socket *so = inp->inp_socket;
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
u_short lport = 0;
int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
if (!in6_ifaddrs) /* XXX broken! */
return (EADDRNOTAVAIL);
if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
return(EINVAL);
if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
wild = 1;
socket_unlock(so, 0); /* keep reference */
lck_rw_lock_exclusive(pcbinfo->mtx);
if (nam) {
sin6 = (struct sockaddr_in6 *)nam;
if (nam->sa_len != sizeof(*sin6)) {
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return(EINVAL);
}
/*
* family check.
*/
if (nam->sa_family != AF_INET6) {
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return(EAFNOSUPPORT);
}
/* KAME hack: embed scopeid */
if (in6_embedscope(&sin6->sin6_addr, sin6, inp, NULL) != 0) {
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return EINVAL;
}
/* this must be cleared for ifa_ifwithaddr() */
sin6->sin6_scope_id = 0;
lport = sin6->sin6_port;
if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
/*
* Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
* allow compepte duplication of binding if
* SO_REUSEPORT is set, or if SO_REUSEADDR is set
* and a multicast address is bound on both
* new and duplicated sockets.
*/
if (so->so_options & SO_REUSEADDR)
reuseport = SO_REUSEADDR|SO_REUSEPORT;
} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
struct ifaddr *ia = NULL;
sin6->sin6_port = 0; /* yech... */
if ((ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == 0) {
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return(EADDRNOTAVAIL);
}
/*
* XXX: bind to an anycast address might accidentally
* cause sending a packet with anycast source address.
* We should allow to bind to a deprecated address, since
* the application dare to use it.
*/
if (ia &&
((struct in6_ifaddr *)ia)->ia6_flags &
(IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
ifafree(ia);
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return(EADDRNOTAVAIL);
}
ifafree(ia);
ia = NULL;
}
if (lport) {
struct inpcb *t;
/* GROSS */
if (ntohs(lport) < IPV6PORT_RESERVED && p &&
((so->so_state & SS_PRIV) == 0)) {
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return(EACCES);
}
if (so->so_uid &&
!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
t = in6_pcblookup_local_and_cleanup(pcbinfo,
&sin6->sin6_addr, lport,
INPLOOKUP_WILDCARD);
if (t &&
(!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
!IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
(t->inp_socket->so_options &
SO_REUSEPORT) == 0) &&
(so->so_uid != t->inp_socket->so_uid) &&
((t->inp_socket->so_flags & SOF_REUSESHAREUID) == 0)) {
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return (EADDRINUSE);
}
if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
struct sockaddr_in sin;
in6_sin6_2_sin(&sin, sin6);
t = in_pcblookup_local_and_cleanup(pcbinfo,
sin.sin_addr, lport,
INPLOOKUP_WILDCARD);
if (t && (t->inp_socket->so_options & SO_REUSEPORT) == 0 &&
(so->so_uid !=
t->inp_socket->so_uid) &&
(ntohl(t->inp_laddr.s_addr) !=
INADDR_ANY ||
INP_SOCKAF(so) ==
INP_SOCKAF(t->inp_socket))) {
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return (EADDRINUSE);
}
}
}
t = in6_pcblookup_local_and_cleanup(pcbinfo, &sin6->sin6_addr,
lport, wild);
if (t && (reuseport & t->inp_socket->so_options) == 0) {
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return(EADDRINUSE);
}
if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
struct sockaddr_in sin;
in6_sin6_2_sin(&sin, sin6);
t = in_pcblookup_local_and_cleanup(pcbinfo, sin.sin_addr,
lport, wild);
if (t &&
(reuseport & t->inp_socket->so_options)
== 0 &&
(ntohl(t->inp_laddr.s_addr)
!= INADDR_ANY ||
INP_SOCKAF(so) ==
INP_SOCKAF(t->inp_socket))) {
lck_rw_done(pcbinfo->mtx);
socket_lock(so, 0);
return (EADDRINUSE);
}
}
}
inp->in6p_laddr = sin6->sin6_addr;
}
socket_lock(so, 0);
if (lport == 0) {
int e;
if ((e = in6_pcbsetport(&inp->in6p_laddr, inp, p, 1)) != 0) {
lck_rw_done(pcbinfo->mtx);
return(e);
}
}
else {
inp->inp_lport = lport;
if (in_pcbinshash(inp, 1) != 0) {
inp->in6p_laddr = in6addr_any;
inp->inp_lport = 0;
lck_rw_done(pcbinfo->mtx);
return (EAGAIN);
}
}
lck_rw_done(pcbinfo->mtx);
sflt_notify(so, sock_evt_bound, NULL);
return(0);
}
int
udp6_output(struct in6pcb *in6p, struct mbuf *m, struct sockaddr *addr6,
struct mbuf *control, struct thread *td)
{
u_int32_t ulen = m->m_pkthdr.len;
u_int32_t plen = sizeof(struct udphdr) + ulen;
struct ip6_hdr *ip6;
struct udphdr *udp6;
struct in6_addr *laddr, *faddr;
u_short fport;
int error = 0;
struct ip6_pktopts opt, *stickyopt = in6p->in6p_outputopts;
int priv;
int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
int flags;
struct sockaddr_in6 tmp;
priv = !priv_check(td, PRIV_ROOT); /* 1 if privileged, 0 if not */
if (control) {
if ((error = ip6_setpktoptions(control, &opt,
in6p->in6p_outputopts,
IPPROTO_UDP, priv)) != 0)
goto release;
in6p->in6p_outputopts = &opt;
}
if (addr6) {
/*
* IPv4 version of udp_output calls in_pcbconnect in this case,
* which needs splnet and affects performance.
* Since we saw no essential reason for calling in_pcbconnect,
* we get rid of such kind of logic, and call in6_selectsrc
* and in6_pcbsetport in order to fill in the local address
* and the local port.
*/
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)addr6;
if (sin6->sin6_port == 0) {
error = EADDRNOTAVAIL;
goto release;
}
if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
/* how about ::ffff:0.0.0.0 case? */
error = EISCONN;
goto release;
}
if (!prison_remote_ip(td, addr6)) {
error = EAFNOSUPPORT; /* IPv4 only jail */
goto release;
}
/* protect *sin6 from overwrites */
tmp = *sin6;
sin6 = &tmp;
faddr = &sin6->sin6_addr;
fport = sin6->sin6_port; /* allow 0 port */
if (IN6_IS_ADDR_V4MAPPED(faddr)) {
if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY)) {
/*
* I believe we should explicitly discard the
* packet when mapped addresses are disabled,
* rather than send the packet as an IPv6 one.
* If we chose the latter approach, the packet
* might be sent out on the wire based on the
* default route, the situation which we'd
* probably want to avoid.
* (20010421 [email protected])
*/
error = EINVAL;
goto release;
} else
af = AF_INET;
}
/* KAME hack: embed scopeid */
if (in6_embedscope(&sin6->sin6_addr, sin6, in6p, NULL) != 0) {
error = EINVAL;
goto release;
}
if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
laddr = in6_selectsrc(sin6, in6p->in6p_outputopts,
in6p->in6p_moptions,
&in6p->in6p_route,
&in6p->in6p_laddr, &error, NULL);
} else
laddr = &in6p->in6p_laddr; /* XXX */
if (laddr == NULL) {
if (error == 0)
error = EADDRNOTAVAIL;
goto release;
}
if (in6p->in6p_lport == 0 &&
(error = in6_pcbsetport(laddr, in6p, td)) != 0)
goto release;
} else {
if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
error = ENOTCONN;
goto release;
}
if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY)) {
/*
* XXX: this case would happen when the
* application sets the V6ONLY flag after
* connecting the foreign address.
* Such applications should be fixed,
* so we bark here.
*/
log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
"option was set for a connected socket\n");
error = EINVAL;
goto release;
} else
af = AF_INET;
}
laddr = &in6p->in6p_laddr;
faddr = &in6p->in6p_faddr;
fport = in6p->in6p_fport;
}
if (af == AF_INET)
hlen = sizeof(struct ip);
/*
* Calculate data length and get a mbuf
* for UDP and IP6 headers.
*/
M_PREPEND(m, hlen + sizeof(struct udphdr), MB_DONTWAIT);
if (m == NULL) {
error = ENOBUFS;
goto release;
}
/*
* Stuff checksum and output datagram.
*/
udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
udp6->uh_sport = in6p->in6p_lport; /* lport is always set in the PCB */
udp6->uh_dport = fport;
if (plen <= 0xffff)
udp6->uh_ulen = htons((u_short)plen);
else
udp6->uh_ulen = 0;
udp6->uh_sum = 0;
switch (af) {
case AF_INET6:
ip6 = mtod(m, struct ip6_hdr *);
ip6->ip6_flow = in6p->in6p_flowinfo & IPV6_FLOWINFO_MASK;
ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
ip6->ip6_vfc |= IPV6_VERSION;
#if 0 /* ip6_plen will be filled in ip6_output. */
ip6->ip6_plen = htons((u_short)plen);
#endif
ip6->ip6_nxt = IPPROTO_UDP;
ip6->ip6_hlim = in6_selecthlim(in6p,
in6p->in6p_route.ro_rt ?
in6p->in6p_route.ro_rt->rt_ifp : NULL);
ip6->ip6_src = *laddr;
ip6->ip6_dst = *faddr;
if ((udp6->uh_sum = in6_cksum(m, IPPROTO_UDP,
sizeof(struct ip6_hdr), plen)) == 0) {
udp6->uh_sum = 0xffff;
}
flags = 0;
udp6stat.udp6s_opackets++;
error = ip6_output(m, in6p->in6p_outputopts, &in6p->in6p_route,
flags, in6p->in6p_moptions, NULL, in6p);
break;
case AF_INET:
error = EAFNOSUPPORT;
goto release;
}
goto releaseopt;
release:
m_freem(m);
releaseopt:
if (control) {
ip6_clearpktopts(in6p->in6p_outputopts, -1);
in6p->in6p_outputopts = stickyopt;
m_freem(control);
}
return (error);
}
/*
* Bind an address (or at least a port) to an PF_INET6 socket.
*/
int
in6_pcbbind(struct inpcb *inp, struct mbuf *nam, struct proc *p)
{
struct socket *so = inp->inp_socket;
struct inpcbtable *head = inp->inp_table;
struct sockaddr_in6 *sin6;
u_short lport = 0;
int wild = INPLOOKUP_IPV6, reuseport = (so->so_options & SO_REUSEPORT);
int error;
/*
* REMINDER: Once up to speed, flow label processing should go here,
* too. (Same with in6_pcbconnect.)
*/
if (in6_ifaddr == 0)
return EADDRNOTAVAIL;
if (inp->inp_lport != 0 || !IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6))
return EINVAL; /* If already bound, EINVAL! */
if ((so->so_options & (SO_REUSEADDR | SO_REUSEPORT)) == 0 &&
((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
(so->so_options & SO_ACCEPTCONN) == 0))
wild |= INPLOOKUP_WILDCARD;
/*
* If I did get a sockaddr passed in...
*/
if (nam) {
sin6 = mtod(nam, struct sockaddr_in6 *);
if (nam->m_len != sizeof (*sin6))
return EINVAL;
/*
* Unlike v4, I have no qualms about EAFNOSUPPORT if the
* wretched family is not filled in!
*/
if (sin6->sin6_family != AF_INET6)
return EAFNOSUPPORT;
/* KAME hack: embed scopeid */
if (in6_embedscope(&sin6->sin6_addr, sin6, inp, NULL) != 0)
return EINVAL;
/* this must be cleared for ifa_ifwithaddr() */
sin6->sin6_scope_id = 0;
lport = sin6->sin6_port;
/* reject IPv4 mapped address, we have no support for it */
if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
return EADDRNOTAVAIL;
if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
/*
* Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
* allow complete duplication of binding if
* SO_REUSEPORT is set, or if SO_REUSEADDR is set
* and a multicast address is bound on both
* new and duplicated sockets.
*/
if (so->so_options & SO_REUSEADDR)
reuseport = SO_REUSEADDR | SO_REUSEPORT;
} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
struct ifaddr *ia = NULL;
sin6->sin6_port = 0; /*
* Yechhhh, because of upcoming
* call to ifa_ifwithaddr(), which
* does bcmp's over the PORTS as
* well. (What about flow?)
*/
sin6->sin6_flowinfo = 0;
if (!(so->so_options & SO_BINDANY) &&
((ia = ifa_ifwithaddr((struct sockaddr *)sin6))
== NULL))
return EADDRNOTAVAIL;
/*
* bind to an anycast address might accidentally
* cause sending a packet with an anycast source
* address, so we forbid it.
*
* We should allow to bind to a deprecated address,
* since the application dare to use it.
* But, can we assume that they are careful enough
* to check if the address is deprecated or not?
* Maybe, as a safeguard, we should have a setsockopt
* flag to control the bind(2) behavior against
* deprecated addresses (default: forbid bind(2)).
*/
if (ia &&
((struct in6_ifaddr *)ia)->ia6_flags &
(IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED))
return (EADDRNOTAVAIL);
}
if (lport) {
struct inpcb *t;
/*
* Question: Do we wish to continue the Berkeley
* tradition of ports < IPPORT_RESERVED be only for
* root?
* Answer: For now yes, but IMHO, it should be REMOVED!
* OUCH: One other thing, is there no better way of
* finding a process for a socket instead of using
* curproc? (Marked with BSD's {in,}famous XXX ?
*/
if (ntohs(lport) < IPPORT_RESERVED &&
(error = suser(p, 0)))
return error;
t = in_pcblookup(head,
(struct in_addr *)&zeroin6_addr, 0,
(struct in_addr *)&sin6->sin6_addr, lport,
wild);
if (t && (reuseport & t->inp_socket->so_options) == 0)
return EADDRINUSE;
}
inp->inp_laddr6 = sin6->sin6_addr;
}
if (lport == 0) {
error = in6_pcbsetport(&inp->inp_laddr6, inp, p);
if (error != 0)
return error;
} else {
inp->inp_lport = lport;
in_pcbrehash(inp);
}
return 0;
}
udp6_output(struct in6pcb *in6p, struct mbuf *m, struct mbuf *addr6,
struct mbuf *control)
#endif
{
u_int32_t plen = sizeof(struct udphdr) + m->m_pkthdr.len;
struct ip6_hdr *ip6;
struct udphdr *udp6;
struct in6_addr *laddr6 = NULL, *faddr6 = NULL;
struct sockaddr_in6 *fsa6 = NULL;
struct ifnet *oifp = NULL;
int scope_ambiguous = 0;
#ifndef __FreeBSD__
struct sockaddr_in6 lsa6_mapped; /* XXX ugly */
#endif
u_int16_t fport;
int error = 0;
struct ip6_pktopts *optp, opt;
int priv;
int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
#ifdef INET
#if defined(__NetBSD__)
struct ip *ip;
struct udpiphdr *ui;
#endif
#endif
int flags = 0;
struct sockaddr_in6 tmp;
#if defined(__OpenBSD__)
struct proc *p = curproc; /* XXX */
#endif
priv = 0;
#if defined(__NetBSD__)
if (p && !suser(p->p_ucred, &p->p_acflag))
priv = 1;
#elif defined(__FreeBSD__)
if (p && !suser(p))
priv = 1;
#else
if ((in6p->in6p_socket->so_state & SS_PRIV) != 0)
priv = 1;
#endif
if (addr6) {
#ifdef __FreeBSD__
/* addr6 has been validated in udp6_send(). */
fsa6 = (struct sockaddr_in6 *)addr6;
#else
fsa6 = mtod(addr6, struct sockaddr_in6 *);
if (addr6->m_len != sizeof(*fsa6))
return (EINVAL);
if (fsa6->sin6_family != AF_INET6)
return (EAFNOSUPPORT);
#endif
/* protect *sin6 from overwrites */
tmp = *fsa6;
fsa6 = &tmp;
/*
* Application should provide a proper zone ID or the use of
* default zone IDs should be enabled. Unfortunately, some
* applications do not behave as it should, so we need a
* workaround. Even if an appropriate ID is not determined,
* we'll see if we can determine the outgoing interface. If we
* can, determine the zone ID based on the interface below.
*/
if (fsa6->sin6_scope_id == 0 && !ip6_use_defzone)
scope_ambiguous = 1;
if ((error = sa6_embedscope(fsa6, ip6_use_defzone)) != 0)
return (error);
}
if (control) {
if ((error = ip6_setpktopts(control, &opt,
in6p->in6p_outputopts, priv, IPPROTO_UDP)) != 0)
goto release;
optp = &opt;
} else
optp = in6p->in6p_outputopts;
if (fsa6) {
faddr6 = &fsa6->sin6_addr;
/*
* IPv4 version of udp_output calls in_pcbconnect in this case,
* which needs splnet and affects performance.
* Since we saw no essential reason for calling in_pcbconnect,
* we get rid of such kind of logic, and call in6_selectsrc
* and in6_pcbsetport in order to fill in the local address
* and the local port.
*/
if (fsa6->sin6_port == 0) {
error = EADDRNOTAVAIL;
goto release;
}
if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
/* how about ::ffff:0.0.0.0 case? */
error = EISCONN;
goto release;
}
fport = fsa6->sin6_port; /* allow 0 port */
if (IN6_IS_ADDR_V4MAPPED(faddr6)) {
#ifdef __OpenBSD__ /* does not support mapped addresses */
if (1)
#else
if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY))
#endif
{
/*
* I believe we should explicitly discard the
* packet when mapped addresses are disabled,
* rather than send the packet as an IPv6 one.
* If we chose the latter approach, the packet
* might be sent out on the wire based on the
* default route, the situation which we'd
* probably want to avoid.
* (20010421 [email protected])
*/
error = EINVAL;
goto release;
}
if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
!IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) {
/*
* when remote addr is an IPv4-mapped address,
* local addr should not be an IPv6 address,
* since you cannot determine how to map IPv6
* source address to IPv4.
*/
error = EINVAL;
goto release;
}
af = AF_INET;
}
if (!IN6_IS_ADDR_V4MAPPED(faddr6)) {
laddr6 = in6_selectsrc(fsa6, optp,
in6p->in6p_moptions, &in6p->in6p_route,
&in6p->in6p_laddr, &oifp, &error);
if (oifp && scope_ambiguous &&
(error = in6_setscope(&fsa6->sin6_addr,
oifp, NULL))) {
goto release;
}
} else {
#ifndef __FreeBSD__
/*
* XXX: freebsd[34] does not have in_selectsrc, but
* we can omit the whole part because freebsd4 calls
* udp_output() directly in this case, and thus we'll
* never see this path.
*/
if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
struct sockaddr_in *sinp, sin_dst;
bzero(&sin_dst, sizeof(sin_dst));
sin_dst.sin_family = AF_INET;
sin_dst.sin_len = sizeof(sin_dst);
bcopy(&fsa6->sin6_addr.s6_addr[12],
&sin_dst.sin_addr,
sizeof(sin_dst.sin_addr));
sinp = in_selectsrc(&sin_dst,
(struct route *)&in6p->in6p_route,
in6p->in6p_socket->so_options,
NULL, &error);
if (sinp == NULL) {
if (error == 0)
error = EADDRNOTAVAIL;
goto release;
}
bzero(&lsa6_mapped, sizeof(lsa6_mapped));
lsa6_mapped.sin6_family = AF_INET6;
lsa6_mapped.sin6_len = sizeof(lsa6_mapped);
/* ugly */
lsa6_mapped.sin6_addr.s6_addr16[5] = 0xffff;
bcopy(&sinp->sin_addr,
&lsa6_mapped.sin6_addr.s6_addr[12],
sizeof(sinp->sin_addr));
laddr6 = &lsa6_mapped.sin6_addr;
} else
#endif /* !freebsd */
{
laddr6 = &in6p->in6p_laddr;
}
}
if (laddr6 == NULL) {
if (error == 0)
error = EADDRNOTAVAIL;
goto release;
}
if (in6p->in6p_lport == 0 &&
(error = in6_pcbsetport(laddr6, in6p,
#ifdef __FreeBSD__
p->td_ucred
#else
p
#endif
)) != 0)
goto release;
} else {
if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
error = ENOTCONN;
goto release;
}
if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
#ifdef __OpenBSD__ /* does not support mapped addresses */
if (1)
#else
if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY))
#endif
{
/*
* XXX: this case would happen when the
* application sets the V6ONLY flag after
* connecting the foreign address.
* Such applications should be fixed,
* so we bark here.
*/
log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
"option was set for a connected socket\n");
error = EINVAL;
goto release;
} else
af = AF_INET;
}
laddr6 = &in6p->in6p_laddr;
faddr6 = &in6p->in6p_faddr;
fport = in6p->in6p_fport;
}
if (af == AF_INET)
hlen = sizeof(struct ip);
/*
* Calculate data length and get a mbuf
* for UDP and IP6 headers.
*/
M_PREPEND(m, hlen + sizeof(struct udphdr), M_DONTWAIT);
if (m == 0) {
error = ENOBUFS;
goto release;
}
/*
* Stuff checksum and output datagram.
*/
udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
udp6->uh_sport = in6p->in6p_lport; /* lport is always set in the PCB */
udp6->uh_dport = fport;
if (plen <= 0xffff)
udp6->uh_ulen = htons((u_int16_t)plen);
else
udp6->uh_ulen = 0;
udp6->uh_sum = 0;
switch (af) {
case AF_INET6:
ip6 = mtod(m, struct ip6_hdr *);
ip6->ip6_flow = in6p->in6p_flowinfo & IPV6_FLOWINFO_MASK;
ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
ip6->ip6_vfc |= IPV6_VERSION;
#if 0 /* ip6_plen will be filled in ip6_output. */
ip6->ip6_plen = htons((u_int16_t)plen);
#endif
ip6->ip6_nxt = IPPROTO_UDP;
ip6->ip6_hlim = in6_selecthlim(in6p, oifp);
ip6->ip6_src = *laddr6;
ip6->ip6_dst = *faddr6;
if ((udp6->uh_sum = in6_cksum(m, IPPROTO_UDP,
sizeof(struct ip6_hdr), plen)) == 0) {
udp6->uh_sum = 0xffff;
}
udp6stat.udp6s_opackets++;
#ifdef IPSEC
if (ipsec_setsocket(m, in6p->in6p_socket) != 0) {
error = ENOBUFS;
goto release;
}
#endif /* IPSEC */
#ifdef __FreeBSD__
error = ip6_output(m, optp, &in6p->in6p_route,
flags, in6p->in6p_moptions, NULL, NULL);
#elif defined(__NetBSD__)
error = ip6_output(m, optp, &in6p->in6p_route,
flags, in6p->in6p_moptions, in6p->in6p_socket, NULL);
#else
error = ip6_output(m, optp, &in6p->in6p_route,
flags, in6p->in6p_moptions, NULL);
#endif
break;
case AF_INET:
#if defined(INET) && defined(__NetBSD__)
/* can't transmit jumbogram over IPv4 */
if (plen > 0xffff) {
error = EMSGSIZE;
goto release;
}
ip = mtod(m, struct ip *);
ui = (struct udpiphdr *)ip;
bzero(ui->ui_x1, sizeof ui->ui_x1);
ui->ui_pr = IPPROTO_UDP;
ui->ui_len = htons(plen);
bcopy(&laddr6->s6_addr[12], &ui->ui_src, sizeof(ui->ui_src));
ui->ui_ulen = ui->ui_len;
#ifdef __NetBSD__
flags = (in6p->in6p_socket->so_options &
(SO_DONTROUTE | SO_BROADCAST));
bcopy(&fsa6->sin6_addr.s6_addr[12],
&ui->ui_dst, sizeof(ui->ui_dst));
udp6->uh_sum = in_cksum(m, hlen + plen);
#endif
if (udp6->uh_sum == 0)
udp6->uh_sum = 0xffff;
ip->ip_len = hlen + plen;
ip->ip_ttl = in6_selecthlim(in6p, NULL); /* XXX */
ip->ip_tos = 0; /* XXX */
ip->ip_len = hlen + plen; /* XXX */
udpstat.udps_opackets++;
#ifdef IPSEC
(void)ipsec_setsocket(m, NULL); /* XXX */
#endif /* IPSEC */
#ifdef __NetBSD__
error = ip_output(m, NULL, &in6p->in6p_route, flags /* XXX */);
#endif
break;
#else
error = EAFNOSUPPORT;
goto release;
#endif
}
goto releaseopt;
release:
m_freem(m);
releaseopt:
if (control) {
ip6_clearpktopts(&opt, -1);
m_freem(control);
}
return (error);
}
static int
udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
struct mbuf *control, struct thread *td)
{
u_int32_t ulen = m->m_pkthdr.len;
u_int32_t plen = sizeof(struct udphdr) + ulen;
struct ip6_hdr *ip6;
struct udphdr *udp6;
struct in6_addr *laddr, *faddr, in6a;
struct sockaddr_in6 *sin6 = NULL;
int cscov_partial = 0;
int scope_ambiguous = 0;
u_short fport;
int error = 0;
uint8_t nxt;
uint16_t cscov = 0;
struct ip6_pktopts *optp, opt;
int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
int flags;
struct sockaddr_in6 tmp;
INP_WLOCK_ASSERT(inp);
INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
if (addr6) {
/* addr6 has been validated in udp6_send(). */
sin6 = (struct sockaddr_in6 *)addr6;
/* protect *sin6 from overwrites */
tmp = *sin6;
sin6 = &tmp;
/*
* Application should provide a proper zone ID or the use of
* default zone IDs should be enabled. Unfortunately, some
* applications do not behave as it should, so we need a
* workaround. Even if an appropriate ID is not determined,
* we'll see if we can determine the outgoing interface. If we
* can, determine the zone ID based on the interface below.
*/
if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
scope_ambiguous = 1;
if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
return (error);
}
nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
IPPROTO_UDP : IPPROTO_UDPLITE;
if (control) {
if ((error = ip6_setpktopts(control, &opt,
inp->in6p_outputopts, td->td_ucred, nxt)) != 0)
goto release;
optp = &opt;
} else
optp = inp->in6p_outputopts;
if (sin6) {
faddr = &sin6->sin6_addr;
/*
* Since we saw no essential reason for calling in_pcbconnect,
* we get rid of such kind of logic, and call in6_selectsrc
* and in6_pcbsetport in order to fill in the local address
* and the local port.
*/
if (sin6->sin6_port == 0) {
error = EADDRNOTAVAIL;
goto release;
}
if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
/* how about ::ffff:0.0.0.0 case? */
error = EISCONN;
goto release;
}
fport = sin6->sin6_port; /* allow 0 port */
if (IN6_IS_ADDR_V4MAPPED(faddr)) {
if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
/*
* I believe we should explicitly discard the
* packet when mapped addresses are disabled,
* rather than send the packet as an IPv6 one.
* If we chose the latter approach, the packet
* might be sent out on the wire based on the
* default route, the situation which we'd
* probably want to avoid.
* (20010421 [email protected])
*/
error = EINVAL;
goto release;
}
if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
!IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
/*
* when remote addr is an IPv4-mapped address,
* local addr should not be an IPv6 address,
* since you cannot determine how to map IPv6
* source address to IPv4.
*/
error = EINVAL;
goto release;
}
af = AF_INET;
}
if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
error = in6_selectsrc_socket(sin6, optp, inp,
td->td_ucred, scope_ambiguous, &in6a, NULL);
if (error)
goto release;
laddr = &in6a;
} else
laddr = &inp->in6p_laddr; /* XXX */
if (laddr == NULL) {
if (error == 0)
error = EADDRNOTAVAIL;
goto release;
}
if (inp->inp_lport == 0 &&
(error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) {
/* Undo an address bind that may have occurred. */
inp->in6p_laddr = in6addr_any;
goto release;
}
} else {
if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
error = ENOTCONN;
goto release;
}
if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
/*
* XXX: this case would happen when the
* application sets the V6ONLY flag after
* connecting the foreign address.
* Such applications should be fixed,
* so we bark here.
*/
log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
"option was set for a connected socket\n");
error = EINVAL;
goto release;
} else
af = AF_INET;
}
laddr = &inp->in6p_laddr;
faddr = &inp->in6p_faddr;
fport = inp->inp_fport;
}
if (af == AF_INET)
hlen = sizeof(struct ip);
/*
* Calculate data length and get a mbuf
* for UDP and IP6 headers.
*/
M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
if (m == NULL) {
error = ENOBUFS;
goto release;
}
/*
* Stuff checksum and output datagram.
*/
udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
udp6->uh_dport = fport;
if (nxt == IPPROTO_UDPLITE) {
struct udpcb *up;
up = intoudpcb(inp);
cscov = up->u_txcslen;
if (cscov >= plen)
cscov = 0;
udp6->uh_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 if (plen <= 0xffff)
udp6->uh_ulen = htons((u_short)plen);
else
udp6->uh_ulen = 0;
udp6->uh_sum = 0;
switch (af) {
case AF_INET6:
ip6 = mtod(m, struct ip6_hdr *);
ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
ip6->ip6_vfc |= IPV6_VERSION;
ip6->ip6_plen = htons((u_short)plen);
ip6->ip6_nxt = nxt;
ip6->ip6_hlim = in6_selecthlim(inp, NULL);
ip6->ip6_src = *laddr;
ip6->ip6_dst = *faddr;
if (cscov_partial) {
if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
sizeof(struct ip6_hdr), plen, cscov)) == 0)
udp6->uh_sum = 0xffff;
} else {
udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
}
#ifdef RSS
{
uint32_t hash_val, hash_type;
uint8_t pr;
pr = inp->inp_socket->so_proto->pr_protocol;
/*
* Calculate an appropriate RSS hash for UDP and
* UDP Lite.
*
* The called function will take care of figuring out
* whether a 2-tuple or 4-tuple hash is required based
* on the currently configured scheme.
*
* Later later on connected socket values should be
* cached in the inpcb and reused, rather than constantly
* re-calculating it.
*
* UDP Lite is a different protocol number and will
* likely end up being hashed as a 2-tuple until
* RSS / NICs grow UDP Lite protocol awareness.
*/
if (rss_proto_software_hash_v6(faddr, laddr, fport,
inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
m->m_pkthdr.flowid = hash_val;
M_HASHTYPE_SET(m, hash_type);
}
}
#endif
flags = 0;
#ifdef RSS
/*
* Don't override with the inp cached flowid.
*
* Until the whole UDP path is vetted, it may actually
* be incorrect.
*/
flags |= IP_NODEFAULTFLOWID;
#endif
UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
UDPSTAT_INC(udps_opackets);
error = ip6_output(m, optp, &inp->inp_route6, flags,
inp->in6p_moptions, NULL, inp);
break;
case AF_INET:
error = EAFNOSUPPORT;
goto release;
}
goto releaseopt;
release:
m_freem(m);
releaseopt:
if (control) {
ip6_clearpktopts(&opt, -1);
m_freem(control);
}
return (error);
}
/*
* Bind port from sin6 to in6p.
*/
static int
in6_pcbbind_port(struct in6pcb *in6p, struct sockaddr_in6 *sin6, struct lwp *l)
{
struct inpcbtable *table = in6p->in6p_table;
struct socket *so = in6p->in6p_socket;
int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
int error;
if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
(so->so_options & SO_ACCEPTCONN) == 0))
wild = 1;
if (sin6->sin6_port != 0) {
enum kauth_network_req req;
#ifndef IPNOPRIVPORTS
if (ntohs(sin6->sin6_port) < IPV6PORT_RESERVED)
req = KAUTH_REQ_NETWORK_BIND_PRIVPORT;
else
#endif /* IPNOPRIVPORTS */
req = KAUTH_REQ_NETWORK_BIND_PORT;
error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_BIND,
req, so, sin6, NULL);
if (error)
return (EACCES);
}
if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
/*
* Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
* allow compepte duplication of binding if
* SO_REUSEPORT is set, or if SO_REUSEADDR is set
* and a multicast address is bound on both
* new and duplicated sockets.
*/
if (so->so_options & SO_REUSEADDR)
reuseport = SO_REUSEADDR|SO_REUSEPORT;
}
if (sin6->sin6_port != 0) {
if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
#ifdef INET
struct inpcb *t;
struct vestigial_inpcb vestige;
t = in_pcblookup_port(table,
*(struct in_addr *)&sin6->sin6_addr.s6_addr32[3],
sin6->sin6_port, wild, &vestige);
if (t && (reuseport & t->inp_socket->so_options) == 0)
return (EADDRINUSE);
if (!t
&& vestige.valid
&& !(reuseport && vestige.reuse_port))
return EADDRINUSE;
#else
return (EADDRNOTAVAIL);
#endif
}
{
struct in6pcb *t;
struct vestigial_inpcb vestige;
t = in6_pcblookup_port(table, &sin6->sin6_addr,
sin6->sin6_port, wild, &vestige);
if (t && (reuseport & t->in6p_socket->so_options) == 0)
return (EADDRINUSE);
if (!t
&& vestige.valid
&& !(reuseport && vestige.reuse_port))
return EADDRINUSE;
}
}
if (sin6->sin6_port == 0) {
int e;
e = in6_pcbsetport(sin6, in6p, l);
if (e != 0)
return (e);
} else {
in6p->in6p_lport = sin6->sin6_port;
in6_pcbstate(in6p, IN6P_BOUND);
}
LIST_REMOVE(&in6p->in6p_head, inph_lhash);
LIST_INSERT_HEAD(IN6PCBHASH_PORT(table, in6p->in6p_lport),
&in6p->in6p_head, inph_lhash);
return (0);
}
int
in6_pcbbind(void *v, struct mbuf *nam, struct lwp *l)
{
struct in6pcb *in6p = v;
struct socket *so = in6p->in6p_socket;
struct inpcbtable *table = in6p->in6p_table;
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
u_int16_t lport = 0;
int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
if (in6p->in6p_af != AF_INET6)
return (EINVAL);
if (in6p->in6p_lport || !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
return (EINVAL);
if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
(so->so_options & SO_ACCEPTCONN) == 0))
wild = 1;
if (nam) {
int error;
sin6 = mtod(nam, struct sockaddr_in6 *);
if (nam->m_len != sizeof(*sin6))
return (EINVAL);
/*
* We should check the family, but old programs
* incorrectly fail to intialize it.
*/
if (sin6->sin6_family != AF_INET6)
return (EAFNOSUPPORT);
#ifndef INET
if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
return (EADDRNOTAVAIL);
#endif
if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
return (error);
lport = sin6->sin6_port;
if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
/*
* Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
* allow compepte duplication of binding if
* SO_REUSEPORT is set, or if SO_REUSEADDR is set
* and a multicast address is bound on both
* new and duplicated sockets.
*/
if (so->so_options & SO_REUSEADDR)
reuseport = SO_REUSEADDR|SO_REUSEPORT;
} else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
return (EINVAL);
if (sin6->sin6_addr.s6_addr32[3]) {
struct sockaddr_in sin;
bzero(&sin, sizeof(sin));
sin.sin_len = sizeof(sin);
sin.sin_family = AF_INET;
bcopy(&sin6->sin6_addr.s6_addr32[3],
&sin.sin_addr, sizeof(sin.sin_addr));
if (ifa_ifwithaddr((struct sockaddr *)&sin) == 0)
return EADDRNOTAVAIL;
}
} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
struct ifaddr *ia = NULL;
sin6->sin6_port = 0; /* yech... */
if ((in6p->in6p_flags & IN6P_FAITH) == 0 &&
(ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == 0)
return (EADDRNOTAVAIL);
/*
* bind to an anycast address might accidentally
* cause sending a packet with an anycast source
* address, so we forbid it.
*
* We should allow to bind to a deprecated address,
* since the application dare to use it.
* But, can we assume that they are careful enough
* to check if the address is deprecated or not?
* Maybe, as a safeguard, we should have a setsockopt
* flag to control the bind(2) behavior against
* deprecated addresses (default: forbid bind(2)).
*/
if (ia &&
((struct in6_ifaddr *)ia)->ia6_flags &
(IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED))
return (EADDRNOTAVAIL);
}
if (lport) {
#ifndef IPNOPRIVPORTS
int priv;
/*
* NOTE: all operating systems use suser() for
* privilege check! do not rewrite it into SS_PRIV.
*/
priv = (l && !kauth_authorize_generic(l->l_cred,
KAUTH_GENERIC_ISSUSER, NULL)) ? 1 : 0;
/* GROSS */
if (ntohs(lport) < IPV6PORT_RESERVED && !priv)
return (EACCES);
#endif
if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
#ifdef INET
struct inpcb *t;
t = in_pcblookup_port(table,
*(struct in_addr *)&sin6->sin6_addr.s6_addr32[3],
lport, wild);
if (t && (reuseport & t->inp_socket->so_options) == 0)
return (EADDRINUSE);
#else
return (EADDRNOTAVAIL);
#endif
}
{
struct in6pcb *t;
t = in6_pcblookup_port(table, &sin6->sin6_addr,
lport, wild);
if (t && (reuseport & t->in6p_socket->so_options) == 0)
return (EADDRINUSE);
}
}
in6p->in6p_laddr = sin6->sin6_addr;
}
if (lport == 0) {
int e;
e = in6_pcbsetport(&in6p->in6p_laddr, in6p, l);
if (e != 0)
return (e);
} else {
in6p->in6p_lport = lport;
in6_pcbstate(in6p, IN6P_BOUND);
}
LIST_REMOVE(&in6p->in6p_head, inph_lhash);
LIST_INSERT_HEAD(IN6PCBHASH_PORT(table, in6p->in6p_lport),
&in6p->in6p_head, inph_lhash);
#if 0
in6p->in6p_flowinfo = 0; /* XXX */
#endif
return (0);
}
