/*
* Outer subroutine:
* Connect from a socket to a specified address.
* Both address and port must be specified in argument sin.
* If don't have a local address for this socket yet,
* then pick one.
*/
int
in6_pcbconnect(
struct inpcb *inp,
struct sockaddr *nam,
struct proc *p)
{
struct in6_addr addr6;
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
struct inpcb *pcb;
int error;
unsigned int outif = 0;
/*
* Call inner routine, to assign local interface address.
* in6_pcbladdr() may automatically fill in sin6_scope_id.
*/
if ((error = in6_pcbladdr(inp, nam, &addr6, &outif)) != 0)
return(error);
socket_unlock(inp->inp_socket, 0);
pcb = in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr,
sin6->sin6_port,
IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
? &addr6 : &inp->in6p_laddr,
inp->inp_lport, 0, NULL);
socket_lock(inp->inp_socket, 0);
if (pcb != NULL) {
in_pcb_checkstate(pcb, WNT_RELEASE, pcb == inp ? 1 : 0);
return (EADDRINUSE);
}
if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
if (inp->inp_lport == 0) {
error = in6_pcbbind(inp, (struct sockaddr *)0, p);
if (error)
return (error);
}
inp->in6p_laddr = addr6;
inp->in6p_last_outif = outif;
}
if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->mtx)) {
/*lock inversion issue, mostly with udp multicast packets */
socket_unlock(inp->inp_socket, 0);
lck_rw_lock_exclusive(inp->inp_pcbinfo->mtx);
socket_lock(inp->inp_socket, 0);
}
inp->in6p_faddr = sin6->sin6_addr;
inp->inp_fport = sin6->sin6_port;
/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
if (inp->in6p_flags & IN6P_AUTOFLOWLABEL)
inp->in6p_flowinfo |=
(htonl(ip6_flow_seq++) & IPV6_FLOWLABEL_MASK);
in_pcbrehash(inp);
lck_rw_done(inp->inp_pcbinfo->mtx);
return (0);
}
/*
* Vnode op for read
*/
int
fifo_read(struct vnop_read_args *ap)
{
struct uio *uio = ap->a_uio;
struct socket *rso = ap->a_vp->v_fifoinfo->fi_readsock;
user_ssize_t startresid;
int error;
int rflags;
#if DIAGNOSTIC
if (uio->uio_rw != UIO_READ)
panic("fifo_read mode");
#endif
if (uio_resid(uio) == 0)
return (0);
rflags = (ap->a_ioflag & IO_NDELAY) ? MSG_NBIO : 0;
startresid = uio_resid(uio);
/* fifo conformance - if we have a reader open on the fifo but no
* writers then we need to make sure we do not block. We do that by
* checking the receive buffer and if empty set error to EWOULDBLOCK.
* If error is set to EWOULDBLOCK we skip the call into soreceive
*/
error = 0;
if (ap->a_vp->v_fifoinfo->fi_writers < 1) {
socket_lock(rso, 1);
error = (rso->so_rcv.sb_cc == 0) ? EWOULDBLOCK : 0;
socket_unlock(rso, 1);
}
/* skip soreceive to avoid blocking when we have no writers */
if (error != EWOULDBLOCK) {
error = soreceive(rso, (struct sockaddr **)0, uio, (struct mbuf **)0,
(struct mbuf **)0, &rflags);
if (error == 0)
lock_vnode_and_post(ap->a_vp, 0);
}
else {
/* clear EWOULDBLOCK and return EOF (zero) */
error = 0;
}
/*
* Clear EOF indication after first such return.
*/
if (uio_resid(uio) == startresid) {
socket_lock(rso, 1);
rso->so_state &= ~SS_CANTRCVMORE;
socket_unlock(rso, 1);
}
return (error);
}
// ---------------------------------------------------------------------------
/// Waits for send to complete.
// ---------------------------------------------------------------------------
void
syncMesh_waitSend (const connection_t *slot)
{
socket_t *s;
while ((s = socket_seekWait (&slot->channel, mc_channel_outcome)))
socket_unlock (s);
}
int
soo_select(struct fileproc *fp, int which, void *wql, vfs_context_t ctx)
{
struct socket *so = (struct socket *)fp->f_fglob->fg_data;
int retnum = 0;
proc_t procp;
if (so == NULL || so == (struct socket *)-1)
return (0);
procp = vfs_context_proc(ctx);
#if CONFIG_MACF_SOCKET
if (mac_socket_check_select(vfs_context_ucred(ctx), so, which) != 0)
return (0);
#endif /* CONFIG_MACF_SOCKET */
socket_lock(so, 1);
switch (which) {
case FREAD:
so->so_rcv.sb_flags |= SB_SEL;
if (soreadable(so)) {
retnum = 1;
so->so_rcv.sb_flags &= ~SB_SEL;
goto done;
}
selrecord(procp, &so->so_rcv.sb_sel, wql);
break;
case FWRITE:
so->so_snd.sb_flags |= SB_SEL;
if (sowriteable(so)) {
retnum = 1;
so->so_snd.sb_flags &= ~SB_SEL;
goto done;
}
selrecord(procp, &so->so_snd.sb_sel, wql);
break;
case 0:
so->so_rcv.sb_flags |= SB_SEL;
if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) {
retnum = 1;
so->so_rcv.sb_flags &= ~SB_SEL;
goto done;
}
selrecord(procp, &so->so_rcv.sb_sel, wql);
break;
}
done:
socket_unlock(so, 1);
return (retnum);
}
/* You'd certainly better have an iocount on the vnode! */
int
fifo_freespace(struct vnode *vp, long *count)
{
struct socket *rsock;
rsock = vp->v_fifoinfo->fi_readsock;
socket_lock(rsock, 1);
*count = sbspace(&rsock->so_rcv);
socket_unlock(rsock, 1);
return 0;
}
int
soo_stat(struct socket *so, void *ub, int isstat64)
{
int ret;
/* warning avoidance ; protected by isstat64 */
struct stat *sb = (struct stat *)0;
/* warning avoidance ; protected by isstat64 */
struct stat64 *sb64 = (struct stat64 *)0;
#if CONFIG_MACF_SOCKET
ret = mac_socket_check_stat(kauth_cred_get(), so);
if (ret)
return (ret);
#endif
if (isstat64 != 0) {
sb64 = (struct stat64 *)ub;
bzero((caddr_t)sb64, sizeof (*sb64));
} else {
sb = (struct stat *)ub;
bzero((caddr_t)sb, sizeof (*sb));
}
socket_lock(so, 1);
if (isstat64 != 0) {
sb64->st_mode = S_IFSOCK;
if ((so->so_state & SS_CANTRCVMORE) == 0 ||
so->so_rcv.sb_cc != 0)
sb64->st_mode |= S_IRUSR | S_IRGRP | S_IROTH;
if ((so->so_state & SS_CANTSENDMORE) == 0)
sb64->st_mode |= S_IWUSR | S_IWGRP | S_IWOTH;
sb64->st_size = so->so_rcv.sb_cc - so->so_rcv.sb_ctl;
sb64->st_uid = kauth_cred_getuid(so->so_cred);
sb64->st_gid = kauth_cred_getgid(so->so_cred);
} else {
sb->st_mode = S_IFSOCK;
if ((so->so_state & SS_CANTRCVMORE) == 0 ||
so->so_rcv.sb_cc != 0)
sb->st_mode |= S_IRUSR | S_IRGRP | S_IROTH;
if ((so->so_state & SS_CANTSENDMORE) == 0)
sb->st_mode |= S_IWUSR | S_IWGRP | S_IWOTH;
sb->st_size = so->so_rcv.sb_cc - so->so_rcv.sb_ctl;
sb->st_uid = kauth_cred_getuid(so->so_cred);
sb->st_gid = kauth_cred_getgid(so->so_cred);
}
ret = (*so->so_proto->pr_usrreqs->pru_sense)(so, ub, isstat64);
socket_unlock(so, 1);
return (ret);
}
/*
* Function: dhcp_timeout
* Purpose:
* Wakeup the process waiting for something on a socket.
*/
static void
dhcp_timeout(void * arg)
{
struct socket * * timer_arg = (struct socket * *)arg;
struct socket * so = *timer_arg;
dprintf(("dhcp: timeout\n"));
*timer_arg = NULL;
socket_lock(so, 1);
sowakeup(so, &so->so_rcv);
socket_unlock(so, 1);
return;
}
__private_extern__ uint32_t
inpcb_count_opportunistic(unsigned int ifindex, struct inpcbinfo *pcbinfo,
u_int32_t flags)
{
uint32_t opportunistic = 0;
struct inpcb *inp;
inp_gen_t gencnt;
lck_rw_lock_shared(pcbinfo->ipi_lock);
gencnt = pcbinfo->ipi_gencnt;
for (inp = LIST_FIRST(pcbinfo->ipi_listhead);
inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
if (inp->inp_gencnt <= gencnt &&
inp->inp_state != INPCB_STATE_DEAD &&
inp->inp_socket != NULL &&
so_get_opportunistic(inp->inp_socket) &&
inp->inp_last_outifp != NULL &&
ifindex == inp->inp_last_outifp->if_index) {
opportunistic++;
struct socket *so = inp->inp_socket;
if ((flags & INPCB_OPPORTUNISTIC_SETCMD) &&
(so->so_state & SS_ISCONNECTED)) {
socket_lock(so, 1);
if (flags & INPCB_OPPORTUNISTIC_THROTTLEON) {
so->so_flags |= SOF_SUSPENDED;
soevent(so,
(SO_FILT_HINT_LOCKED |
SO_FILT_HINT_SUSPEND));
} else {
so->so_flags &= ~(SOF_SUSPENDED);
soevent(so,
(SO_FILT_HINT_LOCKED |
SO_FILT_HINT_RESUME));
}
SOTHROTTLELOG(("throttle[%d]: so 0x%llx "
"[%d,%d] %s\n", so->last_pid,
(uint64_t)VM_KERNEL_ADDRPERM(so),
SOCK_DOM(so), SOCK_TYPE(so),
(so->so_flags & SOF_SUSPENDED) ?
"SUSPENDED" : "RESUMED"));
socket_unlock(so, 1);
}
}
}
lck_rw_done(pcbinfo->ipi_lock);
return (opportunistic);
}
static int
soo_drain(struct fileproc *fp, __unused vfs_context_t ctx)
{
int error = 0;
struct socket *so = (struct socket *)fp->f_fglob->fg_data;
if (so) {
socket_lock(so, 1);
so->so_state |= SS_DRAINING;
wakeup((caddr_t)&so->so_timeo);
sorwakeup(so);
sowwakeup(so);
socket_unlock(so, 1);
}
return (error);
}
void
in6_pcbdisconnect(
struct inpcb *inp)
{
if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->mtx)) {
/*lock inversion issue, mostly with udp multicast packets */
socket_unlock(inp->inp_socket, 0);
lck_rw_lock_exclusive(inp->inp_pcbinfo->mtx);
socket_lock(inp->inp_socket, 0);
}
bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
inp->inp_fport = 0;
/* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
in_pcbrehash(inp);
lck_rw_done(inp->inp_pcbinfo->mtx);
if (inp->inp_socket->so_state & SS_NOFDREF)
in6_pcbdetach(inp);
}
/*
* Wakeup processes waiting on a socket buffer.
* Do asynchronous notification via SIGIO
* if the socket has the SS_ASYNC flag set.
*/
void
sowakeup(struct socket *so, struct sockbuf *sb)
{
sb->sb_flags &= ~SB_SEL;
selwakeup(&sb->sb_sel);
if (sb->sb_flags & SB_WAIT) {
sb->sb_flags &= ~SB_WAIT;
wakeup((caddr_t)&sb->sb_cc);
}
#if 0 // Intentially commented
if (so->so_state & SS_ASYNC) {
if (so->so_pgid < 0)
gsignal(-so->so_pgid, SIGIO);
else if (so->so_pgid > 0)
proc_signal(so->so_pgid, SIGIO);
}
if (sb->sb_flags & SB_KNOTE) {
KNOTE(&sb->sb_sel.si_note, SO_FILT_HINT_LOCKED);
}
if (sb->sb_flags & SB_UPCALL) {
void (*so_upcall)(struct socket *, caddr_t, int);
caddr_t so_upcallarg;
so_upcall = so->so_upcall;
so_upcallarg = so->so_upcallarg;
/* Let close know that we're about to do an upcall */
so->so_flags |= SOF_UPCALLINUSE;
socket_unlock(so, 0);
(*so_upcall)(so, so_upcallarg, M_DONTWAIT);
socket_lock(so, 0);
so->so_flags &= ~SOF_UPCALLINUSE;
/* Tell close that it's safe to proceed */
if (so->so_flags & SOF_CLOSEWAIT)
wakeup((caddr_t)&so->so_upcall);
}
#endif
}
/*
* Detach the raw connection block and discard
* socket resources.
*/
void
raw_detach(struct rawcb *rp)
{
struct socket *so = rp->rcb_socket;
so->so_pcb = 0;
so->so_flags |= SOF_PCBCLEARING;
sofree(so);
if (!lck_mtx_try_lock(raw_mtx)) {
socket_unlock(so, 0);
lck_mtx_lock(raw_mtx);
socket_lock(so, 0);
}
LIST_REMOVE(rp, list);
lck_mtx_unlock(raw_mtx);
#ifdef notdef
if (rp->rcb_laddr)
m_freem(dtom(rp->rcb_laddr));
rp->rcb_laddr = 0;
#endif
rp->rcb_socket = NULL;
FREE((caddr_t)(rp), M_PCB);
}
int
fifo_close_internal(vnode_t vp, int fflag, __unused vfs_context_t context, int locked)
{
struct fifoinfo *fip = vp->v_fifoinfo;
int error1, error2;
struct socket *rso;
struct socket *wso;
if (!locked)
vnode_lock(vp);
if ((fip->fi_flags & FIFO_CREATED) == 0) {
if (!locked)
vnode_unlock(vp);
return(0);
}
if (fflag & FREAD) {
fip->fi_readers--;
if (fip->fi_readers == 0){
socket_lock(fip->fi_writesock, 1);
socantsendmore(fip->fi_writesock);
socket_unlock(fip->fi_writesock, 1);
}
}
if (fflag & FWRITE) {
fip->fi_writers--;
if (fip->fi_writers == 0) {
socket_lock(fip->fi_readsock, 1);
socantrcvmore(fip->fi_readsock);
socket_unlock(fip->fi_readsock, 1);
}
}
#if 0
if (vnode_isinuse_locked(vp, 0, 1)) {
if (!locked)
vnode_unlock(vp);
return (0);
}
#endif
if (fip->fi_writers || fip->fi_readers) {
if (!locked)
vnode_unlock(vp);
return (0);
}
wso = fip->fi_writesock;
rso = fip->fi_readsock;
fip->fi_readsock = NULL;
fip->fi_writesock = NULL;
fip->fi_flags &= ~FIFO_CREATED;
if (!locked)
vnode_unlock(vp);
error1 = soclose(rso);
error2 = soclose(wso);
if (error1)
return (error1);
return (error2);
}
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);
}
static int
dhcp_get_ack(struct dhcp_context * context, int wait_ticks)
{
int error = 0;
const struct in_addr * ip;
int len;
int n;
struct dhcp * reply;
struct in_addr server_id;
struct socket * timer_arg;
timer_arg = context->so;
reply = dhcp_context_reply(context);
timeout((timeout_fcn_t)dhcp_timeout, &timer_arg, wait_ticks);
while (1) {
error = receive_packet(context->so, context->reply,
sizeof(context->reply), &n);
if (error == 0) {
dhcp_msgtype_t msg;
dhcpol_t options;
dprintf(("\ndhcp: received packet length %d\n", n));
if (n < (int)sizeof(struct dhcp)) {
dprintf(("dhcp: packet is too short %d < %d\n",
n, (int)sizeof(struct dhcp)));
continue;
}
if (ntohl(reply->dp_xid) != context->xid
|| bcmp(reply->dp_chaddr, link_address(context->dl_p),
link_address_length(context->dl_p)) != 0) {
/* not for us */
continue;
}
(void)dhcpol_parse_packet(&options, reply, n);
server_id.s_addr = 0;
ip = (const struct in_addr *)
dhcpol_find(&options,
dhcptag_server_identifier_e, &len, NULL);
if (ip != NULL && len >= (int)sizeof(*ip)) {
server_id = *ip;
}
msg = get_dhcp_msgtype(&options);
if (msg == dhcp_msgtype_nak_e
&& server_id.s_addr == context->server_id.s_addr) {
/* server NAK'd us, start over */
dhcpol_free(&options);
error = EPROTO;
untimeout((timeout_fcn_t)dhcp_timeout, &timer_arg);
break;
}
if (msg != dhcp_msgtype_ack_e
|| reply->dp_yiaddr.s_addr == 0
|| reply->dp_yiaddr.s_addr == INADDR_BROADCAST) {
/* ignore the packet */
goto next_packet;
}
printf("dhcp: received ACK: server " IP_FORMAT
" IP address " IP_FORMAT "\n",
IP_LIST(&server_id), IP_LIST(&reply->dp_yiaddr));
context->iaddr = reply->dp_yiaddr;
ip = (const struct in_addr *)
dhcpol_find(&options,
dhcptag_subnet_mask_e, &len, NULL);
if (ip != NULL && len >= (int)sizeof(*ip)) {
context->netmask = *ip;
}
ip = (const struct in_addr *)
dhcpol_find(&options, dhcptag_router_e, &len, NULL);
if (ip != NULL && len >= (int)sizeof(*ip)) {
context->router = *ip;
}
dhcpol_free(&options);
untimeout((timeout_fcn_t)dhcp_timeout, &timer_arg);
break;
next_packet:
dhcpol_free(&options);
}
else if ((error != EWOULDBLOCK)) {
/* if some other error occurred, we're done */
untimeout((timeout_fcn_t)dhcp_timeout, &timer_arg);
break;
}
else if (timer_arg == NULL) {
/* timed out */
break;
}
else {
/* wait for a wait to arrive, or a timeout to occur */
socket_lock(context->so, 1);
error = sbwait(&context->so->so_rcv);
socket_unlock(context->so, 1);
}
}
return (error);
}
errno_t
sock_accept(socket_t sock, struct sockaddr *from, int fromlen, int flags,
sock_upcall callback, void *cookie, socket_t *new_sock)
{
struct sockaddr *sa;
struct socket *new_so;
lck_mtx_t *mutex_held;
int dosocklock;
errno_t error = 0;
if (sock == NULL || new_sock == NULL)
return (EINVAL);
socket_lock(sock, 1);
if ((sock->so_options & SO_ACCEPTCONN) == 0) {
socket_unlock(sock, 1);
return (EINVAL);
}
if ((flags & ~(MSG_DONTWAIT)) != 0) {
socket_unlock(sock, 1);
return (ENOTSUP);
}
if (((flags & MSG_DONTWAIT) != 0 || (sock->so_state & SS_NBIO) != 0) &&
sock->so_comp.tqh_first == NULL) {
socket_unlock(sock, 1);
return (EWOULDBLOCK);
}
if (sock->so_proto->pr_getlock != NULL) {
mutex_held = (*sock->so_proto->pr_getlock)(sock, 0);
dosocklock = 1;
} else {
mutex_held = sock->so_proto->pr_domain->dom_mtx;
dosocklock = 0;
}
while (TAILQ_EMPTY(&sock->so_comp) && sock->so_error == 0) {
if (sock->so_state & SS_CANTRCVMORE) {
sock->so_error = ECONNABORTED;
break;
}
error = msleep((caddr_t)&sock->so_timeo, mutex_held,
PSOCK | PCATCH, "sock_accept", NULL);
if (error != 0) {
socket_unlock(sock, 1);
return (error);
}
}
if (sock->so_error != 0) {
error = sock->so_error;
sock->so_error = 0;
socket_unlock(sock, 1);
return (error);
}
new_so = TAILQ_FIRST(&sock->so_comp);
TAILQ_REMOVE(&sock->so_comp, new_so, so_list);
sock->so_qlen--;
/*
* Pass the pre-accepted socket to any interested socket filter(s).
* Upon failure, the socket would have been closed by the callee.
*/
if (new_so->so_filt != NULL) {
/*
* Temporarily drop the listening socket's lock before we
* hand off control over to the socket filter(s), but keep
* a reference so that it won't go away. We'll grab it
* again once we're done with the filter(s).
*/
socket_unlock(sock, 0);
if ((error = soacceptfilter(new_so)) != 0) {
/* Drop reference on listening socket */
sodereference(sock);
return (error);
}
socket_lock(sock, 0);
}
if (dosocklock) {
lck_mtx_assert(new_so->so_proto->pr_getlock(new_so, 0),
LCK_MTX_ASSERT_NOTOWNED);
socket_lock(new_so, 1);
}
new_so->so_state &= ~SS_COMP;
new_so->so_head = NULL;
(void) soacceptlock(new_so, &sa, 0);
socket_unlock(sock, 1); /* release the head */
/* see comments in sock_setupcall() */
if (callback != NULL) {
sock_setupcalls_common(new_so, callback, cookie, NULL, NULL);
}
if (sa != NULL && from != NULL) {
if (fromlen > sa->sa_len)
fromlen = sa->sa_len;
memcpy(from, sa, fromlen);
}
if (sa != NULL)
FREE(sa, M_SONAME);
/*
* If the socket has been marked as inactive by sosetdefunct(),
* disallow further operations on it.
*/
if (new_so->so_flags & SOF_DEFUNCT) {
(void) sodefunct(current_proc(), new_so,
SHUTDOWN_SOCKET_LEVEL_DISCONNECT_INTERNAL);
}
*new_sock = new_so;
if (dosocklock)
socket_unlock(new_so, 1);
return (error);
}
static struct dhcp_context *
dhcp_context_create(struct ifnet * ifp, int max_try,
struct proc * procp, int * error_p)
{
struct dhcp_context * context = NULL;
struct sockaddr_dl * dl_p;
struct in_addr lo_addr;
struct in_addr lo_mask;
int error;
struct sockaddr_in sin;
/* get the hardware address from the interface */
dl_p = link_from_ifnet(ifp);
if (dl_p == NULL) {
printf("dhcp: can't get link address\n");
error = ENXIO;
goto failed;
}
printf("dhcp: h/w addr ");
link_print(dl_p);
if (dl_p->sdl_type != IFT_ETHER) {
printf("dhcp: hardware type %d not supported\n",
dl_p->sdl_type);
error = ENXIO;
goto failed;
}
context = (struct dhcp_context *)kalloc(sizeof(*context));
if (context == NULL) {
printf("dhcp: failed to allocate context\n");
error = ENOMEM;
goto failed;
}
bzero(context, sizeof(*context));
/* get a socket */
error = socreate(AF_INET, &context->so, SOCK_DGRAM, 0);
if (error != 0) {
printf("dhcp: socreate failed %d\n", error);
goto failed;
}
/* assign 127.0.0.1 to lo0 so that the bind will succeed */
lo_addr.s_addr = htonl(INADDR_LOOPBACK);
lo_mask.s_addr = htonl(IN_CLASSA_NET);
error = inet_aifaddr(context->so, "lo0", &lo_addr, &lo_mask, NULL);
if (error != 0) {
printf("dhcp: assigning loopback address failed %d\n", error);
}
/* enable reception of DHCP packets before an address is assigned */
snprintf(context->ifr.ifr_name,
sizeof(context->ifr.ifr_name), "%s%d", ifp->if_name,
ifp->if_unit);
context->ifr.ifr_intval = 1;
error = ifioctl(context->so, SIOCAUTOADDR, (caddr_t)&context->ifr, procp);
if (error) {
printf("dhcp: SIOCAUTOADDR failed: %d\n", error);
goto failed;
}
dprintf(("dhcp: SIOCAUTOADDR done\n"));
error = ifioctl(context->so, SIOCPROTOATTACH, (caddr_t)&context->ifr,
procp);
if (error) {
printf("dhcp: SIOCPROTOATTACH failed: %d\n", error);
goto failed;
}
dprintf(("dhcp: SIOCPROTOATTACH done\n"));
/* bind the socket */
sin.sin_len = sizeof(sin);
sin.sin_family = AF_INET;
sin.sin_port = htons(IPPORT_BOOTPC);
sin.sin_addr.s_addr = INADDR_ANY;
error = sobind(context->so, (struct sockaddr *)&sin);
if (error) {
printf("dhcp: sobind failed, %d\n", error);
goto failed;
}
/* make it non-blocking I/O */
socket_lock(context->so, 1);
context->so->so_state |= SS_NBIO;
socket_unlock(context->so, 1);
/* save passed-in information */
context->max_try = max_try;
context->dl_p = dl_p;
context->ifp = ifp;
/* get a random transaction id */
context->xid = random();
return (context);
failed:
dhcp_context_free(context, procp);
*error_p = error;
return (NULL);
}
static int
dhcp_get_offer(struct dhcp_context * context, int wait_ticks)
{
int error = 0;
int gather_count = 0;
const struct in_addr * ip;
int last_rating = 0;
int len;
int n;
int rating;
struct dhcp * reply;
struct in_addr server_id;
struct socket * timer_arg;
timer_arg = context->so;
reply = dhcp_context_reply(context);
timeout((timeout_fcn_t)dhcp_timeout, &timer_arg, wait_ticks);
while (1) {
error = receive_packet(context->so, context->reply,
sizeof(context->reply), &n);
if (error == 0) {
dhcpol_t options;
dprintf(("\ndhcp: received packet length %d\n", n));
if (n < (int)sizeof(struct dhcp)) {
dprintf(("dhcp: packet is too short %d < %d\n",
n, (int)sizeof(struct dhcp)));
continue;
}
if (ntohl(reply->dp_xid) != context->xid
|| reply->dp_yiaddr.s_addr == 0
|| reply->dp_yiaddr.s_addr == INADDR_BROADCAST
|| bcmp(reply->dp_chaddr,
link_address(context->dl_p),
link_address_length(context->dl_p)) != 0) {
/* not for us */
continue;
}
(void)dhcpol_parse_packet(&options, reply, n);
if (get_dhcp_msgtype(&options) != dhcp_msgtype_offer_e) {
/* not an offer */
goto next_packet;
}
ip = (const struct in_addr *)
dhcpol_find(&options,
dhcptag_server_identifier_e, &len, NULL);
if (ip == NULL || len < (int)sizeof(*ip)) {
/* missing/invalid server identifier */
goto next_packet;
}
printf("dhcp: received OFFER: server " IP_FORMAT
" IP address " IP_FORMAT "\n",
IP_LIST(ip), IP_LIST(&reply->dp_yiaddr));
server_id = *ip;
rating = rate_packet(&options);
if (rating > last_rating) {
context->iaddr = reply->dp_yiaddr;
ip = (const struct in_addr *)
dhcpol_find(&options,
dhcptag_subnet_mask_e, &len, NULL);
if (ip != NULL && len >= (int)sizeof(*ip)) {
context->netmask = *ip;
}
ip = (const struct in_addr *)
dhcpol_find(&options, dhcptag_router_e, &len, NULL);
if (ip != NULL && len >= (int)sizeof(*ip)) {
context->router = *ip;
}
context->server_id = server_id;
}
if (rating >= GOOD_RATING) {
dhcpol_free(&options);
/* packet is good enough */
untimeout((timeout_fcn_t)dhcp_timeout, &timer_arg);
break;
}
if (gather_count == 0) {
untimeout((timeout_fcn_t)dhcp_timeout, &timer_arg);
timer_arg = context->so;
timeout((timeout_fcn_t)dhcp_timeout, &timer_arg,
hz * GATHER_TIME_SECS);
}
gather_count = 1;
next_packet:
dhcpol_free(&options);
}
else if ((error != EWOULDBLOCK)) {
untimeout((timeout_fcn_t)dhcp_timeout, &timer_arg);
break;
}
else if (timer_arg == NULL) { /* timed out */
if (gather_count != 0) {
dprintf(("dhcp: gathering time has expired\n"));
error = 0;
}
break;
}
else {
socket_lock(context->so, 1);
error = sbwait(&context->so->so_rcv);
socket_unlock(context->so, 1);
}
}
return (error);
}
__private_extern__ int
soioctl(struct socket *so, u_long cmd, caddr_t data, struct proc *p)
{
int error = 0;
int int_arg;
socket_lock(so, 1);
/* call the socket filter's ioctl handler anything but ours */
if (IOCGROUP(cmd) != 'i' && IOCGROUP(cmd) != 'r') {
switch (cmd) {
case SIOCGASSOCIDS32:
case SIOCGASSOCIDS64:
case SIOCGCONNIDS32:
case SIOCGCONNIDS64:
case SIOCGCONNINFO32:
case SIOCGCONNINFO64:
case SIOCSCONNORDER:
case SIOCGCONNORDER:
/* don't pass to filter */
break;
default:
error = sflt_ioctl(so, cmd, data);
if (error != 0)
goto out;
break;
}
}
switch (cmd) {
case FIONBIO: /* int */
bcopy(data, &int_arg, sizeof (int_arg));
if (int_arg)
so->so_state |= SS_NBIO;
else
so->so_state &= ~SS_NBIO;
goto out;
case FIOASYNC: /* int */
bcopy(data, &int_arg, sizeof (int_arg));
if (int_arg) {
so->so_state |= SS_ASYNC;
so->so_rcv.sb_flags |= SB_ASYNC;
so->so_snd.sb_flags |= SB_ASYNC;
} else {
so->so_state &= ~SS_ASYNC;
so->so_rcv.sb_flags &= ~SB_ASYNC;
so->so_snd.sb_flags &= ~SB_ASYNC;
}
goto out;
case FIONREAD: /* int */
bcopy(&so->so_rcv.sb_cc, data, sizeof (u_int32_t));
goto out;
case SIOCSPGRP: /* int */
bcopy(data, &so->so_pgid, sizeof (pid_t));
goto out;
case SIOCGPGRP: /* int */
bcopy(&so->so_pgid, data, sizeof (pid_t));
goto out;
case SIOCATMARK: /* int */
int_arg = (so->so_state & SS_RCVATMARK) != 0;
bcopy(&int_arg, data, sizeof (int_arg));
goto out;
case SIOCSETOT: /* int; deprecated */
error = EOPNOTSUPP;
goto out;
case SIOCGASSOCIDS32: /* so_aidreq32 */
case SIOCGASSOCIDS64: /* so_aidreq64 */
case SIOCGCONNIDS32: /* so_cidreq32 */
case SIOCGCONNIDS64: /* so_cidreq64 */
case SIOCGCONNINFO32: /* so_cinforeq32 */
case SIOCGCONNINFO64: /* so_cinforeq64 */
case SIOCSCONNORDER: /* so_cordreq */
case SIOCGCONNORDER: /* so_cordreq */
error = (*so->so_proto->pr_usrreqs->pru_control)(so,
cmd, data, NULL, p);
goto out;
}
/*
* Interface/routing/protocol specific ioctls:
* interface and routing ioctls should have a
* different entry since a socket's unnecessary
*/
if (IOCGROUP(cmd) == 'i') {
error = ifioctllocked(so, cmd, data, p);
} else {
if (IOCGROUP(cmd) == 'r')
error = rtioctl(cmd, data, p);
else
error = (*so->so_proto->pr_usrreqs->pru_control)(so,
cmd, data, NULL, p);
}
out:
socket_unlock(so, 1);
if (error == EJUSTRETURN)
error = 0;
return (error);
}
errno_t
fill_socketinfo(struct socket *so, struct socket_info *si)
{
errno_t error = 0;
int domain;
short type;
short protocol;
socket_lock(so, 0);
si->soi_kind = SOCKINFO_GENERIC;
fill_common_sockinfo(so, si);
if (so->so_pcb == NULL || so->so_proto == 0 ||
so->so_proto->pr_domain == NULL)
goto out;
/*
* The kind of socket is determined by the triplet
* {domain, type, protocol}
*/
domain = SOCK_DOM(so);
type = SOCK_TYPE(so);
protocol = SOCK_PROTO(so);
switch (domain) {
case PF_INET:
case PF_INET6: {
struct in_sockinfo *insi = &si->soi_proto.pri_in;
struct inpcb *inp = (struct inpcb *)so->so_pcb;
si->soi_kind = SOCKINFO_IN;
insi->insi_fport = inp->inp_fport;
insi->insi_lport = inp->inp_lport;
insi->insi_gencnt = inp->inp_gencnt;
insi->insi_flags = inp->inp_flags;
insi->insi_vflag = inp->inp_vflag;
insi->insi_ip_ttl = inp->inp_ip_ttl;
insi->insi_faddr.ina_6 = inp->inp_dependfaddr.inp6_foreign;
insi->insi_laddr.ina_6 = inp->inp_dependladdr.inp6_local;
insi->insi_v4.in4_tos = inp->inp_depend4.inp4_ip_tos;
insi->insi_v6.in6_hlim = 0;
insi->insi_v6.in6_cksum = inp->inp_depend6.inp6_cksum;
insi->insi_v6.in6_ifindex = 0;
insi->insi_v6.in6_hops = inp->inp_depend6.inp6_hops;
if (type == SOCK_STREAM && (protocol == 0 ||
protocol == IPPROTO_TCP) && inp->inp_ppcb != NULL) {
struct tcp_sockinfo *tcpsi = &si->soi_proto.pri_tcp;
struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb;
si->soi_kind = SOCKINFO_TCP;
tcpsi->tcpsi_state = tp->t_state;
tcpsi->tcpsi_timer[TCPT_REXMT] =
tp->t_timer[TCPT_REXMT];
tcpsi->tcpsi_timer[TCPT_PERSIST] =
tp->t_timer[TCPT_PERSIST];
tcpsi->tcpsi_timer[TCPT_KEEP] = tp->t_timer[TCPT_KEEP];
tcpsi->tcpsi_timer[TCPT_2MSL] = tp->t_timer[TCPT_2MSL];
tcpsi->tcpsi_mss = tp->t_maxseg;
tcpsi->tcpsi_flags = tp->t_flags;
tcpsi->tcpsi_tp =
(u_int64_t)VM_KERNEL_ADDRPERM(tp);
}
break;
}
case PF_UNIX: {
struct unpcb *unp = (struct unpcb *)so->so_pcb;
struct un_sockinfo *unsi = &si->soi_proto.pri_un;
si->soi_kind = SOCKINFO_UN;
unsi->unsi_conn_pcb =
(uint64_t)VM_KERNEL_ADDRPERM(unp->unp_conn);
if (unp->unp_conn)
unsi->unsi_conn_so = (uint64_t)
VM_KERNEL_ADDRPERM(unp->unp_conn->unp_socket);
if (unp->unp_addr) {
size_t addrlen = unp->unp_addr->sun_len;
if (addrlen > SOCK_MAXADDRLEN)
addrlen = SOCK_MAXADDRLEN;
bcopy(unp->unp_addr, &unsi->unsi_addr, addrlen);
}
if (unp->unp_conn && unp->unp_conn->unp_addr) {
size_t addrlen = unp->unp_conn->unp_addr->sun_len;
if (addrlen > SOCK_MAXADDRLEN)
addrlen = SOCK_MAXADDRLEN;
bcopy(unp->unp_conn->unp_addr, &unsi->unsi_caddr,
addrlen);
}
break;
}
case PF_NDRV: {
struct ndrv_cb *ndrv_cb = (struct ndrv_cb *)so->so_pcb;
struct ndrv_info *ndrvsi = &si->soi_proto.pri_ndrv;
si->soi_kind = SOCKINFO_NDRV;
/* TDB lock ifnet ???? */
if (ndrv_cb->nd_if != 0) {
struct ifnet *ifp = ndrv_cb->nd_if;
ndrvsi->ndrvsi_if_family = ifp->if_family;
ndrvsi->ndrvsi_if_unit = ifp->if_unit;
strlcpy(ndrvsi->ndrvsi_if_name, ifp->if_name, IFNAMSIZ);
}
break;
}
case PF_SYSTEM:
if (SOCK_PROTO(so) == SYSPROTO_EVENT) {
struct kern_event_pcb *ev_pcb =
(struct kern_event_pcb *)so->so_pcb;
struct kern_event_info *kesi =
&si->soi_proto.pri_kern_event;
si->soi_kind = SOCKINFO_KERN_EVENT;
kesi->kesi_vendor_code_filter =
ev_pcb->evp_vendor_code_filter;
kesi->kesi_class_filter = ev_pcb->evp_class_filter;
kesi->kesi_subclass_filter = ev_pcb->evp_subclass_filter;
} else if (SOCK_PROTO(so) == SYSPROTO_CONTROL) {
struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
struct kern_ctl_info *kcsi =
&si->soi_proto.pri_kern_ctl;
struct kctl *kctl = kcb->kctl;
si->soi_kind = SOCKINFO_KERN_CTL;
if (kctl == 0)
break;
kcsi->kcsi_id = kctl->id;
kcsi->kcsi_reg_unit = kctl->id;
kcsi->kcsi_flags = kctl->flags;
kcsi->kcsi_recvbufsize = kctl->recvbufsize;
kcsi->kcsi_sendbufsize = kctl->sendbufsize;
kcsi->kcsi_unit = kcb->unit;
strlcpy(kcsi->kcsi_name, kctl->name, MAX_KCTL_NAME);
}
break;
case PF_ROUTE:
case PF_PPP:
default:
break;
}
out:
socket_unlock(so, 0);
return (error);
}
__private_extern__ int
soioctl(struct socket *so, u_long cmd, caddr_t data, struct proc *p)
{
struct sockopt sopt;
int error = 0;
int dropsockref = -1;
socket_lock(so, 1);
sopt.sopt_level = cmd;
sopt.sopt_name = (int)data;
sopt.sopt_p = p;
/* Call the socket filter's ioctl handler for most ioctls */
if (IOCGROUP(cmd) != 'i' && IOCGROUP(cmd) != 'r') {
int filtered = 0;
struct socket_filter_entry *filter;
for (filter = so->so_filt; filter && error == 0;
filter = filter->sfe_next_onsocket) {
if (filter->sfe_filter->sf_filter.sf_ioctl) {
if (filtered == 0) {
sflt_use(so);
socket_unlock(so, 0);
filtered = 1;
}
error = filter->sfe_filter->sf_filter.
sf_ioctl(filter->sfe_cookie, so, cmd, data);
}
}
if (filtered) {
socket_lock(so, 0);
sflt_unuse(so);
}
if (error != 0)
goto out;
}
switch (cmd) {
case FIONBIO:
if (*(int *)data)
so->so_state |= SS_NBIO;
else
so->so_state &= ~SS_NBIO;
goto out;
case FIOASYNC:
if (*(int *)data) {
so->so_state |= SS_ASYNC;
so->so_rcv.sb_flags |= SB_ASYNC;
so->so_snd.sb_flags |= SB_ASYNC;
} else {
so->so_state &= ~SS_ASYNC;
so->so_rcv.sb_flags &= ~SB_ASYNC;
so->so_snd.sb_flags &= ~SB_ASYNC;
}
goto out;
case FIONREAD:
*(int *)data = so->so_rcv.sb_cc;
goto out;
case SIOCSPGRP:
so->so_pgid = *(int *)data;
goto out;
case SIOCGPGRP:
*(int *)data = so->so_pgid;
goto out;
case SIOCATMARK:
*(int *)data = (so->so_state&SS_RCVATMARK) != 0;
goto out;
case SIOCSETOT: {
/*
* Set socket level options here and then call protocol
* specific routine.
*/
struct socket *cloned_so = NULL;
int cloned_fd = *(int *)data;
/* let's make sure it's either -1 or a valid file descriptor */
if (cloned_fd != -1) {
error = file_socket(cloned_fd, &cloned_so);
if (error) {
goto out;
}
dropsockref = cloned_fd;
}
/* Always set socket non-blocking for OT */
so->so_state |= SS_NBIO;
so->so_options |= SO_DONTTRUNC | SO_WANTMORE;
so->so_flags |= SOF_NOSIGPIPE;
if (cloned_so && so != cloned_so) {
/* Flags options */
so->so_options |=
cloned_so->so_options & ~SO_ACCEPTCONN;
/* SO_LINGER */
if (so->so_options & SO_LINGER)
so->so_linger = cloned_so->so_linger;
/* SO_SNDBUF, SO_RCVBUF */
if (cloned_so->so_snd.sb_hiwat > 0) {
if (sbreserve(&so->so_snd,
cloned_so->so_snd.sb_hiwat) == 0) {
error = ENOBUFS;
goto out;
}
}
if (cloned_so->so_rcv.sb_hiwat > 0) {
if (sbreserve(&so->so_rcv,
cloned_so->so_rcv.sb_hiwat) == 0) {
error = ENOBUFS;
goto out;
}
}
/* SO_SNDLOWAT, SO_RCVLOWAT */
so->so_snd.sb_lowat =
(cloned_so->so_snd.sb_lowat > so->so_snd.sb_hiwat) ?
so->so_snd.sb_hiwat : cloned_so->so_snd.sb_lowat;
so->so_rcv.sb_lowat =
(cloned_so->so_rcv.sb_lowat > so->so_rcv.sb_hiwat) ?
so->so_rcv.sb_hiwat : cloned_so->so_rcv.sb_lowat;
/* SO_SNDTIMEO, SO_RCVTIMEO */
so->so_snd.sb_timeo = cloned_so->so_snd.sb_timeo;
so->so_rcv.sb_timeo = cloned_so->so_rcv.sb_timeo;
}
error = (*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
data, 0, p);
/* Just ignore protocols that do not understand it */
if (error == EOPNOTSUPP)
error = 0;
goto out;
}
}
/*
* Interface/routing/protocol specific ioctls:
* interface and routing ioctls should have a
* different entry since a socket's unnecessary
*/
if (IOCGROUP(cmd) == 'i') {
error = ifioctllocked(so, cmd, data, p);
} else {
if (IOCGROUP(cmd) == 'r')
error = rtioctl(cmd, data, p);
else
error = (*so->so_proto->pr_usrreqs->pru_control)(so,
cmd, data, 0, p);
}
out:
if (dropsockref != -1)
file_drop(dropsockref);
socket_unlock(so, 1);
if (error == EJUSTRETURN)
error = 0;
return (error);
}
ni_status checkserves(unsigned long masteraddr, ni_name mastertag, unsigned long myaddr, ni_name mytag)
{
struct sockaddr_in mastersin;
struct in_addr myinaddr;
char myserves[MAXPATHLEN + 16];
ni_index where;
int status;
int sock;
CLIENT *cl;
struct timeval tv;
nibind_getregister_res res;
ni_id_res id_res;
ni_lookup_res lu_res;
ni_lookup_args childdir;
ni_namelist_res prop_res;
ni_prop_args pa;
mastersin.sin_addr.s_addr = masteraddr;
mastersin.sin_port = 0;
mastersin.sin_family = AF_INET;
myinaddr.s_addr = myaddr;
/* connect to master hosts's nibindd */
sock = socket_connect(&mastersin, NIBIND_PROG, NIBIND_VERS);
if (sock < 0)
{
system_log(LOG_WARNING,
"sanitycheck can't connect to %s/%s - %m",
inet_ntoa(mastersin.sin_addr), mastertag);
return NI_FAILED;
}
FD_SET(sock, &clnt_fdset); /* protect client socket */
cl = clnttcp_create(&mastersin, NIBIND_PROG, NIBIND_VERS, &sock, 0, 0);
if (cl == NULL) {
socket_close(sock);
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
return NI_FAILED;
}
tv.tv_sec = READALL_TIMEOUT;
tv.tv_usec = 0;
/* get the ports for master's tag */
bzero((char *)&res, sizeof(res));
if (clnt_call(cl, NIBIND_GETREGISTER,
xdr_ni_name, &mastertag,
xdr_nibind_getregister_res, &res, tv) != RPC_SUCCESS) {
/* call failed */
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
return NI_FAILED;
}
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
if (res.status != NI_OK) {
/* no server for the master's tag */
system_log(LOG_ERR,
"tag %s: master's tag %s is not served at master's addresss",
mytag, mastertag);
return res.status;
}
/* connect to the master */
mastersin.sin_port = htons(res.nibind_getregister_res_u.addrs.tcp_port);
xdr_free(xdr_nibind_getregister_res, (void *)&res);
sock = socket_connect(&mastersin, NI_PROG, NI_VERS);
if (sock < 0) {
system_log(LOG_WARNING, "sanitycheck can't connect to "
"%s/%s - %m",
inet_ntoa(mastersin.sin_addr), mastertag);
return NI_FAILED;
}
FD_SET(sock, &clnt_fdset); /* protect client socket */
cl = clnttcp_create(&mastersin, NI_PROG, NI_VERS, &sock, 0, 0);
if (cl == NULL) {
/* can't connect to master */
socket_close(sock);
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
return NI_FAILED;
}
FD_SET(sock, &clnt_fdset); /* protect client socket */
/* get root directory */
bzero((char *)&id_res, sizeof(id_res));
status = clnt_call(cl, _NI_ROOT,
xdr_void, NULL, xdr_ni_id_res, &id_res, tv);
if (status != NI_OK) {
/* can't get root! */
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: can't get master's root directory",
mytag);
return status;
}
if (id_res.status != NI_OK) {
/* can't get root! */
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: master has no root directory",
mytag);
return status;
}
childdir.key = malloc(16);
childdir.value = malloc(16);
/* get machines subdirectory */
childdir.id = id_res.ni_id_res_u.id;
xdr_free(xdr_ni_id_res, (void *)&id_res);
strcpy(childdir.key,"name");
strcpy(childdir.value,"machines");
bzero((char *)&lu_res, sizeof(lu_res));
status = clnt_call(cl, _NI_LOOKUP,
xdr_ni_lookup_args, &childdir, xdr_ni_lookup_res, &lu_res, tv);
if (status != NI_OK) {
/* can't get /machines! */
free(childdir.key);
free(childdir.value);
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: can't get master's /machines directory",
mytag);
return status;
}
if (lu_res.status != NI_OK) {
/* no /machines! */
free(childdir.key);
free(childdir.value);
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: can't get master's /machines directory",
mytag);
return lu_res.status;
}
if (lu_res.ni_lookup_res_u.stuff.idlist.ni_idlist_len == 0) {
/* no /machines! */
xdr_free(xdr_ni_lookup_res, (void *)&lu_res);
free(childdir.key);
free(childdir.value);
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: master has no /machines directory",
mytag);
return NI_NODIR;
}
/* get my subdirectory */
childdir.id.nii_object = lu_res.ni_lookup_res_u.stuff.idlist.ni_idlist_val[0];
xdr_free(xdr_ni_lookup_res, (void *)&lu_res);
strcpy(childdir.key,"ip_address");
strcpy(childdir.value,(char *)inet_ntoa(myinaddr));
bzero((char *)&lu_res, sizeof(lu_res));
status = clnt_call(cl, _NI_LOOKUP,
xdr_ni_lookup_args, &childdir, xdr_ni_lookup_res, &lu_res, tv);
if (status != NI_OK) {
/* can't get /machines/ip_address=<myaddr>! */
free(childdir.key);
free(childdir.value);
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: can't get master's /machines/ip_address=%s directory",
mytag,(char *)inet_ntoa(myinaddr));
return status;
}
if (lu_res.status != NI_OK) {
/* no /machines! */
free(childdir.key);
free(childdir.value);
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: can't get master's /machines/ip_address=%s directory",
mytag,(char *)inet_ntoa(myinaddr));
return lu_res.status;
}
if (lu_res.ni_lookup_res_u.stuff.idlist.ni_idlist_len == 0) {
/* can't get /machines/ip_address=<myaddr>! */
xdr_free(xdr_ni_lookup_res, (void *)&lu_res);
free(childdir.key);
free(childdir.value);
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: master has no /machines/ip_address=%s directory",
mytag,(char *)inet_ntoa(myinaddr));
return NI_NODIR;
}
/* list properties */
prop_res.ni_namelist_res_u.stuff.values.ni_namelist_val = NULL;
childdir.id.nii_object = lu_res.ni_lookup_res_u.stuff.idlist.ni_idlist_val[0];
xdr_free(xdr_ni_lookup_res, (void *)&lu_res);
bzero((char *)&prop_res, sizeof(prop_res));
status = clnt_call(cl, _NI_LISTPROPS,
xdr_ni_id, &childdir.id, xdr_ni_namelist_res, &prop_res, tv);
if (status != NI_OK) {
/* can't get proplist! */
free(childdir.key);
free(childdir.value);
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: can't get master's property list for /machines/ip_address=%s",
mytag,(char *)inet_ntoa(myinaddr));
return status;
}
free(childdir.key);
free(childdir.value);
if (prop_res.status != NI_OK) {
/* no /machines! */
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: can't get master's property list for /machines/ip_address=%s",
mytag,(char *)inet_ntoa(myinaddr));
return prop_res.status;
}
if (prop_res.ni_namelist_res_u.stuff.values.ni_namelist_len == 0) {
/* can't get proplist! */
xdr_free(xdr_ni_namelist_res, (void *)&prop_res);
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: master has no property list for /machines/ip_address=%s",
mytag,(char *)inet_ntoa(myinaddr));
return NI_NOPROP;
}
/* find "serves" property */
where = ni_namelist_match(prop_res.ni_namelist_res_u.stuff.values, "serves");
xdr_free(xdr_ni_namelist_res, (void *)&prop_res);
if (where == NI_INDEX_NULL) {
/* no serves property! */
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: master has no serves property for /machines/ip_address=%s",
mytag,(char *)inet_ntoa(myinaddr));
return NI_NOPROP;
}
/* fetch serves property */
pa.id = childdir.id;
pa.prop_index = where;
bzero((char *)&prop_res, sizeof(prop_res));
status = clnt_call(cl, _NI_READPROP,
xdr_ni_prop_args, &pa, xdr_ni_namelist_res, &prop_res, tv);
if (status != NI_OK) {
/* can't get proplist! */
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: can't get master's serves property for /machines/ip_address=%s",
mytag,(char *)inet_ntoa(myinaddr));
return status;
}
if (prop_res.status != NI_OK) {
/* can't get proplist! */
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: can't get master's serves property for /machines/ip_address=%s",
mytag,(char *)inet_ntoa(myinaddr));
return prop_res.status;
}
if (prop_res.ni_namelist_res_u.stuff.values.ni_namelist_len == 0) {
/* no values in serves property! */
xdr_free(xdr_ni_namelist_res, (void *)&prop_res);
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: master has no values for serves property in /machines/ip_address=%s",
mytag,(char *)inet_ntoa(myinaddr));
return NI_NOPROP;
}
/* find my "serves" property */
sprintf(myserves, "./%s", mytag);
where = ni_namelist_match(prop_res.ni_namelist_res_u.stuff.values, myserves);
xdr_free(xdr_ni_namelist_res, (void *)&prop_res);
if (where == NI_INDEX_NULL) {
/* no serves property! */
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
system_log(LOG_ERR,
"tag %s: master has no serves ./%s property in /machines/ip_address=%s",
mytag,mytag,(char *)inet_ntoa(myinaddr));
return NI_NONAME;
}
socket_lock();
clnt_destroy(cl);
(void)close(sock);
socket_unlock();
FD_CLR(sock, &clnt_fdset); /* unprotect client socket */
return NI_OK;
}
/* ARGSUSED */
int
fifo_open(struct vnop_open_args *ap)
{
struct vnode *vp = ap->a_vp;
struct fifoinfo *fip;
struct socket *rso, *wso;
int error;
vnode_lock(vp);
retry:
fip = vp->v_fifoinfo;
if (fip == (struct fifoinfo *)0)
panic("fifo_open with no fifoinfo");
if ((fip->fi_flags & FIFO_CREATED) == 0) {
if (fip->fi_flags & FIFO_INCREATE) {
fip->fi_flags |= FIFO_CREATEWAIT;
error = msleep(&fip->fi_flags, &vp->v_lock, PRIBIO | PCATCH, "fifocreatewait", NULL);
if (error) {
vnode_unlock(vp);
return(error);
}
goto retry;
} else {
fip->fi_flags |= FIFO_INCREATE;
vnode_unlock(vp);
if ( (error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) ) {
goto bad1;
}
if ( (error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) ) {
(void)soclose(rso);
goto bad1;
}
if ( (error = soconnect2(wso, rso)) ) {
(void)soclose(wso);
(void)soclose(rso);
goto bad1;
}
fip->fi_readers = fip->fi_writers = 0;
/* Lock ordering between wso and rso does not matter here
* because they are just created and no one has a reference to them
*/
socket_lock(wso, 1);
wso->so_state |= SS_CANTRCVMORE;
wso->so_snd.sb_lowat = PIPE_BUF;
socket_unlock(wso, 1);
socket_lock(rso, 1);
rso->so_state |= SS_CANTSENDMORE;
socket_unlock(rso, 1);
vnode_lock(vp);
fip->fi_readsock = rso;
fip->fi_writesock = wso;
fip->fi_flags |= FIFO_CREATED;
fip->fi_flags &= ~FIFO_INCREATE;
if ((fip->fi_flags & FIFO_CREATEWAIT)) {
fip->fi_flags &= ~FIFO_CREATEWAIT;
wakeup(&fip->fi_flags);
}
/* vnode lock is held to process further */
}
}
/* vnode is locked at this point */
/* fifo in created already */
if (ap->a_mode & FREAD) {
fip->fi_readers++;
if (fip->fi_readers == 1) {
socket_lock(fip->fi_writesock, 1);
fip->fi_writesock->so_state &= ~SS_CANTSENDMORE;
socket_unlock(fip->fi_writesock, 1);
if (fip->fi_writers > 0)
wakeup((caddr_t)&fip->fi_writers);
}
}
if (ap->a_mode & FWRITE) {
fip->fi_writers++;
if (fip->fi_writers == 1) {
socket_lock(fip->fi_readsock, 1);
fip->fi_readsock->so_state &= ~SS_CANTRCVMORE;
socket_unlock(fip->fi_readsock, 1);
if (fip->fi_readers > 0)
wakeup((caddr_t)&fip->fi_readers);
}
}
if ((ap->a_mode & FREAD) && (ap->a_mode & O_NONBLOCK) == 0) {
if (fip->fi_writers == 0) {
error = msleep((caddr_t)&fip->fi_readers, &vp->v_lock,
PCATCH | PSOCK, "fifoor", NULL);
if (error)
goto bad;
if (fip->fi_readers == 1) {
if (fip->fi_writers > 0)
wakeup((caddr_t)&fip->fi_writers);
}
}
}
if (ap->a_mode & FWRITE) {
if (ap->a_mode & O_NONBLOCK) {
if (fip->fi_readers == 0) {
error = ENXIO;
goto bad;
}
} else {
if (fip->fi_readers == 0) {
error = msleep((caddr_t)&fip->fi_writers,&vp->v_lock,
PCATCH | PSOCK, "fifoow", NULL);
if (error)
goto bad;
if (fip->fi_writers == 1) {
if (fip->fi_readers > 0)
wakeup((caddr_t)&fip->fi_readers);
}
}
}
}
vnode_unlock(vp);
return (0);
bad:
fifo_close_internal(vp, ap->a_mode, ap->a_context, 1);
vnode_unlock(vp);
return (error);
bad1:
vnode_lock(vp);
fip->fi_flags &= ~FIFO_INCREATE;
if ((fip->fi_flags & FIFO_CREATEWAIT)) {
fip->fi_flags &= ~FIFO_CREATEWAIT;
wakeup(&fip->fi_flags);
}
vnode_unlock(vp);
return (error);
}
/*
* XXX: this is borrowed from in6_pcbbind(). If possible, we should
* share this function by all *bsd*...
*/
int
in6_pcbsetport(
__unused struct in6_addr *laddr,
struct inpcb *inp,
struct proc *p,
int locked)
{
struct socket *so = inp->inp_socket;
u_int16_t lport = 0, first, last, *lastport;
int count, error = 0, wild = 0;
struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
kauth_cred_t cred;
if (!locked) { /* Make sure we don't run into a deadlock: 4052373 */
if (!lck_rw_try_lock_exclusive(pcbinfo->mtx)) {
socket_unlock(inp->inp_socket, 0);
lck_rw_lock_exclusive(pcbinfo->mtx);
socket_lock(inp->inp_socket, 0);
}
}
/* XXX: this is redundant when called from in6_pcbbind */
if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
wild = INPLOOKUP_WILDCARD;
inp->inp_flags |= INP_ANONPORT;
if (inp->inp_flags & INP_HIGHPORT) {
first = ipport_hifirstauto; /* sysctl */
last = ipport_hilastauto;
lastport = &pcbinfo->lasthi;
} else if (inp->inp_flags & INP_LOWPORT) {
cred = kauth_cred_proc_ref(p);
error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
kauth_cred_unref(&cred);
if (error != 0) {
if (!locked)
lck_rw_done(pcbinfo->mtx);
return error;
}
first = ipport_lowfirstauto; /* 1023 */
last = ipport_lowlastauto; /* 600 */
lastport = &pcbinfo->lastlow;
} else {
first = ipport_firstauto; /* sysctl */
last = ipport_lastauto;
lastport = &pcbinfo->lastport;
}
/*
* Simple check to ensure all ports are not used up causing
* a deadlock here.
*
* We split the two cases (up and down) so that the direction
* is not being tested on each round of the loop.
*/
if (first > last) {
/*
* counting down
*/
count = first - last;
do {
if (count-- < 0) { /* completely used? */
/*
* Undo any address bind that may have
* occurred above.
*/
inp->in6p_laddr = in6addr_any;
inp->in6p_last_outif = 0;
if (!locked)
lck_rw_done(pcbinfo->mtx);
return (EAGAIN);
}
--*lastport;
if (*lastport > first || *lastport < last)
*lastport = first;
lport = htons(*lastport);
} while (in6_pcblookup_local(pcbinfo,
&inp->in6p_laddr, lport, wild));
} else {
/*
* counting up
*/
count = last - first;
do {
if (count-- < 0) { /* completely used? */
/*
* Undo any address bind that may have
* occurred above.
*/
inp->in6p_laddr = in6addr_any;
inp->in6p_last_outif = 0;
if (!locked)
lck_rw_done(pcbinfo->mtx);
return (EAGAIN);
}
++*lastport;
if (*lastport < first || *lastport > last)
*lastport = first;
lport = htons(*lastport);
} while (in6_pcblookup_local(pcbinfo,
&inp->in6p_laddr, lport, wild));
}
inp->inp_lport = lport;
if (in_pcbinshash(inp, 1) != 0) {
inp->in6p_laddr = in6addr_any;
inp->inp_lport = 0;
inp->in6p_last_outif = 0;
if (!locked)
lck_rw_done(pcbinfo->mtx);
return (EAGAIN);
}
if (!locked)
lck_rw_done(pcbinfo->mtx);
return(0);
}
static int
shutdown_sockets_on_interface_proc_callout(proc_t p, void *arg)
{
struct filedesc *fdp;
int i;
struct ifnet *ifp = (struct ifnet *)arg;
if (ifp == NULL)
return (PROC_RETURNED);
proc_fdlock(p);
fdp = p->p_fd;
for (i = 0; i < fdp->fd_nfiles; i++) {
struct fileproc *fp = fdp->fd_ofiles[i];
struct fileglob *fg;
struct socket *so;
struct inpcb *inp;
struct ifnet *inp_ifp;
int error;
if (fp == NULL || (fdp->fd_ofileflags[i] & UF_RESERVED) != 0) {
continue;
}
fg = fp->f_fglob;
if (FILEGLOB_DTYPE(fg) != DTYPE_SOCKET)
continue;
so = (struct socket *)fp->f_fglob->fg_data;
if (SOCK_DOM(so) != PF_INET && SOCK_DOM(so) != PF_INET6)
continue;
inp = (struct inpcb *)so->so_pcb;
if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING)
continue;
socket_lock(so, 1);
if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
socket_unlock(so, 1);
continue;
}
if (inp->inp_boundifp != NULL) {
inp_ifp = inp->inp_boundifp;
} else if (inp->inp_last_outifp != NULL) {
inp_ifp = inp->inp_last_outifp;
} else {
socket_unlock(so, 1);
continue;
}
if (inp_ifp != ifp && inp_ifp->if_delegated.ifp != ifp) {
socket_unlock(so, 1);
continue;
}
error = sosetdefunct(p, so, 0, TRUE);
if (error != 0) {
log(LOG_ERR, "%s: sosetdefunct() error %d",
__func__, error);
} else {
error = sodefunct(p, so, 0);
if (error != 0) {
log(LOG_ERR, "%s: sodefunct() error %d",
__func__, error);
}
}
socket_unlock(so, 1);
}
proc_fdunlock(p);
return (PROC_RETURNED);
}
