int
udp_attach(PNATState pData, struct socket *so)
{
struct sockaddr_in *addr;
struct sockaddr sa_addr;
socklen_t socklen = sizeof(struct sockaddr);
int status;
int opt = 1;
/* We attaching some olready attched socket ??? */
Assert(so->so_type == 0);
if ((so->s = socket(AF_INET, SOCK_DGRAM, 0)) == -1)
goto error;
/*
* Here, we bind() the socket. Although not really needed
* (sendto() on an unbound socket will bind it), it's done
* here so that emulation of ytalk etc. don't have to do it
*/
memset(&sa_addr, 0, sizeof(struct sockaddr));
addr = (struct sockaddr_in *)&sa_addr;
#ifdef RT_OS_DARWIN
addr->sin_len = sizeof(struct sockaddr_in);
#endif
addr->sin_family = AF_INET;
addr->sin_addr.s_addr = pData->bindIP.s_addr;
fd_nonblock(so->s);
if (bind(so->s, &sa_addr, sizeof(struct sockaddr_in)) < 0)
{
int lasterrno = errno;
closesocket(so->s);
so->s = -1;
#ifdef RT_OS_WINDOWS
WSASetLastError(lasterrno);
#else
errno = lasterrno;
#endif
goto error;
}
/* success, insert in queue */
so->so_expire = curtime + SO_EXPIRE;
/* enable broadcast for later use */
setsockopt(so->s, SOL_SOCKET, SO_BROADCAST, (const char *)&opt, sizeof(opt));
status = getsockname(so->s, &sa_addr, &socklen);
Assert(status == 0 && sa_addr.sa_family == AF_INET);
so->so_hlport = ((struct sockaddr_in *)&sa_addr)->sin_port;
so->so_hladdr.s_addr = ((struct sockaddr_in *)&sa_addr)->sin_addr.s_addr;
SOCKET_LOCK_CREATE(so);
QSOCKET_LOCK(udb);
insque(pData, so, &udb);
NSOCK_INC();
QSOCKET_UNLOCK(udb);
so->so_type = IPPROTO_UDP;
return so->s;
error:
Log2(("NAT: can't create datagramm socket\n"));
return -1;
}
void
udp_detach(PNATState pData, struct socket *so)
{
if (so != &pData->icmp_socket)
{
Assert(so->so_type == IPPROTO_UDP);
QSOCKET_LOCK(udb);
SOCKET_LOCK(so);
QSOCKET_UNLOCK(udb);
closesocket(so->s);
sofree(pData, so);
SOCKET_UNLOCK(so);
}
}
int
udp_attach(PNATState pData, struct socket *so)
{
struct sockaddr sa_addr;
socklen_t socklen = sizeof(struct sockaddr);
int status;
int opt = 1;
AssertReturn(so->so_type == 0, -1);
so->so_type = IPPROTO_UDP;
so->s = socket(AF_INET, SOCK_DGRAM, 0);
if (so->s == -1)
goto error;
fd_nonblock(so->s);
so->so_sottl = 0;
so->so_sotos = 0;
so->so_sodf = -1;
status = sobind(pData, so);
if (status != 0)
return status;
/* success, insert in queue */
so->so_expire = curtime + SO_EXPIRE;
/* enable broadcast for later use */
setsockopt(so->s, SOL_SOCKET, SO_BROADCAST, (const char *)&opt, sizeof(opt));
status = getsockname(so->s, &sa_addr, &socklen);
if (status == 0)
{
Assert(sa_addr.sa_family == AF_INET);
so->so_hlport = ((struct sockaddr_in *)&sa_addr)->sin_port;
so->so_hladdr.s_addr = ((struct sockaddr_in *)&sa_addr)->sin_addr.s_addr;
}
SOCKET_LOCK_CREATE(so);
QSOCKET_LOCK(udb);
insque(pData, so, &udb);
NSOCK_INC();
QSOCKET_UNLOCK(udb);
return so->s;
error:
Log2(("NAT: can't create datagram socket\n"));
return -1;
}
void
udp_detach(PNATState pData, struct socket *so)
{
if (so != &pData->icmp_socket)
{
Assert(so->so_type == IPPROTO_UDP);
QSOCKET_LOCK(udb);
SOCKET_LOCK(so);
QSOCKET_UNLOCK(udb);
#ifdef VBOX_WITH_NAT_UDP_SOCKET_CLONE
if (so->so_cloneOf)
so->so_cloneOf->so_cCloneCounter--;
else if (so->so_cCloneCounter > 0)
{
/* we can't close socket yet */
SOCKET_UNLOCK(so);
return;
}
#endif
closesocket(so->s);
sofree(pData, so);
SOCKET_UNLOCK(so);
}
}
/*
* Read from so's socket into sb_snd, updating all relevant sbuf fields
* NOTE: This will only be called if it is select()ed for reading, so
* a read() of 0 (or less) means it's disconnected
*/
int
soread(PNATState pData, struct socket *so)
{
int n, nn, lss, total;
struct sbuf *sb = &so->so_snd;
u_int len = sb->sb_datalen - sb->sb_cc;
struct iovec iov[2];
int mss = so->so_tcpcb->t_maxseg;
int sockerr;
STAM_PROFILE_START(&pData->StatIOread, a);
STAM_COUNTER_RESET(&pData->StatIORead_in_1);
STAM_COUNTER_RESET(&pData->StatIORead_in_2);
QSOCKET_LOCK(tcb);
SOCKET_LOCK(so);
QSOCKET_UNLOCK(tcb);
LogFlow(("soread: so = %R[natsock]\n", so));
Log2(("%s: so = %R[natsock] so->so_snd = %R[sbuf]\n", RT_GCC_EXTENSION __PRETTY_FUNCTION__, so, sb));
/*
* No need to check if there's enough room to read.
* soread wouldn't have been called if there weren't
*/
len = sb->sb_datalen - sb->sb_cc;
iov[0].iov_base = sb->sb_wptr;
iov[1].iov_base = 0;
iov[1].iov_len = 0;
if (sb->sb_wptr < sb->sb_rptr)
{
iov[0].iov_len = sb->sb_rptr - sb->sb_wptr;
/* Should never succeed, but... */
if (iov[0].iov_len > len)
iov[0].iov_len = len;
if (iov[0].iov_len > mss)
iov[0].iov_len -= iov[0].iov_len%mss;
n = 1;
}
else
{
iov[0].iov_len = (sb->sb_data + sb->sb_datalen) - sb->sb_wptr;
/* Should never succeed, but... */
if (iov[0].iov_len > len)
iov[0].iov_len = len;
len -= iov[0].iov_len;
if (len)
{
iov[1].iov_base = sb->sb_data;
iov[1].iov_len = sb->sb_rptr - sb->sb_data;
if (iov[1].iov_len > len)
iov[1].iov_len = len;
total = iov[0].iov_len + iov[1].iov_len;
if (total > mss)
{
lss = total % mss;
if (iov[1].iov_len > lss)
{
iov[1].iov_len -= lss;
n = 2;
}
else
{
lss -= iov[1].iov_len;
iov[0].iov_len -= lss;
n = 1;
}
}
else
n = 2;
}
else
{
if (iov[0].iov_len > mss)
iov[0].iov_len -= iov[0].iov_len%mss;
n = 1;
}
}
#ifdef HAVE_READV
nn = readv(so->s, (struct iovec *)iov, n);
#else
nn = recv(so->s, iov[0].iov_base, iov[0].iov_len, (so->so_tcpcb->t_force? MSG_OOB:0));
#endif
if (nn < 0)
sockerr = errno; /* save it, as it may be clobbered by logging */
else
sockerr = 0;
Log2(("%s: read(1) nn = %d bytes\n", RT_GCC_EXTENSION __PRETTY_FUNCTION__, nn));
Log2(("%s: so = %R[natsock] so->so_snd = %R[sbuf]\n", RT_GCC_EXTENSION __PRETTY_FUNCTION__, so, sb));
if (nn <= 0)
{
#ifdef RT_OS_WINDOWS
/*
* Windows reports ESHUTDOWN after SHUT_RD (SD_RECEIVE)
* instead of just returning EOF indication.
*/
if (nn < 0 && sockerr == ESHUTDOWN)
{
nn = 0;
sockerr = 0;
}
#endif
if (nn == 0) /* XXX: should this be inside #if defined(RT_OS_WINDOWS)? */
{
/*
* Special case for WSAEnumNetworkEvents: If we receive 0 bytes that
* _could_ mean that the connection is closed. But we will receive an
* FD_CLOSE event later if the connection was _really_ closed. With
* www.youtube.com I see this very often. Closing the socket too early
* would be dangerous.
*/
int status;
unsigned long pending = 0;
status = ioctlsocket(so->s, FIONREAD, &pending);
if (status < 0)
Log(("NAT:%s: error in WSAIoctl: %d\n", RT_GCC_EXTENSION __PRETTY_FUNCTION__, errno));
if (pending != 0)
{
SOCKET_UNLOCK(so);
STAM_PROFILE_STOP(&pData->StatIOread, a);
return 0;
}
}
if ( nn < 0
&& soIgnorableErrorCode(sockerr))
{
SOCKET_UNLOCK(so);
STAM_PROFILE_STOP(&pData->StatIOread, a);
return 0;
}
else
{
int fUninitializedTemplate = 0;
int shuterr;
fUninitializedTemplate = RT_BOOL(( sototcpcb(so)
&& ( sototcpcb(so)->t_template.ti_src.s_addr == INADDR_ANY
|| sototcpcb(so)->t_template.ti_dst.s_addr == INADDR_ANY)));
/* nn == 0 means peer has performed an orderly shutdown */
Log2(("%s: disconnected, nn = %d, errno = %d (%s)\n",
RT_GCC_EXTENSION __PRETTY_FUNCTION__, nn, sockerr, strerror(sockerr)));
shuterr = sofcantrcvmore(so);
if (!sockerr && !shuterr && !fUninitializedTemplate)
tcp_sockclosed(pData, sototcpcb(so));
else
{
LogRel2(("NAT: sockerr %d, shuterr %d - %R[natsock]\n", sockerr, shuterr, so));
tcp_drop(pData, sototcpcb(so), sockerr);
}
SOCKET_UNLOCK(so);
STAM_PROFILE_STOP(&pData->StatIOread, a);
return -1;
}
}
STAM_STATS(
if (n == 1)
{
STAM_COUNTER_INC(&pData->StatIORead_in_1);
STAM_COUNTER_ADD(&pData->StatIORead_in_1_bytes, nn);
}
else
{
STAM_COUNTER_INC(&pData->StatIORead_in_2);
STAM_COUNTER_ADD(&pData->StatIORead_in_2_1st_bytes, nn);
}
);
struct socket *
udp_listen(PNATState pData, u_int32_t bind_addr, u_int port, u_int32_t laddr, u_int lport, int flags)
{
struct sockaddr_in addr;
struct socket *so;
socklen_t addrlen = sizeof(struct sockaddr_in);
int opt = 1;
LogFlowFunc(("ENTER: bind_addr:%RTnaipv4, port:%d, laddr:%RTnaipv4, lport:%d, flags:%x\n",
bind_addr, RT_N2H_U16(port), laddr, RT_N2H_U16(lport), flags));
if ((so = socreate()) == NULL)
{
LogFlowFunc(("LEAVE: NULL\n"));
return NULL;
}
so->s = socket(AF_INET, SOCK_DGRAM, 0);
if (so->s == -1)
{
LogRel(("NAT: can't create datagram socket\n"));
RTMemFree(so);
LogFlowFunc(("LEAVE: NULL\n"));
return NULL;
}
so->so_expire = curtime + SO_EXPIRE;
so->so_type = IPPROTO_UDP;
fd_nonblock(so->s);
so->so_sottl = 0;
so->so_sotos = 0;
so->so_sodf = -1;
SOCKET_LOCK_CREATE(so);
QSOCKET_LOCK(udb);
insque(pData, so, &udb);
NSOCK_INC();
QSOCKET_UNLOCK(udb);
memset(&addr, 0, sizeof(addr));
#ifdef RT_OS_DARWIN
addr.sin_len = sizeof(addr);
#endif
addr.sin_family = AF_INET;
addr.sin_addr.s_addr = bind_addr;
addr.sin_port = port;
if (bind(so->s,(struct sockaddr *)&addr, addrlen) < 0)
{
LogRel(("NAT: udp bind to %RTnaipv4:%d failed, error %d\n",
addr.sin_addr, RT_N2H_U16(port), errno));
udp_detach(pData, so);
LogFlowFunc(("LEAVE: NULL\n"));
return NULL;
}
setsockopt(so->s, SOL_SOCKET, SO_REUSEADDR,(char *)&opt, sizeof(int));
/* setsockopt(so->s, SOL_SOCKET, SO_OOBINLINE,(char *)&opt, sizeof(int)); */
getsockname(so->s,(struct sockaddr *)&addr,&addrlen);
so->so_hladdr = addr.sin_addr;
so->so_hlport = addr.sin_port;
/* XXX: wtf are we setting so_faddr/so_fport here? */
so->so_fport = addr.sin_port;
#if 0
/* The original check was completely broken, as the commented out
* if statement was always true (INADDR_ANY=0). */
/** @todo vvl - alias_addr should be set (if required)
* later by liabalias module.
*/
if (addr.sin_addr.s_addr == 0 || addr.sin_addr.s_addr == loopback_addr.s_addr)
so->so_faddr = alias_addr;
else
#endif
so->so_faddr = addr.sin_addr;
so->so_lport = lport;
so->so_laddr.s_addr = laddr;
if (flags != SS_FACCEPTONCE)
so->so_expire = 0;
so->so_state = SS_ISFCONNECTED;
LogFlowFunc(("LEAVE: %R[natsock]\n", so));
return so;
}