/* m->m_data points at ip packet header
* m->m_len length ip packet
* ip->ip_len length data (IPDU)
*/
void
udp_input(PNATState pData, register struct mbuf *m, int iphlen)
{
register struct ip *ip;
register struct udphdr *uh;
int len;
struct ip save_ip;
struct socket *so;
int ret;
int ttl, tos;
LogFlowFunc(("ENTER: m = %p, iphlen = %d\n", m, iphlen));
ip = mtod(m, struct ip *);
Log2(("%RTnaipv4 iphlen = %d\n", ip->ip_dst, iphlen));
udpstat.udps_ipackets++;
/*
* Strip IP options, if any; should skip this,
* make available to user, and use on returned packets,
* but we don't yet have a way to check the checksum
* with options still present.
*/
if (iphlen > sizeof(struct ip))
{
ip_stripoptions(m, (struct mbuf *)0);
iphlen = sizeof(struct ip);
}
/*
* Get IP and UDP header together in first mbuf.
*/
ip = mtod(m, struct ip *);
uh = (struct udphdr *)((caddr_t)ip + iphlen);
/*
* Make mbuf data length reflect UDP length.
* If not enough data to reflect UDP length, drop.
*/
len = RT_N2H_U16((u_int16_t)uh->uh_ulen);
Assert(ip->ip_len + iphlen == (ssize_t)m_length(m, NULL));
if (ip->ip_len != len)
{
if (len > ip->ip_len)
{
udpstat.udps_badlen++;
Log3(("NAT: IP(id: %hd) has bad size\n", ip->ip_id));
goto bad_free_mbuf;
}
m_adj(m, len - ip->ip_len);
ip->ip_len = len;
}
/*
* Save a copy of the IP header in case we want restore it
* for sending an ICMP error message in response.
*/
save_ip = *ip;
save_ip.ip_len+= iphlen; /* tcp_input subtracts this */
/*
* Checksum extended UDP header and data.
*/
if (udpcksum && uh->uh_sum)
{
memset(((struct ipovly *)ip)->ih_x1, 0, 9);
((struct ipovly *)ip)->ih_len = uh->uh_ulen;
#if 0
/* keep uh_sum for ICMP reply */
uh->uh_sum = cksum(m, len + sizeof (struct ip));
if (uh->uh_sum)
{
#endif
if (cksum(m, len + iphlen))
{
udpstat.udps_badsum++;
Log3(("NAT: IP(id: %hd) has bad (udp) cksum\n", ip->ip_id));
goto bad_free_mbuf;
}
}
#if 0
}
#endif
/*
* handle DHCP/BOOTP
*/
if (uh->uh_dport == RT_H2N_U16_C(BOOTP_SERVER))
{
bootp_input(pData, m);
goto done_free_mbuf;
}
LogFunc(("uh src: %RTnaipv4:%d, dst: %RTnaipv4:%d\n",
ip->ip_src.s_addr, RT_N2H_U16(uh->uh_sport),
ip->ip_dst.s_addr, RT_N2H_U16(uh->uh_dport)));
/*
* handle DNS host resolver without creating a socket
*/
if ( pData->fUseHostResolver
&& uh->uh_dport == RT_H2N_U16_C(53)
&& CTL_CHECK(ip->ip_dst.s_addr, CTL_DNS))
{
struct sockaddr_in dst, src;
src.sin_addr.s_addr = ip->ip_dst.s_addr;
src.sin_port = uh->uh_dport;
dst.sin_addr.s_addr = ip->ip_src.s_addr;
dst.sin_port = uh->uh_sport;
m_adj(m, sizeof(struct udpiphdr));
m = hostresolver(pData, m, ip->ip_src.s_addr, uh->uh_sport);
if (m == NULL)
goto done_free_mbuf;
slirpMbufTagService(pData, m, CTL_DNS);
udp_output2(pData, NULL, m, &src, &dst, IPTOS_LOWDELAY);
LogFlowFuncLeave();
return;
}
/*
* handle TFTP
*/
if ( uh->uh_dport == RT_H2N_U16_C(TFTP_SERVER)
&& CTL_CHECK(ip->ip_dst.s_addr, CTL_TFTP))
{
if (pData->pvTftpSessions)
slirpTftpInput(pData, m);
goto done_free_mbuf;
}
/*
* XXX: DNS proxy currently relies on the fact that each socket
* only serves one request.
*/
if ( pData->fUseDnsProxy
&& CTL_CHECK(ip->ip_dst.s_addr, CTL_DNS)
&& (uh->uh_dport == RT_H2N_U16_C(53)))
{
so = NULL;
goto new_socket;
}
/*
* Locate pcb for datagram.
*/
so = udp_last_so;
if ( so->so_lport != uh->uh_sport
|| so->so_laddr.s_addr != ip->ip_src.s_addr)
{
struct socket *tmp;
for (tmp = udb.so_next; tmp != &udb; tmp = tmp->so_next)
{
if ( tmp->so_lport == uh->uh_sport
&& tmp->so_laddr.s_addr == ip->ip_src.s_addr)
{
so = tmp;
break;
}
}
if (tmp == &udb)
so = NULL;
else
{
udpstat.udpps_pcbcachemiss++;
udp_last_so = so;
}
}
new_socket:
if (so == NULL)
{
/*
* If there's no socket for this packet,
* create one
*/
if ((so = socreate()) == NULL)
{
Log2(("NAT: IP(id: %hd) failed to create socket\n", ip->ip_id));
goto bad_free_mbuf;
}
/*
* Setup fields
*/
so->so_laddr = ip->ip_src;
so->so_lport = uh->uh_sport;
so->so_iptos = ip->ip_tos;
if (udp_attach(pData, so) <= 0)
{
Log2(("NAT: IP(id: %hd) udp_attach errno = %d (%s)\n",
ip->ip_id, errno, strerror(errno)));
sofree(pData, so);
goto bad_free_mbuf;
}
/* udp_last_so = so; */
/*
* XXXXX Here, check if it's in udpexec_list,
* and if it is, do the fork_exec() etc.
*/
}
so->so_faddr = ip->ip_dst; /* XXX */
so->so_fport = uh->uh_dport; /* XXX */
Assert(so->so_type == IPPROTO_UDP);
/*
* DNS proxy
*/
if ( pData->fUseDnsProxy
&& CTL_CHECK(ip->ip_dst.s_addr, CTL_DNS)
&& (uh->uh_dport == RT_H2N_U16_C(53)))
{
dnsproxy_query(pData, so, m, iphlen);
goto done_free_mbuf;
}
iphlen += sizeof(struct udphdr);
m->m_len -= iphlen;
m->m_data += iphlen;
ttl = ip->ip_ttl = save_ip.ip_ttl;
if (ttl != so->so_sottl) {
ret = setsockopt(so->s, IPPROTO_IP, IP_TTL,
(char *)&ttl, sizeof(ttl));
if (RT_LIKELY(ret == 0))
so->so_sottl = ttl;
}
tos = save_ip.ip_tos;
if (tos != so->so_sotos) {
ret = setsockopt(so->s, IPPROTO_IP, IP_TOS,
(char *)&tos, sizeof(tos));
if (RT_LIKELY(ret == 0))
so->so_sotos = tos;
}
{
/*
* Different OSes have different socket options for DF. We
* can't use IP_HDRINCL here as it's only valid for SOCK_RAW.
*/
# define USE_DF_OPTION(_Optname) \
const int dfopt = _Optname
#if defined(IP_MTU_DISCOVER)
USE_DF_OPTION(IP_MTU_DISCOVER);
#elif defined(IP_DONTFRAG) /* Solaris 11+, FreeBSD */
USE_DF_OPTION(IP_DONTFRAG);
#elif defined(IP_DONTFRAGMENT) /* Windows */
USE_DF_OPTION(IP_DONTFRAGMENT);
#else
USE_DF_OPTION(0);
#endif
if (dfopt) {
int df = (save_ip.ip_off & IP_DF) != 0;
#if defined(IP_MTU_DISCOVER)
df = df ? IP_PMTUDISC_DO : IP_PMTUDISC_DONT;
#endif
if (df != so->so_sodf) {
ret = setsockopt(so->s, IPPROTO_IP, dfopt,
(char *)&df, sizeof(df));
if (RT_LIKELY(ret == 0))
so->so_sodf = df;
}
}
}
if ( sosendto(pData, so, m) == -1
&& ( !soIgnorableErrorCode(errno)
&& errno != ENOTCONN))
{
m->m_len += iphlen;
m->m_data -= iphlen;
*ip = save_ip;
Log2(("NAT: UDP tx errno = %d (%s) on sent to %RTnaipv4\n",
errno, strerror(errno), ip->ip_dst));
icmp_error(pData, m, ICMP_UNREACH, ICMP_UNREACH_NET, 0, strerror(errno));
so->so_m = NULL;
LogFlowFuncLeave();
return;
}
if (so->so_m)
m_freem(pData, so->so_m); /* used for ICMP if error on sorecvfrom */
/* restore the orig mbuf packet */
m->m_len += iphlen;
m->m_data -= iphlen;
*ip = save_ip;
so->so_m = m; /* ICMP backup */
LogFlowFuncLeave();
return;
bad_free_mbuf:
Log2(("NAT: UDP(id: %hd) datagram to %RTnaipv4 with size(%d) claimed as bad\n",
ip->ip_id, &ip->ip_dst, ip->ip_len));
done_free_mbuf:
/* some services like bootp(built-in), dns(buildt-in) and dhcp don't need sockets
* and create new m'buffers to send them to guest, so we'll free their incomming
* buffers here.
*/
if (m != NULL)
m_freem(pData, m);
LogFlowFuncLeave();
return;
}
/*
* 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);
}
);
/* m->m_data points at ip packet header
* m->m_len length ip packet
* ip->ip_len length data (IPDU)
*/
void
udp_input(PNATState pData, register struct mbuf *m, int iphlen)
{
register struct ip *ip;
register struct udphdr *uh;
int len;
struct ip save_ip;
struct socket *so;
int ret;
int ttl;
LogFlowFunc(("ENTER: m = %p, iphlen = %d\n", m, iphlen));
ip = mtod(m, struct ip *);
Log2(("%RTnaipv4 iphlen = %d\n", ip->ip_dst, iphlen));
udpstat.udps_ipackets++;
/*
* Strip IP options, if any; should skip this,
* make available to user, and use on returned packets,
* but we don't yet have a way to check the checksum
* with options still present.
*/
if (iphlen > sizeof(struct ip))
{
ip_stripoptions(m, (struct mbuf *)0);
iphlen = sizeof(struct ip);
}
/*
* Get IP and UDP header together in first mbuf.
*/
ip = mtod(m, struct ip *);
uh = (struct udphdr *)((caddr_t)ip + iphlen);
/*
* Make mbuf data length reflect UDP length.
* If not enough data to reflect UDP length, drop.
*/
len = RT_N2H_U16((u_int16_t)uh->uh_ulen);
Assert((ip->ip_len == len));
Assert((ip->ip_len + iphlen == m_length(m, NULL)));
if (ip->ip_len != len)
{
if (len > ip->ip_len)
{
udpstat.udps_badlen++;
Log3(("NAT: IP(id: %hd) has bad size\n", ip->ip_id));
}
m_adj(m, len - ip->ip_len);
ip->ip_len = len;
}
/*
* Save a copy of the IP header in case we want restore it
* for sending an ICMP error message in response.
*/
save_ip = *ip;
save_ip.ip_len+= iphlen; /* tcp_input subtracts this */
/*
* Checksum extended UDP header and data.
*/
if (udpcksum && uh->uh_sum)
{
memset(((struct ipovly *)ip)->ih_x1, 0, 9);
((struct ipovly *)ip)->ih_len = uh->uh_ulen;
#if 0
/* keep uh_sum for ICMP reply */
uh->uh_sum = cksum(m, len + sizeof (struct ip));
if (uh->uh_sum)
{
#endif
if (cksum(m, len + iphlen))
{
udpstat.udps_badsum++;
Log3(("NAT: IP(id: %hd) has bad (udp) cksum\n", ip->ip_id));
goto bad_free_mbuf;
}
}
#if 0
}
#endif
/*
* handle DHCP/BOOTP
*/
if (uh->uh_dport == RT_H2N_U16_C(BOOTP_SERVER))
{
bootp_input(pData, m);
goto done_free_mbuf;
}
LogFunc(("uh src: %RTnaipv4:%d, dst: %RTnaipv4:%d\n", ip->ip_src, RT_H2N_U16_C(uh->uh_sport), ip->ip_dst, RT_H2N_U16_C(uh->uh_dport)));
if ( pData->fUseHostResolver
&& uh->uh_dport == RT_H2N_U16_C(53)
&& CTL_CHECK(ip->ip_dst.s_addr, CTL_DNS))
{
struct sockaddr_in dst, src;
src.sin_addr.s_addr = ip->ip_dst.s_addr;
src.sin_port = uh->uh_dport;
dst.sin_addr.s_addr = ip->ip_src.s_addr;
dst.sin_port = uh->uh_sport;
slirpMbufTagService(pData, m, CTL_DNS);
/* udp_output2() expects a pointer to the body of UDP packet. */
m->m_data += sizeof(struct udpiphdr);
m->m_len -= sizeof(struct udpiphdr);
udp_output2(pData, NULL, m, &src, &dst, IPTOS_LOWDELAY);
LogFlowFuncLeave();
return;
}
/*
* handle TFTP
*/
if ( uh->uh_dport == RT_H2N_U16_C(TFTP_SERVER)
&& CTL_CHECK(ip->ip_dst.s_addr, CTL_TFTP))
{
if (pData->pvTftpSessions)
slirpTftpInput(pData, m);
goto done_free_mbuf;
}
/*
* Locate pcb for datagram.
*/
so = udp_last_so;
if ( so->so_lport != uh->uh_sport
|| so->so_laddr.s_addr != ip->ip_src.s_addr)
{
struct socket *tmp;
for (tmp = udb.so_next; tmp != &udb; tmp = tmp->so_next)
{
if ( tmp->so_lport == uh->uh_sport
&& tmp->so_laddr.s_addr == ip->ip_src.s_addr)
{
so = tmp;
break;
}
}
if (tmp == &udb)
so = NULL;
else
{
udpstat.udpps_pcbcachemiss++;
udp_last_so = so;
}
}
if (so == NULL)
{
/*
* If there's no socket for this packet,
* create one
*/
if ((so = socreate()) == NULL)
{
Log2(("NAT: IP(id: %hd) failed to create socket\n", ip->ip_id));
goto bad_free_mbuf;
}
if (udp_attach(pData, so) <= 0)
{
Log2(("NAT: IP(id: %hd) udp_attach errno = %d (%s)\n",
ip->ip_id, errno, strerror(errno)));
sofree(pData, so);
goto bad_free_mbuf;
}
/*
* Setup fields
*/
/* udp_last_so = so; */
so->so_laddr = ip->ip_src;
so->so_lport = uh->uh_sport;
so->so_iptos = ip->ip_tos;
/*
* XXXXX Here, check if it's in udpexec_list,
* and if it is, do the fork_exec() etc.
*/
}
so->so_faddr = ip->ip_dst; /* XXX */
so->so_fport = uh->uh_dport; /* XXX */
Assert(so->so_type == IPPROTO_UDP);
/*
* DNS proxy
*/
if ( pData->fUseDnsProxy
&& CTL_CHECK(ip->ip_dst.s_addr, CTL_DNS)
&& (uh->uh_dport == RT_H2N_U16_C(53)))
{
dnsproxy_query(pData, so, m, iphlen);
goto done_free_mbuf;
}
iphlen += sizeof(struct udphdr);
m->m_len -= iphlen;
m->m_data += iphlen;
ttl = ip->ip_ttl = save_ip.ip_ttl;
ret = setsockopt(so->s, IPPROTO_IP, IP_TTL, (const char*)&ttl, sizeof(ttl));
if (ret < 0)
LogRel(("NAT: Error (%s) occurred while setting TTL(%d) attribute "
"of IP packet to socket %R[natsock]\n", strerror(errno), ip->ip_ttl, so));
if ( sosendto(pData, so, m) == -1
&& ( !soIgnorableErrorCode(errno)
&& errno != ENOTCONN))
{
m->m_len += iphlen;
m->m_data -= iphlen;
*ip = save_ip;
Log2(("NAT: UDP tx errno = %d (%s) on sent to %RTnaipv4\n",
errno, strerror(errno), ip->ip_dst));
#if 0
/* ICMP_SOURCEQUENCH haven't got any effect, the idea here
* inform guest about the exosting NAT resources with assumption that
* that guest reduce traffic. But it doesn't work
*/
if( errno == EAGAIN
|| errno == EWOULDBLOCK
|| errno == EINPROGRESS
|| errno == ENOTCONN)
icmp_error(pData, m, ICMP_SOURCEQUENCH, 0, 1, strerror(errno));
else
#endif
icmp_error(pData, m, ICMP_UNREACH, ICMP_UNREACH_NET, 0, strerror(errno));
so->so_m = NULL;
LogFlowFuncLeave();
return;
}
if (so->so_m)
m_freem(pData, so->so_m); /* used for ICMP if error on sorecvfrom */
/* restore the orig mbuf packet */
m->m_len += iphlen;
m->m_data -= iphlen;
*ip = save_ip;
so->so_m = m; /* ICMP backup */
LogFlowFuncLeave();
return;
bad_free_mbuf:
Log2(("NAT: UDP(id: %hd) datagram to %RTnaipv4 with size(%d) claimed as bad\n",
ip->ip_id, &ip->ip_dst, ip->ip_len));
done_free_mbuf:
/* some services like bootp(built-in), dns(buildt-in) and dhcp don't need sockets
* and create new m'buffers to send them to guest, so we'll free their incomming
* buffers here.
*/
m_freem(pData, m);
LogFlowFuncLeave();
return;
}
