2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved.
5 * This code is derived from software contributed to The DragonFly Project
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * $DragonFly: src/sys/netinet/ip_demux.c,v 1.43 2008/10/28 04:35:12 sephe Exp $
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/kernel.h>
41 #include <sys/socket.h>
42 #include <sys/socketvar.h>
43 #include <sys/thread.h>
44 #include <sys/sysctl.h>
45 #include <sys/globaldata.h>
48 #include <net/netisr.h>
50 #include <netinet/in_systm.h>
51 #include <netinet/in.h>
52 #include <netinet/in_var.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/ip.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/tcp.h>
57 #include <netinet/tcpip.h>
58 #include <netinet/tcp_var.h>
59 #include <netinet/udp.h>
60 #include <netinet/udp_var.h>
62 extern struct thread netisr_cpu[];
63 extern int udp_mpsafe_thread;
65 static struct thread tcp_thread[MAXCPU];
66 static struct thread udp_thread[MAXCPU];
69 INP_MPORT_HASH(in_addr_t faddr, in_addr_t laddr,
70 in_port_t fport, in_port_t lport)
73 * Use low order bytes.
76 #if (BYTE_ORDER == LITTLE_ENDIAN)
77 KASSERT(ncpus2 < 256, ("need different hash function")); /* XXX JH */
78 return (((faddr >> 24) ^ (fport >> 8) ^ (laddr >> 24) ^ (lport >> 8)) &
81 return ((faddr ^ fport ^ laddr ^ lport) & ncpus2_mask);
86 * If the packet is a valid IP datagram, upon returning of this function
87 * following things are promised:
89 * o IP header (including any possible IP options) is in one mbuf (m_len).
90 * o IP header length is not less than the minimum (sizeof(struct ip)).
91 * o IP total length is not less than IP header length.
93 * If the packet is a UDP datagram,
94 * o IP header (including any possible IP options) and UDP header are in
96 * o IP total length is not less than (IP header length + UDP header length).
98 * If the packet is a TCP segment,
99 * o IP header (including any possible IP options) and TCP header (including
100 * any possible TCP options) are in one mbuf (m_len).
101 * o TCP header length is not less than the minimum (sizeof(struct tcphdr)).
102 * o IP total length is not less than (IP header length + TCP header length).
105 ip_lengthcheck(struct mbuf **mp)
107 struct mbuf *m = *mp;
111 int thoff; /* TCP data offset */
113 /* The packet must be at least the size of an IP header. */
114 if (m->m_pkthdr.len < sizeof(struct ip)) {
115 ipstat.ips_tooshort++;
119 /* The fixed IP header must reside completely in the first mbuf. */
120 if (m->m_len < sizeof(struct ip)) {
121 m = m_pullup(m, sizeof(struct ip));
123 ipstat.ips_toosmall++;
128 ip = mtod(m, struct ip *);
130 /* Bound check the packet's stated IP header length. */
131 iphlen = ip->ip_hl << 2;
132 if (iphlen < sizeof(struct ip)) { /* minimum header length */
133 ipstat.ips_badhlen++;
137 /* The full IP header must reside completely in the one mbuf. */
138 if (m->m_len < iphlen) {
139 m = m_pullup(m, iphlen);
141 ipstat.ips_badhlen++;
144 ip = mtod(m, struct ip *);
147 iplen = ntohs(ip->ip_len);
150 * Fragments other than the first fragment don't have much
151 * length information.
153 if (ntohs(ip->ip_off) & IP_OFFMASK)
157 * The TCP/IP or UDP/IP header must be entirely contained within
158 * the first fragment of a packet. Packet filters will break if they
161 * Since the packet will be trimmed to ip_len we must also make sure
162 * the potentially trimmed down length is still sufficient to hold
167 if (iplen < iphlen + sizeof(struct tcphdr)) {
168 ++tcpstat.tcps_rcvshort;
171 if (m->m_len < iphlen + sizeof(struct tcphdr)) {
172 m = m_pullup(m, iphlen + sizeof(struct tcphdr));
174 tcpstat.tcps_rcvshort++;
177 ip = mtod(m, struct ip *);
179 th = (struct tcphdr *)((caddr_t)ip + iphlen);
180 thoff = th->th_off << 2;
181 if (thoff < sizeof(struct tcphdr) ||
182 thoff + iphlen > ntohs(ip->ip_len)) {
183 tcpstat.tcps_rcvbadoff++;
186 if (m->m_len < iphlen + thoff) {
187 m = m_pullup(m, iphlen + thoff);
189 tcpstat.tcps_rcvshort++;
195 if (iplen < iphlen + sizeof(struct udphdr)) {
196 ++udpstat.udps_hdrops;
199 if (m->m_len < iphlen + sizeof(struct udphdr)) {
200 m = m_pullup(m, iphlen + sizeof(struct udphdr));
202 udpstat.udps_hdrops++;
209 if (iplen < iphlen) {
227 * Map a packet to a protocol processing thread and return the thread's port.
228 * If an error occurs, the passed mbuf will be freed, *mptr will be set
229 * to NULL, and NULL will be returned. If no error occurs, the passed mbuf
230 * may be modified and a port pointer will be returned.
233 ip_mport(struct mbuf **mptr, int dir)
240 int thoff; /* TCP data offset */
244 if (!ip_lengthcheck(mptr))
248 ip = mtod(m, struct ip *);
249 iphlen = ip->ip_hl << 2;
252 * XXX generic packet handling defrag on CPU 0 for now.
254 if (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK))
255 return (&netisr_cpu[0].td_msgport);
259 th = (struct tcphdr *)((caddr_t)ip + iphlen);
260 thoff = th->th_off << 2;
261 cpu = INP_MPORT_HASH(ip->ip_src.s_addr, ip->ip_dst.s_addr,
262 th->th_sport, th->th_dport);
263 port = &tcp_thread[cpu].td_msgport;
266 uh = (struct udphdr *)((caddr_t)ip + iphlen);
268 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
269 (dir == IP_MPORT_IN &&
270 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))) {
273 cpu = INP_MPORT_HASH(ip->ip_src.s_addr,
274 ip->ip_dst.s_addr, uh->uh_sport, uh->uh_dport);
276 port = &udp_thread[cpu].td_msgport;
279 port = &netisr_cpu[0].td_msgport;
287 ip_mport_in(struct mbuf **mptr)
289 return ip_mport(mptr, IP_MPORT_IN);
293 * Map a TCP socket to a protocol processing thread.
296 tcp_soport(struct socket *so, struct sockaddr *nam __unused,
297 struct mbuf **dummy __unused, int req)
301 /* The following processing all take place on Protocol Thread 0. */
302 if (req == PRU_BIND || req == PRU_CONNECT || req == PRU_ATTACH ||
304 return (&tcp_thread[0].td_msgport);
307 if (!inp) /* connection reset by peer */
308 return (&tcp_thread[0].td_msgport);
311 * Already bound and connected or listening. For TCP connections,
312 * the (faddr, fport, laddr, lport) association cannot change now.
314 * Note: T/TCP code needs some reorganization to fit into
317 * Rely on type-stable memory and check in protocol handler
318 * to fix race condition here w/ deallocation of inp. XXX JH
320 return (&tcp_thread[INP_MPORT_HASH(inp->inp_faddr.s_addr,
321 inp->inp_laddr.s_addr, inp->inp_fport, inp->inp_lport)].td_msgport);
325 * Used to route icmp messages to the proper protocol thread for ctlinput
329 tcp_ctlport(int cmd, struct sockaddr *sa, void *vip)
333 struct in_addr faddr;
336 faddr = ((struct sockaddr_in *)sa)->sin_addr;
337 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
342 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
343 cpu = tcp_addrcpu(faddr.s_addr, th->th_dport,
344 ip->ip_src.s_addr, th->th_sport);
346 return(&tcp_thread[cpu].td_msgport);
350 tcp_addrport(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
352 return (&tcp_thread[tcp_addrcpu(faddr, fport,
353 laddr, lport)].td_msgport);
357 * Map a UDP socket to a protocol processing thread.
360 udp_soport(struct socket *so, struct sockaddr *nam __unused,
361 struct mbuf **dummy __unused, int req)
366 * The following processing all take place on Protocol Thread 0:
368 * attach() has a null socket parameter
369 * Fast and slow timeouts pass in null socket parameter
371 if (req == PRU_BIND || so == NULL)
372 return (&udp_thread[0].td_msgport);
376 if (IN_MULTICAST(ntohl(inp->inp_laddr.s_addr)))
377 return (&udp_thread[0].td_msgport);
380 * Rely on type-stable memory and check in protocol handler
381 * to fix race condition here w/ deallocation of inp. XXX JH
384 return (&udp_thread[INP_MPORT_HASH(inp->inp_faddr.s_addr,
385 inp->inp_laddr.s_addr, inp->inp_fport, inp->inp_lport)].td_msgport);
389 * Used to route icmp messages to the proper protocol thread for ctlinput
393 udp_ctlport(int cmd, struct sockaddr *sa, void *vip)
397 struct in_addr faddr;
400 faddr = ((struct sockaddr_in *)sa)->sin_addr;
401 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
406 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
408 cpu = INP_MPORT_HASH(faddr.s_addr, ip->ip_src.s_addr,
409 uh->uh_dport, uh->uh_sport);
411 return (&udp_thread[cpu].td_msgport);
415 * Map a network address to a processor.
418 tcp_addrcpu(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
420 return (INP_MPORT_HASH(faddr, laddr, fport, lport));
424 udp_addrcpu(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
426 if (IN_MULTICAST(ntohl(laddr)))
429 return (INP_MPORT_HASH(faddr, laddr, fport, lport));
433 * Return LWKT port for cpu.
438 return (&tcp_thread[cpu].td_msgport);
442 tcp_thread_init(void)
446 for (cpu = 0; cpu < ncpus2; cpu++) {
447 lwkt_create(tcpmsg_service_loop, NULL, NULL,
448 &tcp_thread[cpu], TDF_NETWORK | TDF_MPSAFE, cpu,
449 "tcp_thread %d", cpu);
450 netmsg_service_port_init(&tcp_thread[cpu].td_msgport);
455 udp_thread_init(void)
459 for (cpu = 0; cpu < ncpus2; cpu++) {
460 lwkt_create(netmsg_service_loop, &udp_mpsafe_thread, NULL,
461 &udp_thread[cpu], TDF_NETWORK | TDF_MPSAFE, cpu,
462 "udp_thread %d", cpu);
463 netmsg_service_port_init(&udp_thread[cpu].td_msgport);