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 1.45 2008/11/11 10:46:58 sephe Exp $
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
42 #include <sys/socket.h>
43 #include <sys/socketvar.h>
44 #include <sys/thread.h>
45 #include <sys/sysctl.h>
46 #include <sys/globaldata.h>
49 #include <net/netisr.h>
50 #include <net/toeplitz2.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/in.h>
54 #include <netinet/in_var.h>
55 #include <netinet/in_pcb.h>
56 #include <netinet/ip.h>
57 #include <netinet/ip_var.h>
58 #include <netinet/tcp.h>
59 #include <netinet/tcpip.h>
60 #include <netinet/tcp_var.h>
61 #include <netinet/udp.h>
62 #include <netinet/udp_var.h>
64 extern int udp_mpsafe_thread;
67 * Toeplitz hash functions - the idea is to match the hardware.
70 INP_MPORT_HASH_UDP(in_addr_t faddr, in_addr_t laddr,
71 in_port_t fport, in_port_t lport)
73 return toeplitz_hash(toeplitz_rawhash_addr(faddr, laddr));
77 INP_MPORT_HASH_TCP(in_addr_t faddr, in_addr_t laddr,
78 in_port_t fport, in_port_t lport)
81 toeplitz_rawhash_addrport(faddr, laddr, fport, lport));
85 * Map a network address to a processor.
88 tcp_addrcpu(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
90 return (INP_MPORT_HASH_TCP(faddr, laddr, fport, lport));
94 * Not implemented yet, use protocol thread 0
97 udp_addrcpu(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
100 return (INP_MPORT_HASH_UDP(faddr, laddr, fport, lport));
107 * If the packet is a valid IP datagram, upon returning of this function
108 * following things are promised:
110 * o IP header (including any possible IP options) and any data preceding
111 * IP header (usually linker layer header) are in one mbuf (m_len).
112 * o IP header length is not less than the minimum (sizeof(struct ip)).
113 * o IP total length is not less than IP header length.
114 * o IP datagram resides completely in the mbuf chain,
115 * i.e. pkthdr.len >= IP total length.
117 * If the packet is a UDP datagram,
118 * o IP header (including any possible IP options) and UDP header are in
120 * o IP total length is not less than (IP header length + UDP header length).
122 * If the packet is a TCP segment,
123 * o IP header (including any possible IP options) and TCP header (including
124 * any possible TCP options) are in one mbuf (m_len).
125 * o TCP header length is not less than the minimum (sizeof(struct tcphdr)).
126 * o IP total length is not less than (IP header length + TCP header length).
129 ip_lengthcheck(struct mbuf **mp, int hoff)
131 struct mbuf *m = *mp;
133 int len, iphlen, iplen;
135 int thoff; /* TCP data offset */
137 len = hoff + sizeof(struct ip);
139 /* The packet must be at least the size of an IP header. */
140 if (m->m_pkthdr.len < len) {
141 ipstat.ips_tooshort++;
145 /* The fixed IP header must reside completely in the first mbuf. */
146 if (m->m_len < len) {
147 m = m_pullup(m, len);
149 ipstat.ips_toosmall++;
154 ip = mtodoff(m, struct ip *, hoff);
156 /* Bound check the packet's stated IP header length. */
157 iphlen = ip->ip_hl << 2;
158 if (iphlen < sizeof(struct ip)) { /* minimum header length */
159 ipstat.ips_badhlen++;
163 /* The full IP header must reside completely in the one mbuf. */
164 if (m->m_len < hoff + iphlen) {
165 m = m_pullup(m, hoff + iphlen);
167 ipstat.ips_badhlen++;
170 ip = mtodoff(m, struct ip *, hoff);
173 iplen = ntohs(ip->ip_len);
176 * Check that the amount of data in the buffers is as
177 * at least much as the IP header would have us expect.
179 if (m->m_pkthdr.len < hoff + iplen) {
180 ipstat.ips_tooshort++;
185 * Fragments other than the first fragment don't have much
186 * length information.
188 if (ntohs(ip->ip_off) & IP_OFFMASK)
192 * The TCP/IP or UDP/IP header must be entirely contained within
193 * the first fragment of a packet. Packet filters will break if they
196 * Since the packet will be trimmed to ip_len we must also make sure
197 * the potentially trimmed down length is still sufficient to hold
202 if (iplen < iphlen + sizeof(struct tcphdr)) {
203 ++tcpstat.tcps_rcvshort;
206 if (m->m_len < hoff + iphlen + sizeof(struct tcphdr)) {
207 m = m_pullup(m, hoff + iphlen + sizeof(struct tcphdr));
209 tcpstat.tcps_rcvshort++;
212 ip = mtodoff(m, struct ip *, hoff);
214 th = (struct tcphdr *)((caddr_t)ip + iphlen);
215 thoff = th->th_off << 2;
216 if (thoff < sizeof(struct tcphdr) ||
217 thoff + iphlen > ntohs(ip->ip_len)) {
218 tcpstat.tcps_rcvbadoff++;
221 if (m->m_len < hoff + iphlen + thoff) {
222 m = m_pullup(m, hoff + iphlen + thoff);
224 tcpstat.tcps_rcvshort++;
230 if (iplen < iphlen + sizeof(struct udphdr)) {
231 ++udpstat.udps_hdrops;
234 if (m->m_len < hoff + iphlen + sizeof(struct udphdr)) {
235 m = m_pullup(m, hoff + iphlen + sizeof(struct udphdr));
237 udpstat.udps_hdrops++;
244 if (iplen < iphlen) {
251 m->m_flags |= M_LENCHECKED;
263 * Assign a protocol processing thread to a packet. The IP header is at
264 * offset (hoff) in the packet (i.e. the mac header might still be intact).
266 * This function can blow away the mbuf if the packet is malformed.
269 ip_cpufn(struct mbuf **mptr, int hoff, int dir)
276 int thoff; /* TCP data offset */
279 if (!ip_lengthcheck(mptr, hoff))
283 ip = mtodoff(m, struct ip *, hoff);
284 iphlen = ip->ip_hl << 2;
287 * XXX generic packet handling defrag on CPU 0 for now.
289 if (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) {
296 th = (struct tcphdr *)((caddr_t)ip + iphlen);
297 thoff = th->th_off << 2;
298 cpu = INP_MPORT_HASH_TCP(ip->ip_src.s_addr,
305 uh = (struct udphdr *)((caddr_t)ip + iphlen);
307 cpu = INP_MPORT_HASH_UDP(ip->ip_src.s_addr,
318 m->m_flags |= M_HASH;
319 m->m_pkthdr.hash = cpu;
323 ip_cpufn_in(struct mbuf **mptr, int hoff)
325 ip_cpufn(mptr, hoff, IP_MPORT_IN);
329 * Verify and adjust the hash value of the packet.
331 * Unlike ip_cpufn(), the packet content is not accessed. The packet info
332 * (pi) and the hash of the packet (m_pkthdr.hash) is used instead.
334 * Caller has already made sure that m_pkthdr.hash is valid, i.e. m_flags
338 ip_hashcheck(struct mbuf *m, const struct pktinfo *pi)
340 KASSERT((m->m_flags & M_HASH), ("no valid packet hash\n"));
341 KASSERT(m->m_pkthdr.hash < ncpus2,
342 ("invalid packet hash %#x\n", m->m_pkthdr.hash));
345 * XXX generic packet handling defrag on CPU 0 for now.
347 if (pi->pi_flags & PKTINFO_FLAG_FRAG) {
348 m->m_pkthdr.hash = 0;
352 switch (pi->pi_l3proto) {
358 /* Let software calculate the hash */
359 m->m_flags &= ~M_HASH;
365 * This is used to map a socket to a message port for sendmsg() and friends.
366 * It is not called for any other purpose. In the case of TCP we just return
367 * the port already installed in the socket.
370 tcp_soport(struct socket *so, struct sockaddr *nam,
371 struct mbuf **dummy __unused)
377 * Used to route icmp messages to the proper protocol thread for ctlinput
381 tcp_ctlport(int cmd, struct sockaddr *sa, void *vip)
385 struct in_addr faddr;
388 faddr = ((struct sockaddr_in *)sa)->sin_addr;
389 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
391 if (ip == NULL || PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
393 * A new message will be allocated later to save necessary
394 * information and will be forwarded to all network protocol
395 * threads in the following way:
397 * (the the thread owns the msgport that we return here)
404 * [msg is kmalloc()ed]
407 * Later on, when the msg is received by netisr0:
409 * forwardmsg forwardmsg
410 * netisr0 ---------> netisr1 ---------> netisrN
413 return cpu0_ctlport(cmd, sa, vip);
415 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
416 cpu = tcp_addrcpu(faddr.s_addr, th->th_dport,
417 ip->ip_src.s_addr, th->th_sport);
419 return(cpu_portfn(cpu));
423 tcp_addrport(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
425 return(cpu_portfn(tcp_addrcpu(faddr, fport, laddr, lport)));
431 return(cpu_portfn(0));
435 udp_addrport(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
437 return(cpu_portfn(udp_addrcpu(faddr, fport, laddr, lport)));
441 * Used to route icmp messages to the proper protocol thread for ctlinput
445 udp_ctlport(int cmd, struct sockaddr *sa, void *vip)
449 struct in_addr faddr;
452 faddr = ((struct sockaddr_in *)sa)->sin_addr;
453 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
455 if (PRC_IS_REDIRECT(cmd)) {
457 * See the comment in tcp_ctlport; the only difference
458 * is that message is forwarded to UDP protocol theads.
460 return cpu0_ctlport(cmd, sa, vip);
461 } else if (ip == NULL || cmd == PRC_HOSTDEAD) {
464 * Once UDP inpcbs are CPU localized, we should do
465 * the same forwarding as PRC_IS_REDIRECT(cmd)
469 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
471 cpu = udp_addrcpu(faddr.s_addr, ip->ip_src.s_addr,
472 uh->uh_dport, uh->uh_sport);
474 return (cpu_portfn(cpu));