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
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/kernel.h>
40 #include <sys/socket.h>
41 #include <sys/socketvar.h>
42 #include <sys/thread.h>
43 #include <sys/sysctl.h>
44 #include <sys/globaldata.h>
47 #include <net/netisr2.h>
48 #include <net/toeplitz2.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>
63 * Toeplitz hash functions - the idea is to match the hardware.
66 INP_MPORT_HASH_UDP(in_addr_t faddr, in_addr_t laddr,
67 in_port_t fport, in_port_t lport)
69 return toeplitz_hash(toeplitz_rawhash_addr(faddr, laddr));
73 INP_MPORT_HASH_TCP(in_addr_t faddr, in_addr_t laddr,
74 in_port_t fport, in_port_t lport)
77 toeplitz_rawhash_addrport(faddr, laddr, fport, lport));
81 * Map a network address to a processor.
84 tcp_addrcpu(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
86 return (netisr_hashcpu(INP_MPORT_HASH_TCP(faddr, laddr, fport, lport)));
90 * Not implemented yet, use protocol thread 0
93 udp_addrcpu(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
96 return (netisr_hashcpu(INP_MPORT_HASH_UDP(faddr, laddr, fport, lport)));
103 * If the packet is a valid IP datagram, upon returning of this function
104 * following things are promised:
106 * o IP header (including any possible IP options) and any data preceding
107 * IP header (usually linker layer header) are in one mbuf (m_len).
108 * o IP header length is not less than the minimum (sizeof(struct ip)).
109 * o IP total length is not less than IP header length.
110 * o IP datagram resides completely in the mbuf chain,
111 * i.e. pkthdr.len >= IP total length.
113 * If the packet is a UDP datagram,
114 * o IP header (including any possible IP options) and UDP header are in
116 * o IP total length is not less than (IP header length + UDP header length).
118 * If the packet is a TCP segment,
119 * o IP header (including any possible IP options) and TCP header (including
120 * any possible TCP options) are in one mbuf (m_len).
121 * o TCP header length is not less than the minimum (sizeof(struct tcphdr)).
122 * o IP total length is not less than (IP header length + TCP header length).
125 ip_lengthcheck(struct mbuf **mp, int hoff)
127 struct mbuf *m = *mp;
129 int len, iphlen, iplen;
131 int thoff; /* TCP data offset */
133 len = hoff + sizeof(struct ip);
135 /* The packet must be at least the size of an IP header. */
136 if (m->m_pkthdr.len < len) {
137 ipstat.ips_tooshort++;
141 /* The fixed IP header must reside completely in the first mbuf. */
142 if (m->m_len < len) {
143 m = m_pullup(m, len);
145 ipstat.ips_toosmall++;
150 ip = mtodoff(m, struct ip *, hoff);
152 /* Bound check the packet's stated IP header length. */
153 iphlen = ip->ip_hl << 2;
154 if (iphlen < sizeof(struct ip)) { /* minimum header length */
155 ipstat.ips_badhlen++;
159 /* The full IP header must reside completely in the one mbuf. */
160 if (m->m_len < hoff + iphlen) {
161 m = m_pullup(m, hoff + iphlen);
163 ipstat.ips_badhlen++;
166 ip = mtodoff(m, struct ip *, hoff);
169 iplen = ntohs(ip->ip_len);
172 * Check that the amount of data in the buffers is as
173 * at least much as the IP header would have us expect.
175 if (m->m_pkthdr.len < hoff + iplen) {
176 ipstat.ips_tooshort++;
181 * Fragments other than the first fragment don't have much
182 * length information.
184 if (ntohs(ip->ip_off) & IP_OFFMASK)
188 * The TCP/IP or UDP/IP header must be entirely contained within
189 * the first fragment of a packet. Packet filters will break if they
192 * Since the packet will be trimmed to ip_len we must also make sure
193 * the potentially trimmed down length is still sufficient to hold
198 if (iplen < iphlen + sizeof(struct tcphdr)) {
199 ++tcpstat.tcps_rcvshort;
202 if (m->m_len < hoff + iphlen + sizeof(struct tcphdr)) {
203 m = m_pullup(m, hoff + iphlen + sizeof(struct tcphdr));
205 tcpstat.tcps_rcvshort++;
208 ip = mtodoff(m, struct ip *, hoff);
210 th = (struct tcphdr *)((caddr_t)ip + iphlen);
211 thoff = th->th_off << 2;
212 if (thoff < sizeof(struct tcphdr) ||
213 thoff + iphlen > ntohs(ip->ip_len)) {
214 tcpstat.tcps_rcvbadoff++;
217 if (m->m_len < hoff + iphlen + thoff) {
218 m = m_pullup(m, hoff + iphlen + thoff);
220 tcpstat.tcps_rcvshort++;
226 if (iplen < iphlen + sizeof(struct udphdr)) {
227 ++udp_stat.udps_hdrops;
230 if (m->m_len < hoff + iphlen + sizeof(struct udphdr)) {
231 m = m_pullup(m, hoff + iphlen + sizeof(struct udphdr));
233 udp_stat.udps_hdrops++;
240 if (iplen < iphlen) {
247 m->m_flags |= M_LENCHECKED;
259 * Assign a protocol processing thread to a packet. The IP header is at
260 * offset (hoff) in the packet (i.e. the mac header might still be intact).
262 * This function can blow away the mbuf if the packet is malformed.
265 ip_hashfn(struct mbuf **mptr, int hoff, int dir)
274 if (!ip_lengthcheck(mptr, hoff))
278 ip = mtodoff(m, struct ip *, hoff);
279 iphlen = ip->ip_hl << 2;
282 * XXX generic packet handling defrag on CPU 0 for now.
284 if (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) {
291 th = (struct tcphdr *)((caddr_t)ip + iphlen);
292 hash = INP_MPORT_HASH_TCP(ip->ip_src.s_addr, ip->ip_dst.s_addr,
293 th->th_sport, th->th_dport);
297 uh = (struct udphdr *)((caddr_t)ip + iphlen);
298 hash = INP_MPORT_HASH_UDP(ip->ip_src.s_addr, ip->ip_dst.s_addr,
299 uh->uh_sport, uh->uh_dport);
307 m->m_flags |= M_HASH;
308 m->m_pkthdr.hash = hash;
312 ip_hashfn_in(struct mbuf **mptr, int hoff)
314 ip_hashfn(mptr, hoff, IP_MPORT_IN);
318 * Verify and adjust the hash value of the packet.
320 * Unlike ip_hashfn(), the packet content is not accessed. The packet info
321 * (pi) and the hash of the packet (m_pkthdr.hash) is used instead.
323 * Caller has already made sure that m_pkthdr.hash is valid, i.e. m_flags
327 ip_hashcheck(struct mbuf *m, const struct pktinfo *pi)
329 KASSERT((m->m_flags & M_HASH), ("no valid packet hash"));
332 * XXX generic packet handling defrag on CPU 0 for now.
334 if (pi->pi_flags & PKTINFO_FLAG_FRAG) {
335 m->m_pkthdr.hash = 0;
339 switch (pi->pi_l3proto) {
345 /* Let software calculate the hash */
346 m->m_flags &= ~M_HASH;
352 * This is used to map a socket to a message port for sendmsg() and friends.
353 * It is not called for any other purpose. In the case of TCP we just return
354 * the port already installed in the socket.
357 tcp_soport(struct socket *so, struct sockaddr *nam,
358 struct mbuf **dummy __unused)
364 * Used to route icmp messages to the proper protocol thread for ctlinput
368 tcp_ctlport(int cmd, struct sockaddr *sa, void *vip)
372 struct in_addr faddr;
375 faddr = ((struct sockaddr_in *)sa)->sin_addr;
376 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
378 if (ip == NULL || PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
380 * A new message will be allocated later to save necessary
381 * information and will be forwarded to all network protocol
382 * threads in the following way:
384 * (the the thread owns the msgport that we return here)
391 * [msg is kmalloc()ed]
394 * Later on, when the msg is received by netisr0:
396 * forwardmsg forwardmsg
397 * netisr0 ---------> netisr1 ---------> netisrN
400 return cpu0_ctlport(cmd, sa, vip);
402 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
403 cpu = tcp_addrcpu(faddr.s_addr, th->th_dport,
404 ip->ip_src.s_addr, th->th_sport);
406 return(netisr_cpuport(cpu));
410 tcp_addrport(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
412 return(netisr_cpuport(tcp_addrcpu(faddr, fport, laddr, lport)));
418 return(netisr_cpuport(0));
422 udp_addrport(in_addr_t faddr, in_port_t fport, in_addr_t laddr, in_port_t lport)
424 return(netisr_cpuport(udp_addrcpu(faddr, fport, laddr, lport)));
428 * Used to route icmp messages to the proper protocol thread for ctlinput
432 udp_ctlport(int cmd, struct sockaddr *sa, void *vip)
436 struct in_addr faddr;
439 faddr = ((struct sockaddr_in *)sa)->sin_addr;
440 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
442 if (PRC_IS_REDIRECT(cmd)) {
444 * See the comment in tcp_ctlport; the only difference
445 * is that message is forwarded to UDP protocol theads.
447 return cpu0_ctlport(cmd, sa, vip);
448 } else if (ip == NULL || cmd == PRC_HOSTDEAD) {
451 * Once UDP inpcbs are CPU localized, we should do
452 * the same forwarding as PRC_IS_REDIRECT(cmd)
456 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
458 cpu = udp_addrcpu(faddr.s_addr, ip->ip_src.s_addr,
459 uh->uh_dport, uh->uh_sport);
461 return (netisr_cpuport(cpu));