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 udp_addrcpu(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
93 * NOTE: laddr could be multicast, since UDP socket could be
94 * bound to multicast address.
96 if (IN_MULTICAST(ntohl(faddr
)) || IN_MULTICAST(ntohl(laddr
))) {
97 /* XXX handle multicast on CPU0 for now */
100 return (netisr_hashcpu(INP_MPORT_HASH_UDP(faddr
, laddr
, fport
, lport
)));
104 * If the packet is a valid IP datagram, upon returning of this function
105 * following things are promised:
107 * o IP header (including any possible IP options) and any data preceding
108 * IP header (usually linker layer header) are in one mbuf (m_len).
109 * o IP header length is not less than the minimum (sizeof(struct ip)).
110 * o IP total length is not less than IP header length.
111 * o IP datagram resides completely in the mbuf chain,
112 * i.e. pkthdr.len >= IP total length.
114 * If the packet is a UDP datagram,
115 * o IP header (including any possible IP options) and UDP header are in
117 * o IP total length is not less than (IP header length + UDP header length).
119 * If the packet is a TCP segment,
120 * o IP header (including any possible IP options) and TCP header (including
121 * any possible TCP options) are in one mbuf (m_len).
122 * o TCP header length is not less than the minimum (sizeof(struct tcphdr)).
123 * o IP total length is not less than (IP header length + TCP header length).
126 ip_lengthcheck(struct mbuf
**mp
, int hoff
)
128 struct mbuf
*m
= *mp
;
130 int len
, iphlen
, iplen
;
132 int thoff
; /* TCP data offset */
134 len
= hoff
+ sizeof(struct ip
);
136 /* The packet must be at least the size of an IP header. */
137 if (m
->m_pkthdr
.len
< len
) {
138 ipstat
.ips_tooshort
++;
142 /* The fixed IP header must reside completely in the first mbuf. */
143 if (m
->m_len
< len
) {
144 m
= m_pullup(m
, len
);
146 ipstat
.ips_toosmall
++;
151 ip
= mtodoff(m
, struct ip
*, hoff
);
153 /* Bound check the packet's stated IP header length. */
154 iphlen
= ip
->ip_hl
<< 2;
155 if (iphlen
< sizeof(struct ip
)) { /* minimum header length */
156 ipstat
.ips_badhlen
++;
160 /* The full IP header must reside completely in the one mbuf. */
161 if (m
->m_len
< hoff
+ iphlen
) {
162 m
= m_pullup(m
, hoff
+ iphlen
);
164 ipstat
.ips_badhlen
++;
167 ip
= mtodoff(m
, struct ip
*, hoff
);
170 iplen
= ntohs(ip
->ip_len
);
173 * Check that the amount of data in the buffers is as
174 * at least much as the IP header would have us expect.
176 if (m
->m_pkthdr
.len
< hoff
+ iplen
) {
177 ipstat
.ips_tooshort
++;
182 * Fragments other than the first fragment don't have much
183 * length information.
185 if (ntohs(ip
->ip_off
) & IP_OFFMASK
)
189 * The TCP/IP or UDP/IP header must be entirely contained within
190 * the first fragment of a packet. Packet filters will break if they
193 * Since the packet will be trimmed to ip_len we must also make sure
194 * the potentially trimmed down length is still sufficient to hold
199 if (iplen
< iphlen
+ sizeof(struct tcphdr
)) {
200 ++tcpstat
.tcps_rcvshort
;
203 if (m
->m_len
< hoff
+ iphlen
+ sizeof(struct tcphdr
)) {
204 m
= m_pullup(m
, hoff
+ iphlen
+ sizeof(struct tcphdr
));
206 tcpstat
.tcps_rcvshort
++;
209 ip
= mtodoff(m
, struct ip
*, hoff
);
211 th
= (struct tcphdr
*)((caddr_t
)ip
+ iphlen
);
212 thoff
= th
->th_off
<< 2;
213 if (thoff
< sizeof(struct tcphdr
) ||
214 thoff
+ iphlen
> ntohs(ip
->ip_len
)) {
215 tcpstat
.tcps_rcvbadoff
++;
218 if (m
->m_len
< hoff
+ iphlen
+ thoff
) {
219 m
= m_pullup(m
, hoff
+ iphlen
+ thoff
);
221 tcpstat
.tcps_rcvshort
++;
227 if (iplen
< iphlen
+ sizeof(struct udphdr
)) {
228 ++udp_stat
.udps_hdrops
;
231 if (m
->m_len
< hoff
+ iphlen
+ sizeof(struct udphdr
)) {
232 m
= m_pullup(m
, hoff
+ iphlen
+ sizeof(struct udphdr
));
234 udp_stat
.udps_hdrops
++;
241 if (iplen
< iphlen
) {
248 m
->m_flags
|= M_LENCHECKED
;
260 * Assign a protocol processing thread to a packet. The IP header is at
261 * offset (hoff) in the packet (i.e. the mac header might still be intact).
263 * This function can blow away the mbuf if the packet is malformed.
266 ip_hashfn(struct mbuf
**mptr
, int hoff
)
275 if (!ip_lengthcheck(mptr
, hoff
))
279 ip
= mtodoff(m
, struct ip
*, hoff
);
280 iphlen
= ip
->ip_hl
<< 2;
282 if (ntohs(ip
->ip_off
) & (IP_MF
| IP_OFFMASK
)) {
283 hash
= toeplitz_hash(toeplitz_rawhash_addr(
284 ip
->ip_src
.s_addr
, ip
->ip_dst
.s_addr
));
290 th
= (struct tcphdr
*)((caddr_t
)ip
+ iphlen
);
291 hash
= INP_MPORT_HASH_TCP(ip
->ip_src
.s_addr
, ip
->ip_dst
.s_addr
,
292 th
->th_sport
, th
->th_dport
);
296 if (IN_MULTICAST(ntohl(ip
->ip_dst
.s_addr
))) {
297 /* XXX handle multicast on CPU0 for now */
301 uh
= (struct udphdr
*)((caddr_t
)ip
+ iphlen
);
302 hash
= INP_MPORT_HASH_UDP(ip
->ip_src
.s_addr
, ip
->ip_dst
.s_addr
,
303 uh
->uh_sport
, uh
->uh_dport
);
311 m
->m_flags
|= M_HASH
;
312 m
->m_pkthdr
.hash
= hash
;
316 * Verify and adjust the hash value of the packet.
318 * Unlike ip_hashfn(), the packet content is not accessed. The packet info
319 * (pi) and the hash of the packet (m_pkthdr.hash) is used instead.
321 * Caller has already made sure that m_pkthdr.hash is valid, i.e. m_flags
325 ip_hashcheck(struct mbuf
*m
, const struct pktinfo
*pi
)
327 KASSERT((m
->m_flags
& M_HASH
), ("no valid packet hash"));
329 switch (pi
->pi_l3proto
) {
335 /* Let software calculate the hash */
336 m
->m_flags
&= ~M_HASH
;
342 * This is used to map a socket to a message port for sendmsg() and friends.
343 * It is not called for any other purpose. In the case of TCP we just return
344 * the port already installed in the socket.
347 tcp_soport(struct socket
*so
, struct sockaddr
*nam
,
348 struct mbuf
**dummy __unused
)
354 * Used to route icmp messages to the proper protocol thread for ctlinput
358 tcp_ctlport(int cmd
, struct sockaddr
*sa
, void *vip
, int *cpuid
)
364 notify
= tcp_get_inpnotify(cmd
, sa
, &arg
, &ip
, cpuid
);
368 if (*cpuid
== ncpus
) {
370 * Go through all CPUs.
372 * A new message will be allocated later to save necessary
373 * information and will be forwarded to all network protocol
374 * threads in the following way:
376 * (the the thread owns the msgport that we return here)
383 * [msg is kmalloc()ed]
386 * Later on, when the msg is received by netisr0:
388 * forwardmsg forwardmsg
389 * netisr0 ---------> netisr1 ---------> netisrN
392 return netisr_cpuport(0);
394 return netisr_cpuport(*cpuid
);
399 tcp_addrport(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
401 return(netisr_cpuport(tcp_addrcpu(faddr
, fport
, laddr
, lport
)));
407 return(netisr_cpuport(0));
411 udp_addrport(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
413 return(netisr_cpuport(udp_addrcpu(faddr
, fport
, laddr
, lport
)));
417 * Used to route icmp messages to the proper protocol thread for ctlinput
421 udp_ctlport(int cmd
, struct sockaddr
*sa
, void *vip
, int *cpuid
)
426 notify
= udp_get_inpnotify(cmd
, sa
, &ip
, cpuid
);
430 if (*cpuid
== ncpus
) {
432 * Go through all CPUs.
434 * See the comment in tcp_ctlport.
436 return netisr_cpuport(0);
438 return netisr_cpuport(*cpuid
);
445 return netisr_cpuport(mycpuid
& ncpus2_mask
);
451 return netisr_cpuport(mycpuid
& ncpus2_mask
);