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>
62 struct initport_index
{
65 static struct initport_index initport_indices
[MAXCPU
];
68 * Toeplitz hash functions - the idea is to match the hardware.
71 INP_MPORT_HASH_UDP(in_addr_t faddr
, in_addr_t laddr
,
72 in_port_t fport
, in_port_t lport
)
74 return toeplitz_hash(toeplitz_rawhash_addr(faddr
, laddr
));
78 INP_MPORT_HASH_TCP(in_addr_t faddr
, in_addr_t laddr
,
79 in_port_t fport
, in_port_t lport
)
82 toeplitz_rawhash_addrport(faddr
, laddr
, fport
, lport
));
86 * Map a network address to a processor.
89 tcp_addrcpu(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
91 return (netisr_hashcpu(INP_MPORT_HASH_TCP(faddr
, laddr
, fport
, lport
)));
95 udp_addrcpu(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
98 * NOTE: laddr could be multicast, since UDP socket could be
99 * bound to multicast address.
101 if (IN_MULTICAST(ntohl(faddr
)) || IN_MULTICAST(ntohl(laddr
))) {
102 /* XXX handle multicast on CPU0 for now */
105 return (netisr_hashcpu(INP_MPORT_HASH_UDP(faddr
, laddr
, fport
, lport
)));
109 * If the packet is a valid IP datagram, upon returning of this function
110 * following things are promised:
112 * o IP header (including any possible IP options) and any data preceding
113 * IP header (usually linker layer header) are in one mbuf (m_len).
114 * o IP header length is not less than the minimum (sizeof(struct ip)).
115 * o IP total length is not less than IP header length.
116 * o IP datagram resides completely in the mbuf chain,
117 * i.e. pkthdr.len >= IP total length.
119 * If the packet is a UDP datagram,
120 * o IP header (including any possible IP options) and UDP header are in
122 * o IP total length is not less than (IP header length + UDP header length).
124 * If the packet is a TCP segment,
125 * o IP header (including any possible IP options) and TCP header (including
126 * any possible TCP options) are in one mbuf (m_len).
127 * o TCP header length is not less than the minimum (sizeof(struct tcphdr)).
128 * o IP total length is not less than (IP header length + TCP header length).
131 ip_lengthcheck(struct mbuf
**mp
, int hoff
)
133 struct mbuf
*m
= *mp
;
135 int len
, iphlen
, iplen
;
137 int thoff
; /* TCP data offset */
139 len
= hoff
+ sizeof(struct ip
);
141 /* The packet must be at least the size of an IP header. */
142 if (m
->m_pkthdr
.len
< len
) {
143 ipstat
.ips_tooshort
++;
147 /* The fixed IP header must reside completely in the first mbuf. */
148 if (m
->m_len
< len
) {
149 m
= m_pullup(m
, len
);
151 ipstat
.ips_toosmall
++;
156 ip
= mtodoff(m
, struct ip
*, hoff
);
158 /* Bound check the packet's stated IP header length. */
159 iphlen
= ip
->ip_hl
<< 2;
160 if (iphlen
< sizeof(struct ip
)) { /* minimum header length */
161 ipstat
.ips_badhlen
++;
165 /* The full IP header must reside completely in the one mbuf. */
166 if (m
->m_len
< hoff
+ iphlen
) {
167 m
= m_pullup(m
, hoff
+ iphlen
);
169 ipstat
.ips_badhlen
++;
172 ip
= mtodoff(m
, struct ip
*, hoff
);
175 iplen
= ntohs(ip
->ip_len
);
178 * Check that the amount of data in the buffers is as
179 * at least much as the IP header would have us expect.
181 if (m
->m_pkthdr
.len
< hoff
+ iplen
) {
182 ipstat
.ips_tooshort
++;
187 * Fragments other than the first fragment don't have much
188 * length information.
190 if (ntohs(ip
->ip_off
) & IP_OFFMASK
)
194 * The TCP/IP or UDP/IP header must be entirely contained within
195 * the first fragment of a packet. Packet filters will break if they
198 * Since the packet will be trimmed to ip_len we must also make sure
199 * the potentially trimmed down length is still sufficient to hold
204 if (iplen
< iphlen
+ sizeof(struct tcphdr
)) {
205 ++tcpstat
.tcps_rcvshort
;
208 if (m
->m_len
< hoff
+ iphlen
+ sizeof(struct tcphdr
)) {
209 m
= m_pullup(m
, hoff
+ iphlen
+ sizeof(struct tcphdr
));
211 tcpstat
.tcps_rcvshort
++;
214 ip
= mtodoff(m
, struct ip
*, hoff
);
216 th
= (struct tcphdr
*)((caddr_t
)ip
+ iphlen
);
217 thoff
= th
->th_off
<< 2;
218 if (thoff
< sizeof(struct tcphdr
) ||
219 thoff
+ iphlen
> ntohs(ip
->ip_len
)) {
220 tcpstat
.tcps_rcvbadoff
++;
223 if (m
->m_len
< hoff
+ iphlen
+ thoff
) {
224 m
= m_pullup(m
, hoff
+ iphlen
+ thoff
);
226 tcpstat
.tcps_rcvshort
++;
232 if (iplen
< iphlen
+ sizeof(struct udphdr
)) {
233 ++udp_stat
.udps_hdrops
;
236 if (m
->m_len
< hoff
+ iphlen
+ sizeof(struct udphdr
)) {
237 m
= m_pullup(m
, hoff
+ iphlen
+ sizeof(struct udphdr
));
239 udp_stat
.udps_hdrops
++;
246 if (iplen
< iphlen
) {
253 m
->m_flags
|= M_LENCHECKED
;
265 * Assign a protocol processing thread to a packet. The IP header is at
266 * offset (hoff) in the packet (i.e. the mac header might still be intact).
268 * This function can blow away the mbuf if the packet is malformed.
271 ip_hashfn(struct mbuf
**mptr
, int hoff
)
280 if (!ip_lengthcheck(mptr
, hoff
))
284 ip
= mtodoff(m
, struct ip
*, hoff
);
285 iphlen
= ip
->ip_hl
<< 2;
287 if (ntohs(ip
->ip_off
) & (IP_MF
| IP_OFFMASK
)) {
288 hash
= toeplitz_hash(toeplitz_rawhash_addr(
289 ip
->ip_src
.s_addr
, ip
->ip_dst
.s_addr
));
295 th
= (struct tcphdr
*)((caddr_t
)ip
+ iphlen
);
296 hash
= INP_MPORT_HASH_TCP(ip
->ip_src
.s_addr
, ip
->ip_dst
.s_addr
,
297 th
->th_sport
, th
->th_dport
);
301 if (IN_MULTICAST(ntohl(ip
->ip_dst
.s_addr
))) {
302 /* XXX handle multicast on CPU0 for now */
306 uh
= (struct udphdr
*)((caddr_t
)ip
+ iphlen
);
307 hash
= INP_MPORT_HASH_UDP(ip
->ip_src
.s_addr
, ip
->ip_dst
.s_addr
,
308 uh
->uh_sport
, uh
->uh_dport
);
316 m
->m_flags
|= M_HASH
;
317 m
->m_pkthdr
.hash
= hash
;
321 * Verify and adjust the hash value of the packet.
323 * Unlike ip_hashfn(), the packet content is not accessed. The packet info
324 * (pi) and the hash of the packet (m_pkthdr.hash) is used instead.
326 * Caller has already made sure that m_pkthdr.hash is valid, i.e. m_flags
330 ip_hashcheck(struct mbuf
*m
, const struct pktinfo
*pi
)
332 KASSERT((m
->m_flags
& M_HASH
), ("no valid packet hash"));
334 switch (pi
->pi_l3proto
) {
340 /* Let software calculate the hash */
341 m
->m_flags
&= ~M_HASH
;
347 * This is used to map a socket to a message port for sendmsg() and friends.
348 * It is not called for any other purpose. In the case of TCP we just return
349 * the port already installed in the socket.
352 tcp_soport(struct socket
*so
, struct sockaddr
*nam
,
353 struct mbuf
**dummy __unused
)
359 * Used to route icmp messages to the proper protocol thread for ctlinput
363 tcp_ctlport(int cmd
, struct sockaddr
*sa
, void *vip
, int *cpuid
)
369 notify
= tcp_get_inpnotify(cmd
, sa
, &arg
, &ip
, cpuid
);
373 if (*cpuid
== ncpus
) {
375 * Go through all CPUs.
377 * A new message will be allocated later to save necessary
378 * information and will be forwarded to all network protocol
379 * threads in the following way:
381 * (the the thread owns the msgport that we return here)
388 * [msg is kmalloc()ed]
391 * Later on, when the msg is received by netisr0:
393 * forwardmsg forwardmsg
394 * netisr0 ---------> netisr1 ---------> netisrN
397 return netisr_cpuport(0);
399 return netisr_cpuport(*cpuid
);
404 tcp_addrport(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
406 return(netisr_cpuport(tcp_addrcpu(faddr
, fport
, laddr
, lport
)));
412 return(netisr_cpuport(0));
416 udp_addrport(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
418 return(netisr_cpuport(udp_addrcpu(faddr
, fport
, laddr
, lport
)));
422 * Used to route icmp messages to the proper protocol thread for ctlinput
426 udp_ctlport(int cmd
, struct sockaddr
*sa
, void *vip
, int *cpuid
)
431 notify
= udp_get_inpnotify(cmd
, sa
, &ip
, cpuid
);
435 if (*cpuid
== ncpus
) {
437 * Go through all CPUs.
439 * See the comment in tcp_ctlport.
441 return netisr_cpuport(0);
443 return netisr_cpuport(*cpuid
);
447 static __inline
struct lwkt_port
*
448 initport_ncpus2(void)
453 return netisr_cpuport(cpu
);
455 return netisr_cpuport(
456 ((initport_indices
[cpu
].port_index
++) + (uint32_t)cpu
) &
464 return initport_ncpus2();
470 return initport_ncpus2();