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
)
75 * NOTE: laddr could be multicast, since UDP socket could be
76 * bound to multicast address.
78 if (IN_MULTICAST(ntohl(faddr
)) || IN_MULTICAST(ntohl(laddr
))) {
79 /* XXX handle multicast on CPU0 for now */
82 return toeplitz_hash(toeplitz_rawhash_addr(faddr
, laddr
));
86 INP_MPORT_HASH_TCP(in_addr_t faddr
, in_addr_t laddr
,
87 in_port_t fport
, in_port_t lport
)
90 toeplitz_rawhash_addrport(faddr
, laddr
, fport
, lport
));
94 * Hash for the network address.
97 tcp_addrhash(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
99 return (INP_MPORT_HASH_TCP(faddr
, laddr
, fport
, lport
));
103 udp_addrhash(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
105 return (INP_MPORT_HASH_UDP(faddr
, laddr
, fport
, lport
));
109 * Map a network address to a processor.
112 tcp_addrcpu(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
114 return (netisr_hashcpu(INP_MPORT_HASH_TCP(faddr
, laddr
, fport
, lport
)));
118 udp_addrcpu(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
120 return (netisr_hashcpu(INP_MPORT_HASH_UDP(faddr
, laddr
, fport
, lport
)));
124 * If the packet is a valid IP datagram, upon returning of this function
125 * following things are promised:
127 * o IP header (including any possible IP options) and any data preceding
128 * IP header (usually linker layer header) are in one mbuf (m_len).
129 * o IP header length is not less than the minimum (sizeof(struct ip)).
130 * o IP total length is not less than IP header length.
131 * o IP datagram resides completely in the mbuf chain,
132 * i.e. pkthdr.len >= IP total length.
134 * If the packet is a UDP datagram,
135 * o IP header (including any possible IP options) and UDP header are in
137 * o IP total length is not less than (IP header length + UDP header length).
139 * If the packet is a TCP segment,
140 * o IP header (including any possible IP options) and TCP header (including
141 * any possible TCP options) are in one mbuf (m_len).
142 * o TCP header length is not less than the minimum (sizeof(struct tcphdr)).
143 * o IP total length is not less than (IP header length + TCP header length).
146 ip_lengthcheck(struct mbuf
**mp
, int hoff
)
148 struct mbuf
*m
= *mp
;
150 int len
, iphlen
, iplen
;
152 int thoff
; /* TCP data offset */
154 len
= hoff
+ sizeof(struct ip
);
156 /* The packet must be at least the size of an IP header. */
157 if (m
->m_pkthdr
.len
< len
) {
158 ipstat
.ips_tooshort
++;
162 /* The fixed IP header must reside completely in the first mbuf. */
163 if (m
->m_len
< len
) {
164 m
= m_pullup(m
, len
);
166 ipstat
.ips_toosmall
++;
171 ip
= mtodoff(m
, struct ip
*, hoff
);
173 /* Bound check the packet's stated IP header length. */
174 iphlen
= ip
->ip_hl
<< 2;
175 if (iphlen
< sizeof(struct ip
)) { /* minimum header length */
176 ipstat
.ips_badhlen
++;
180 /* The full IP header must reside completely in the one mbuf. */
181 if (m
->m_len
< hoff
+ iphlen
) {
182 m
= m_pullup(m
, hoff
+ iphlen
);
184 ipstat
.ips_badhlen
++;
187 ip
= mtodoff(m
, struct ip
*, hoff
);
190 iplen
= ntohs(ip
->ip_len
);
193 * Check that the amount of data in the buffers is as
194 * at least much as the IP header would have us expect.
196 if (m
->m_pkthdr
.len
< hoff
+ iplen
) {
197 ipstat
.ips_tooshort
++;
202 * Fragments other than the first fragment don't have much
203 * length information.
205 if (ntohs(ip
->ip_off
) & IP_OFFMASK
)
209 * The TCP/IP or UDP/IP header must be entirely contained within
210 * the first fragment of a packet. Packet filters will break if they
213 * Since the packet will be trimmed to ip_len we must also make sure
214 * the potentially trimmed down length is still sufficient to hold
219 if (iplen
< iphlen
+ sizeof(struct tcphdr
)) {
220 ++tcpstat
.tcps_rcvshort
;
223 if (m
->m_len
< hoff
+ iphlen
+ sizeof(struct tcphdr
)) {
224 m
= m_pullup(m
, hoff
+ iphlen
+ sizeof(struct tcphdr
));
226 tcpstat
.tcps_rcvshort
++;
229 ip
= mtodoff(m
, struct ip
*, hoff
);
231 th
= (struct tcphdr
*)((caddr_t
)ip
+ iphlen
);
232 thoff
= th
->th_off
<< 2;
233 if (thoff
< sizeof(struct tcphdr
) ||
234 thoff
+ iphlen
> ntohs(ip
->ip_len
)) {
235 tcpstat
.tcps_rcvbadoff
++;
238 if (m
->m_len
< hoff
+ iphlen
+ thoff
) {
239 m
= m_pullup(m
, hoff
+ iphlen
+ thoff
);
241 tcpstat
.tcps_rcvshort
++;
247 if (iplen
< iphlen
+ sizeof(struct udphdr
)) {
248 ++udp_stat
.udps_hdrops
;
251 if (m
->m_len
< hoff
+ iphlen
+ sizeof(struct udphdr
)) {
252 m
= m_pullup(m
, hoff
+ iphlen
+ sizeof(struct udphdr
));
254 udp_stat
.udps_hdrops
++;
261 if (iplen
< iphlen
) {
268 m
->m_flags
|= M_LENCHECKED
;
280 * Assign a protocol processing thread to a packet. The IP header is at
281 * offset (hoff) in the packet (i.e. the mac header might still be intact).
283 * This function can blow away the mbuf if the packet is malformed.
286 ip_hashfn(struct mbuf
**mptr
, int hoff
)
295 if (!ip_lengthcheck(mptr
, hoff
))
299 ip
= mtodoff(m
, struct ip
*, hoff
);
300 iphlen
= ip
->ip_hl
<< 2;
302 if (ntohs(ip
->ip_off
) & (IP_MF
| IP_OFFMASK
)) {
303 hash
= toeplitz_hash(toeplitz_rawhash_addr(
304 ip
->ip_src
.s_addr
, ip
->ip_dst
.s_addr
));
310 th
= (struct tcphdr
*)((caddr_t
)ip
+ iphlen
);
311 hash
= INP_MPORT_HASH_TCP(ip
->ip_src
.s_addr
, ip
->ip_dst
.s_addr
,
312 th
->th_sport
, th
->th_dport
);
316 uh
= (struct udphdr
*)((caddr_t
)ip
+ iphlen
);
317 hash
= INP_MPORT_HASH_UDP(ip
->ip_src
.s_addr
, ip
->ip_dst
.s_addr
,
318 uh
->uh_sport
, uh
->uh_dport
);
330 * Verify and adjust the hash value of the packet.
332 * Unlike ip_hashfn(), the packet content is not accessed. The packet info
333 * (pi) and the hash of the packet (m_pkthdr.hash) is used instead.
335 * Caller has already made sure that m_pkthdr.hash is valid, i.e. m_flags
339 ip_hashcheck(struct mbuf
*m
, const struct pktinfo
*pi
)
341 KASSERT((m
->m_flags
& M_HASH
), ("no valid packet hash"));
343 switch (pi
->pi_l3proto
) {
349 /* Let software calculate the hash */
350 m
->m_flags
&= ~M_HASH
;
356 * This is used to map a socket to a message port for sendmsg() and friends.
357 * It is not called for any other purpose. In the case of TCP we just return
358 * the port already installed in the socket.
361 tcp_soport(struct socket
*so
, struct sockaddr
*nam
,
362 struct mbuf
**dummy __unused
)
368 * Used to route icmp messages to the proper protocol thread for ctlinput
372 tcp_ctlport(int cmd
, struct sockaddr
*sa
, void *vip
, int *cpuid
)
378 notify
= tcp_get_inpnotify(cmd
, sa
, &arg
, &ip
, cpuid
);
382 if (*cpuid
== ncpus
) {
384 * Go through all CPUs.
386 * A new message will be allocated later to save necessary
387 * information and will be forwarded to all network protocol
388 * threads in the following way:
390 * (the the thread owns the msgport that we return here)
397 * [msg is kmalloc()ed]
400 * Later on, when the msg is received by netisr0:
402 * forwardmsg forwardmsg
403 * netisr0 ---------> netisr1 ---------> netisrN
406 return netisr_cpuport(0);
408 return netisr_cpuport(*cpuid
);
413 tcp_addrport(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
415 return(netisr_cpuport(tcp_addrcpu(faddr
, fport
, laddr
, lport
)));
421 return(netisr_cpuport(0));
425 udp_addrport(in_addr_t faddr
, in_port_t fport
, in_addr_t laddr
, in_port_t lport
)
427 return(netisr_cpuport(udp_addrcpu(faddr
, fport
, laddr
, lport
)));
431 * Used to route icmp messages to the proper protocol thread for ctlinput
435 udp_ctlport(int cmd
, struct sockaddr
*sa
, void *vip
, int *cpuid
)
440 notify
= udp_get_inpnotify(cmd
, sa
, &ip
, cpuid
);
444 if (*cpuid
== ncpus
) {
446 * Go through all CPUs.
448 * See the comment in tcp_ctlport.
450 return netisr_cpuport(0);
452 return netisr_cpuport(*cpuid
);
456 static __inline
struct lwkt_port
*
457 initport_ncpus2(void)
462 return netisr_cpuport(cpu
);
464 return netisr_cpuport(
465 ((initport_indices
[cpu
].port_index
++) + (uint32_t)cpu
) &
473 return initport_ncpus2();
479 return initport_ncpus2();