2 * Copyright (c) 1998 The NetBSD Foundation, Inc.
5 * This code is derived from software contributed to The NetBSD Foundation
6 * by the 3am Software Foundry ("3am"). It was developed by Matt Thomas.
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. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the NetBSD
19 * Foundation, Inc. and its contributors.
20 * 4. Neither the name of The NetBSD Foundation nor the names of its
21 * contributors may be used to endorse or promote products derived
22 * from this software without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
25 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
26 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
27 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
28 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
36 * $FreeBSD: src/sys/netinet/ip_flow.c,v 1.9.2.2 2001/11/04 17:35:31 luigi Exp $
37 * $DragonFly: src/sys/netinet/ip_flow.c,v 1.11 2008/03/30 10:15:46 sephe Exp $
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
44 #include <sys/globaldata.h>
45 #include <sys/thread.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/kernel.h>
50 #include <sys/sysctl.h>
51 #include <sys/thread2.h>
54 #include <net/route.h>
56 #include <netinet/in.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/ip.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip_var.h>
61 #include <netinet/ip_flow.h>
63 #define IPFLOW_TIMER (5 * PR_SLOWHZ)
64 #define IPFLOW_HASHBITS 6 /* should not be a multiple of 8 */
65 #define IPFLOW_HASHSIZE (1 << IPFLOW_HASHBITS)
66 #define IPFLOW_MAX 256
68 static LIST_HEAD(ipflowhead, ipflow) ipflows[IPFLOW_HASHSIZE];
69 static int ipflow_inuse;
70 static int ipflow_active = 0;
72 SYSCTL_INT(_net_inet_ip, IPCTL_FASTFORWARDING, fastforwarding, CTLFLAG_RW,
73 &ipflow_active, 0, "Enable flow-based IP forwarding");
75 static MALLOC_DEFINE(M_IPFLOW, "ip_flow", "IP flow");
78 ipflow_hash(struct in_addr dst, struct in_addr src, unsigned tos)
83 for (idx = 0; idx < 32; idx += IPFLOW_HASHBITS)
84 hash += (dst.s_addr >> (32 - idx)) + (src.s_addr >> idx);
85 return hash & (IPFLOW_HASHSIZE-1);
88 static struct ipflow *
89 ipflow_lookup(const struct ip *ip)
94 hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
96 ipf = LIST_FIRST(&ipflows[hash]);
98 if (ip->ip_dst.s_addr == ipf->ipf_dst.s_addr &&
99 ip->ip_src.s_addr == ipf->ipf_src.s_addr &&
100 ip->ip_tos == ipf->ipf_tos)
102 ipf = LIST_NEXT(ipf, ipf_next);
108 ipflow_fastforward(struct mbuf *m, lwkt_serialize_t serializer)
113 struct sockaddr *dst;
118 * Are we forwarding packets? Big enough for an IP packet?
120 if (!ipforwarding || !ipflow_active || m->m_len < sizeof(struct ip))
124 * IP header with no option and valid version and length
126 ip = mtod(m, struct ip *);
127 if (ip->ip_v != IPVERSION || ip->ip_hl != (sizeof(struct ip) >> 2) ||
128 ntohs(ip->ip_len) > m->m_pkthdr.len)
134 ipf = ipflow_lookup(ip);
139 * Route and interface still up?
141 rt = ipf->ipf_ro.ro_rt;
142 if ((rt->rt_flags & RTF_UP) == 0 ||
143 (rt->rt_ifp->if_flags & IFF_UP) == 0)
148 * Packet size OK? TTL?
150 if (m->m_pkthdr.len > ifp->if_mtu || ip->ip_ttl <= IPTTLDEC)
154 * Everything checks out and so we can forward this packet.
155 * Modify the TTL and incrementally change the checksum.
157 ip->ip_ttl -= IPTTLDEC;
158 if (ip->ip_sum >= htons(0xffff - (IPTTLDEC << 8)))
159 ip->ip_sum += htons(IPTTLDEC << 8) + 1;
161 ip->ip_sum += htons(IPTTLDEC << 8);
164 * Send the packet on its way. All we can get back is ENOBUFS
167 ipf->ipf_timer = IPFLOW_TIMER;
169 if (rt->rt_flags & RTF_GATEWAY)
170 dst = rt->rt_gateway;
172 dst = &ipf->ipf_ro.ro_dst;
175 lwkt_serialize_exit(serializer);
177 lwkt_serialize_enter(ifp->if_serializer);
178 error = ifp->if_output(ifp, m, dst, rt);
180 if (error == ENOBUFS)
185 lwkt_serialize_exit(ifp->if_serializer);
188 lwkt_serialize_enter(serializer);
193 ipflow_addstats(struct ipflow *ipf)
195 ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
196 ipstat.ips_cantforward += ipf->ipf_errors + ipf->ipf_dropped;
197 ipstat.ips_forward += ipf->ipf_uses;
198 ipstat.ips_fastforward += ipf->ipf_uses;
202 ipflow_free(struct ipflow *ipf)
205 * Remove the flow from the hash table (at elevated IPL).
206 * Once it's off the list, we can deal with it at normal
210 LIST_REMOVE(ipf, ipf_next);
213 ipflow_addstats(ipf);
214 RTFREE(ipf->ipf_ro.ro_rt);
216 kfree(ipf, M_IPFLOW);
219 static struct ipflow *
222 struct ipflow *ipf, *maybe_ipf = NULL;
225 for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
226 ipf = LIST_FIRST(&ipflows[idx]);
227 while (ipf != NULL) {
229 * If this no longer points to a valid route
232 if ((ipf->ipf_ro.ro_rt->rt_flags & RTF_UP) == 0)
235 * choose the one that's been least recently used
236 * or has had the least uses in the last 1.5
239 if (maybe_ipf == NULL ||
240 ipf->ipf_timer < maybe_ipf->ipf_timer ||
241 (ipf->ipf_timer == maybe_ipf->ipf_timer &&
242 ipf->ipf_last_uses + ipf->ipf_uses <
243 maybe_ipf->ipf_last_uses + maybe_ipf->ipf_uses))
245 ipf = LIST_NEXT(ipf, ipf_next);
251 * Remove the entry from the flow table.
254 LIST_REMOVE(ipf, ipf_next);
257 ipflow_addstats(ipf);
258 RTFREE(ipf->ipf_ro.ro_rt);
263 ipflow_slowtimo(void)
268 for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
269 ipf = LIST_FIRST(&ipflows[idx]);
270 while (ipf != NULL) {
271 struct ipflow *next_ipf = LIST_NEXT(ipf, ipf_next);
272 if (--ipf->ipf_timer == 0) {
275 ipf->ipf_last_uses = ipf->ipf_uses;
276 ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
277 ipstat.ips_forward += ipf->ipf_uses;
278 ipstat.ips_fastforward += ipf->ipf_uses;
287 ipflow_create(const struct route *ro, struct mbuf *m)
289 const struct ip *const ip = mtod(m, struct ip *);
294 * Don't create cache entries for ICMP messages.
296 if (!ipflow_active || ip->ip_p == IPPROTO_ICMP)
300 * See if an existing flow struct exists. If so remove it from it's
301 * list and free the old route. If not, try to malloc a new one
302 * (if we aren't at our limit).
304 ipf = ipflow_lookup(ip);
306 if (ipflow_inuse == IPFLOW_MAX) {
309 ipf = kmalloc(sizeof(*ipf), M_IPFLOW,
310 M_INTWAIT | M_NULLOK);
315 bzero(ipf, sizeof(*ipf));
318 LIST_REMOVE(ipf, ipf_next);
321 ipflow_addstats(ipf);
322 RTFREE(ipf->ipf_ro.ro_rt);
323 ipf->ipf_uses = ipf->ipf_last_uses = 0;
324 ipf->ipf_errors = ipf->ipf_dropped = 0;
328 * Fill in the updated information.
331 ro->ro_rt->rt_refcnt++;
332 ipf->ipf_dst = ip->ip_dst;
333 ipf->ipf_src = ip->ip_src;
334 ipf->ipf_tos = ip->ip_tos;
335 ipf->ipf_timer = IPFLOW_TIMER;
338 * Insert into the approriate bucket of the flow table.
340 hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
342 LIST_INSERT_HEAD(&ipflows[hash], ipf, ipf_next);