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 1.2 2003/06/17 04:28:51 dillon Exp $
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
44 #include <sys/protosw.h>
45 #include <sys/socket.h>
46 #include <sys/kernel.h>
48 #include <sys/sysctl.h>
51 #include <net/route.h>
53 #include <netinet/in.h>
54 #include <netinet/in_systm.h>
55 #include <netinet/ip.h>
56 #include <netinet/in_var.h>
57 #include <netinet/ip_var.h>
58 #include <netinet/ip_flow.h>
60 #define IPFLOW_TIMER (5 * PR_SLOWHZ)
61 #define IPFLOW_HASHBITS 6 /* should not be a multiple of 8 */
62 #define IPFLOW_HASHSIZE (1 << IPFLOW_HASHBITS)
63 static LIST_HEAD(ipflowhead, ipflow) ipflows[IPFLOW_HASHSIZE];
64 static int ipflow_inuse;
65 #define IPFLOW_MAX 256
67 static int ipflow_active = 0;
68 SYSCTL_INT(_net_inet_ip, IPCTL_FASTFORWARDING, fastforwarding, CTLFLAG_RW,
69 &ipflow_active, 0, "Enable flow-based IP forwarding");
71 static MALLOC_DEFINE(M_IPFLOW, "ip_flow", "IP flow");
81 for (idx = 0; idx < 32; idx += IPFLOW_HASHBITS)
82 hash += (dst.s_addr >> (32 - idx)) + (src.s_addr >> idx);
83 return hash & (IPFLOW_HASHSIZE-1);
86 static struct ipflow *
93 hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
95 ipf = LIST_FIRST(&ipflows[hash]);
97 if (ip->ip_dst.s_addr == ipf->ipf_dst.s_addr
98 && ip->ip_src.s_addr == ipf->ipf_src.s_addr
99 && ip->ip_tos == ipf->ipf_tos)
101 ipf = LIST_NEXT(ipf, ipf_next);
113 struct sockaddr *dst;
117 * Are we forwarding packets? Big enough for an IP packet?
119 if (!ipforwarding || !ipflow_active || m->m_len < sizeof(struct ip))
122 * IP header with no option and valid version and length
124 ip = mtod(m, struct ip *);
125 if (ip->ip_v != IPVERSION || ip->ip_hl != (sizeof(struct ip) >> 2)
126 || ntohs(ip->ip_len) > m->m_pkthdr.len)
131 if ((ipf = ipflow_lookup(ip)) == NULL)
135 * Route and interface still up?
137 rt = ipf->ipf_ro.ro_rt;
138 if ((rt->rt_flags & RTF_UP) == 0 || (rt->rt_ifp->if_flags & IFF_UP) == 0)
142 * Packet size OK? TTL?
144 if (m->m_pkthdr.len > rt->rt_ifp->if_mtu || ip->ip_ttl <= IPTTLDEC)
148 * Everything checks out and so we can forward this packet.
149 * Modify the TTL and incrementally change the checksum.
151 ip->ip_ttl -= IPTTLDEC;
152 if (ip->ip_sum >= htons(0xffff - (IPTTLDEC << 8))) {
153 ip->ip_sum += htons(IPTTLDEC << 8) + 1;
155 ip->ip_sum += htons(IPTTLDEC << 8);
159 * Send the packet on its way. All we can get back is ENOBUFS
162 ipf->ipf_timer = IPFLOW_TIMER;
164 if (rt->rt_flags & RTF_GATEWAY)
165 dst = rt->rt_gateway;
167 dst = &ipf->ipf_ro.ro_dst;
168 if ((error = (*rt->rt_ifp->if_output)(rt->rt_ifp, m, dst, rt)) != 0) {
169 if (error == ENOBUFS)
181 ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
182 ipstat.ips_cantforward += ipf->ipf_errors + ipf->ipf_dropped;
183 ipstat.ips_forward += ipf->ipf_uses;
184 ipstat.ips_fastforward += ipf->ipf_uses;
193 * Remove the flow from the hash table (at elevated IPL).
194 * Once it's off the list, we can deal with it at normal
198 LIST_REMOVE(ipf, ipf_next);
200 ipflow_addstats(ipf);
201 RTFREE(ipf->ipf_ro.ro_rt);
206 static struct ipflow *
210 struct ipflow *ipf, *maybe_ipf = NULL;
214 for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
215 ipf = LIST_FIRST(&ipflows[idx]);
216 while (ipf != NULL) {
218 * If this no longer points to a valid route
221 if ((ipf->ipf_ro.ro_rt->rt_flags & RTF_UP) == 0)
224 * choose the one that's been least recently used
225 * or has had the least uses in the last 1.5
228 if (maybe_ipf == NULL
229 || ipf->ipf_timer < maybe_ipf->ipf_timer
230 || (ipf->ipf_timer == maybe_ipf->ipf_timer
231 && ipf->ipf_last_uses + ipf->ipf_uses <
232 maybe_ipf->ipf_last_uses +
233 maybe_ipf->ipf_uses))
235 ipf = LIST_NEXT(ipf, ipf_next);
241 * Remove the entry from the flow table.
244 LIST_REMOVE(ipf, ipf_next);
246 ipflow_addstats(ipf);
247 RTFREE(ipf->ipf_ro.ro_rt);
258 for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
259 ipf = LIST_FIRST(&ipflows[idx]);
260 while (ipf != NULL) {
261 struct ipflow *next_ipf = LIST_NEXT(ipf, ipf_next);
262 if (--ipf->ipf_timer == 0) {
265 ipf->ipf_last_uses = ipf->ipf_uses;
266 ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
267 ipstat.ips_forward += ipf->ipf_uses;
268 ipstat.ips_fastforward += ipf->ipf_uses;
278 const struct route *ro,
281 const struct ip *const ip = mtod(m, struct ip *);
287 * Don't create cache entries for ICMP messages.
289 if (!ipflow_active || ip->ip_p == IPPROTO_ICMP)
292 * See if an existing flow struct exists. If so remove it from it's
293 * list and free the old route. If not, try to malloc a new one
294 * (if we aren't at our limit).
296 ipf = ipflow_lookup(ip);
298 if (ipflow_inuse == IPFLOW_MAX) {
301 ipf = (struct ipflow *) malloc(sizeof(*ipf), M_IPFLOW,
307 bzero((caddr_t) ipf, sizeof(*ipf));
310 LIST_REMOVE(ipf, ipf_next);
312 ipflow_addstats(ipf);
313 RTFREE(ipf->ipf_ro.ro_rt);
314 ipf->ipf_uses = ipf->ipf_last_uses = 0;
315 ipf->ipf_errors = ipf->ipf_dropped = 0;
319 * Fill in the updated information.
322 ro->ro_rt->rt_refcnt++;
323 ipf->ipf_dst = ip->ip_dst;
324 ipf->ipf_src = ip->ip_src;
325 ipf->ipf_tos = ip->ip_tos;
326 ipf->ipf_timer = IPFLOW_TIMER;
328 * Insert into the approriate bucket of the flow table.
330 hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
332 LIST_INSERT_HEAD(&ipflows[hash], ipf, ipf_next);