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 1.5 2004/12/21 02:54:15 hsu 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>
53 #include <net/route.h>
55 #include <netinet/in.h>
56 #include <netinet/in_systm.h>
57 #include <netinet/ip.h>
58 #include <netinet/in_var.h>
59 #include <netinet/ip_var.h>
60 #include <netinet/ip_flow.h>
62 #define IPFLOW_TIMER (5 * PR_SLOWHZ)
63 #define IPFLOW_HASHBITS 6 /* should not be a multiple of 8 */
64 #define IPFLOW_HASHSIZE (1 << IPFLOW_HASHBITS)
65 static LIST_HEAD(ipflowhead
, ipflow
) ipflows
[IPFLOW_HASHSIZE
];
66 static int ipflow_inuse
;
67 #define IPFLOW_MAX 256
69 static int ipflow_active
= 0;
70 SYSCTL_INT(_net_inet_ip
, IPCTL_FASTFORWARDING
, fastforwarding
, CTLFLAG_RW
,
71 &ipflow_active
, 0, "Enable flow-based IP forwarding");
73 static MALLOC_DEFINE(M_IPFLOW
, "ip_flow", "IP flow");
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
*
95 hash
= ipflow_hash(ip
->ip_dst
, ip
->ip_src
, ip
->ip_tos
);
97 ipf
= LIST_FIRST(&ipflows
[hash
]);
99 if (ip
->ip_dst
.s_addr
== ipf
->ipf_dst
.s_addr
100 && ip
->ip_src
.s_addr
== ipf
->ipf_src
.s_addr
101 && ip
->ip_tos
== ipf
->ipf_tos
)
103 ipf
= LIST_NEXT(ipf
, ipf_next
);
115 struct sockaddr
*dst
;
119 * Are we forwarding packets? Big enough for an IP packet?
121 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
)
133 if ((ipf
= ipflow_lookup(ip
)) == NULL
)
137 * Route and interface still up?
139 rt
= ipf
->ipf_ro
.ro_rt
;
140 if ((rt
->rt_flags
& RTF_UP
) == 0 || (rt
->rt_ifp
->if_flags
& IFF_UP
) == 0)
144 * Packet size OK? TTL?
146 if (m
->m_pkthdr
.len
> rt
->rt_ifp
->if_mtu
|| ip
->ip_ttl
<= IPTTLDEC
)
150 * Everything checks out and so we can forward this packet.
151 * Modify the TTL and incrementally change the checksum.
153 ip
->ip_ttl
-= IPTTLDEC
;
154 if (ip
->ip_sum
>= htons(0xffff - (IPTTLDEC
<< 8))) {
155 ip
->ip_sum
+= htons(IPTTLDEC
<< 8) + 1;
157 ip
->ip_sum
+= htons(IPTTLDEC
<< 8);
161 * Send the packet on its way. All we can get back is ENOBUFS
164 ipf
->ipf_timer
= IPFLOW_TIMER
;
166 if (rt
->rt_flags
& RTF_GATEWAY
)
167 dst
= rt
->rt_gateway
;
169 dst
= &ipf
->ipf_ro
.ro_dst
;
170 if ((error
= (*rt
->rt_ifp
->if_output
)(rt
->rt_ifp
, m
, dst
, rt
)) != 0) {
171 if (error
== ENOBUFS
)
183 ipf
->ipf_ro
.ro_rt
->rt_use
+= ipf
->ipf_uses
;
184 ipstat
.ips_cantforward
+= ipf
->ipf_errors
+ ipf
->ipf_dropped
;
185 ipstat
.ips_forward
+= ipf
->ipf_uses
;
186 ipstat
.ips_fastforward
+= ipf
->ipf_uses
;
196 * Remove the flow from the hash table (at elevated IPL).
197 * Once it's off the list, we can deal with it at normal
201 LIST_REMOVE(ipf
, ipf_next
);
203 ipflow_addstats(ipf
);
204 RTFREE(ipf
->ipf_ro
.ro_rt
);
209 static struct ipflow
*
213 struct ipflow
*ipf
, *maybe_ipf
= NULL
;
217 for (idx
= 0; idx
< IPFLOW_HASHSIZE
; idx
++) {
218 ipf
= LIST_FIRST(&ipflows
[idx
]);
219 while (ipf
!= NULL
) {
221 * If this no longer points to a valid route
224 if ((ipf
->ipf_ro
.ro_rt
->rt_flags
& RTF_UP
) == 0)
227 * choose the one that's been least recently used
228 * or has had the least uses in the last 1.5
231 if (maybe_ipf
== NULL
232 || ipf
->ipf_timer
< maybe_ipf
->ipf_timer
233 || (ipf
->ipf_timer
== maybe_ipf
->ipf_timer
234 && ipf
->ipf_last_uses
+ ipf
->ipf_uses
<
235 maybe_ipf
->ipf_last_uses
+
236 maybe_ipf
->ipf_uses
))
238 ipf
= LIST_NEXT(ipf
, ipf_next
);
244 * Remove the entry from the flow table.
247 LIST_REMOVE(ipf
, ipf_next
);
249 ipflow_addstats(ipf
);
250 RTFREE(ipf
->ipf_ro
.ro_rt
);
261 for (idx
= 0; idx
< IPFLOW_HASHSIZE
; idx
++) {
262 ipf
= LIST_FIRST(&ipflows
[idx
]);
263 while (ipf
!= NULL
) {
264 struct ipflow
*next_ipf
= LIST_NEXT(ipf
, ipf_next
);
265 if (--ipf
->ipf_timer
== 0) {
268 ipf
->ipf_last_uses
= ipf
->ipf_uses
;
269 ipf
->ipf_ro
.ro_rt
->rt_use
+= ipf
->ipf_uses
;
270 ipstat
.ips_forward
+= ipf
->ipf_uses
;
271 ipstat
.ips_fastforward
+= ipf
->ipf_uses
;
280 ipflow_create(const struct route
*ro
, struct mbuf
*m
)
282 const struct ip
*const ip
= mtod(m
, struct ip
*);
288 * Don't create cache entries for ICMP messages.
290 if (!ipflow_active
|| ip
->ip_p
== IPPROTO_ICMP
)
293 * See if an existing flow struct exists. If so remove it from it's
294 * list and free the old route. If not, try to malloc a new one
295 * (if we aren't at our limit).
297 ipf
= ipflow_lookup(ip
);
299 if (ipflow_inuse
== IPFLOW_MAX
) {
302 ipf
= malloc(sizeof(*ipf
), M_IPFLOW
,
303 M_INTWAIT
| M_NULLOK
);
308 bzero((caddr_t
) ipf
, sizeof(*ipf
));
311 LIST_REMOVE(ipf
, ipf_next
);
313 ipflow_addstats(ipf
);
314 RTFREE(ipf
->ipf_ro
.ro_rt
);
315 ipf
->ipf_uses
= ipf
->ipf_last_uses
= 0;
316 ipf
->ipf_errors
= ipf
->ipf_dropped
= 0;
320 * Fill in the updated information.
323 ro
->ro_rt
->rt_refcnt
++;
324 ipf
->ipf_dst
= ip
->ip_dst
;
325 ipf
->ipf_src
= ip
->ip_src
;
326 ipf
->ipf_tos
= ip
->ip_tos
;
327 ipf
->ipf_timer
= IPFLOW_TIMER
;
329 * Insert into the approriate bucket of the flow table.
331 hash
= ipflow_hash(ip
->ip_dst
, ip
->ip_src
, ip
->ip_tos
);
333 LIST_INSERT_HEAD(&ipflows
[hash
], ipf
, ipf_next
);