1 /*- 2 * Copyright (c) 1998 The NetBSD Foundation, Inc. 3 * All rights reserved. 4 * 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. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 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. 23 * 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. 35 * 36 * $FreeBSD: src/sys/netinet/ip_flow.c,v 1.9.2.2 2001/11/04 17:35:31 luigi Exp $ 37 */ 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/protosw.h> 44 #include <sys/socket.h> 45 #include <sys/kernel.h> 46 47 #include <sys/sysctl.h> 48 49 #include <net/if.h> 50 #include <net/route.h> 51 52 #include <netinet/in.h> 53 #include <netinet/in_systm.h> 54 #include <netinet/ip.h> 55 #include <netinet/in_var.h> 56 #include <netinet/ip_var.h> 57 #include <netinet/ip_flow.h> 58 59 #define IPFLOW_TIMER (5 * PR_SLOWHZ) 60 #define IPFLOW_HASHBITS 6 /* should not be a multiple of 8 */ 61 #define IPFLOW_HASHSIZE (1 << IPFLOW_HASHBITS) 62 static LIST_HEAD(ipflowhead, ipflow) ipflows[IPFLOW_HASHSIZE]; 63 static int ipflow_inuse; 64 #define IPFLOW_MAX 256 65 66 static int ipflow_active = 0; 67 SYSCTL_INT(_net_inet_ip, IPCTL_FASTFORWARDING, fastforwarding, CTLFLAG_RW, 68 &ipflow_active, 0, "Enable flow-based IP forwarding"); 69 70 static MALLOC_DEFINE(M_IPFLOW, "ip_flow", "IP flow"); 71 72 static unsigned 73 ipflow_hash( 74 struct in_addr dst, 75 struct in_addr src, 76 unsigned tos) 77 { 78 unsigned hash = tos; 79 int idx; 80 for (idx = 0; idx < 32; idx += IPFLOW_HASHBITS) 81 hash += (dst.s_addr >> (32 - idx)) + (src.s_addr >> idx); 82 return hash & (IPFLOW_HASHSIZE-1); 83 } 84 85 static struct ipflow * 86 ipflow_lookup( 87 const struct ip *ip) 88 { 89 unsigned hash; 90 struct ipflow *ipf; 91 92 hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos); 93 94 ipf = LIST_FIRST(&ipflows[hash]); 95 while (ipf != NULL) { 96 if (ip->ip_dst.s_addr == ipf->ipf_dst.s_addr 97 && ip->ip_src.s_addr == ipf->ipf_src.s_addr 98 && ip->ip_tos == ipf->ipf_tos) 99 break; 100 ipf = LIST_NEXT(ipf, ipf_next); 101 } 102 return ipf; 103 } 104 105 int 106 ipflow_fastforward( 107 struct mbuf *m) 108 { 109 struct ip *ip; 110 struct ipflow *ipf; 111 struct rtentry *rt; 112 struct sockaddr *dst; 113 int error; 114 115 /* 116 * Are we forwarding packets? Big enough for an IP packet? 117 */ 118 if (!ipforwarding || !ipflow_active || m->m_len < sizeof(struct ip)) 119 return 0; 120 /* 121 * IP header with no option and valid version and length 122 */ 123 ip = mtod(m, struct ip *); 124 if (ip->ip_v != IPVERSION || ip->ip_hl != (sizeof(struct ip) >> 2) 125 || ntohs(ip->ip_len) > m->m_pkthdr.len) 126 return 0; 127 /* 128 * Find a flow. 129 */ 130 if ((ipf = ipflow_lookup(ip)) == NULL) 131 return 0; 132 133 /* 134 * Route and interface still up? 135 */ 136 rt = ipf->ipf_ro.ro_rt; 137 if ((rt->rt_flags & RTF_UP) == 0 || (rt->rt_ifp->if_flags & IFF_UP) == 0) 138 return 0; 139 140 /* 141 * Packet size OK? TTL? 142 */ 143 if (m->m_pkthdr.len > rt->rt_ifp->if_mtu || ip->ip_ttl <= IPTTLDEC) 144 return 0; 145 146 /* 147 * Everything checks out and so we can forward this packet. 148 * Modify the TTL and incrementally change the checksum. 149 */ 150 ip->ip_ttl -= IPTTLDEC; 151 if (ip->ip_sum >= htons(0xffff - (IPTTLDEC << 8))) { 152 ip->ip_sum += htons(IPTTLDEC << 8) + 1; 153 } else { 154 ip->ip_sum += htons(IPTTLDEC << 8); 155 } 156 157 /* 158 * Send the packet on its way. All we can get back is ENOBUFS 159 */ 160 ipf->ipf_uses++; 161 ipf->ipf_timer = IPFLOW_TIMER; 162 163 if (rt->rt_flags & RTF_GATEWAY) 164 dst = rt->rt_gateway; 165 else 166 dst = &ipf->ipf_ro.ro_dst; 167 if ((error = (*rt->rt_ifp->if_output)(rt->rt_ifp, m, dst, rt)) != 0) { 168 if (error == ENOBUFS) 169 ipf->ipf_dropped++; 170 else 171 ipf->ipf_errors++; 172 } 173 return 1; 174 } 175 176 static void 177 ipflow_addstats( 178 struct ipflow *ipf) 179 { 180 ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses; 181 ipstat.ips_cantforward += ipf->ipf_errors + ipf->ipf_dropped; 182 ipstat.ips_forward += ipf->ipf_uses; 183 ipstat.ips_fastforward += ipf->ipf_uses; 184 } 185 186 static void 187 ipflow_free( 188 struct ipflow *ipf) 189 { 190 int s; 191 /* 192 * Remove the flow from the hash table (at elevated IPL). 193 * Once it's off the list, we can deal with it at normal 194 * network IPL. 195 */ 196 s = splimp(); 197 LIST_REMOVE(ipf, ipf_next); 198 splx(s); 199 ipflow_addstats(ipf); 200 RTFREE(ipf->ipf_ro.ro_rt); 201 ipflow_inuse--; 202 free(ipf, M_IPFLOW); 203 } 204 205 static struct ipflow * 206 ipflow_reap( 207 void) 208 { 209 struct ipflow *ipf, *maybe_ipf = NULL; 210 int idx; 211 int s; 212 213 for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) { 214 ipf = LIST_FIRST(&ipflows[idx]); 215 while (ipf != NULL) { 216 /* 217 * If this no longer points to a valid route 218 * reclaim it. 219 */ 220 if ((ipf->ipf_ro.ro_rt->rt_flags & RTF_UP) == 0) 221 goto done; 222 /* 223 * choose the one that's been least recently used 224 * or has had the least uses in the last 1.5 225 * intervals. 226 */ 227 if (maybe_ipf == NULL 228 || ipf->ipf_timer < maybe_ipf->ipf_timer 229 || (ipf->ipf_timer == maybe_ipf->ipf_timer 230 && ipf->ipf_last_uses + ipf->ipf_uses < 231 maybe_ipf->ipf_last_uses + 232 maybe_ipf->ipf_uses)) 233 maybe_ipf = ipf; 234 ipf = LIST_NEXT(ipf, ipf_next); 235 } 236 } 237 ipf = maybe_ipf; 238 done: 239 /* 240 * Remove the entry from the flow table. 241 */ 242 s = splimp(); 243 LIST_REMOVE(ipf, ipf_next); 244 splx(s); 245 ipflow_addstats(ipf); 246 RTFREE(ipf->ipf_ro.ro_rt); 247 return ipf; 248 } 249 250 void 251 ipflow_slowtimo( 252 void) 253 { 254 struct ipflow *ipf; 255 int idx; 256 257 for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) { 258 ipf = LIST_FIRST(&ipflows[idx]); 259 while (ipf != NULL) { 260 struct ipflow *next_ipf = LIST_NEXT(ipf, ipf_next); 261 if (--ipf->ipf_timer == 0) { 262 ipflow_free(ipf); 263 } else { 264 ipf->ipf_last_uses = ipf->ipf_uses; 265 ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses; 266 ipstat.ips_forward += ipf->ipf_uses; 267 ipstat.ips_fastforward += ipf->ipf_uses; 268 ipf->ipf_uses = 0; 269 } 270 ipf = next_ipf; 271 } 272 } 273 } 274 275 void 276 ipflow_create( 277 const struct route *ro, 278 struct mbuf *m) 279 { 280 const struct ip *const ip = mtod(m, struct ip *); 281 struct ipflow *ipf; 282 unsigned hash; 283 int s; 284 285 /* 286 * Don't create cache entries for ICMP messages. 287 */ 288 if (!ipflow_active || ip->ip_p == IPPROTO_ICMP) 289 return; 290 /* 291 * See if an existing flow struct exists. If so remove it from it's 292 * list and free the old route. If not, try to malloc a new one 293 * (if we aren't at our limit). 294 */ 295 ipf = ipflow_lookup(ip); 296 if (ipf == NULL) { 297 if (ipflow_inuse == IPFLOW_MAX) { 298 ipf = ipflow_reap(); 299 } else { 300 ipf = (struct ipflow *) malloc(sizeof(*ipf), M_IPFLOW, 301 M_NOWAIT); 302 if (ipf == NULL) 303 return; 304 ipflow_inuse++; 305 } 306 bzero((caddr_t) ipf, sizeof(*ipf)); 307 } else { 308 s = splimp(); 309 LIST_REMOVE(ipf, ipf_next); 310 splx(s); 311 ipflow_addstats(ipf); 312 RTFREE(ipf->ipf_ro.ro_rt); 313 ipf->ipf_uses = ipf->ipf_last_uses = 0; 314 ipf->ipf_errors = ipf->ipf_dropped = 0; 315 } 316 317 /* 318 * Fill in the updated information. 319 */ 320 ipf->ipf_ro = *ro; 321 ro->ro_rt->rt_refcnt++; 322 ipf->ipf_dst = ip->ip_dst; 323 ipf->ipf_src = ip->ip_src; 324 ipf->ipf_tos = ip->ip_tos; 325 ipf->ipf_timer = IPFLOW_TIMER; 326 /* 327 * Insert into the approriate bucket of the flow table. 328 */ 329 hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos); 330 s = splimp(); 331 LIST_INSERT_HEAD(&ipflows[hash], ipf, ipf_next); 332 splx(s); 333 } 334