bsd/netinet/ip_flow.c

   1 /*
   2  * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_OSREFERENCE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License.  The rights granted to you under the
  10  * License may not be used to create, or enable the creation or
  11  * redistribution of, unlawful or unlicensed copies of an Apple operating
  12  * system, or to circumvent, violate, or enable the circumvention or
  13  * violation of, any terms of an Apple operating system software license
  14  * agreement.
  15  *
  16  * Please obtain a copy of the License at
  17  * http://www.opensource.apple.com/apsl/ and read it before using this
  18  * file.
  19  *
  20  * The Original Code and all software distributed under the License are
  21  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  22  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  23  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  24  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  25  * Please see the License for the specific language governing rights and
  26  * limitations under the License.
  27  *
  28  * @APPLE_LICENSE_OSREFERENCE_HEADER_END@
  29  */
  30 /*-
  31  * Copyright (c) 1998 The NetBSD Foundation, Inc.
  32  * All rights reserved.
  33  *
  34  * This code is derived from software contributed to The NetBSD Foundation
  35  * by the 3am Software Foundry ("3am").  It was developed by Matt Thomas.
  36  *
  37  * Redistribution and use in source and binary forms, with or without
  38  * modification, are permitted provided that the following conditions
  39  * are met:
  40  * 1. Redistributions of source code must retain the above copyright
  41  *    notice, this list of conditions and the following disclaimer.
  42  * 2. Redistributions in binary form must reproduce the above copyright
  43  *    notice, this list of conditions and the following disclaimer in the
  44  *    documentation and/or other materials provided with the distribution.
  45  * 3. All advertising materials mentioning features or use of this software
  46  *    must display the following acknowledgement:
  47  *      This product includes software developed by the NetBSD
  48  *      Foundation, Inc. and its contributors.
  49  * 4. Neither the name of The NetBSD Foundation nor the names of its
  50  *    contributors may be used to endorse or promote products derived
  51  *    from this software without specific prior written permission.
  52  *
  53  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
  54  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  56  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
  57  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  58  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  59  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  60  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  61  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  62  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  63  * POSSIBILITY OF SUCH DAMAGE.
  64  *
  65  * $FreeBSD: src/sys/netinet/ip_flow.c,v 1.9.2.1 2001/08/08 08:20:35 ru Exp $
  66  */
  67
  68 #include <sys/param.h>
  69 #include <sys/systm.h>
  70 #include <sys/malloc.h>
  71 #include <sys/mbuf.h>
  72 #include <sys/protosw.h>
  73 #include <sys/socket.h>
  74 #include <sys/kernel.h>
  75
  76 #include <sys/sysctl.h>
  77
  78 #include <net/if.h>
  79 #include <net/route.h>
  80
  81 #include <netinet/in.h>
  82 #include <netinet/in_systm.h>
  83 #include <netinet/ip.h>
  84 #include <netinet/in_var.h>
  85 #include <netinet/ip_var.h>
  86 #include <netinet/ip_flow.h>
  87 #include <net/dlil.h>
  88
  89 #define IPFLOW_TIMER            (5 * PR_SLOWHZ)
  90 #define IPFLOW_HASHBITS         6       /* should not be a multiple of 8 */
  91 #define IPFLOW_HASHSIZE         (1 << IPFLOW_HASHBITS)
  92 static LIST_HEAD(ipflowhead, ipflow) ipflows[IPFLOW_HASHSIZE];
  93 static int ipflow_inuse;
  94 #define IPFLOW_MAX              256
  95
  96 #ifdef __APPLE__
  97 #define M_IPFLOW M_TEMP
  98 #endif
  99
 100 static int ipflow_active = 0;
 101 SYSCTL_INT(_net_inet_ip, IPCTL_FASTFORWARDING, fastforwarding, CTLFLAG_RW,
 102     &ipflow_active, 0, "Enable flow-based IP forwarding");
 103
 104 #ifndef __APPLE__
 105 static MALLOC_DEFINE(M_IPFLOW, "ip_flow", "IP flow");
 106 #endif
 107
 108 static unsigned
 109 ipflow_hash(
 110         struct in_addr dst,
 111         struct in_addr src,
 112         unsigned tos)
 113 {
 114         unsigned hash = tos;
 115         int idx;
 116         for (idx = 0; idx < 32; idx += IPFLOW_HASHBITS)
 117                 hash += (dst.s_addr >> (32 - idx)) + (src.s_addr >> idx);
 118         return hash & (IPFLOW_HASHSIZE-1);
 119 }
 120
 121 static struct ipflow *
 122 ipflow_lookup(
 123         const struct ip *ip)
 124 {
 125         unsigned hash;
 126         struct ipflow *ipf;
 127
 128         hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
 129
 130         ipf = LIST_FIRST(&ipflows[hash]);
 131         while (ipf != NULL) {
 132                 if (ip->ip_dst.s_addr == ipf->ipf_dst.s_addr
 133                     && ip->ip_src.s_addr == ipf->ipf_src.s_addr
 134                     && ip->ip_tos == ipf->ipf_tos)
 135                         break;
 136                 ipf = LIST_NEXT(ipf, ipf_next);
 137         }
 138         return ipf;
 139 }
 140
 141 int
 142 ipflow_fastforward(
 143         struct mbuf *m)
 144 {
 145         struct ip *ip;
 146         struct ipflow *ipf;
 147         struct rtentry *rt;
 148         struct sockaddr *dst;
 149         int error;
 150
 151         /*
 152          * Are we forwarding packets?  Big enough for an IP packet?
 153          */
 154         if (!ipforwarding || !ipflow_active || m->m_len < sizeof(struct ip))
 155                 return 0;
 156         /*
 157          * IP header with no option and valid version and length
 158          */
 159         ip = mtod(m, struct ip *);
 160         if (ip->ip_v != IPVERSION || ip->ip_hl != (sizeof(struct ip) >> 2)
 161             || ntohs(ip->ip_len) > m->m_pkthdr.len)
 162                 return 0;
 163         /*
 164          * Find a flow.
 165          */
 166         if ((ipf = ipflow_lookup(ip)) == NULL)
 167                 return 0;
 168
 169         /*
 170          * Route and interface still up?
 171          */
 172         rt = ipf->ipf_ro.ro_rt;
 173         if ((rt->rt_flags & RTF_UP) == 0 || (rt->rt_ifp->if_flags & IFF_UP) == 0)
 174                 return 0;
 175
 176         /*
 177          * Packet size OK?  TTL?
 178          */
 179         if (m->m_pkthdr.len > rt->rt_ifp->if_mtu || ip->ip_ttl <= IPTTLDEC)
 180                 return 0;
 181
 182         /*
 183          * Everything checks out and so we can forward this packet.
 184          * Modify the TTL and incrementally change the checksum.
 185          */
 186         ip->ip_ttl -= IPTTLDEC;
 187         if (ip->ip_sum >= htons(0xffff - (IPTTLDEC << 8))) {
 188                 ip->ip_sum += htons(IPTTLDEC << 8) + 1;
 189         } else {
 190                 ip->ip_sum += htons(IPTTLDEC << 8);
 191         }
 192
 193         /*
 194          * Send the packet on its way.  All we can get back is ENOBUFS
 195          */
 196         ipf->ipf_uses++;
 197         ipf->ipf_timer = IPFLOW_TIMER;
 198
 199         if (rt->rt_flags & RTF_GATEWAY)
 200                 dst = rt->rt_gateway;
 201         else
 202                 dst = &ipf->ipf_ro.ro_dst;
 203 #ifdef __APPLE__
 204         /* Not sure the rt_dlt is valid here !! XXX */
 205         if ((error = dlil_output(rt->rt_ifp, PF_INET, m, (caddr_t) rt, dst, 0)) != 0) {
 206
 207 #else
 208         if ((error = (*rt->rt_ifp->if_output)(rt->rt_ifp, m, dst, rt)) != 0) {
 209 #endif
 210                 if (error == ENOBUFS)
 211                         ipf->ipf_dropped++;
 212                 else
 213                         ipf->ipf_errors++;
 214         }
 215         return 1;
 216 }
 217 \f
 218 static void
 219 ipflow_addstats(
 220         struct ipflow *ipf)
 221 {
 222         ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
 223         ipstat.ips_cantforward += ipf->ipf_errors + ipf->ipf_dropped;
 224         ipstat.ips_forward += ipf->ipf_uses;
 225         ipstat.ips_fastforward += ipf->ipf_uses;
 226 }
 227
 228 static void
 229 ipflow_free(
 230         struct ipflow *ipf)
 231 {
 232         int s;
 233         /*
 234          * Remove the flow from the hash table (at elevated IPL).
 235          * Once it's off the list, we can deal with it at normal
 236          * network IPL.
 237          */
 238         s = splimp();
 239         LIST_REMOVE(ipf, ipf_next);
 240         splx(s);
 241         ipflow_addstats(ipf);
 242         rtfree(ipf->ipf_ro.ro_rt);
 243         ipflow_inuse--;
 244         FREE(ipf, M_IPFLOW);
 245 }
 246
 247 static struct ipflow *
 248 ipflow_reap(
 249         void)
 250 {
 251         struct ipflow *ipf, *maybe_ipf = NULL;
 252         int idx;
 253         int s;
 254
 255         for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
 256                 ipf = LIST_FIRST(&ipflows[idx]);
 257                 while (ipf != NULL) {
 258                         /*
 259                          * If this no longer points to a valid route
 260                          * reclaim it.
 261                          */
 262                         if ((ipf->ipf_ro.ro_rt->rt_flags & RTF_UP) == 0)
 263                                 goto done;
 264                         /*
 265                          * choose the one that's been least recently used
 266                          * or has had the least uses in the last 1.5
 267                          * intervals.
 268                          */
 269                         if (maybe_ipf == NULL
 270                             || ipf->ipf_timer < maybe_ipf->ipf_timer
 271                             || (ipf->ipf_timer == maybe_ipf->ipf_timer
 272                                 && ipf->ipf_last_uses + ipf->ipf_uses <
 273                                       maybe_ipf->ipf_last_uses +
 274                                         maybe_ipf->ipf_uses))
 275                                 maybe_ipf = ipf;
 276                         ipf = LIST_NEXT(ipf, ipf_next);
 277                 }
 278         }
 279         ipf = maybe_ipf;
 280     done:
 281         /*
 282          * Remove the entry from the flow table.
 283          */
 284         s = splimp();
 285         LIST_REMOVE(ipf, ipf_next);
 286         splx(s);
 287         ipflow_addstats(ipf);
 288         rtfree(ipf->ipf_ro.ro_rt);
 289         return ipf;
 290 }
 291
 292 void
 293 ipflow_slowtimo(
 294         void)
 295 {
 296         struct ipflow *ipf;
 297         int idx;
 298
 299         for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
 300                 ipf = LIST_FIRST(&ipflows[idx]);
 301                 while (ipf != NULL) {
 302                         struct ipflow *next_ipf = LIST_NEXT(ipf, ipf_next);
 303                         if (--ipf->ipf_timer == 0) {
 304                                 ipflow_free(ipf);
 305                         } else {
 306                                 ipf->ipf_last_uses = ipf->ipf_uses;
 307                                 ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
 308                                 ipstat.ips_forward += ipf->ipf_uses;
 309                                 ipstat.ips_fastforward += ipf->ipf_uses;
 310                                 ipf->ipf_uses = 0;
 311                         }
 312                         ipf = next_ipf;
 313                 }
 314         }
 315 }
 316
 317 void
 318 ipflow_create(
 319         const struct route *ro,
 320         struct mbuf *m)
 321 {
 322         const struct ip *const ip = mtod(m, struct ip *);
 323         struct ipflow *ipf;
 324         unsigned hash;
 325         int s;
 326
 327         /*
 328          * Don't create cache entries for ICMP messages.
 329          */
 330         if (!ipflow_active || ip->ip_p == IPPROTO_ICMP)
 331                 return;
 332         /*
 333          * See if an existing flow struct exists.  If so remove it from it's
 334          * list and free the old route.  If not, try to malloc a new one
 335          * (if we aren't at our limit).
 336          */
 337         ipf = ipflow_lookup(ip);
 338         if (ipf == NULL) {
 339                 if (ipflow_inuse == IPFLOW_MAX) {
 340                         ipf = ipflow_reap();
 341                 } else {
 342                         ipf = (struct ipflow *) _MALLOC(sizeof(*ipf), M_IPFLOW,
 343                                                        M_NOWAIT);
 344                         if (ipf == NULL)
 345                                 return;
 346                         ipflow_inuse++;
 347                 }
 348                 bzero((caddr_t) ipf, sizeof(*ipf));
 349         } else {
 350                 s = splimp();
 351                 LIST_REMOVE(ipf, ipf_next);
 352                 splx(s);
 353                 ipflow_addstats(ipf);
 354                 rtfree(ipf->ipf_ro.ro_rt);
 355                 ipf->ipf_uses = ipf->ipf_last_uses = 0;
 356                 ipf->ipf_errors = ipf->ipf_dropped = 0;
 357         }
 358
 359         /*
 360          * Fill in the updated information.
 361          */
 362         ipf->ipf_ro = *ro;
 363         rtref(ro->ro_rt);               //### LD 5/25/04 needs rt_mtx lock
 364         ipf->ipf_dst = ip->ip_dst;
 365         ipf->ipf_src = ip->ip_src;
 366         ipf->ipf_tos = ip->ip_tos;
 367         ipf->ipf_timer = IPFLOW_TIMER;
 368         /*
 369          * Insert into the approriate bucket of the flow table.
 370          */
 371         hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
 372         s = splimp();
 373         LIST_INSERT_HEAD(&ipflows[hash], ipf, ipf_next);
 374         splx(s);
 375 }