bsd/netinet/ip_flow.c

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