bsd/netinet/ip_flow.c

   1 /*
   2  * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * The contents of this file constitute Original Code as defined in and
   7  * are subject to the Apple Public Source License Version 1.1 (the
   8  * "License").  You may not use this file except in compliance with the
   9  * License.  Please obtain a copy of the License at
  10  * http://www.apple.com/publicsource and read it before using this file.
  11  *
  12  * This Original Code and all software distributed under the License are
  13  * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  14  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  15  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
  17  * License for the specific language governing rights and limitations
  18  * under the License.
  19  *
  20  * @APPLE_LICENSE_HEADER_END@
  21  */
  22 /*-
  23  * Copyright (c) 1998 The NetBSD Foundation, Inc.
  24  * All rights reserved.
  25  *
  26  * This code is derived from software contributed to The NetBSD Foundation
  27  * by the 3am Software Foundry ("3am").  It was developed by Matt Thomas.
  28  *
  29  * Redistribution and use in source and binary forms, with or without
  30  * modification, are permitted provided that the following conditions
  31  * are met:
  32  * 1. Redistributions of source code must retain the above copyright
  33  *    notice, this list of conditions and the following disclaimer.
  34  * 2. Redistributions in binary form must reproduce the above copyright
  35  *    notice, this list of conditions and the following disclaimer in the
  36  *    documentation and/or other materials provided with the distribution.
  37  * 3. All advertising materials mentioning features or use of this software
  38  *    must display the following acknowledgement:
  39  *      This product includes software developed by the NetBSD
  40  *      Foundation, Inc. and its contributors.
  41  * 4. Neither the name of The NetBSD Foundation nor the names of its
  42  *    contributors may be used to endorse or promote products derived
  43  *    from this software without specific prior written permission.
  44  *
  45  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
  46  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  47  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  48  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
  49  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  50  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  51  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  52  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  53  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  54  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  55  * POSSIBILITY OF SUCH DAMAGE.
  56  *
  57  * $FreeBSD: src/sys/netinet/ip_flow.c,v 1.9.2.1 2001/08/08 08:20:35 ru Exp $
  58  */
  59
  60 #include <sys/param.h>
  61 #include <sys/systm.h>
  62 #include <sys/malloc.h>
  63 #include <sys/mbuf.h>
  64 #include <sys/protosw.h>
  65 #include <sys/socket.h>
  66 #include <sys/kernel.h>
  67
  68 #include <sys/sysctl.h>
  69
  70 #include <net/if.h>
  71 #include <net/route.h>
  72
  73 #include <netinet/in.h>
  74 #include <netinet/in_systm.h>
  75 #include <netinet/ip.h>
  76 #include <netinet/in_var.h>
  77 #include <netinet/ip_var.h>
  78 #include <netinet/ip_flow.h>
  79 #include <net/dlil.h>
  80
  81 #define IPFLOW_TIMER            (5 * PR_SLOWHZ)
  82 #define IPFLOW_HASHBITS         6       /* should not be a multiple of 8 */
  83 #define IPFLOW_HASHSIZE         (1 << IPFLOW_HASHBITS)
  84 static LIST_HEAD(ipflowhead, ipflow) ipflows[IPFLOW_HASHSIZE];
  85 static int ipflow_inuse;
  86 #define IPFLOW_MAX              256
  87
  88 #ifdef __APPLE__
  89 #define M_IPFLOW M_TEMP
  90 #endif
  91
  92 static int ipflow_active = 0;
  93 SYSCTL_INT(_net_inet_ip, IPCTL_FASTFORWARDING, fastforwarding, CTLFLAG_RW,
  94     &ipflow_active, 0, "Enable flow-based IP forwarding");
  95
  96 #ifndef __APPLE__
  97 static MALLOC_DEFINE(M_IPFLOW, "ip_flow", "IP flow");
  98 #endif
  99
 100 static unsigned
 101 ipflow_hash(
 102         struct in_addr dst,
 103         struct in_addr src,
 104         unsigned tos)
 105 {
 106         unsigned hash = tos;
 107         int idx;
 108         for (idx = 0; idx < 32; idx += IPFLOW_HASHBITS)
 109                 hash += (dst.s_addr >> (32 - idx)) + (src.s_addr >> idx);
 110         return hash & (IPFLOW_HASHSIZE-1);
 111 }
 112
 113 static struct ipflow *
 114 ipflow_lookup(
 115         const struct ip *ip)
 116 {
 117         unsigned hash;
 118         struct ipflow *ipf;
 119
 120         hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
 121
 122         ipf = LIST_FIRST(&ipflows[hash]);
 123         while (ipf != NULL) {
 124                 if (ip->ip_dst.s_addr == ipf->ipf_dst.s_addr
 125                     && ip->ip_src.s_addr == ipf->ipf_src.s_addr
 126                     && ip->ip_tos == ipf->ipf_tos)
 127                         break;
 128                 ipf = LIST_NEXT(ipf, ipf_next);
 129         }
 130         return ipf;
 131 }
 132
 133 int
 134 ipflow_fastforward(
 135         struct mbuf *m)
 136 {
 137         struct ip *ip;
 138         struct ipflow *ipf;
 139         struct rtentry *rt;
 140         struct sockaddr *dst;
 141         int error;
 142
 143         /*
 144          * Are we forwarding packets?  Big enough for an IP packet?
 145          */
 146         if (!ipforwarding || !ipflow_active || m->m_len < sizeof(struct ip))
 147                 return 0;
 148         /*
 149          * IP header with no option and valid version and length
 150          */
 151         ip = mtod(m, struct ip *);
 152         if (ip->ip_v != IPVERSION || ip->ip_hl != (sizeof(struct ip) >> 2)
 153             || ntohs(ip->ip_len) > m->m_pkthdr.len)
 154                 return 0;
 155         /*
 156          * Find a flow.
 157          */
 158         if ((ipf = ipflow_lookup(ip)) == NULL)
 159                 return 0;
 160
 161         /*
 162          * Route and interface still up?
 163          */
 164         rt = ipf->ipf_ro.ro_rt;
 165         if ((rt->rt_flags & RTF_UP) == 0 || (rt->rt_ifp->if_flags & IFF_UP) == 0)
 166                 return 0;
 167
 168         /*
 169          * Packet size OK?  TTL?
 170          */
 171         if (m->m_pkthdr.len > rt->rt_ifp->if_mtu || ip->ip_ttl <= IPTTLDEC)
 172                 return 0;
 173
 174         /*
 175          * Everything checks out and so we can forward this packet.
 176          * Modify the TTL and incrementally change the checksum.
 177          */
 178         ip->ip_ttl -= IPTTLDEC;
 179         if (ip->ip_sum >= htons(0xffff - (IPTTLDEC << 8))) {
 180                 ip->ip_sum += htons(IPTTLDEC << 8) + 1;
 181         } else {
 182                 ip->ip_sum += htons(IPTTLDEC << 8);
 183         }
 184
 185         /*
 186          * Send the packet on its way.  All we can get back is ENOBUFS
 187          */
 188         ipf->ipf_uses++;
 189         ipf->ipf_timer = IPFLOW_TIMER;
 190
 191         if (rt->rt_flags & RTF_GATEWAY)
 192                 dst = rt->rt_gateway;
 193         else
 194                 dst = &ipf->ipf_ro.ro_dst;
 195 #ifdef __APPLE__
 196         /* Not sure the rt_dlt is valid here !! XXX */
 197         if ((error = dlil_output(rt->rt_ifp, PF_INET, m, (caddr_t) rt, dst, 0)) != 0) {
 198
 199 #else
 200         if ((error = (*rt->rt_ifp->if_output)(rt->rt_ifp, m, dst, rt)) != 0) {
 201 #endif
 202                 if (error == ENOBUFS)
 203                         ipf->ipf_dropped++;
 204                 else
 205                         ipf->ipf_errors++;
 206         }
 207         return 1;
 208 }
 209 \f
 210 static void
 211 ipflow_addstats(
 212         struct ipflow *ipf)
 213 {
 214         ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
 215         ipstat.ips_cantforward += ipf->ipf_errors + ipf->ipf_dropped;
 216         ipstat.ips_forward += ipf->ipf_uses;
 217         ipstat.ips_fastforward += ipf->ipf_uses;
 218 }
 219
 220 static void
 221 ipflow_free(
 222         struct ipflow *ipf)
 223 {
 224         int s;
 225         /*
 226          * Remove the flow from the hash table (at elevated IPL).
 227          * Once it's off the list, we can deal with it at normal
 228          * network IPL.
 229          */
 230         s = splimp();
 231         LIST_REMOVE(ipf, ipf_next);
 232         splx(s);
 233         ipflow_addstats(ipf);
 234         rtfree(ipf->ipf_ro.ro_rt);
 235         ipflow_inuse--;
 236         FREE(ipf, M_IPFLOW);
 237 }
 238
 239 static struct ipflow *
 240 ipflow_reap(
 241         void)
 242 {
 243         struct ipflow *ipf, *maybe_ipf = NULL;
 244         int idx;
 245         int s;
 246
 247         for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
 248                 ipf = LIST_FIRST(&ipflows[idx]);
 249                 while (ipf != NULL) {
 250                         /*
 251                          * If this no longer points to a valid route
 252                          * reclaim it.
 253                          */
 254                         if ((ipf->ipf_ro.ro_rt->rt_flags & RTF_UP) == 0)
 255                                 goto done;
 256                         /*
 257                          * choose the one that's been least recently used
 258                          * or has had the least uses in the last 1.5
 259                          * intervals.
 260                          */
 261                         if (maybe_ipf == NULL
 262                             || ipf->ipf_timer < maybe_ipf->ipf_timer
 263                             || (ipf->ipf_timer == maybe_ipf->ipf_timer
 264                                 && ipf->ipf_last_uses + ipf->ipf_uses <
 265                                       maybe_ipf->ipf_last_uses +
 266                                         maybe_ipf->ipf_uses))
 267                                 maybe_ipf = ipf;
 268                         ipf = LIST_NEXT(ipf, ipf_next);
 269                 }
 270         }
 271         ipf = maybe_ipf;
 272     done:
 273         /*
 274          * Remove the entry from the flow table.
 275          */
 276         s = splimp();
 277         LIST_REMOVE(ipf, ipf_next);
 278         splx(s);
 279         ipflow_addstats(ipf);
 280         rtfree(ipf->ipf_ro.ro_rt);
 281         return ipf;
 282 }
 283
 284 void
 285 ipflow_slowtimo(
 286         void)
 287 {
 288         struct ipflow *ipf;
 289         int idx;
 290
 291         for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
 292                 ipf = LIST_FIRST(&ipflows[idx]);
 293                 while (ipf != NULL) {
 294                         struct ipflow *next_ipf = LIST_NEXT(ipf, ipf_next);
 295                         if (--ipf->ipf_timer == 0) {
 296                                 ipflow_free(ipf);
 297                         } else {
 298                                 ipf->ipf_last_uses = ipf->ipf_uses;
 299                                 ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
 300                                 ipstat.ips_forward += ipf->ipf_uses;
 301                                 ipstat.ips_fastforward += ipf->ipf_uses;
 302                                 ipf->ipf_uses = 0;
 303                         }
 304                         ipf = next_ipf;
 305                 }
 306         }
 307 }
 308
 309 void
 310 ipflow_create(
 311         const struct route *ro,
 312         struct mbuf *m)
 313 {
 314         const struct ip *const ip = mtod(m, struct ip *);
 315         struct ipflow *ipf;
 316         unsigned hash;
 317         int s;
 318
 319         /*
 320          * Don't create cache entries for ICMP messages.
 321          */
 322         if (!ipflow_active || ip->ip_p == IPPROTO_ICMP)
 323                 return;
 324         /*
 325          * See if an existing flow struct exists.  If so remove it from it's
 326          * list and free the old route.  If not, try to malloc a new one
 327          * (if we aren't at our limit).
 328          */
 329         ipf = ipflow_lookup(ip);
 330         if (ipf == NULL) {
 331                 if (ipflow_inuse == IPFLOW_MAX) {
 332                         ipf = ipflow_reap();
 333                 } else {
 334                         ipf = (struct ipflow *) _MALLOC(sizeof(*ipf), M_IPFLOW,
 335                                                        M_NOWAIT);
 336                         if (ipf == NULL)
 337                                 return;
 338                         ipflow_inuse++;
 339                 }
 340                 bzero((caddr_t) ipf, sizeof(*ipf));
 341         } else {
 342                 s = splimp();
 343                 LIST_REMOVE(ipf, ipf_next);
 344                 splx(s);
 345                 ipflow_addstats(ipf);
 346                 rtfree(ipf->ipf_ro.ro_rt);
 347                 ipf->ipf_uses = ipf->ipf_last_uses = 0;
 348                 ipf->ipf_errors = ipf->ipf_dropped = 0;
 349         }
 350
 351         /*
 352          * Fill in the updated information.
 353          */
 354         ipf->ipf_ro = *ro;
 355         rtref(ro->ro_rt);               //### LD 5/25/04 needs rt_mtx lock
 356         ipf->ipf_dst = ip->ip_dst;
 357         ipf->ipf_src = ip->ip_src;
 358         ipf->ipf_tos = ip->ip_tos;
 359         ipf->ipf_timer = IPFLOW_TIMER;
 360         /*
 361          * Insert into the approriate bucket of the flow table.
 362          */
 363         hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
 364         s = splimp();
 365         LIST_INSERT_HEAD(&ipflows[hash], ipf, ipf_next);
 366         splx(s);
 367 }