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  */
  58
  59 #include <sys/param.h>
  60 #include <sys/systm.h>
  61 #include <sys/malloc.h>
  62 #include <sys/mbuf.h>
  63 #include <sys/protosw.h>
  64 #include <sys/socket.h>
  65 #include <sys/kernel.h>
  66
  67 #include <sys/sysctl.h>
  68
  69 #include <net/if.h>
  70 #include <net/route.h>
  71
  72 #include <netinet/in.h>
  73 #include <netinet/in_systm.h>
  74 #include <netinet/ip.h>
  75 #include <netinet/in_var.h>
  76 #include <netinet/ip_var.h>
  77 #include <netinet/ip_flow.h>
  78 #include <net/dlil.h>
  79
  80 #define IPFLOW_TIMER            (5 * PR_SLOWHZ)
  81 #define IPFLOW_HASHBITS         6       /* should not be a multiple of 8 */
  82 #define IPFLOW_HASHSIZE         (1 << IPFLOW_HASHBITS)
  83 static LIST_HEAD(ipflowhead, ipflow) ipflows[IPFLOW_HASHSIZE];
  84 static int ipflow_inuse;
  85 #define IPFLOW_MAX              256
  86
  87 #if ISFB31
  88 #else
  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, "");
  95
  96 MALLOC_DEFINE(M_IPFLOW, "ip_flow", "IP flow");
  97
  98 static unsigned
  99 ipflow_hash(
 100         struct in_addr dst,
 101         struct in_addr src,
 102         unsigned tos)
 103 {
 104         unsigned hash = tos;
 105         int idx;
 106         for (idx = 0; idx < 32; idx += IPFLOW_HASHBITS)
 107                 hash += (dst.s_addr >> (32 - idx)) + (src.s_addr >> idx);
 108         return hash & (IPFLOW_HASHSIZE-1);
 109 }
 110
 111 static struct ipflow *
 112 ipflow_lookup(
 113         const struct ip *ip)
 114 {
 115         unsigned hash;
 116         struct ipflow *ipf;
 117
 118         hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
 119
 120         ipf = LIST_FIRST(&ipflows[hash]);
 121         while (ipf != NULL) {
 122                 if (ip->ip_dst.s_addr == ipf->ipf_dst.s_addr
 123                     && ip->ip_src.s_addr == ipf->ipf_src.s_addr
 124                     && ip->ip_tos == ipf->ipf_tos)
 125                         break;
 126                 ipf = LIST_NEXT(ipf, ipf_next);
 127         }
 128         return ipf;
 129 }
 130
 131 int
 132 ipflow_fastforward(
 133         struct mbuf *m)
 134 {
 135         struct ip *ip;
 136         struct ipflow *ipf;
 137         struct rtentry *rt;
 138         int error;
 139
 140         /*
 141          * Are we forwarding packets?  Big enough for an IP packet?
 142          */
 143         if (!ipforwarding || !ipflow_active || m->m_len < sizeof(struct ip))
 144                 return 0;
 145         /*
 146          * IP header with no option and valid version and length
 147          */
 148         ip = mtod(m, struct ip *);
 149         if (ip->ip_v != IPVERSION || ip->ip_hl != (sizeof(struct ip) >> 2)
 150             || ntohs(ip->ip_len) > m->m_pkthdr.len)
 151                 return 0;
 152         /*
 153          * Find a flow.
 154          */
 155         if ((ipf = ipflow_lookup(ip)) == NULL)
 156                 return 0;
 157
 158         /*
 159          * Route and interface still up?
 160          */
 161         rt = ipf->ipf_ro.ro_rt;
 162         if ((rt->rt_flags & RTF_UP) == 0 || (rt->rt_ifp->if_flags & IFF_UP) == 0)
 163                 return 0;
 164
 165         /*
 166          * Packet size OK?  TTL?
 167          */
 168         if (m->m_pkthdr.len > rt->rt_ifp->if_mtu || ip->ip_ttl <= IPTTLDEC)
 169                 return 0;
 170
 171         /*
 172          * Everything checks out and so we can forward this packet.
 173          * Modify the TTL and incrementally change the checksum.
 174          */
 175         ip->ip_ttl -= IPTTLDEC;
 176         if (ip->ip_sum >= htons(0xffff - (IPTTLDEC << 8))) {
 177                 ip->ip_sum += htons(IPTTLDEC << 8) + 1;
 178         } else {
 179                 ip->ip_sum += htons(IPTTLDEC << 8);
 180         }
 181
 182         /*
 183          * Send the packet on its way.  All we can get back is ENOBUFS
 184          */
 185         ipf->ipf_uses++;
 186         ipf->ipf_timer = IPFLOW_TIMER;
 187
 188         /* Not sure the rt_dlt is valid here !! XXX */
 189         if ((error = dlil_output((u_long)rt->rt_dlt, m, (caddr_t) rt, &ipf->ipf_ro.ro_dst, 0)) != 0) {
 190                 if (error == ENOBUFS)
 191                         ipf->ipf_dropped++;
 192                 else
 193                         ipf->ipf_errors++;
 194         }
 195         return 1;
 196 }
 197 \f
 198 static void
 199 ipflow_addstats(
 200         struct ipflow *ipf)
 201 {
 202         ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
 203         ipstat.ips_cantforward += ipf->ipf_errors + ipf->ipf_dropped;
 204         ipstat.ips_forward += ipf->ipf_uses;
 205         ipstat.ips_fastforward += ipf->ipf_uses;
 206 }
 207
 208 static void
 209 ipflow_free(
 210         struct ipflow *ipf)
 211 {
 212         int s;
 213         /*
 214          * Remove the flow from the hash table (at elevated IPL).
 215          * Once it's off the list, we can deal with it at normal
 216          * network IPL.
 217          */
 218         s = splimp();
 219         LIST_REMOVE(ipf, ipf_next);
 220         splx(s);
 221         ipflow_addstats(ipf);
 222         RTFREE(ipf->ipf_ro.ro_rt);
 223         ipflow_inuse--;
 224         FREE(ipf, M_IPFLOW);
 225 }
 226
 227 static struct ipflow *
 228 ipflow_reap(
 229         void)
 230 {
 231         struct ipflow *ipf, *maybe_ipf = NULL;
 232         int idx;
 233         int s;
 234
 235         for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
 236                 ipf = LIST_FIRST(&ipflows[idx]);
 237                 while (ipf != NULL) {
 238                         /*
 239                          * If this no longer points to a valid route
 240                          * reclaim it.
 241                          */
 242                         if ((ipf->ipf_ro.ro_rt->rt_flags & RTF_UP) == 0)
 243                                 goto done;
 244                         /*
 245                          * choose the one that's been least recently used
 246                          * or has had the least uses in the last 1.5
 247                          * intervals.
 248                          */
 249                         if (maybe_ipf == NULL
 250                             || ipf->ipf_timer < maybe_ipf->ipf_timer
 251                             || (ipf->ipf_timer == maybe_ipf->ipf_timer
 252                                 && ipf->ipf_last_uses + ipf->ipf_uses <
 253                                       maybe_ipf->ipf_last_uses +
 254                                         maybe_ipf->ipf_uses))
 255                                 maybe_ipf = ipf;
 256                         ipf = LIST_NEXT(ipf, ipf_next);
 257                 }
 258         }
 259         ipf = maybe_ipf;
 260     done:
 261         /*
 262          * Remove the entry from the flow table.
 263          */
 264         s = splimp();
 265         LIST_REMOVE(ipf, ipf_next);
 266         splx(s);
 267         ipflow_addstats(ipf);
 268         RTFREE(ipf->ipf_ro.ro_rt);
 269         return ipf;
 270 }
 271
 272 void
 273 ipflow_slowtimo(
 274         void)
 275 {
 276         struct ipflow *ipf;
 277         int idx;
 278
 279         for (idx = 0; idx < IPFLOW_HASHSIZE; idx++) {
 280                 ipf = LIST_FIRST(&ipflows[idx]);
 281                 while (ipf != NULL) {
 282                         struct ipflow *next_ipf = LIST_NEXT(ipf, ipf_next);
 283                         if (--ipf->ipf_timer == 0) {
 284                                 ipflow_free(ipf);
 285                         } else {
 286                                 ipf->ipf_last_uses = ipf->ipf_uses;
 287                                 ipf->ipf_ro.ro_rt->rt_use += ipf->ipf_uses;
 288                                 ipstat.ips_forward += ipf->ipf_uses;
 289                                 ipstat.ips_fastforward += ipf->ipf_uses;
 290                                 ipf->ipf_uses = 0;
 291                         }
 292                         ipf = next_ipf;
 293                 }
 294         }
 295 }
 296
 297 void
 298 ipflow_create(
 299         const struct route *ro,
 300         struct mbuf *m)
 301 {
 302         const struct ip *const ip = mtod(m, struct ip *);
 303         struct ipflow *ipf;
 304         unsigned hash;
 305         int s;
 306
 307         /*
 308          * Don't create cache entries for ICMP messages.
 309          */
 310         if (!ipflow_active || ip->ip_p == IPPROTO_ICMP)
 311                 return;
 312         /*
 313          * See if an existing flow struct exists.  If so remove it from it's
 314          * list and free the old route.  If not, try to malloc a new one
 315          * (if we aren't at our limit).
 316          */
 317         ipf = ipflow_lookup(ip);
 318         if (ipf == NULL) {
 319                 if (ipflow_inuse == IPFLOW_MAX) {
 320                         ipf = ipflow_reap();
 321                 } else {
 322                         ipf = (struct ipflow *) _MALLOC(sizeof(*ipf), M_IPFLOW,
 323                                                        M_NOWAIT);
 324                         if (ipf == NULL)
 325                                 return;
 326                         ipflow_inuse++;
 327                 }
 328                 bzero((caddr_t) ipf, sizeof(*ipf));
 329         } else {
 330                 s = splimp();
 331                 LIST_REMOVE(ipf, ipf_next);
 332                 splx(s);
 333                 ipflow_addstats(ipf);
 334                 RTFREE(ipf->ipf_ro.ro_rt);
 335                 ipf->ipf_uses = ipf->ipf_last_uses = 0;
 336                 ipf->ipf_errors = ipf->ipf_dropped = 0;
 337         }
 338
 339         /*
 340          * Fill in the updated information.
 341          */
 342         ipf->ipf_ro = *ro;
 343         ro->ro_rt->rt_refcnt++;
 344         ipf->ipf_dst = ip->ip_dst;
 345         ipf->ipf_src = ip->ip_src;
 346         ipf->ipf_tos = ip->ip_tos;
 347         ipf->ipf_timer = IPFLOW_TIMER;
 348         /*
 349          * Insert into the approriate bucket of the flow table.
 350          */
 351         hash = ipflow_hash(ip->ip_dst, ip->ip_src, ip->ip_tos);
 352         s = splimp();
 353         LIST_INSERT_HEAD(&ipflows[hash], ipf, ipf_next);
 354         splx(s);
 355 }