contrib/pf/pfctl/pfctl_optimize.c

   1 /*      $OpenBSD: pfctl_optimize.c,v 1.17 2008/05/06 03:45:21 mpf Exp $ */
   2
   3 /*
   4  * Copyright (c) 2004 Mike Frantzen <frantzen@openbsd.org>
   5  *
   6  * Permission to use, copy, modify, and distribute this software for any
   7  * purpose with or without fee is hereby granted, provided that the above
   8  * copyright notice and this permission notice appear in all copies.
   9  *
  10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
  13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  14  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  15  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  16  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  17  */
  18
  19 #include <sys/cdefs.h>
  20 __FBSDID("$FreeBSD$");
  21
  22 #include <sys/types.h>
  23 #include <sys/ioctl.h>
  24 #include <sys/socket.h>
  25
  26 #include <net/if.h>
  27 #include <net/pfvar.h>
  28
  29 #include <netinet/in.h>
  30 #include <arpa/inet.h>
  31
  32 #include <assert.h>
  33 #include <ctype.h>
  34 #include <err.h>
  35 #include <errno.h>
  36 #include <stddef.h>
  37 #include <stdio.h>
  38 #include <stdlib.h>
  39 #include <string.h>
  40
  41 #include "pfctl_parser.h"
  42 #include "pfctl.h"
  43
  44 /* The size at which a table becomes faster than individual rules */
  45 #define TABLE_THRESHOLD         6
  46
  47
  48 /* #define OPT_DEBUG    1 */
  49 #ifdef OPT_DEBUG
  50 # define DEBUG(str, v...) \
  51         printf("%s: " str "\n", __FUNCTION__ , ## v)
  52 #else
  53 # define DEBUG(str, v...) ((void)0)
  54 #endif
  55
  56
  57 /*
  58  * A container that lets us sort a superblock to optimize the skip step jumps
  59  */
  60 struct pf_skip_step {
  61         int                             ps_count;       /* number of items */
  62         TAILQ_HEAD( , pf_opt_rule)      ps_rules;
  63         TAILQ_ENTRY(pf_skip_step)       ps_entry;
  64 };
  65
  66
  67 /*
  68  * A superblock is a block of adjacent rules of similar action.  If there
  69  * are five PASS rules in a row, they all become members of a superblock.
  70  * Once we have a superblock, we are free to re-order any rules within it
  71  * in order to improve performance; if a packet is passed, it doesn't matter
  72  * who passed it.
  73  */
  74 struct superblock {
  75         TAILQ_HEAD( , pf_opt_rule)               sb_rules;
  76         TAILQ_ENTRY(superblock)                  sb_entry;
  77         struct superblock                       *sb_profiled_block;
  78         TAILQ_HEAD(skiplist, pf_skip_step)       sb_skipsteps[PF_SKIP_COUNT];
  79 };
  80 TAILQ_HEAD(superblocks, superblock);
  81
  82
  83 /*
  84  * Description of the PF rule structure.
  85  */
  86 enum {
  87     BARRIER,    /* the presence of the field puts the rule in it's own block */
  88     BREAK,      /* the field may not differ between rules in a superblock */
  89     NOMERGE,    /* the field may not differ between rules when combined */
  90     COMBINED,   /* the field may itself be combined with other rules */
  91     DC,         /* we just don't care about the field */
  92     NEVER};     /* we should never see this field set?!? */
  93 struct pf_rule_field {
  94         const char      *prf_name;
  95         int              prf_type;
  96         size_t           prf_offset;
  97         size_t           prf_size;
  98 } pf_rule_desc[] = {
  99 #define PF_RULE_FIELD(field, ty)        \
 100     {#field,                            \
 101     ty,                                 \
 102     offsetof(struct pf_rule, field),    \
 103     sizeof(((struct pf_rule *)0)->field)}
 104
 105
 106     /*
 107      * The presence of these fields in a rule put the rule in it's own
 108      * superblock.  Thus it will not be optimized.  It also prevents the
 109      * rule from being re-ordered at all.
 110      */
 111     PF_RULE_FIELD(label,                BARRIER),
 112     PF_RULE_FIELD(prob,                 BARRIER),
 113     PF_RULE_FIELD(max_states,           BARRIER),
 114     PF_RULE_FIELD(max_src_nodes,        BARRIER),
 115     PF_RULE_FIELD(max_src_states,       BARRIER),
 116     PF_RULE_FIELD(max_src_conn,         BARRIER),
 117     PF_RULE_FIELD(max_src_conn_rate,    BARRIER),
 118     PF_RULE_FIELD(anchor,               BARRIER),       /* for now */
 119
 120     /*
 121      * These fields must be the same between all rules in the same superblock.
 122      * These rules are allowed to be re-ordered but only among like rules.
 123      * For instance we can re-order all 'tag "foo"' rules because they have the
 124      * same tag.  But we can not re-order between a 'tag "foo"' and a
 125      * 'tag "bar"' since that would change the meaning of the ruleset.
 126      */
 127     PF_RULE_FIELD(tagname,              BREAK),
 128     PF_RULE_FIELD(keep_state,           BREAK),
 129     PF_RULE_FIELD(qname,                BREAK),
 130     PF_RULE_FIELD(pqname,               BREAK),
 131     PF_RULE_FIELD(rt,                   BREAK),
 132     PF_RULE_FIELD(allow_opts,           BREAK),
 133     PF_RULE_FIELD(rule_flag,            BREAK),
 134     PF_RULE_FIELD(action,               BREAK),
 135     PF_RULE_FIELD(log,                  BREAK),
 136     PF_RULE_FIELD(quick,                BREAK),
 137     PF_RULE_FIELD(return_ttl,           BREAK),
 138     PF_RULE_FIELD(overload_tblname,     BREAK),
 139     PF_RULE_FIELD(flush,                BREAK),
 140     PF_RULE_FIELD(rpool,                BREAK),
 141     PF_RULE_FIELD(logif,                BREAK),
 142
 143     /*
 144      * Any fields not listed in this structure act as BREAK fields
 145      */
 146
 147
 148     /*
 149      * These fields must not differ when we merge two rules together but
 150      * their difference isn't enough to put the rules in different superblocks.
 151      * There are no problems re-ordering any rules with these fields.
 152      */
 153     PF_RULE_FIELD(af,                   NOMERGE),
 154     PF_RULE_FIELD(ifnot,                NOMERGE),
 155     PF_RULE_FIELD(ifname,               NOMERGE),       /* hack for IF groups */
 156     PF_RULE_FIELD(match_tag_not,        NOMERGE),
 157     PF_RULE_FIELD(match_tagname,        NOMERGE),
 158     PF_RULE_FIELD(os_fingerprint,       NOMERGE),
 159     PF_RULE_FIELD(timeout,              NOMERGE),
 160     PF_RULE_FIELD(return_icmp,          NOMERGE),
 161     PF_RULE_FIELD(return_icmp6,         NOMERGE),
 162     PF_RULE_FIELD(uid,                  NOMERGE),
 163     PF_RULE_FIELD(gid,                  NOMERGE),
 164     PF_RULE_FIELD(direction,            NOMERGE),
 165     PF_RULE_FIELD(proto,                NOMERGE),
 166     PF_RULE_FIELD(type,                 NOMERGE),
 167     PF_RULE_FIELD(code,                 NOMERGE),
 168     PF_RULE_FIELD(flags,                NOMERGE),
 169     PF_RULE_FIELD(flagset,              NOMERGE),
 170     PF_RULE_FIELD(tos,                  NOMERGE),
 171     PF_RULE_FIELD(src.port,             NOMERGE),
 172     PF_RULE_FIELD(dst.port,             NOMERGE),
 173     PF_RULE_FIELD(src.port_op,          NOMERGE),
 174     PF_RULE_FIELD(dst.port_op,          NOMERGE),
 175     PF_RULE_FIELD(src.neg,              NOMERGE),
 176     PF_RULE_FIELD(dst.neg,              NOMERGE),
 177
 178     /* These fields can be merged */
 179     PF_RULE_FIELD(src.addr,             COMBINED),
 180     PF_RULE_FIELD(dst.addr,             COMBINED),
 181
 182     /* We just don't care about these fields.  They're set by the kernel */
 183     PF_RULE_FIELD(skip,                 DC),
 184     PF_RULE_FIELD(evaluations,          DC),
 185     PF_RULE_FIELD(packets,              DC),
 186     PF_RULE_FIELD(bytes,                DC),
 187     PF_RULE_FIELD(kif,                  DC),
 188     PF_RULE_FIELD(states_cur,           DC),
 189     PF_RULE_FIELD(states_tot,           DC),
 190     PF_RULE_FIELD(src_nodes,            DC),
 191     PF_RULE_FIELD(nr,                   DC),
 192     PF_RULE_FIELD(entries,              DC),
 193     PF_RULE_FIELD(qid,                  DC),
 194     PF_RULE_FIELD(pqid,                 DC),
 195     PF_RULE_FIELD(anchor_relative,      DC),
 196     PF_RULE_FIELD(anchor_wildcard,      DC),
 197     PF_RULE_FIELD(tag,                  DC),
 198     PF_RULE_FIELD(match_tag,            DC),
 199     PF_RULE_FIELD(overload_tbl,         DC),
 200
 201     /* These fields should never be set in a PASS/BLOCK rule */
 202     PF_RULE_FIELD(natpass,              NEVER),
 203     PF_RULE_FIELD(max_mss,              NEVER),
 204     PF_RULE_FIELD(min_ttl,              NEVER),
 205     PF_RULE_FIELD(set_tos,              NEVER),
 206 };
 207
 208
 209
 210 int     add_opt_table(struct pfctl *, struct pf_opt_tbl **, sa_family_t,
 211             struct pf_rule_addr *);
 212 int     addrs_combineable(struct pf_rule_addr *, struct pf_rule_addr *);
 213 int     addrs_equal(struct pf_rule_addr *, struct pf_rule_addr *);
 214 int     block_feedback(struct pfctl *, struct superblock *);
 215 int     combine_rules(struct pfctl *, struct superblock *);
 216 void    comparable_rule(struct pf_rule *, const struct pf_rule *, int);
 217 int     construct_superblocks(struct pfctl *, struct pf_opt_queue *,
 218             struct superblocks *);
 219 void    exclude_supersets(struct pf_rule *, struct pf_rule *);
 220 int     interface_group(const char *);
 221 int     load_feedback_profile(struct pfctl *, struct superblocks *);
 222 int     optimize_superblock(struct pfctl *, struct superblock *);
 223 int     pf_opt_create_table(struct pfctl *, struct pf_opt_tbl *);
 224 void    remove_from_skipsteps(struct skiplist *, struct superblock *,
 225             struct pf_opt_rule *, struct pf_skip_step *);
 226 int     remove_identical_rules(struct pfctl *, struct superblock *);
 227 int     reorder_rules(struct pfctl *, struct superblock *, int);
 228 int     rules_combineable(struct pf_rule *, struct pf_rule *);
 229 void    skip_append(struct superblock *, int, struct pf_skip_step *,
 230             struct pf_opt_rule *);
 231 int     skip_compare(int, struct pf_skip_step *, struct pf_opt_rule *);
 232 void    skip_init(void);
 233 int     skip_cmp_af(struct pf_rule *, struct pf_rule *);
 234 int     skip_cmp_dir(struct pf_rule *, struct pf_rule *);
 235 int     skip_cmp_dst_addr(struct pf_rule *, struct pf_rule *);
 236 int     skip_cmp_dst_port(struct pf_rule *, struct pf_rule *);
 237 int     skip_cmp_ifp(struct pf_rule *, struct pf_rule *);
 238 int     skip_cmp_proto(struct pf_rule *, struct pf_rule *);
 239 int     skip_cmp_src_addr(struct pf_rule *, struct pf_rule *);
 240 int     skip_cmp_src_port(struct pf_rule *, struct pf_rule *);
 241 int     superblock_inclusive(struct superblock *, struct pf_opt_rule *);
 242 void    superblock_free(struct pfctl *, struct superblock *);
 243
 244
 245 int (*skip_comparitors[PF_SKIP_COUNT])(struct pf_rule *, struct pf_rule *);
 246 const char *skip_comparitors_names[PF_SKIP_COUNT];
 247 #define PF_SKIP_COMPARITORS {                           \
 248     { "ifp", PF_SKIP_IFP, skip_cmp_ifp },               \
 249     { "dir", PF_SKIP_DIR, skip_cmp_dir },               \
 250     { "af", PF_SKIP_AF, skip_cmp_af },                  \
 251     { "proto", PF_SKIP_PROTO, skip_cmp_proto },         \
 252     { "saddr", PF_SKIP_SRC_ADDR, skip_cmp_src_addr },   \
 253     { "sport", PF_SKIP_SRC_PORT, skip_cmp_src_port },   \
 254     { "daddr", PF_SKIP_DST_ADDR, skip_cmp_dst_addr },   \
 255     { "dport", PF_SKIP_DST_PORT, skip_cmp_dst_port }    \
 256 }
 257
 258 struct pfr_buffer table_buffer;
 259 int table_identifier;
 260
 261
 262 int
 263 pfctl_optimize_ruleset(struct pfctl *pf, struct pf_ruleset *rs)
 264 {
 265         struct superblocks superblocks;
 266         struct pf_opt_queue opt_queue;
 267         struct superblock *block;
 268         struct pf_opt_rule *por;
 269         struct pf_rule *r;
 270         struct pf_rulequeue *old_rules;
 271
 272         DEBUG("optimizing ruleset");
 273         memset(&table_buffer, 0, sizeof(table_buffer));
 274         skip_init();
 275         TAILQ_INIT(&opt_queue);
 276
 277         old_rules = rs->rules[PF_RULESET_FILTER].active.ptr;
 278         rs->rules[PF_RULESET_FILTER].active.ptr =
 279             rs->rules[PF_RULESET_FILTER].inactive.ptr;
 280         rs->rules[PF_RULESET_FILTER].inactive.ptr = old_rules;
 281
 282         /*
 283          * XXX expanding the pf_opt_rule format throughout pfctl might allow
 284          * us to avoid all this copying.
 285          */
 286         while ((r = TAILQ_FIRST(rs->rules[PF_RULESET_FILTER].inactive.ptr))
 287             != NULL) {
 288                 TAILQ_REMOVE(rs->rules[PF_RULESET_FILTER].inactive.ptr, r,
 289                     entries);
 290                 if ((por = calloc(1, sizeof(*por))) == NULL)
 291                         err(1, "calloc");
 292                 memcpy(&por->por_rule, r, sizeof(*r));
 293                 if (TAILQ_FIRST(&r->rpool.list) != NULL) {
 294                         TAILQ_INIT(&por->por_rule.rpool.list);
 295                         pfctl_move_pool(&r->rpool, &por->por_rule.rpool);
 296                 } else
 297                         bzero(&por->por_rule.rpool,
 298                             sizeof(por->por_rule.rpool));
 299
 300
 301                 TAILQ_INSERT_TAIL(&opt_queue, por, por_entry);
 302         }
 303
 304         TAILQ_INIT(&superblocks);
 305         if (construct_superblocks(pf, &opt_queue, &superblocks))
 306                 goto error;
 307
 308         if (pf->optimize & PF_OPTIMIZE_PROFILE) {
 309                 if (load_feedback_profile(pf, &superblocks))
 310                         goto error;
 311         }
 312
 313         TAILQ_FOREACH(block, &superblocks, sb_entry) {
 314                 if (optimize_superblock(pf, block))
 315                         goto error;
 316         }
 317
 318         rs->anchor->refcnt = 0;
 319         while ((block = TAILQ_FIRST(&superblocks))) {
 320                 TAILQ_REMOVE(&superblocks, block, sb_entry);
 321
 322                 while ((por = TAILQ_FIRST(&block->sb_rules))) {
 323                         TAILQ_REMOVE(&block->sb_rules, por, por_entry);
 324                         por->por_rule.nr = rs->anchor->refcnt++;
 325                         if ((r = calloc(1, sizeof(*r))) == NULL)
 326                                 err(1, "calloc");
 327                         memcpy(r, &por->por_rule, sizeof(*r));
 328                         TAILQ_INIT(&r->rpool.list);
 329                         pfctl_move_pool(&por->por_rule.rpool, &r->rpool);
 330                         TAILQ_INSERT_TAIL(
 331                             rs->rules[PF_RULESET_FILTER].active.ptr,
 332                             r, entries);
 333                         free(por);
 334                 }
 335                 free(block);
 336         }
 337
 338         return (0);
 339
 340 error:
 341         while ((por = TAILQ_FIRST(&opt_queue))) {
 342                 TAILQ_REMOVE(&opt_queue, por, por_entry);
 343                 if (por->por_src_tbl) {
 344                         pfr_buf_clear(por->por_src_tbl->pt_buf);
 345                         free(por->por_src_tbl->pt_buf);
 346                         free(por->por_src_tbl);
 347                 }
 348                 if (por->por_dst_tbl) {
 349                         pfr_buf_clear(por->por_dst_tbl->pt_buf);
 350                         free(por->por_dst_tbl->pt_buf);
 351                         free(por->por_dst_tbl);
 352                 }
 353                 free(por);
 354         }
 355         while ((block = TAILQ_FIRST(&superblocks))) {
 356                 TAILQ_REMOVE(&superblocks, block, sb_entry);
 357                 superblock_free(pf, block);
 358         }
 359         return (1);
 360 }
 361
 362
 363 /*
 364  * Go ahead and optimize a superblock
 365  */
 366 int
 367 optimize_superblock(struct pfctl *pf, struct superblock *block)
 368 {
 369 #ifdef OPT_DEBUG
 370         struct pf_opt_rule *por;
 371 #endif /* OPT_DEBUG */
 372
 373         /* We have a few optimization passes:
 374          *   1) remove duplicate rules or rules that are a subset of other
 375          *      rules
 376          *   2) combine otherwise identical rules with different IP addresses
 377          *      into a single rule and put the addresses in a table.
 378          *   3) re-order the rules to improve kernel skip steps
 379          *   4) re-order the 'quick' rules based on feedback from the
 380          *      active ruleset statistics
 381          *
 382          * XXX combine_rules() doesn't combine v4 and v6 rules.  would just
 383          *     have to keep af in the table container, make af 'COMBINE' and
 384          *     twiddle the af on the merged rule
 385          * XXX maybe add a weighting to the metric on skipsteps when doing
 386          *     reordering.  sometimes two sequential tables will be better
 387          *     that four consecutive interfaces.
 388          * XXX need to adjust the skipstep count of everything after PROTO,
 389          *     since they aren't actually checked on a proto mismatch in
 390          *     pf_test_{tcp, udp, icmp}()
 391          * XXX should i treat proto=0, af=0 or dir=0 special in skepstep
 392          *     calculation since they are a DC?
 393          * XXX keep last skiplist of last superblock to influence this
 394          *     superblock.  '5 inet6 log' should make '3 inet6' come before '4
 395          *     inet' in the next superblock.
 396          * XXX would be useful to add tables for ports
 397          * XXX we can also re-order some mutually exclusive superblocks to
 398          *     try merging superblocks before any of these optimization passes.
 399          *     for instance a single 'log in' rule in the middle of non-logging
 400          *     out rules.
 401          */
 402
 403         /* shortcut.  there will be a lot of 1-rule superblocks */
 404         if (!TAILQ_NEXT(TAILQ_FIRST(&block->sb_rules), por_entry))
 405                 return (0);
 406
 407 #ifdef OPT_DEBUG
 408         printf("--- Superblock ---\n");
 409         TAILQ_FOREACH(por, &block->sb_rules, por_entry) {
 410                 printf("  ");
 411                 print_rule(&por->por_rule, por->por_rule.anchor ?
 412                     por->por_rule.anchor->name : "", 1, 0);
 413         }
 414 #endif /* OPT_DEBUG */
 415
 416
 417         if (remove_identical_rules(pf, block))
 418                 return (1);
 419         if (combine_rules(pf, block))
 420                 return (1);
 421         if ((pf->optimize & PF_OPTIMIZE_PROFILE) &&
 422             TAILQ_FIRST(&block->sb_rules)->por_rule.quick &&
 423             block->sb_profiled_block) {
 424                 if (block_feedback(pf, block))
 425                         return (1);
 426         } else if (reorder_rules(pf, block, 0)) {
 427                 return (1);
 428         }
 429
 430         /*
 431          * Don't add any optimization passes below reorder_rules().  It will
 432          * have divided superblocks into smaller blocks for further refinement
 433          * and doesn't put them back together again.  What once was a true
 434          * superblock might have been split into multiple superblocks.
 435          */
 436
 437 #ifdef OPT_DEBUG
 438         printf("--- END Superblock ---\n");
 439 #endif /* OPT_DEBUG */
 440         return (0);
 441 }
 442
 443
 444 /*
 445  * Optimization pass #1: remove identical rules
 446  */
 447 int
 448 remove_identical_rules(struct pfctl *pf, struct superblock *block)
 449 {
 450         struct pf_opt_rule *por1, *por2, *por_next, *por2_next;
 451         struct pf_rule a, a2, b, b2;
 452
 453         for (por1 = TAILQ_FIRST(&block->sb_rules); por1; por1 = por_next) {
 454                 por_next = TAILQ_NEXT(por1, por_entry);
 455                 for (por2 = por_next; por2; por2 = por2_next) {
 456                         por2_next = TAILQ_NEXT(por2, por_entry);
 457                         comparable_rule(&a, &por1->por_rule, DC);
 458                         comparable_rule(&b, &por2->por_rule, DC);
 459                         memcpy(&a2, &a, sizeof(a2));
 460                         memcpy(&b2, &b, sizeof(b2));
 461
 462                         exclude_supersets(&a, &b);
 463                         exclude_supersets(&b2, &a2);
 464                         if (memcmp(&a, &b, sizeof(a)) == 0) {
 465                                 DEBUG("removing identical rule  nr%d = *nr%d*",
 466                                     por1->por_rule.nr, por2->por_rule.nr);
 467                                 TAILQ_REMOVE(&block->sb_rules, por2, por_entry);
 468                                 if (por_next == por2)
 469                                         por_next = TAILQ_NEXT(por1, por_entry);
 470                                 free(por2);
 471                         } else if (memcmp(&a2, &b2, sizeof(a2)) == 0) {
 472                                 DEBUG("removing identical rule  *nr%d* = nr%d",
 473                                     por1->por_rule.nr, por2->por_rule.nr);
 474                                 TAILQ_REMOVE(&block->sb_rules, por1, por_entry);
 475                                 free(por1);
 476                                 break;
 477                         }
 478                 }
 479         }
 480
 481         return (0);
 482 }
 483
 484
 485 /*
 486  * Optimization pass #2: combine similar rules with different addresses
 487  * into a single rule and a table
 488  */
 489 int
 490 combine_rules(struct pfctl *pf, struct superblock *block)
 491 {
 492         struct pf_opt_rule *p1, *p2, *por_next;
 493         int src_eq, dst_eq;
 494
 495         if ((pf->loadopt & PFCTL_FLAG_TABLE) == 0) {
 496                 warnx("Must enable table loading for optimizations");
 497                 return (1);
 498         }
 499
 500         /* First we make a pass to combine the rules.  O(n log n) */
 501         TAILQ_FOREACH(p1, &block->sb_rules, por_entry) {
 502                 for (p2 = TAILQ_NEXT(p1, por_entry); p2; p2 = por_next) {
 503                         por_next = TAILQ_NEXT(p2, por_entry);
 504
 505                         src_eq = addrs_equal(&p1->por_rule.src,
 506                             &p2->por_rule.src);
 507                         dst_eq = addrs_equal(&p1->por_rule.dst,
 508                             &p2->por_rule.dst);
 509
 510                         if (src_eq && !dst_eq && p1->por_src_tbl == NULL &&
 511                             p2->por_dst_tbl == NULL &&
 512                             p2->por_src_tbl == NULL &&
 513                             rules_combineable(&p1->por_rule, &p2->por_rule) &&
 514                             addrs_combineable(&p1->por_rule.dst,
 515                             &p2->por_rule.dst)) {
 516                                 DEBUG("can combine rules  nr%d = nr%d",
 517                                     p1->por_rule.nr, p2->por_rule.nr);
 518                                 if (p1->por_dst_tbl == NULL &&
 519                                     add_opt_table(pf, &p1->por_dst_tbl,
 520                                     p1->por_rule.af, &p1->por_rule.dst))
 521                                         return (1);
 522                                 if (add_opt_table(pf, &p1->por_dst_tbl,
 523                                     p1->por_rule.af, &p2->por_rule.dst))
 524                                         return (1);
 525                                 p2->por_dst_tbl = p1->por_dst_tbl;
 526                                 if (p1->por_dst_tbl->pt_rulecount >=
 527                                     TABLE_THRESHOLD) {
 528                                         TAILQ_REMOVE(&block->sb_rules, p2,
 529                                             por_entry);
 530                                         free(p2);
 531                                 }
 532                         } else if (!src_eq && dst_eq && p1->por_dst_tbl == NULL
 533                             && p2->por_src_tbl == NULL &&
 534                             p2->por_dst_tbl == NULL &&
 535                             rules_combineable(&p1->por_rule, &p2->por_rule) &&
 536                             addrs_combineable(&p1->por_rule.src,
 537                             &p2->por_rule.src)) {
 538                                 DEBUG("can combine rules  nr%d = nr%d",
 539                                     p1->por_rule.nr, p2->por_rule.nr);
 540                                 if (p1->por_src_tbl == NULL &&
 541                                     add_opt_table(pf, &p1->por_src_tbl,
 542                                     p1->por_rule.af, &p1->por_rule.src))
 543                                         return (1);
 544                                 if (add_opt_table(pf, &p1->por_src_tbl,
 545                                     p1->por_rule.af, &p2->por_rule.src))
 546                                         return (1);
 547                                 p2->por_src_tbl = p1->por_src_tbl;
 548                                 if (p1->por_src_tbl->pt_rulecount >=
 549                                     TABLE_THRESHOLD) {
 550                                         TAILQ_REMOVE(&block->sb_rules, p2,
 551                                             por_entry);
 552                                         free(p2);
 553                                 }
 554                         }
 555                 }
 556         }
 557
 558
 559         /*
 560          * Then we make a final pass to create a valid table name and
 561          * insert the name into the rules.
 562          */
 563         for (p1 = TAILQ_FIRST(&block->sb_rules); p1; p1 = por_next) {
 564                 por_next = TAILQ_NEXT(p1, por_entry);
 565                 assert(p1->por_src_tbl == NULL || p1->por_dst_tbl == NULL);
 566
 567                 if (p1->por_src_tbl && p1->por_src_tbl->pt_rulecount >=
 568                     TABLE_THRESHOLD) {
 569                         if (p1->por_src_tbl->pt_generated) {
 570                                 /* This rule is included in a table */
 571                                 TAILQ_REMOVE(&block->sb_rules, p1, por_entry);
 572                                 free(p1);
 573                                 continue;
 574                         }
 575                         p1->por_src_tbl->pt_generated = 1;
 576
 577                         if ((pf->opts & PF_OPT_NOACTION) == 0 &&
 578                             pf_opt_create_table(pf, p1->por_src_tbl))
 579                                 return (1);
 580
 581                         pf->tdirty = 1;
 582
 583                         if (pf->opts & PF_OPT_VERBOSE)
 584                                 print_tabledef(p1->por_src_tbl->pt_name,
 585                                     PFR_TFLAG_CONST, 1,
 586                                     &p1->por_src_tbl->pt_nodes);
 587
 588                         memset(&p1->por_rule.src.addr, 0,
 589                             sizeof(p1->por_rule.src.addr));
 590                         p1->por_rule.src.addr.type = PF_ADDR_TABLE;
 591                         strlcpy(p1->por_rule.src.addr.v.tblname,
 592                             p1->por_src_tbl->pt_name,
 593                             sizeof(p1->por_rule.src.addr.v.tblname));
 594
 595                         pfr_buf_clear(p1->por_src_tbl->pt_buf);
 596                         free(p1->por_src_tbl->pt_buf);
 597                         p1->por_src_tbl->pt_buf = NULL;
 598                 }
 599                 if (p1->por_dst_tbl && p1->por_dst_tbl->pt_rulecount >=
 600                     TABLE_THRESHOLD) {
 601                         if (p1->por_dst_tbl->pt_generated) {
 602                                 /* This rule is included in a table */
 603                                 TAILQ_REMOVE(&block->sb_rules, p1, por_entry);
 604                                 free(p1);
 605                                 continue;
 606                         }
 607                         p1->por_dst_tbl->pt_generated = 1;
 608
 609                         if ((pf->opts & PF_OPT_NOACTION) == 0 &&
 610                             pf_opt_create_table(pf, p1->por_dst_tbl))
 611                                 return (1);
 612                         pf->tdirty = 1;
 613
 614                         if (pf->opts & PF_OPT_VERBOSE)
 615                                 print_tabledef(p1->por_dst_tbl->pt_name,
 616                                     PFR_TFLAG_CONST, 1,
 617                                     &p1->por_dst_tbl->pt_nodes);
 618
 619                         memset(&p1->por_rule.dst.addr, 0,
 620                             sizeof(p1->por_rule.dst.addr));
 621                         p1->por_rule.dst.addr.type = PF_ADDR_TABLE;
 622                         strlcpy(p1->por_rule.dst.addr.v.tblname,
 623                             p1->por_dst_tbl->pt_name,
 624                             sizeof(p1->por_rule.dst.addr.v.tblname));
 625
 626                         pfr_buf_clear(p1->por_dst_tbl->pt_buf);
 627                         free(p1->por_dst_tbl->pt_buf);
 628                         p1->por_dst_tbl->pt_buf = NULL;
 629                 }
 630         }
 631
 632         return (0);
 633 }
 634
 635
 636 /*
 637  * Optimization pass #3: re-order rules to improve skip steps
 638  */
 639 int
 640 reorder_rules(struct pfctl *pf, struct superblock *block, int depth)
 641 {
 642         struct superblock *newblock;
 643         struct pf_skip_step *skiplist;
 644         struct pf_opt_rule *por;
 645         int i, largest, largest_list, rule_count = 0;
 646         TAILQ_HEAD( , pf_opt_rule) head;
 647
 648         /*
 649          * Calculate the best-case skip steps.  We put each rule in a list
 650          * of other rules with common fields
 651          */
 652         for (i = 0; i < PF_SKIP_COUNT; i++) {
 653                 TAILQ_FOREACH(por, &block->sb_rules, por_entry) {
 654                         TAILQ_FOREACH(skiplist, &block->sb_skipsteps[i],
 655                             ps_entry) {
 656                                 if (skip_compare(i, skiplist, por) == 0)
 657                                         break;
 658                         }
 659                         if (skiplist == NULL) {
 660                                 if ((skiplist = calloc(1, sizeof(*skiplist))) ==
 661                                     NULL)
 662                                         err(1, "calloc");
 663                                 TAILQ_INIT(&skiplist->ps_rules);
 664                                 TAILQ_INSERT_TAIL(&block->sb_skipsteps[i],
 665                                     skiplist, ps_entry);
 666                         }
 667                         skip_append(block, i, skiplist, por);
 668                 }
 669         }
 670
 671         TAILQ_FOREACH(por, &block->sb_rules, por_entry)
 672                 rule_count++;
 673
 674         /*
 675          * Now we're going to ignore any fields that are identical between
 676          * all of the rules in the superblock and those fields which differ
 677          * between every rule in the superblock.
 678          */
 679         largest = 0;
 680         for (i = 0; i < PF_SKIP_COUNT; i++) {
 681                 skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]);
 682                 if (skiplist->ps_count == rule_count) {
 683                         DEBUG("(%d) original skipstep '%s' is all rules",
 684                             depth, skip_comparitors_names[i]);
 685                         skiplist->ps_count = 0;
 686                 } else if (skiplist->ps_count == 1) {
 687                         skiplist->ps_count = 0;
 688                 } else {
 689                         DEBUG("(%d) original skipstep '%s' largest jump is %d",
 690                             depth, skip_comparitors_names[i],
 691                             skiplist->ps_count);
 692                         if (skiplist->ps_count > largest)
 693                                 largest = skiplist->ps_count;
 694                 }
 695         }
 696         if (largest == 0) {
 697                 /* Ugh.  There is NO commonality in the superblock on which
 698                  * optimize the skipsteps optimization.
 699                  */
 700                 goto done;
 701         }
 702
 703         /*
 704          * Now we're going to empty the superblock rule list and re-create
 705          * it based on a more optimal skipstep order.
 706          */
 707         TAILQ_INIT(&head);
 708         while ((por = TAILQ_FIRST(&block->sb_rules))) {
 709                 TAILQ_REMOVE(&block->sb_rules, por, por_entry);
 710                 TAILQ_INSERT_TAIL(&head, por, por_entry);
 711         }
 712
 713
 714         while (!TAILQ_EMPTY(&head)) {
 715                 largest = 1;
 716
 717                 /*
 718                  * Find the most useful skip steps remaining
 719                  */
 720                 for (i = 0; i < PF_SKIP_COUNT; i++) {
 721                         skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]);
 722                         if (skiplist->ps_count > largest) {
 723                                 largest = skiplist->ps_count;
 724                                 largest_list = i;
 725                         }
 726                 }
 727
 728                 if (largest <= 1) {
 729                         /*
 730                          * Nothing useful left.  Leave remaining rules in order.
 731                          */
 732                         DEBUG("(%d) no more commonality for skip steps", depth);
 733                         while ((por = TAILQ_FIRST(&head))) {
 734                                 TAILQ_REMOVE(&head, por, por_entry);
 735                                 TAILQ_INSERT_TAIL(&block->sb_rules, por,
 736                                     por_entry);
 737                         }
 738                 } else {
 739                         /*
 740                          * There is commonality.  Extract those common rules
 741                          * and place them in the ruleset adjacent to each
 742                          * other.
 743                          */
 744                         skiplist = TAILQ_FIRST(&block->sb_skipsteps[
 745                             largest_list]);
 746                         DEBUG("(%d) skipstep '%s' largest jump is %d @ #%d",
 747                             depth, skip_comparitors_names[largest_list],
 748                             largest, TAILQ_FIRST(&TAILQ_FIRST(&block->
 749                             sb_skipsteps [largest_list])->ps_rules)->
 750                             por_rule.nr);
 751                         TAILQ_REMOVE(&block->sb_skipsteps[largest_list],
 752                             skiplist, ps_entry);
 753
 754
 755                         /*
 756                          * There may be further commonality inside these
 757                          * rules.  So we'll split them off into they're own
 758                          * superblock and pass it back into the optimizer.
 759                          */
 760                         if (skiplist->ps_count > 2) {
 761                                 if ((newblock = calloc(1, sizeof(*newblock)))
 762                                     == NULL) {
 763                                         warn("calloc");
 764                                         return (1);
 765                                 }
 766                                 TAILQ_INIT(&newblock->sb_rules);
 767                                 for (i = 0; i < PF_SKIP_COUNT; i++)
 768                                         TAILQ_INIT(&newblock->sb_skipsteps[i]);
 769                                 TAILQ_INSERT_BEFORE(block, newblock, sb_entry);
 770                                 DEBUG("(%d) splitting off %d rules from superblock @ #%d",
 771                                     depth, skiplist->ps_count,
 772                                     TAILQ_FIRST(&skiplist->ps_rules)->
 773                                     por_rule.nr);
 774                         } else {
 775                                 newblock = block;
 776                         }
 777
 778                         while ((por = TAILQ_FIRST(&skiplist->ps_rules))) {
 779                                 TAILQ_REMOVE(&head, por, por_entry);
 780                                 TAILQ_REMOVE(&skiplist->ps_rules, por,
 781                                     por_skip_entry[largest_list]);
 782                                 TAILQ_INSERT_TAIL(&newblock->sb_rules, por,
 783                                     por_entry);
 784
 785                                 /* Remove this rule from all other skiplists */
 786                                 remove_from_skipsteps(&block->sb_skipsteps[
 787                                     largest_list], block, por, skiplist);
 788                         }
 789                         free(skiplist);
 790                         if (newblock != block)
 791                                 if (reorder_rules(pf, newblock, depth + 1))
 792                                         return (1);
 793                 }
 794         }
 795
 796 done:
 797         for (i = 0; i < PF_SKIP_COUNT; i++) {
 798                 while ((skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]))) {
 799                         TAILQ_REMOVE(&block->sb_skipsteps[i], skiplist,
 800                             ps_entry);
 801                         free(skiplist);
 802                 }
 803         }
 804
 805         return (0);
 806 }
 807
 808
 809 /*
 810  * Optimization pass #4: re-order 'quick' rules based on feedback from the
 811  * currently running ruleset
 812  */
 813 int
 814 block_feedback(struct pfctl *pf, struct superblock *block)
 815 {
 816         TAILQ_HEAD( , pf_opt_rule) queue;
 817         struct pf_opt_rule *por1, *por2;
 818         u_int64_t total_count = 0;
 819         struct pf_rule a, b;
 820
 821
 822         /*
 823          * Walk through all of the profiled superblock's rules and copy
 824          * the counters onto our rules.
 825          */
 826         TAILQ_FOREACH(por1, &block->sb_profiled_block->sb_rules, por_entry) {
 827                 comparable_rule(&a, &por1->por_rule, DC);
 828                 total_count += por1->por_rule.packets[0] +
 829                     por1->por_rule.packets[1];
 830                 TAILQ_FOREACH(por2, &block->sb_rules, por_entry) {
 831                         if (por2->por_profile_count)
 832                                 continue;
 833                         comparable_rule(&b, &por2->por_rule, DC);
 834                         if (memcmp(&a, &b, sizeof(a)) == 0) {
 835                                 por2->por_profile_count =
 836                                     por1->por_rule.packets[0] +
 837                                     por1->por_rule.packets[1];
 838                                 break;
 839                         }
 840                 }
 841         }
 842         superblock_free(pf, block->sb_profiled_block);
 843         block->sb_profiled_block = NULL;
 844
 845         /*
 846          * Now we pull all of the rules off the superblock and re-insert them
 847          * in sorted order.
 848          */
 849
 850         TAILQ_INIT(&queue);
 851         while ((por1 = TAILQ_FIRST(&block->sb_rules)) != NULL) {
 852                 TAILQ_REMOVE(&block->sb_rules, por1, por_entry);
 853                 TAILQ_INSERT_TAIL(&queue, por1, por_entry);
 854         }
 855
 856         while ((por1 = TAILQ_FIRST(&queue)) != NULL) {
 857                 TAILQ_REMOVE(&queue, por1, por_entry);
 858 /* XXX I should sort all of the unused rules based on skip steps */
 859                 TAILQ_FOREACH(por2, &block->sb_rules, por_entry) {
 860                         if (por1->por_profile_count > por2->por_profile_count) {
 861                                 TAILQ_INSERT_BEFORE(por2, por1, por_entry);
 862                                 break;
 863                         }
 864                 }
 865 #ifdef __FreeBSD__
 866                 if (por2 == NULL)
 867 #else
 868                 if (por2 == TAILQ_END(&block->sb_rules))
 869 #endif
 870                         TAILQ_INSERT_TAIL(&block->sb_rules, por1, por_entry);
 871         }
 872
 873         return (0);
 874 }
 875
 876
 877 /*
 878  * Load the current ruleset from the kernel and try to associate them with
 879  * the ruleset we're optimizing.
 880  */
 881 int
 882 load_feedback_profile(struct pfctl *pf, struct superblocks *superblocks)
 883 {
 884         struct superblock *block, *blockcur;
 885         struct superblocks prof_superblocks;
 886         struct pf_opt_rule *por;
 887         struct pf_opt_queue queue;
 888         struct pfioc_rule pr;
 889         struct pf_rule a, b;
 890         int nr, mnr;
 891
 892         TAILQ_INIT(&queue);
 893         TAILQ_INIT(&prof_superblocks);
 894
 895         memset(&pr, 0, sizeof(pr));
 896         pr.rule.action = PF_PASS;
 897         if (ioctl(pf->dev, DIOCGETRULES, &pr)) {
 898                 warn("DIOCGETRULES");
 899                 return (1);
 900         }
 901         mnr = pr.nr;
 902
 903         DEBUG("Loading %d active rules for a feedback profile", mnr);
 904         for (nr = 0; nr < mnr; ++nr) {
 905                 struct pf_ruleset *rs;
 906                 if ((por = calloc(1, sizeof(*por))) == NULL) {
 907                         warn("calloc");
 908                         return (1);
 909                 }
 910                 pr.nr = nr;
 911                 if (ioctl(pf->dev, DIOCGETRULE, &pr)) {
 912                         warn("DIOCGETRULES");
 913                         return (1);
 914                 }
 915                 memcpy(&por->por_rule, &pr.rule, sizeof(por->por_rule));
 916                 rs = pf_find_or_create_ruleset(pr.anchor_call);
 917                 por->por_rule.anchor = rs->anchor;
 918                 if (TAILQ_EMPTY(&por->por_rule.rpool.list))
 919                         memset(&por->por_rule.rpool, 0,
 920                             sizeof(por->por_rule.rpool));
 921                 TAILQ_INSERT_TAIL(&queue, por, por_entry);
 922
 923                 /* XXX pfctl_get_pool(pf->dev, &pr.rule.rpool, nr, pr.ticket,
 924                  *         PF_PASS, pf->anchor) ???
 925                  * ... pfctl_clear_pool(&pr.rule.rpool)
 926                  */
 927         }
 928
 929         if (construct_superblocks(pf, &queue, &prof_superblocks))
 930                 return (1);
 931
 932
 933         /*
 934          * Now we try to associate the active ruleset's superblocks with
 935          * the superblocks we're compiling.
 936          */
 937         block = TAILQ_FIRST(superblocks);
 938         blockcur = TAILQ_FIRST(&prof_superblocks);
 939         while (block && blockcur) {
 940                 comparable_rule(&a, &TAILQ_FIRST(&block->sb_rules)->por_rule,
 941                     BREAK);
 942                 comparable_rule(&b, &TAILQ_FIRST(&blockcur->sb_rules)->por_rule,
 943                     BREAK);
 944                 if (memcmp(&a, &b, sizeof(a)) == 0) {
 945                         /* The two superblocks lined up */
 946                         block->sb_profiled_block = blockcur;
 947                 } else {
 948                         DEBUG("superblocks don't line up between #%d and #%d",
 949                             TAILQ_FIRST(&block->sb_rules)->por_rule.nr,
 950                             TAILQ_FIRST(&blockcur->sb_rules)->por_rule.nr);
 951                         break;
 952                 }
 953                 block = TAILQ_NEXT(block, sb_entry);
 954                 blockcur = TAILQ_NEXT(blockcur, sb_entry);
 955         }
 956
 957
 958
 959         /* Free any superblocks we couldn't link */
 960         while (blockcur) {
 961                 block = TAILQ_NEXT(blockcur, sb_entry);
 962                 superblock_free(pf, blockcur);
 963                 blockcur = block;
 964         }
 965         return (0);
 966 }
 967
 968
 969 /*
 970  * Compare a rule to a skiplist to see if the rule is a member
 971  */
 972 int
 973 skip_compare(int skipnum, struct pf_skip_step *skiplist,
 974     struct pf_opt_rule *por)
 975 {
 976         struct pf_rule *a, *b;
 977         if (skipnum >= PF_SKIP_COUNT || skipnum < 0)
 978                 errx(1, "skip_compare() out of bounds");
 979         a = &por->por_rule;
 980         b = &TAILQ_FIRST(&skiplist->ps_rules)->por_rule;
 981
 982         return ((skip_comparitors[skipnum])(a, b));
 983 }
 984
 985
 986 /*
 987  * Add a rule to a skiplist
 988  */
 989 void
 990 skip_append(struct superblock *superblock, int skipnum,
 991     struct pf_skip_step *skiplist, struct pf_opt_rule *por)
 992 {
 993         struct pf_skip_step *prev;
 994
 995         skiplist->ps_count++;
 996         TAILQ_INSERT_TAIL(&skiplist->ps_rules, por, por_skip_entry[skipnum]);
 997
 998         /* Keep the list of skiplists sorted by whichever is larger */
 999         while ((prev = TAILQ_PREV(skiplist, skiplist, ps_entry)) &&
1000             prev->ps_count < skiplist->ps_count) {
1001                 TAILQ_REMOVE(&superblock->sb_skipsteps[skipnum],
1002                     skiplist, ps_entry);
1003                 TAILQ_INSERT_BEFORE(prev, skiplist, ps_entry);
1004         }
1005 }
1006
1007
1008 /*
1009  * Remove a rule from the other skiplist calculations.
1010  */
1011 void
1012 remove_from_skipsteps(struct skiplist *head, struct superblock *block,
1013     struct pf_opt_rule *por, struct pf_skip_step *active_list)
1014 {
1015         struct pf_skip_step *sk, *next;
1016         struct pf_opt_rule *p2;
1017         int i, found;
1018
1019         for (i = 0; i < PF_SKIP_COUNT; i++) {
1020                 sk = TAILQ_FIRST(&block->sb_skipsteps[i]);
1021                 if (sk == NULL || sk == active_list || sk->ps_count <= 1)
1022                         continue;
1023                 found = 0;
1024                 do {
1025                         TAILQ_FOREACH(p2, &sk->ps_rules, por_skip_entry[i])
1026                                 if (p2 == por) {
1027                                         TAILQ_REMOVE(&sk->ps_rules, p2,
1028                                             por_skip_entry[i]);
1029                                         found = 1;
1030                                         sk->ps_count--;
1031                                         break;
1032                                 }
1033                 } while (!found && (sk = TAILQ_NEXT(sk, ps_entry)));
1034                 if (found && sk) {
1035                         /* Does this change the sorting order? */
1036                         while ((next = TAILQ_NEXT(sk, ps_entry)) &&
1037                             next->ps_count > sk->ps_count) {
1038                                 TAILQ_REMOVE(head, sk, ps_entry);
1039                                 TAILQ_INSERT_AFTER(head, next, sk, ps_entry);
1040                         }
1041 #ifdef OPT_DEBUG
1042                         next = TAILQ_NEXT(sk, ps_entry);
1043                         assert(next == NULL || next->ps_count <= sk->ps_count);
1044 #endif /* OPT_DEBUG */
1045                 }
1046         }
1047 }
1048
1049
1050 /* Compare two rules AF field for skiplist construction */
1051 int
1052 skip_cmp_af(struct pf_rule *a, struct pf_rule *b)
1053 {
1054         if (a->af != b->af || a->af == 0)
1055                 return (1);
1056         return (0);
1057 }
1058
1059 /* Compare two rules DIRECTION field for skiplist construction */
1060 int
1061 skip_cmp_dir(struct pf_rule *a, struct pf_rule *b)
1062 {
1063         if (a->direction == 0 || a->direction != b->direction)
1064                 return (1);
1065         return (0);
1066 }
1067
1068 /* Compare two rules DST Address field for skiplist construction */
1069 int
1070 skip_cmp_dst_addr(struct pf_rule *a, struct pf_rule *b)
1071 {
1072         if (a->dst.neg != b->dst.neg ||
1073             a->dst.addr.type != b->dst.addr.type)
1074                 return (1);
1075         /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0
1076          *    && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP ||
1077          *    a->proto == IPPROTO_ICMP
1078          *      return (1);
1079          */
1080         switch (a->dst.addr.type) {
1081         case PF_ADDR_ADDRMASK:
1082                 if (memcmp(&a->dst.addr.v.a.addr, &b->dst.addr.v.a.addr,
1083                     sizeof(a->dst.addr.v.a.addr)) ||
1084                     memcmp(&a->dst.addr.v.a.mask, &b->dst.addr.v.a.mask,
1085                     sizeof(a->dst.addr.v.a.mask)) ||
1086                     (a->dst.addr.v.a.addr.addr32[0] == 0 &&
1087                     a->dst.addr.v.a.addr.addr32[1] == 0 &&
1088                     a->dst.addr.v.a.addr.addr32[2] == 0 &&
1089                     a->dst.addr.v.a.addr.addr32[3] == 0))
1090                         return (1);
1091                 return (0);
1092         case PF_ADDR_DYNIFTL:
1093                 if (strcmp(a->dst.addr.v.ifname, b->dst.addr.v.ifname) != 0 ||
1094                     a->dst.addr.iflags != a->dst.addr.iflags ||
1095                     memcmp(&a->dst.addr.v.a.mask, &b->dst.addr.v.a.mask,
1096                     sizeof(a->dst.addr.v.a.mask)))
1097                         return (1);
1098                 return (0);
1099         case PF_ADDR_NOROUTE:
1100         case PF_ADDR_URPFFAILED:
1101                 return (0);
1102         case PF_ADDR_TABLE:
1103                 return (strcmp(a->dst.addr.v.tblname, b->dst.addr.v.tblname));
1104         }
1105         return (1);
1106 }
1107
1108 /* Compare two rules DST port field for skiplist construction */
1109 int
1110 skip_cmp_dst_port(struct pf_rule *a, struct pf_rule *b)
1111 {
1112         /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0
1113          *    && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP ||
1114          *    a->proto == IPPROTO_ICMP
1115          *      return (1);
1116          */
1117         if (a->dst.port_op == PF_OP_NONE || a->dst.port_op != b->dst.port_op ||
1118             a->dst.port[0] != b->dst.port[0] ||
1119             a->dst.port[1] != b->dst.port[1])
1120                 return (1);
1121         return (0);
1122 }
1123
1124 /* Compare two rules IFP field for skiplist construction */
1125 int
1126 skip_cmp_ifp(struct pf_rule *a, struct pf_rule *b)
1127 {
1128         if (strcmp(a->ifname, b->ifname) || a->ifname[0] == '\0')
1129                 return (1);
1130         return (a->ifnot != b->ifnot);
1131 }
1132
1133 /* Compare two rules PROTO field for skiplist construction */
1134 int
1135 skip_cmp_proto(struct pf_rule *a, struct pf_rule *b)
1136 {
1137         return (a->proto != b->proto || a->proto == 0);
1138 }
1139
1140 /* Compare two rules SRC addr field for skiplist construction */
1141 int
1142 skip_cmp_src_addr(struct pf_rule *a, struct pf_rule *b)
1143 {
1144         if (a->src.neg != b->src.neg ||
1145             a->src.addr.type != b->src.addr.type)
1146                 return (1);
1147         /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0
1148          *    && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP ||
1149          *    a->proto == IPPROTO_ICMP
1150          *      return (1);
1151          */
1152         switch (a->src.addr.type) {
1153         case PF_ADDR_ADDRMASK:
1154                 if (memcmp(&a->src.addr.v.a.addr, &b->src.addr.v.a.addr,
1155                     sizeof(a->src.addr.v.a.addr)) ||
1156                     memcmp(&a->src.addr.v.a.mask, &b->src.addr.v.a.mask,
1157                     sizeof(a->src.addr.v.a.mask)) ||
1158                     (a->src.addr.v.a.addr.addr32[0] == 0 &&
1159                     a->src.addr.v.a.addr.addr32[1] == 0 &&
1160                     a->src.addr.v.a.addr.addr32[2] == 0 &&
1161                     a->src.addr.v.a.addr.addr32[3] == 0))
1162                         return (1);
1163                 return (0);
1164         case PF_ADDR_DYNIFTL:
1165                 if (strcmp(a->src.addr.v.ifname, b->src.addr.v.ifname) != 0 ||
1166                     a->src.addr.iflags != a->src.addr.iflags ||
1167                     memcmp(&a->src.addr.v.a.mask, &b->src.addr.v.a.mask,
1168                     sizeof(a->src.addr.v.a.mask)))
1169                         return (1);
1170                 return (0);
1171         case PF_ADDR_NOROUTE:
1172         case PF_ADDR_URPFFAILED:
1173                 return (0);
1174         case PF_ADDR_TABLE:
1175                 return (strcmp(a->src.addr.v.tblname, b->src.addr.v.tblname));
1176         }
1177         return (1);
1178 }
1179
1180 /* Compare two rules SRC port field for skiplist construction */
1181 int
1182 skip_cmp_src_port(struct pf_rule *a, struct pf_rule *b)
1183 {
1184         if (a->src.port_op == PF_OP_NONE || a->src.port_op != b->src.port_op ||
1185             a->src.port[0] != b->src.port[0] ||
1186             a->src.port[1] != b->src.port[1])
1187                 return (1);
1188         /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0
1189          *    && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP ||
1190          *    a->proto == IPPROTO_ICMP
1191          *      return (1);
1192          */
1193         return (0);
1194 }
1195
1196
1197 void
1198 skip_init(void)
1199 {
1200         struct {
1201                 char *name;
1202                 int skipnum;
1203                 int (*func)(struct pf_rule *, struct pf_rule *);
1204         } comps[] = PF_SKIP_COMPARITORS;
1205         int skipnum, i;
1206
1207         for (skipnum = 0; skipnum < PF_SKIP_COUNT; skipnum++) {
1208                 for (i = 0; i < sizeof(comps)/sizeof(*comps); i++)
1209                         if (comps[i].skipnum == skipnum) {
1210                                 skip_comparitors[skipnum] = comps[i].func;
1211                                 skip_comparitors_names[skipnum] = comps[i].name;
1212                         }
1213         }
1214         for (skipnum = 0; skipnum < PF_SKIP_COUNT; skipnum++)
1215                 if (skip_comparitors[skipnum] == NULL)
1216                         errx(1, "Need to add skip step comparitor to pfctl?!");
1217 }
1218
1219 /*
1220  * Add a host/netmask to a table
1221  */
1222 int
1223 add_opt_table(struct pfctl *pf, struct pf_opt_tbl **tbl, sa_family_t af,
1224     struct pf_rule_addr *addr)
1225 {
1226 #ifdef OPT_DEBUG
1227         char buf[128];
1228 #endif /* OPT_DEBUG */
1229         static int tablenum = 0;
1230         struct node_host node_host;
1231
1232         if (*tbl == NULL) {
1233                 if ((*tbl = calloc(1, sizeof(**tbl))) == NULL ||
1234                     ((*tbl)->pt_buf = calloc(1, sizeof(*(*tbl)->pt_buf))) ==
1235                     NULL)
1236                         err(1, "calloc");
1237                 (*tbl)->pt_buf->pfrb_type = PFRB_ADDRS;
1238                 SIMPLEQ_INIT(&(*tbl)->pt_nodes);
1239
1240                 /* This is just a temporary table name */
1241                 snprintf((*tbl)->pt_name, sizeof((*tbl)->pt_name), "%s%d",
1242                     PF_OPT_TABLE_PREFIX, tablenum++);
1243                 DEBUG("creating table <%s>", (*tbl)->pt_name);
1244         }
1245
1246         memset(&node_host, 0, sizeof(node_host));
1247         node_host.af = af;
1248         node_host.addr = addr->addr;
1249
1250 #ifdef OPT_DEBUG
1251         DEBUG("<%s> adding %s/%d", (*tbl)->pt_name, inet_ntop(af,
1252             &node_host.addr.v.a.addr, buf, sizeof(buf)),
1253             unmask(&node_host.addr.v.a.mask, af));
1254 #endif /* OPT_DEBUG */
1255
1256         if (append_addr_host((*tbl)->pt_buf, &node_host, 0, 0)) {
1257                 warn("failed to add host");
1258                 return (1);
1259         }
1260         if (pf->opts & PF_OPT_VERBOSE) {
1261                 struct node_tinit *ti;
1262
1263                 if ((ti = calloc(1, sizeof(*ti))) == NULL)
1264                         err(1, "malloc");
1265                 if ((ti->host = malloc(sizeof(*ti->host))) == NULL)
1266                         err(1, "malloc");
1267                 memcpy(ti->host, &node_host, sizeof(*ti->host));
1268                 SIMPLEQ_INSERT_TAIL(&(*tbl)->pt_nodes, ti, entries);
1269         }
1270
1271         (*tbl)->pt_rulecount++;
1272         if ((*tbl)->pt_rulecount == TABLE_THRESHOLD)
1273                 DEBUG("table <%s> now faster than skip steps", (*tbl)->pt_name);
1274
1275         return (0);
1276 }
1277
1278
1279 /*
1280  * Do the dirty work of choosing an unused table name and creating it.
1281  * (be careful with the table name, it might already be used in another anchor)
1282  */
1283 int
1284 pf_opt_create_table(struct pfctl *pf, struct pf_opt_tbl *tbl)
1285 {
1286         static int tablenum;
1287         struct pfr_table *t;
1288
1289         if (table_buffer.pfrb_type == 0) {
1290                 /* Initialize the list of tables */
1291                 table_buffer.pfrb_type = PFRB_TABLES;
1292                 for (;;) {
1293                         pfr_buf_grow(&table_buffer, table_buffer.pfrb_size);
1294                         table_buffer.pfrb_size = table_buffer.pfrb_msize;
1295                         if (pfr_get_tables(NULL, table_buffer.pfrb_caddr,
1296                             &table_buffer.pfrb_size, PFR_FLAG_ALLRSETS))
1297                                 err(1, "pfr_get_tables");
1298                         if (table_buffer.pfrb_size <= table_buffer.pfrb_msize)
1299                                 break;
1300                 }
1301                 table_identifier = arc4random();
1302         }
1303
1304         /* XXX would be *really* nice to avoid duplicating identical tables */
1305
1306         /* Now we have to pick a table name that isn't used */
1307 again:
1308         DEBUG("translating temporary table <%s> to <%s%x_%d>", tbl->pt_name,
1309             PF_OPT_TABLE_PREFIX, table_identifier, tablenum);
1310         snprintf(tbl->pt_name, sizeof(tbl->pt_name), "%s%x_%d",
1311             PF_OPT_TABLE_PREFIX, table_identifier, tablenum);
1312         PFRB_FOREACH(t, &table_buffer) {
1313                 if (strcasecmp(t->pfrt_name, tbl->pt_name) == 0) {
1314                         /* Collision.  Try again */
1315                         DEBUG("wow, table <%s> in use.  trying again",
1316                             tbl->pt_name);
1317                         table_identifier = arc4random();
1318                         goto again;
1319                 }
1320         }
1321         tablenum++;
1322
1323
1324         if (pfctl_define_table(tbl->pt_name, PFR_TFLAG_CONST, 1,
1325             pf->astack[0]->name, tbl->pt_buf, pf->astack[0]->ruleset.tticket)) {
1326                 warn("failed to create table %s in %s",
1327                     tbl->pt_name, pf->astack[0]->name);
1328                 return (1);
1329         }
1330         return (0);
1331 }
1332
1333 /*
1334  * Partition the flat ruleset into a list of distinct superblocks
1335  */
1336 int
1337 construct_superblocks(struct pfctl *pf, struct pf_opt_queue *opt_queue,
1338     struct superblocks *superblocks)
1339 {
1340         struct superblock *block = NULL;
1341         struct pf_opt_rule *por;
1342         int i;
1343
1344         while (!TAILQ_EMPTY(opt_queue)) {
1345                 por = TAILQ_FIRST(opt_queue);
1346                 TAILQ_REMOVE(opt_queue, por, por_entry);
1347                 if (block == NULL || !superblock_inclusive(block, por)) {
1348                         if ((block = calloc(1, sizeof(*block))) == NULL) {
1349                                 warn("calloc");
1350                                 return (1);
1351                         }
1352                         TAILQ_INIT(&block->sb_rules);
1353                         for (i = 0; i < PF_SKIP_COUNT; i++)
1354                                 TAILQ_INIT(&block->sb_skipsteps[i]);
1355                         TAILQ_INSERT_TAIL(superblocks, block, sb_entry);
1356                 }
1357                 TAILQ_INSERT_TAIL(&block->sb_rules, por, por_entry);
1358         }
1359
1360         return (0);
1361 }
1362
1363
1364 /*
1365  * Compare two rule addresses
1366  */
1367 int
1368 addrs_equal(struct pf_rule_addr *a, struct pf_rule_addr *b)
1369 {
1370         if (a->neg != b->neg)
1371                 return (0);
1372         return (memcmp(&a->addr, &b->addr, sizeof(a->addr)) == 0);
1373 }
1374
1375
1376 /*
1377  * The addresses are not equal, but can we combine them into one table?
1378  */
1379 int
1380 addrs_combineable(struct pf_rule_addr *a, struct pf_rule_addr *b)
1381 {
1382         if (a->addr.type != PF_ADDR_ADDRMASK ||
1383             b->addr.type != PF_ADDR_ADDRMASK)
1384                 return (0);
1385         if (a->neg != b->neg || a->port_op != b->port_op ||
1386             a->port[0] != b->port[0] || a->port[1] != b->port[1])
1387                 return (0);
1388         return (1);
1389 }
1390
1391
1392 /*
1393  * Are we allowed to combine these two rules
1394  */
1395 int
1396 rules_combineable(struct pf_rule *p1, struct pf_rule *p2)
1397 {
1398         struct pf_rule a, b;
1399
1400         comparable_rule(&a, p1, COMBINED);
1401         comparable_rule(&b, p2, COMBINED);
1402         return (memcmp(&a, &b, sizeof(a)) == 0);
1403 }
1404
1405
1406 /*
1407  * Can a rule be included inside a superblock
1408  */
1409 int
1410 superblock_inclusive(struct superblock *block, struct pf_opt_rule *por)
1411 {
1412         struct pf_rule a, b;
1413         int i, j;
1414
1415         /* First check for hard breaks */
1416         for (i = 0; i < sizeof(pf_rule_desc)/sizeof(*pf_rule_desc); i++) {
1417                 if (pf_rule_desc[i].prf_type == BARRIER) {
1418                         for (j = 0; j < pf_rule_desc[i].prf_size; j++)
1419                                 if (((char *)&por->por_rule)[j +
1420                                     pf_rule_desc[i].prf_offset] != 0)
1421                                         return (0);
1422                 }
1423         }
1424
1425         /* per-rule src-track is also a hard break */
1426         if (por->por_rule.rule_flag & PFRULE_RULESRCTRACK)
1427                 return (0);
1428
1429         /*
1430          * Have to handle interface groups separately.  Consider the following
1431          * rules:
1432          *      block on EXTIFS to any port 22
1433          *      pass  on em0 to any port 22
1434          * (where EXTIFS is an arbitrary interface group)
1435          * The optimizer may decide to re-order the pass rule in front of the
1436          * block rule.  But what if EXTIFS includes em0???  Such a reordering
1437          * would change the meaning of the ruleset.
1438          * We can't just lookup the EXTIFS group and check if em0 is a member
1439          * because the user is allowed to add interfaces to a group during
1440          * runtime.
1441          * Ergo interface groups become a defacto superblock break :-(
1442          */
1443         if (interface_group(por->por_rule.ifname) ||
1444             interface_group(TAILQ_FIRST(&block->sb_rules)->por_rule.ifname)) {
1445                 if (strcasecmp(por->por_rule.ifname,
1446                     TAILQ_FIRST(&block->sb_rules)->por_rule.ifname) != 0)
1447                         return (0);
1448         }
1449
1450         comparable_rule(&a, &TAILQ_FIRST(&block->sb_rules)->por_rule, NOMERGE);
1451         comparable_rule(&b, &por->por_rule, NOMERGE);
1452         if (memcmp(&a, &b, sizeof(a)) == 0)
1453                 return (1);
1454
1455 #ifdef OPT_DEBUG
1456         for (i = 0; i < sizeof(por->por_rule); i++) {
1457                 int closest = -1;
1458                 if (((u_int8_t *)&a)[i] != ((u_int8_t *)&b)[i]) {
1459                         for (j = 0; j < sizeof(pf_rule_desc) /
1460                             sizeof(*pf_rule_desc); j++) {
1461                                 if (i >= pf_rule_desc[j].prf_offset &&
1462                                     i < pf_rule_desc[j].prf_offset +
1463                                     pf_rule_desc[j].prf_size) {
1464                                         DEBUG("superblock break @ %d due to %s",
1465                                             por->por_rule.nr,
1466                                             pf_rule_desc[j].prf_name);
1467                                         return (0);
1468                                 }
1469                                 if (i > pf_rule_desc[j].prf_offset) {
1470                                         if (closest == -1 ||
1471                                             i-pf_rule_desc[j].prf_offset <
1472                                             i-pf_rule_desc[closest].prf_offset)
1473                                                 closest = j;
1474                                 }
1475                         }
1476
1477                         if (closest >= 0)
1478                                 DEBUG("superblock break @ %d on %s+%xh",
1479                                     por->por_rule.nr,
1480                                     pf_rule_desc[closest].prf_name,
1481                                     i - pf_rule_desc[closest].prf_offset -
1482                                     pf_rule_desc[closest].prf_size);
1483                         else
1484                                 DEBUG("superblock break @ %d on field @ %d",
1485                                     por->por_rule.nr, i);
1486                         return (0);
1487                 }
1488         }
1489 #endif /* OPT_DEBUG */
1490
1491         return (0);
1492 }
1493
1494
1495 /*
1496  * Figure out if an interface name is an actual interface or actually a
1497  * group of interfaces.
1498  */
1499 int
1500 interface_group(const char *ifname)
1501 {
1502         if (ifname == NULL || !ifname[0])
1503                 return (0);
1504
1505         /* Real interfaces must end in a number, interface groups do not */
1506         if (isdigit(ifname[strlen(ifname) - 1]))
1507                 return (0);
1508         else
1509                 return (1);
1510 }
1511
1512
1513 /*
1514  * Make a rule that can directly compared by memcmp()
1515  */
1516 void
1517 comparable_rule(struct pf_rule *dst, const struct pf_rule *src, int type)
1518 {
1519         int i;
1520         /*
1521          * To simplify the comparison, we just zero out the fields that are
1522          * allowed to be different and then do a simple memcmp()
1523          */
1524         memcpy(dst, src, sizeof(*dst));
1525         for (i = 0; i < sizeof(pf_rule_desc)/sizeof(*pf_rule_desc); i++)
1526                 if (pf_rule_desc[i].prf_type >= type) {
1527 #ifdef OPT_DEBUG
1528                         assert(pf_rule_desc[i].prf_type != NEVER ||
1529                             *(((char *)dst) + pf_rule_desc[i].prf_offset) == 0);
1530 #endif /* OPT_DEBUG */
1531                         memset(((char *)dst) + pf_rule_desc[i].prf_offset, 0,
1532                             pf_rule_desc[i].prf_size);
1533                 }
1534 }
1535
1536
1537 /*
1538  * Remove superset information from two rules so we can directly compare them
1539  * with memcmp()
1540  */
1541 void
1542 exclude_supersets(struct pf_rule *super, struct pf_rule *sub)
1543 {
1544         if (super->ifname[0] == '\0')
1545                 memset(sub->ifname, 0, sizeof(sub->ifname));
1546         if (super->direction == PF_INOUT)
1547                 sub->direction = PF_INOUT;
1548         if ((super->proto == 0 || super->proto == sub->proto) &&
1549             super->flags == 0 && super->flagset == 0 && (sub->flags ||
1550             sub->flagset)) {
1551                 sub->flags = super->flags;
1552                 sub->flagset = super->flagset;
1553         }
1554         if (super->proto == 0)
1555                 sub->proto = 0;
1556
1557         if (super->src.port_op == 0) {
1558                 sub->src.port_op = 0;
1559                 sub->src.port[0] = 0;
1560                 sub->src.port[1] = 0;
1561         }
1562         if (super->dst.port_op == 0) {
1563                 sub->dst.port_op = 0;
1564                 sub->dst.port[0] = 0;
1565                 sub->dst.port[1] = 0;
1566         }
1567
1568         if (super->src.addr.type == PF_ADDR_ADDRMASK && !super->src.neg &&
1569             !sub->src.neg && super->src.addr.v.a.mask.addr32[0] == 0 &&
1570             super->src.addr.v.a.mask.addr32[1] == 0 &&
1571             super->src.addr.v.a.mask.addr32[2] == 0 &&
1572             super->src.addr.v.a.mask.addr32[3] == 0)
1573                 memset(&sub->src.addr, 0, sizeof(sub->src.addr));
1574         else if (super->src.addr.type == PF_ADDR_ADDRMASK &&
1575             sub->src.addr.type == PF_ADDR_ADDRMASK &&
1576             super->src.neg == sub->src.neg &&
1577             super->af == sub->af &&
1578             unmask(&super->src.addr.v.a.mask, super->af) <
1579             unmask(&sub->src.addr.v.a.mask, sub->af) &&
1580             super->src.addr.v.a.addr.addr32[0] ==
1581             (sub->src.addr.v.a.addr.addr32[0] &
1582             super->src.addr.v.a.mask.addr32[0]) &&
1583             super->src.addr.v.a.addr.addr32[1] ==
1584             (sub->src.addr.v.a.addr.addr32[1] &
1585             super->src.addr.v.a.mask.addr32[1]) &&
1586             super->src.addr.v.a.addr.addr32[2] ==
1587             (sub->src.addr.v.a.addr.addr32[2] &
1588             super->src.addr.v.a.mask.addr32[2]) &&
1589             super->src.addr.v.a.addr.addr32[3] ==
1590             (sub->src.addr.v.a.addr.addr32[3] &
1591             super->src.addr.v.a.mask.addr32[3])) {
1592                 /* sub->src.addr is a subset of super->src.addr/mask */
1593                 memcpy(&sub->src.addr, &super->src.addr, sizeof(sub->src.addr));
1594         }
1595
1596         if (super->dst.addr.type == PF_ADDR_ADDRMASK && !super->dst.neg &&
1597             !sub->dst.neg && super->dst.addr.v.a.mask.addr32[0] == 0 &&
1598             super->dst.addr.v.a.mask.addr32[1] == 0 &&
1599             super->dst.addr.v.a.mask.addr32[2] == 0 &&
1600             super->dst.addr.v.a.mask.addr32[3] == 0)
1601                 memset(&sub->dst.addr, 0, sizeof(sub->dst.addr));
1602         else if (super->dst.addr.type == PF_ADDR_ADDRMASK &&
1603             sub->dst.addr.type == PF_ADDR_ADDRMASK &&
1604             super->dst.neg == sub->dst.neg &&
1605             super->af == sub->af &&
1606             unmask(&super->dst.addr.v.a.mask, super->af) <
1607             unmask(&sub->dst.addr.v.a.mask, sub->af) &&
1608             super->dst.addr.v.a.addr.addr32[0] ==
1609             (sub->dst.addr.v.a.addr.addr32[0] &
1610             super->dst.addr.v.a.mask.addr32[0]) &&
1611             super->dst.addr.v.a.addr.addr32[1] ==
1612             (sub->dst.addr.v.a.addr.addr32[1] &
1613             super->dst.addr.v.a.mask.addr32[1]) &&
1614             super->dst.addr.v.a.addr.addr32[2] ==
1615             (sub->dst.addr.v.a.addr.addr32[2] &
1616             super->dst.addr.v.a.mask.addr32[2]) &&
1617             super->dst.addr.v.a.addr.addr32[3] ==
1618             (sub->dst.addr.v.a.addr.addr32[3] &
1619             super->dst.addr.v.a.mask.addr32[3])) {
1620                 /* sub->dst.addr is a subset of super->dst.addr/mask */
1621                 memcpy(&sub->dst.addr, &super->dst.addr, sizeof(sub->dst.addr));
1622         }
1623
1624         if (super->af == 0)
1625                 sub->af = 0;
1626 }
1627
1628
1629 void
1630 superblock_free(struct pfctl *pf, struct superblock *block)
1631 {
1632         struct pf_opt_rule *por;
1633         while ((por = TAILQ_FIRST(&block->sb_rules))) {
1634                 TAILQ_REMOVE(&block->sb_rules, por, por_entry);
1635                 if (por->por_src_tbl) {
1636                         if (por->por_src_tbl->pt_buf) {
1637                                 pfr_buf_clear(por->por_src_tbl->pt_buf);
1638                                 free(por->por_src_tbl->pt_buf);
1639                         }
1640                         free(por->por_src_tbl);
1641                 }
1642                 if (por->por_dst_tbl) {
1643                         if (por->por_dst_tbl->pt_buf) {
1644                                 pfr_buf_clear(por->por_dst_tbl->pt_buf);
1645                                 free(por->por_dst_tbl->pt_buf);
1646                         }
1647                         free(por->por_dst_tbl);
1648                 }
1649                 free(por);
1650         }
1651         if (block->sb_profiled_block)
1652                 superblock_free(pf, block->sb_profiled_block);
1653         free(block);
1654 }
1655