1 /* $OpenBSD: pfctl_optimize.c,v 1.36 2016/08/03 16:27:25 krw 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/types.h> 20 #include <sys/ioctl.h> 21 #include <sys/socket.h> 22 23 #include <netinet/in.h> 24 #include <arpa/inet.h> 25 #include <net/if.h> 26 #include <net/pfvar.h> 27 28 #include <assert.h> 29 #include <ctype.h> 30 #include <err.h> 31 #include <errno.h> 32 #include <stddef.h> 33 #include <stdio.h> 34 #include <stdlib.h> 35 #include <string.h> 36 37 #include "pfctl_parser.h" 38 #include "pfctl.h" 39 40 /* The size at which a table becomes faster than individual rules */ 41 #define TABLE_THRESHOLD 6 42 43 44 /* #define OPT_DEBUG 1 */ 45 #ifdef OPT_DEBUG 46 # define DEBUG(str, v...) \ 47 printf("%s: " str "\n", __FUNCTION__ , ## v) 48 #else 49 # define DEBUG(str, v...) ((void)0) 50 #endif 51 52 53 /* 54 * A container that lets us sort a superblock to optimize the skip step jumps 55 */ 56 struct pf_skip_step { 57 int ps_count; /* number of items */ 58 TAILQ_HEAD( , pf_opt_rule) ps_rules; 59 TAILQ_ENTRY(pf_skip_step) ps_entry; 60 }; 61 62 63 /* 64 * A superblock is a block of adjacent rules of similar action. If there 65 * are five PASS rules in a row, they all become members of a superblock. 66 * Once we have a superblock, we are free to re-order any rules within it 67 * in order to improve performance; if a packet is passed, it doesn't matter 68 * who passed it. 69 */ 70 struct superblock { 71 TAILQ_HEAD( , pf_opt_rule) sb_rules; 72 TAILQ_ENTRY(superblock) sb_entry; 73 struct superblock *sb_profiled_block; 74 TAILQ_HEAD(skiplist, pf_skip_step) sb_skipsteps[PF_SKIP_COUNT]; 75 }; 76 TAILQ_HEAD(superblocks, superblock); 77 78 79 /* 80 * Description of the PF rule structure. 81 */ 82 enum { 83 BARRIER, /* the presence of the field puts the rule in it's own block */ 84 BREAK, /* the field may not differ between rules in a superblock */ 85 NOMERGE, /* the field may not differ between rules when combined */ 86 COMBINED, /* the field may itself be combined with other rules */ 87 DC, /* we just don't care about the field */ 88 NEVER}; /* we should never see this field set?!? */ 89 struct pf_rule_field { 90 const char *prf_name; 91 int prf_type; 92 size_t prf_offset; 93 size_t prf_size; 94 } pf_rule_desc[] = { 95 #define PF_RULE_FIELD(field, ty) \ 96 {#field, \ 97 ty, \ 98 offsetof(struct pf_rule, field), \ 99 sizeof(((struct pf_rule *)0)->field)} 100 101 102 /* 103 * The presence of these fields in a rule put the rule in it's own 104 * superblock. Thus it will not be optimized. It also prevents the 105 * rule from being re-ordered at all. 106 */ 107 PF_RULE_FIELD(label, BARRIER), 108 PF_RULE_FIELD(prob, BARRIER), 109 PF_RULE_FIELD(max_states, BARRIER), 110 PF_RULE_FIELD(max_src_nodes, BARRIER), 111 PF_RULE_FIELD(max_src_states, BARRIER), 112 PF_RULE_FIELD(max_src_conn, BARRIER), 113 PF_RULE_FIELD(max_src_conn_rate, BARRIER), 114 PF_RULE_FIELD(anchor, BARRIER), /* for now */ 115 116 /* 117 * These fields must be the same between all rules in the same superblock. 118 * These rules are allowed to be re-ordered but only among like rules. 119 * For instance we can re-order all 'tag "foo"' rules because they have the 120 * same tag. But we can not re-order between a 'tag "foo"' and a 121 * 'tag "bar"' since that would change the meaning of the ruleset. 122 */ 123 PF_RULE_FIELD(tagname, BREAK), 124 PF_RULE_FIELD(keep_state, BREAK), 125 PF_RULE_FIELD(qname, BREAK), 126 PF_RULE_FIELD(pqname, BREAK), 127 PF_RULE_FIELD(rt, BREAK), 128 PF_RULE_FIELD(allow_opts, BREAK), 129 PF_RULE_FIELD(rule_flag, BREAK), 130 PF_RULE_FIELD(action, BREAK), 131 PF_RULE_FIELD(log, BREAK), 132 PF_RULE_FIELD(quick, BREAK), 133 PF_RULE_FIELD(return_ttl, BREAK), 134 PF_RULE_FIELD(overload_tblname, BREAK), 135 PF_RULE_FIELD(flush, BREAK), 136 PF_RULE_FIELD(rdr, BREAK), 137 PF_RULE_FIELD(nat, BREAK), 138 PF_RULE_FIELD(logif, BREAK), 139 PF_RULE_FIELD(route, BREAK), 140 PF_RULE_FIELD(rtableid, BREAK), 141 142 /* 143 * Any fields not listed in this structure act as BREAK fields 144 */ 145 146 147 /* 148 * These fields must not differ when we merge two rules together but 149 * their difference isn't enough to put the rules in different superblocks. 150 * There are no problems re-ordering any rules with these fields. 151 */ 152 PF_RULE_FIELD(af, NOMERGE), 153 PF_RULE_FIELD(ifnot, NOMERGE), 154 PF_RULE_FIELD(ifname, NOMERGE), /* hack for IF groups */ 155 PF_RULE_FIELD(match_tag_not, NOMERGE), 156 PF_RULE_FIELD(match_tagname, NOMERGE), 157 PF_RULE_FIELD(os_fingerprint, NOMERGE), 158 PF_RULE_FIELD(timeout, NOMERGE), 159 PF_RULE_FIELD(return_icmp, NOMERGE), 160 PF_RULE_FIELD(return_icmp6, NOMERGE), 161 PF_RULE_FIELD(uid, NOMERGE), 162 PF_RULE_FIELD(gid, NOMERGE), 163 PF_RULE_FIELD(direction, NOMERGE), 164 PF_RULE_FIELD(proto, NOMERGE), 165 PF_RULE_FIELD(type, NOMERGE), 166 PF_RULE_FIELD(code, NOMERGE), 167 PF_RULE_FIELD(flags, NOMERGE), 168 PF_RULE_FIELD(flagset, NOMERGE), 169 PF_RULE_FIELD(tos, NOMERGE), 170 PF_RULE_FIELD(src.port, NOMERGE), 171 PF_RULE_FIELD(dst.port, NOMERGE), 172 PF_RULE_FIELD(src.port_op, NOMERGE), 173 PF_RULE_FIELD(dst.port_op, NOMERGE), 174 PF_RULE_FIELD(src.neg, NOMERGE), 175 PF_RULE_FIELD(dst.neg, NOMERGE), 176 PF_RULE_FIELD(onrdomain, NOMERGE), 177 PF_RULE_FIELD(naf, NOMERGE), 178 179 /* These fields can be merged */ 180 PF_RULE_FIELD(src.addr, COMBINED), 181 PF_RULE_FIELD(dst.addr, COMBINED), 182 183 /* We just don't care about these fields. They're set by the kernel */ 184 PF_RULE_FIELD(skip, DC), 185 PF_RULE_FIELD(evaluations, DC), 186 PF_RULE_FIELD(packets, DC), 187 PF_RULE_FIELD(bytes, DC), 188 PF_RULE_FIELD(kif, DC), 189 PF_RULE_FIELD(states_cur, DC), 190 PF_RULE_FIELD(states_tot, DC), 191 PF_RULE_FIELD(src_nodes, DC), 192 PF_RULE_FIELD(nr, DC), 193 PF_RULE_FIELD(entries, DC), 194 PF_RULE_FIELD(qid, DC), 195 PF_RULE_FIELD(pqid, DC), 196 PF_RULE_FIELD(anchor_relative, DC), 197 PF_RULE_FIELD(anchor_wildcard, DC), 198 PF_RULE_FIELD(tag, DC), 199 PF_RULE_FIELD(match_tag, DC), 200 PF_RULE_FIELD(overload_tbl, DC), 201 202 /* These fields should never be set in a PASS/BLOCK rule XXX fix*/ 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 addrs_combineable(struct pf_rule_addr *, struct pf_rule_addr *); 211 int addrs_equal(struct pf_rule_addr *, struct pf_rule_addr *); 212 int block_feedback(struct pfctl *, struct superblock *); 213 int combine_rules(struct pfctl *, struct superblock *); 214 void comparable_rule(struct pf_rule *, const struct pf_rule *, int); 215 int construct_superblocks(struct pfctl *, struct pf_opt_queue *, 216 struct superblocks *); 217 void exclude_supersets(struct pf_rule *, struct pf_rule *); 218 int interface_group(const char *); 219 int load_feedback_profile(struct pfctl *, struct superblocks *); 220 int optimize_superblock(struct pfctl *, struct superblock *); 221 void remove_from_skipsteps(struct skiplist *, struct superblock *, 222 struct pf_opt_rule *, struct pf_skip_step *); 223 int remove_identical_rules(struct pfctl *, struct superblock *); 224 int reorder_rules(struct pfctl *, struct superblock *, int); 225 int rules_combineable(struct pf_rule *, struct pf_rule *); 226 void skip_append(struct superblock *, int, struct pf_skip_step *, 227 struct pf_opt_rule *); 228 int skip_compare(int, struct pf_skip_step *, struct pf_opt_rule *); 229 void skip_init(void); 230 int skip_cmp_af(struct pf_rule *, struct pf_rule *); 231 int skip_cmp_dir(struct pf_rule *, struct pf_rule *); 232 int skip_cmp_rdom(struct pf_rule *, struct pf_rule *); 233 int skip_cmp_dst_addr(struct pf_rule *, struct pf_rule *); 234 int skip_cmp_dst_port(struct pf_rule *, struct pf_rule *); 235 int skip_cmp_ifp(struct pf_rule *, struct pf_rule *); 236 int skip_cmp_proto(struct pf_rule *, struct pf_rule *); 237 int skip_cmp_src_addr(struct pf_rule *, struct pf_rule *); 238 int skip_cmp_src_port(struct pf_rule *, struct pf_rule *); 239 int superblock_inclusive(struct superblock *, struct pf_opt_rule *); 240 void superblock_free(struct pfctl *, struct superblock *); 241 242 243 int (*skip_comparitors[PF_SKIP_COUNT])(struct pf_rule *, struct pf_rule *); 244 const char *skip_comparitors_names[PF_SKIP_COUNT]; 245 #define PF_SKIP_COMPARITORS { \ 246 { "ifp", PF_SKIP_IFP, skip_cmp_ifp }, \ 247 { "dir", PF_SKIP_DIR, skip_cmp_dir }, \ 248 { "rdomain", PF_SKIP_RDOM, skip_cmp_rdom }, \ 249 { "af", PF_SKIP_AF, skip_cmp_af }, \ 250 { "proto", PF_SKIP_PROTO, skip_cmp_proto }, \ 251 { "saddr", PF_SKIP_SRC_ADDR, skip_cmp_src_addr }, \ 252 { "daddr", PF_SKIP_DST_ADDR, skip_cmp_dst_addr }, \ 253 { "sport", PF_SKIP_SRC_PORT, skip_cmp_src_port }, \ 254 { "dport", PF_SKIP_DST_PORT, skip_cmp_dst_port } \ 255 } 256 257 struct pfr_buffer table_buffer; 258 int table_identifier; 259 260 261 int 262 pfctl_optimize_ruleset(struct pfctl *pf, struct pf_ruleset *rs) 263 { 264 struct superblocks superblocks; 265 struct pf_opt_queue opt_queue; 266 struct superblock *block; 267 struct pf_opt_rule *por; 268 struct pf_rule *r; 269 struct pf_rulequeue *old_rules; 270 271 DEBUG("optimizing ruleset"); 272 memset(&table_buffer, 0, sizeof(table_buffer)); 273 skip_init(); 274 TAILQ_INIT(&opt_queue); 275 276 old_rules = rs->rules.active.ptr; 277 rs->rules.active.ptr = rs->rules.inactive.ptr; 278 rs->rules.inactive.ptr = old_rules; 279 280 /* 281 * XXX expanding the pf_opt_rule format throughout pfctl might allow 282 * us to avoid all this copying. 283 */ 284 while ((r = TAILQ_FIRST(rs->rules.inactive.ptr)) != NULL) { 285 TAILQ_REMOVE(rs->rules.inactive.ptr, r, entries); 286 if ((por = calloc(1, sizeof(*por))) == NULL) 287 err(1, "calloc"); 288 memcpy(&por->por_rule, r, sizeof(*r)); 289 290 TAILQ_INSERT_TAIL(&opt_queue, por, por_entry); 291 } 292 293 TAILQ_INIT(&superblocks); 294 if (construct_superblocks(pf, &opt_queue, &superblocks)) 295 goto error; 296 297 if (pf->optimize & PF_OPTIMIZE_PROFILE) { 298 if (load_feedback_profile(pf, &superblocks)) 299 goto error; 300 } 301 302 TAILQ_FOREACH(block, &superblocks, sb_entry) { 303 if (optimize_superblock(pf, block)) 304 goto error; 305 } 306 307 rs->anchor->refcnt = 0; 308 while ((block = TAILQ_FIRST(&superblocks))) { 309 TAILQ_REMOVE(&superblocks, block, sb_entry); 310 311 while ((por = TAILQ_FIRST(&block->sb_rules))) { 312 TAILQ_REMOVE(&block->sb_rules, por, por_entry); 313 por->por_rule.nr = rs->anchor->refcnt++; 314 if ((r = calloc(1, sizeof(*r))) == NULL) 315 err(1, "calloc"); 316 memcpy(r, &por->por_rule, sizeof(*r)); 317 TAILQ_INSERT_TAIL(rs->rules.active.ptr, r, entries); 318 free(por); 319 } 320 free(block); 321 } 322 323 return (0); 324 325 error: 326 while ((por = TAILQ_FIRST(&opt_queue))) { 327 TAILQ_REMOVE(&opt_queue, por, por_entry); 328 if (por->por_src_tbl) { 329 pfr_buf_clear(por->por_src_tbl->pt_buf); 330 free(por->por_src_tbl->pt_buf); 331 free(por->por_src_tbl); 332 } 333 if (por->por_dst_tbl) { 334 pfr_buf_clear(por->por_dst_tbl->pt_buf); 335 free(por->por_dst_tbl->pt_buf); 336 free(por->por_dst_tbl); 337 } 338 free(por); 339 } 340 while ((block = TAILQ_FIRST(&superblocks))) { 341 TAILQ_REMOVE(&superblocks, block, sb_entry); 342 superblock_free(pf, block); 343 } 344 return (1); 345 } 346 347 348 /* 349 * Go ahead and optimize a superblock 350 */ 351 int 352 optimize_superblock(struct pfctl *pf, struct superblock *block) 353 { 354 #ifdef OPT_DEBUG 355 struct pf_opt_rule *por; 356 #endif /* OPT_DEBUG */ 357 358 /* We have a few optimization passes: 359 * 1) remove duplicate rules or rules that are a subset of other 360 * rules 361 * 2) combine otherwise identical rules with different IP addresses 362 * into a single rule and put the addresses in a table. 363 * 3) re-order the rules to improve kernel skip steps 364 * 4) re-order the 'quick' rules based on feedback from the 365 * active ruleset statistics 366 * 367 * XXX combine_rules() doesn't combine v4 and v6 rules. would just 368 * have to keep af in the table container, make af 'COMBINE' and 369 * twiddle the af on the merged rule 370 * XXX maybe add a weighting to the metric on skipsteps when doing 371 * reordering. sometimes two sequential tables will be better 372 * that four consecutive interfaces. 373 * XXX need to adjust the skipstep count of everything after PROTO, 374 * since they aren't actually checked on a proto mismatch in 375 * pf_test_{tcp, udp, icmp}() 376 * XXX should i treat proto=0, af=0 or dir=0 special in skepstep 377 * calculation since they are a DC? 378 * XXX keep last skiplist of last superblock to influence this 379 * superblock. '5 inet6 log' should make '3 inet6' come before '4 380 * inet' in the next superblock. 381 * XXX would be useful to add tables for ports 382 * XXX we can also re-order some mutually exclusive superblocks to 383 * try merging superblocks before any of these optimization passes. 384 * for instance a single 'log in' rule in the middle of non-logging 385 * out rules. 386 */ 387 388 /* shortcut. there will be a lot of 1-rule superblocks */ 389 if (!TAILQ_NEXT(TAILQ_FIRST(&block->sb_rules), por_entry)) 390 return (0); 391 392 #ifdef OPT_DEBUG 393 printf("--- Superblock ---\n"); 394 TAILQ_FOREACH(por, &block->sb_rules, por_entry) { 395 printf(" "); 396 print_rule(&por->por_rule, por->por_rule.anchor ? 397 por->por_rule.anchor->name : "", PF_OPT_DEBUG); 398 } 399 #endif /* OPT_DEBUG */ 400 401 402 if (remove_identical_rules(pf, block)) 403 return (1); 404 if (combine_rules(pf, block)) 405 return (1); 406 if ((pf->optimize & PF_OPTIMIZE_PROFILE) && 407 TAILQ_FIRST(&block->sb_rules)->por_rule.quick && 408 block->sb_profiled_block) { 409 if (block_feedback(pf, block)) 410 return (1); 411 } else if (reorder_rules(pf, block, 0)) { 412 return (1); 413 } 414 415 /* 416 * Don't add any optimization passes below reorder_rules(). It will 417 * have divided superblocks into smaller blocks for further refinement 418 * and doesn't put them back together again. What once was a true 419 * superblock might have been split into multiple superblocks. 420 */ 421 422 #ifdef OPT_DEBUG 423 printf("--- END Superblock ---\n"); 424 #endif /* OPT_DEBUG */ 425 return (0); 426 } 427 428 429 /* 430 * Optimization pass #1: remove identical rules 431 */ 432 int 433 remove_identical_rules(struct pfctl *pf, struct superblock *block) 434 { 435 struct pf_opt_rule *por1, *por2, *por_next, *por2_next; 436 struct pf_rule a, a2, b, b2; 437 438 for (por1 = TAILQ_FIRST(&block->sb_rules); por1; por1 = por_next) { 439 por_next = TAILQ_NEXT(por1, por_entry); 440 for (por2 = por_next; por2; por2 = por2_next) { 441 por2_next = TAILQ_NEXT(por2, por_entry); 442 comparable_rule(&a, &por1->por_rule, DC); 443 comparable_rule(&b, &por2->por_rule, DC); 444 memcpy(&a2, &a, sizeof(a2)); 445 memcpy(&b2, &b, sizeof(b2)); 446 447 exclude_supersets(&a, &b); 448 exclude_supersets(&b2, &a2); 449 if (memcmp(&a, &b, sizeof(a)) == 0) { 450 DEBUG("removing identical rule nr%d = *nr%d*", 451 por1->por_rule.nr, por2->por_rule.nr); 452 TAILQ_REMOVE(&block->sb_rules, por2, por_entry); 453 if (por_next == por2) 454 por_next = TAILQ_NEXT(por1, por_entry); 455 free(por2); 456 } else if (memcmp(&a2, &b2, sizeof(a2)) == 0) { 457 DEBUG("removing identical rule *nr%d* = nr%d", 458 por1->por_rule.nr, por2->por_rule.nr); 459 TAILQ_REMOVE(&block->sb_rules, por1, por_entry); 460 free(por1); 461 break; 462 } 463 } 464 } 465 466 return (0); 467 } 468 469 470 /* 471 * Optimization pass #2: combine similar rules with different addresses 472 * into a single rule and a table 473 */ 474 int 475 combine_rules(struct pfctl *pf, struct superblock *block) 476 { 477 struct pf_opt_rule *p1, *p2, *por_next; 478 int src_eq, dst_eq; 479 480 /* First we make a pass to combine the rules. O(n log n) */ 481 TAILQ_FOREACH(p1, &block->sb_rules, por_entry) { 482 for (p2 = TAILQ_NEXT(p1, por_entry); p2; p2 = por_next) { 483 por_next = TAILQ_NEXT(p2, por_entry); 484 485 src_eq = addrs_equal(&p1->por_rule.src, 486 &p2->por_rule.src); 487 dst_eq = addrs_equal(&p1->por_rule.dst, 488 &p2->por_rule.dst); 489 490 if (src_eq && !dst_eq && p1->por_src_tbl == NULL && 491 p2->por_dst_tbl == NULL && 492 p2->por_src_tbl == NULL && 493 rules_combineable(&p1->por_rule, &p2->por_rule) && 494 addrs_combineable(&p1->por_rule.dst, 495 &p2->por_rule.dst)) { 496 DEBUG("can combine rules nr%d = nr%d", 497 p1->por_rule.nr, p2->por_rule.nr); 498 if (p1->por_dst_tbl == NULL && 499 add_opt_table(pf, &p1->por_dst_tbl, 500 p1->por_rule.af, &p1->por_rule.dst, NULL)) 501 return (1); 502 if (add_opt_table(pf, &p1->por_dst_tbl, 503 p1->por_rule.af, &p2->por_rule.dst, NULL)) 504 return (1); 505 p2->por_dst_tbl = p1->por_dst_tbl; 506 if (p1->por_dst_tbl->pt_rulecount >= 507 TABLE_THRESHOLD) { 508 TAILQ_REMOVE(&block->sb_rules, p2, 509 por_entry); 510 free(p2); 511 } 512 } else if (!src_eq && dst_eq && p1->por_dst_tbl == NULL 513 && p2->por_src_tbl == NULL && 514 p2->por_dst_tbl == NULL && 515 rules_combineable(&p1->por_rule, &p2->por_rule) && 516 addrs_combineable(&p1->por_rule.src, 517 &p2->por_rule.src)) { 518 DEBUG("can combine rules nr%d = nr%d", 519 p1->por_rule.nr, p2->por_rule.nr); 520 if (p1->por_src_tbl == NULL && 521 add_opt_table(pf, &p1->por_src_tbl, 522 p1->por_rule.af, &p1->por_rule.src, NULL)) 523 return (1); 524 if (add_opt_table(pf, &p1->por_src_tbl, 525 p1->por_rule.af, &p2->por_rule.src, NULL)) 526 return (1); 527 p2->por_src_tbl = p1->por_src_tbl; 528 if (p1->por_src_tbl->pt_rulecount >= 529 TABLE_THRESHOLD) { 530 TAILQ_REMOVE(&block->sb_rules, p2, 531 por_entry); 532 free(p2); 533 } 534 } 535 } 536 } 537 538 539 /* 540 * Then we make a final pass to create a valid table name and 541 * insert the name into the rules. 542 * Convert translation/routing mapping pools to tables as well. 543 */ 544 for (p1 = TAILQ_FIRST(&block->sb_rules); p1; p1 = por_next) { 545 por_next = TAILQ_NEXT(p1, por_entry); 546 assert(p1->por_src_tbl == NULL || p1->por_dst_tbl == NULL); 547 548 if (p1->por_src_tbl && p1->por_src_tbl->pt_rulecount >= 549 TABLE_THRESHOLD) { 550 if (p1->por_src_tbl->pt_generated) { 551 /* This rule is included in a table */ 552 TAILQ_REMOVE(&block->sb_rules, p1, por_entry); 553 free(p1); 554 continue; 555 } 556 p1->por_src_tbl->pt_generated = 1; 557 558 if ((pf->opts & PF_OPT_NOACTION) == 0 && 559 pf_opt_create_table(pf, p1->por_src_tbl)) 560 return (1); 561 562 pf->tdirty = 1; 563 564 if (pf->opts & PF_OPT_VERBOSE) 565 print_tabledef(p1->por_src_tbl->pt_name, 566 PFR_TFLAG_CONST, 1, 567 &p1->por_src_tbl->pt_nodes); 568 569 memset(&p1->por_rule.src.addr, 0, 570 sizeof(p1->por_rule.src.addr)); 571 p1->por_rule.src.addr.type = PF_ADDR_TABLE; 572 strlcpy(p1->por_rule.src.addr.v.tblname, 573 p1->por_src_tbl->pt_name, 574 sizeof(p1->por_rule.src.addr.v.tblname)); 575 576 pfr_buf_clear(p1->por_src_tbl->pt_buf); 577 free(p1->por_src_tbl->pt_buf); 578 p1->por_src_tbl->pt_buf = NULL; 579 } 580 if (p1->por_dst_tbl && p1->por_dst_tbl->pt_rulecount >= 581 TABLE_THRESHOLD) { 582 if (p1->por_dst_tbl->pt_generated) { 583 /* This rule is included in a table */ 584 TAILQ_REMOVE(&block->sb_rules, p1, por_entry); 585 free(p1); 586 continue; 587 } 588 p1->por_dst_tbl->pt_generated = 1; 589 590 if ((pf->opts & PF_OPT_NOACTION) == 0 && 591 pf_opt_create_table(pf, p1->por_dst_tbl)) 592 return (1); 593 pf->tdirty = 1; 594 595 if (pf->opts & PF_OPT_VERBOSE) 596 print_tabledef(p1->por_dst_tbl->pt_name, 597 PFR_TFLAG_CONST, 1, 598 &p1->por_dst_tbl->pt_nodes); 599 600 memset(&p1->por_rule.dst.addr, 0, 601 sizeof(p1->por_rule.dst.addr)); 602 p1->por_rule.dst.addr.type = PF_ADDR_TABLE; 603 strlcpy(p1->por_rule.dst.addr.v.tblname, 604 p1->por_dst_tbl->pt_name, 605 sizeof(p1->por_rule.dst.addr.v.tblname)); 606 607 pfr_buf_clear(p1->por_dst_tbl->pt_buf); 608 free(p1->por_dst_tbl->pt_buf); 609 p1->por_dst_tbl->pt_buf = NULL; 610 } 611 } 612 613 return (0); 614 } 615 616 617 /* 618 * Optimization pass #3: re-order rules to improve skip steps 619 */ 620 int 621 reorder_rules(struct pfctl *pf, struct superblock *block, int depth) 622 { 623 struct superblock *newblock; 624 struct pf_skip_step *skiplist; 625 struct pf_opt_rule *por; 626 int i, largest, largest_list, rule_count = 0; 627 TAILQ_HEAD( , pf_opt_rule) head; 628 629 /* 630 * Calculate the best-case skip steps. We put each rule in a list 631 * of other rules with common fields 632 */ 633 for (i = 0; i < PF_SKIP_COUNT; i++) { 634 TAILQ_FOREACH(por, &block->sb_rules, por_entry) { 635 TAILQ_FOREACH(skiplist, &block->sb_skipsteps[i], 636 ps_entry) { 637 if (skip_compare(i, skiplist, por) == 0) 638 break; 639 } 640 if (skiplist == NULL) { 641 if ((skiplist = calloc(1, sizeof(*skiplist))) == 642 NULL) 643 err(1, "calloc"); 644 TAILQ_INIT(&skiplist->ps_rules); 645 TAILQ_INSERT_TAIL(&block->sb_skipsteps[i], 646 skiplist, ps_entry); 647 } 648 skip_append(block, i, skiplist, por); 649 } 650 } 651 652 TAILQ_FOREACH(por, &block->sb_rules, por_entry) 653 rule_count++; 654 655 /* 656 * Now we're going to ignore any fields that are identical between 657 * all of the rules in the superblock and those fields which differ 658 * between every rule in the superblock. 659 */ 660 largest = 0; 661 for (i = 0; i < PF_SKIP_COUNT; i++) { 662 skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]); 663 if (skiplist->ps_count == rule_count) { 664 DEBUG("(%d) original skipstep '%s' is all rules", 665 depth, skip_comparitors_names[i]); 666 skiplist->ps_count = 0; 667 } else if (skiplist->ps_count == 1) { 668 skiplist->ps_count = 0; 669 } else { 670 DEBUG("(%d) original skipstep '%s' largest jump is %d", 671 depth, skip_comparitors_names[i], 672 skiplist->ps_count); 673 if (skiplist->ps_count > largest) 674 largest = skiplist->ps_count; 675 } 676 } 677 if (largest == 0) { 678 /* Ugh. There is NO commonality in the superblock on which 679 * optimize the skipsteps optimization. 680 */ 681 goto done; 682 } 683 684 /* 685 * Now we're going to empty the superblock rule list and re-create 686 * it based on a more optimal skipstep order. 687 */ 688 TAILQ_INIT(&head); 689 while ((por = TAILQ_FIRST(&block->sb_rules))) { 690 TAILQ_REMOVE(&block->sb_rules, por, por_entry); 691 TAILQ_INSERT_TAIL(&head, por, por_entry); 692 } 693 694 695 while (!TAILQ_EMPTY(&head)) { 696 largest = 1; 697 698 /* 699 * Find the most useful skip steps remaining 700 */ 701 for (i = 0; i < PF_SKIP_COUNT; i++) { 702 skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]); 703 if (skiplist->ps_count > largest) { 704 largest = skiplist->ps_count; 705 largest_list = i; 706 } 707 } 708 709 if (largest <= 1) { 710 /* 711 * Nothing useful left. Leave remaining rules in order. 712 */ 713 DEBUG("(%d) no more commonality for skip steps", depth); 714 while ((por = TAILQ_FIRST(&head))) { 715 TAILQ_REMOVE(&head, por, por_entry); 716 TAILQ_INSERT_TAIL(&block->sb_rules, por, 717 por_entry); 718 } 719 } else { 720 /* 721 * There is commonality. Extract those common rules 722 * and place them in the ruleset adjacent to each 723 * other. 724 */ 725 skiplist = TAILQ_FIRST(&block->sb_skipsteps[ 726 largest_list]); 727 DEBUG("(%d) skipstep '%s' largest jump is %d @ #%d", 728 depth, skip_comparitors_names[largest_list], 729 largest, TAILQ_FIRST(&TAILQ_FIRST(&block-> 730 sb_skipsteps [largest_list])->ps_rules)-> 731 por_rule.nr); 732 TAILQ_REMOVE(&block->sb_skipsteps[largest_list], 733 skiplist, ps_entry); 734 735 736 /* 737 * There may be further commonality inside these 738 * rules. So we'll split them off into they're own 739 * superblock and pass it back into the optimizer. 740 */ 741 if (skiplist->ps_count > 2) { 742 if ((newblock = calloc(1, sizeof(*newblock))) 743 == NULL) { 744 warn("calloc"); 745 return (1); 746 } 747 TAILQ_INIT(&newblock->sb_rules); 748 for (i = 0; i < PF_SKIP_COUNT; i++) 749 TAILQ_INIT(&newblock->sb_skipsteps[i]); 750 TAILQ_INSERT_BEFORE(block, newblock, sb_entry); 751 DEBUG("(%d) splitting off %d rules from superblock @ #%d", 752 depth, skiplist->ps_count, 753 TAILQ_FIRST(&skiplist->ps_rules)-> 754 por_rule.nr); 755 } else { 756 newblock = block; 757 } 758 759 while ((por = TAILQ_FIRST(&skiplist->ps_rules))) { 760 TAILQ_REMOVE(&head, por, por_entry); 761 TAILQ_REMOVE(&skiplist->ps_rules, por, 762 por_skip_entry[largest_list]); 763 TAILQ_INSERT_TAIL(&newblock->sb_rules, por, 764 por_entry); 765 766 /* Remove this rule from all other skiplists */ 767 remove_from_skipsteps(&block->sb_skipsteps[ 768 largest_list], block, por, skiplist); 769 } 770 free(skiplist); 771 if (newblock != block) 772 if (reorder_rules(pf, newblock, depth + 1)) 773 return (1); 774 } 775 } 776 777 done: 778 for (i = 0; i < PF_SKIP_COUNT; i++) { 779 while ((skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]))) { 780 TAILQ_REMOVE(&block->sb_skipsteps[i], skiplist, 781 ps_entry); 782 free(skiplist); 783 } 784 } 785 786 return (0); 787 } 788 789 790 /* 791 * Optimization pass #4: re-order 'quick' rules based on feedback from the 792 * currently running ruleset 793 */ 794 int 795 block_feedback(struct pfctl *pf, struct superblock *block) 796 { 797 TAILQ_HEAD( , pf_opt_rule) queue; 798 struct pf_opt_rule *por1, *por2; 799 u_int64_t total_count = 0; 800 struct pf_rule a, b; 801 802 803 /* 804 * Walk through all of the profiled superblock's rules and copy 805 * the counters onto our rules. 806 */ 807 TAILQ_FOREACH(por1, &block->sb_profiled_block->sb_rules, por_entry) { 808 comparable_rule(&a, &por1->por_rule, DC); 809 total_count += por1->por_rule.packets[0] + 810 por1->por_rule.packets[1]; 811 TAILQ_FOREACH(por2, &block->sb_rules, por_entry) { 812 if (por2->por_profile_count) 813 continue; 814 comparable_rule(&b, &por2->por_rule, DC); 815 if (memcmp(&a, &b, sizeof(a)) == 0) { 816 por2->por_profile_count = 817 por1->por_rule.packets[0] + 818 por1->por_rule.packets[1]; 819 break; 820 } 821 } 822 } 823 superblock_free(pf, block->sb_profiled_block); 824 block->sb_profiled_block = NULL; 825 826 /* 827 * Now we pull all of the rules off the superblock and re-insert them 828 * in sorted order. 829 */ 830 831 TAILQ_INIT(&queue); 832 while ((por1 = TAILQ_FIRST(&block->sb_rules)) != NULL) { 833 TAILQ_REMOVE(&block->sb_rules, por1, por_entry); 834 TAILQ_INSERT_TAIL(&queue, por1, por_entry); 835 } 836 837 while ((por1 = TAILQ_FIRST(&queue)) != NULL) { 838 TAILQ_REMOVE(&queue, por1, por_entry); 839 /* XXX I should sort all of the unused rules based on skip steps */ 840 TAILQ_FOREACH(por2, &block->sb_rules, por_entry) { 841 if (por1->por_profile_count > por2->por_profile_count) { 842 TAILQ_INSERT_BEFORE(por2, por1, por_entry); 843 break; 844 } 845 } 846 if (por2 == NULL) 847 TAILQ_INSERT_TAIL(&block->sb_rules, por1, por_entry); 848 } 849 850 return (0); 851 } 852 853 854 /* 855 * Load the current ruleset from the kernel and try to associate them with 856 * the ruleset we're optimizing. 857 */ 858 int 859 load_feedback_profile(struct pfctl *pf, struct superblocks *superblocks) 860 { 861 struct superblock *block, *blockcur; 862 struct superblocks prof_superblocks; 863 struct pf_opt_rule *por; 864 struct pf_opt_queue queue; 865 struct pfioc_rule pr; 866 struct pf_rule a, b; 867 int nr, mnr; 868 869 TAILQ_INIT(&queue); 870 TAILQ_INIT(&prof_superblocks); 871 872 memset(&pr, 0, sizeof(pr)); 873 pr.rule.action = PF_PASS; 874 if (ioctl(pf->dev, DIOCGETRULES, &pr)) { 875 warn("DIOCGETRULES"); 876 return (1); 877 } 878 mnr = pr.nr; 879 880 DEBUG("Loading %d active rules for a feedback profile", mnr); 881 for (nr = 0; nr < mnr; ++nr) { 882 struct pf_ruleset *rs; 883 if ((por = calloc(1, sizeof(*por))) == NULL) { 884 warn("calloc"); 885 return (1); 886 } 887 pr.nr = nr; 888 if (ioctl(pf->dev, DIOCGETRULE, &pr)) { 889 warn("DIOCGETRULES"); 890 free(por); 891 return (1); 892 } 893 memcpy(&por->por_rule, &pr.rule, sizeof(por->por_rule)); 894 rs = pf_find_or_create_ruleset(pr.anchor_call); 895 por->por_rule.anchor = rs->anchor; 896 TAILQ_INSERT_TAIL(&queue, por, por_entry); 897 898 /* XXX pfctl_get_pool(pf->dev, &pr.rule.rpool, nr, pr.ticket, 899 * PF_PASS, pf->anchor) ??? 900 * ... pfctl_clear_pool(&pr.rule.rpool) 901 */ 902 } 903 904 if (construct_superblocks(pf, &queue, &prof_superblocks)) 905 return (1); 906 907 908 /* 909 * Now we try to associate the active ruleset's superblocks with 910 * the superblocks we're compiling. 911 */ 912 block = TAILQ_FIRST(superblocks); 913 blockcur = TAILQ_FIRST(&prof_superblocks); 914 while (block && blockcur) { 915 comparable_rule(&a, &TAILQ_FIRST(&block->sb_rules)->por_rule, 916 BREAK); 917 comparable_rule(&b, &TAILQ_FIRST(&blockcur->sb_rules)->por_rule, 918 BREAK); 919 if (memcmp(&a, &b, sizeof(a)) == 0) { 920 /* The two superblocks lined up */ 921 block->sb_profiled_block = blockcur; 922 } else { 923 DEBUG("superblocks don't line up between #%d and #%d", 924 TAILQ_FIRST(&block->sb_rules)->por_rule.nr, 925 TAILQ_FIRST(&blockcur->sb_rules)->por_rule.nr); 926 break; 927 } 928 block = TAILQ_NEXT(block, sb_entry); 929 blockcur = TAILQ_NEXT(blockcur, sb_entry); 930 } 931 932 933 934 /* Free any superblocks we couldn't link */ 935 while (blockcur) { 936 block = TAILQ_NEXT(blockcur, sb_entry); 937 superblock_free(pf, blockcur); 938 blockcur = block; 939 } 940 return (0); 941 } 942 943 944 /* 945 * Compare a rule to a skiplist to see if the rule is a member 946 */ 947 int 948 skip_compare(int skipnum, struct pf_skip_step *skiplist, 949 struct pf_opt_rule *por) 950 { 951 struct pf_rule *a, *b; 952 if (skipnum >= PF_SKIP_COUNT || skipnum < 0) 953 errx(1, "skip_compare() out of bounds"); 954 a = &por->por_rule; 955 b = &TAILQ_FIRST(&skiplist->ps_rules)->por_rule; 956 957 return ((skip_comparitors[skipnum])(a, b)); 958 } 959 960 961 /* 962 * Add a rule to a skiplist 963 */ 964 void 965 skip_append(struct superblock *superblock, int skipnum, 966 struct pf_skip_step *skiplist, struct pf_opt_rule *por) 967 { 968 struct pf_skip_step *prev; 969 970 skiplist->ps_count++; 971 TAILQ_INSERT_TAIL(&skiplist->ps_rules, por, por_skip_entry[skipnum]); 972 973 /* Keep the list of skiplists sorted by whichever is larger */ 974 while ((prev = TAILQ_PREV(skiplist, skiplist, ps_entry)) && 975 prev->ps_count < skiplist->ps_count) { 976 TAILQ_REMOVE(&superblock->sb_skipsteps[skipnum], 977 skiplist, ps_entry); 978 TAILQ_INSERT_BEFORE(prev, skiplist, ps_entry); 979 } 980 } 981 982 983 /* 984 * Remove a rule from the other skiplist calculations. 985 */ 986 void 987 remove_from_skipsteps(struct skiplist *head, struct superblock *block, 988 struct pf_opt_rule *por, struct pf_skip_step *active_list) 989 { 990 struct pf_skip_step *sk, *next; 991 struct pf_opt_rule *p2; 992 int i, found; 993 994 for (i = 0; i < PF_SKIP_COUNT; i++) { 995 sk = TAILQ_FIRST(&block->sb_skipsteps[i]); 996 if (sk == NULL || sk == active_list || sk->ps_count <= 1) 997 continue; 998 found = 0; 999 do { 1000 TAILQ_FOREACH(p2, &sk->ps_rules, por_skip_entry[i]) 1001 if (p2 == por) { 1002 TAILQ_REMOVE(&sk->ps_rules, p2, 1003 por_skip_entry[i]); 1004 found = 1; 1005 sk->ps_count--; 1006 break; 1007 } 1008 } while (!found && (sk = TAILQ_NEXT(sk, ps_entry))); 1009 if (found && sk) { 1010 /* Does this change the sorting order? */ 1011 while ((next = TAILQ_NEXT(sk, ps_entry)) && 1012 next->ps_count > sk->ps_count) { 1013 TAILQ_REMOVE(head, sk, ps_entry); 1014 TAILQ_INSERT_AFTER(head, next, sk, ps_entry); 1015 } 1016 #ifdef OPT_DEBUG 1017 next = TAILQ_NEXT(sk, ps_entry); 1018 assert(next == NULL || next->ps_count <= sk->ps_count); 1019 #endif /* OPT_DEBUG */ 1020 } 1021 } 1022 } 1023 1024 1025 /* Compare two rules AF field for skiplist construction */ 1026 int 1027 skip_cmp_af(struct pf_rule *a, struct pf_rule *b) 1028 { 1029 if (a->af != b->af || a->af == 0) 1030 return (1); 1031 return (0); 1032 } 1033 1034 /* Compare two rules DIRECTION field for skiplist construction */ 1035 int 1036 skip_cmp_dir(struct pf_rule *a, struct pf_rule *b) 1037 { 1038 if (a->direction == 0 || a->direction != b->direction) 1039 return (1); 1040 return (0); 1041 } 1042 1043 /* Compare two rules ON RDOMAIN field for skiplist construction */ 1044 int 1045 skip_cmp_rdom(struct pf_rule *a, struct pf_rule *b) 1046 { 1047 if (a->onrdomain == -1 || a->onrdomain != b->onrdomain) 1048 return (1); 1049 return (a->ifnot != b->ifnot); 1050 } 1051 1052 /* Compare two rules DST Address field for skiplist construction */ 1053 int 1054 skip_cmp_dst_addr(struct pf_rule *a, struct pf_rule *b) 1055 { 1056 if (a->dst.neg != b->dst.neg || 1057 a->dst.addr.type != b->dst.addr.type) 1058 return (1); 1059 /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0 1060 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP || 1061 * a->proto == IPPROTO_ICMP 1062 * return (1); 1063 */ 1064 switch (a->dst.addr.type) { 1065 case PF_ADDR_ADDRMASK: 1066 if (memcmp(&a->dst.addr.v.a.addr, &b->dst.addr.v.a.addr, 1067 sizeof(a->dst.addr.v.a.addr)) || 1068 memcmp(&a->dst.addr.v.a.mask, &b->dst.addr.v.a.mask, 1069 sizeof(a->dst.addr.v.a.mask)) || 1070 (a->dst.addr.v.a.addr.addr32[0] == 0 && 1071 a->dst.addr.v.a.addr.addr32[1] == 0 && 1072 a->dst.addr.v.a.addr.addr32[2] == 0 && 1073 a->dst.addr.v.a.addr.addr32[3] == 0)) 1074 return (1); 1075 return (0); 1076 case PF_ADDR_DYNIFTL: 1077 if (strcmp(a->dst.addr.v.ifname, b->dst.addr.v.ifname) != 0 || 1078 a->dst.addr.iflags != b->dst.addr.iflags || 1079 memcmp(&a->dst.addr.v.a.mask, &b->dst.addr.v.a.mask, 1080 sizeof(a->dst.addr.v.a.mask))) 1081 return (1); 1082 return (0); 1083 case PF_ADDR_NOROUTE: 1084 case PF_ADDR_URPFFAILED: 1085 return (0); 1086 case PF_ADDR_TABLE: 1087 return (strcmp(a->dst.addr.v.tblname, b->dst.addr.v.tblname)); 1088 } 1089 return (1); 1090 } 1091 1092 /* Compare two rules DST port field for skiplist construction */ 1093 int 1094 skip_cmp_dst_port(struct pf_rule *a, struct pf_rule *b) 1095 { 1096 /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0 1097 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP || 1098 * a->proto == IPPROTO_ICMP 1099 * return (1); 1100 */ 1101 if (a->dst.port_op == PF_OP_NONE || a->dst.port_op != b->dst.port_op || 1102 a->dst.port[0] != b->dst.port[0] || 1103 a->dst.port[1] != b->dst.port[1]) 1104 return (1); 1105 return (0); 1106 } 1107 1108 /* Compare two rules IFP field for skiplist construction */ 1109 int 1110 skip_cmp_ifp(struct pf_rule *a, struct pf_rule *b) 1111 { 1112 if (strcmp(a->ifname, b->ifname) || a->ifname[0] == '\0') 1113 return (1); 1114 return (a->ifnot != b->ifnot); 1115 } 1116 1117 /* Compare two rules PROTO field for skiplist construction */ 1118 int 1119 skip_cmp_proto(struct pf_rule *a, struct pf_rule *b) 1120 { 1121 return (a->proto != b->proto || a->proto == 0); 1122 } 1123 1124 /* Compare two rules SRC addr field for skiplist construction */ 1125 int 1126 skip_cmp_src_addr(struct pf_rule *a, struct pf_rule *b) 1127 { 1128 if (a->src.neg != b->src.neg || 1129 a->src.addr.type != b->src.addr.type) 1130 return (1); 1131 /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0 1132 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP || 1133 * a->proto == IPPROTO_ICMP 1134 * return (1); 1135 */ 1136 switch (a->src.addr.type) { 1137 case PF_ADDR_ADDRMASK: 1138 if (memcmp(&a->src.addr.v.a.addr, &b->src.addr.v.a.addr, 1139 sizeof(a->src.addr.v.a.addr)) || 1140 memcmp(&a->src.addr.v.a.mask, &b->src.addr.v.a.mask, 1141 sizeof(a->src.addr.v.a.mask)) || 1142 (a->src.addr.v.a.addr.addr32[0] == 0 && 1143 a->src.addr.v.a.addr.addr32[1] == 0 && 1144 a->src.addr.v.a.addr.addr32[2] == 0 && 1145 a->src.addr.v.a.addr.addr32[3] == 0)) 1146 return (1); 1147 return (0); 1148 case PF_ADDR_DYNIFTL: 1149 if (strcmp(a->src.addr.v.ifname, b->src.addr.v.ifname) != 0 || 1150 a->src.addr.iflags != b->src.addr.iflags || 1151 memcmp(&a->src.addr.v.a.mask, &b->src.addr.v.a.mask, 1152 sizeof(a->src.addr.v.a.mask))) 1153 return (1); 1154 return (0); 1155 case PF_ADDR_NOROUTE: 1156 case PF_ADDR_URPFFAILED: 1157 return (0); 1158 case PF_ADDR_TABLE: 1159 return (strcmp(a->src.addr.v.tblname, b->src.addr.v.tblname)); 1160 } 1161 return (1); 1162 } 1163 1164 /* Compare two rules SRC port field for skiplist construction */ 1165 int 1166 skip_cmp_src_port(struct pf_rule *a, struct pf_rule *b) 1167 { 1168 if (a->src.port_op == PF_OP_NONE || a->src.port_op != b->src.port_op || 1169 a->src.port[0] != b->src.port[0] || 1170 a->src.port[1] != b->src.port[1]) 1171 return (1); 1172 /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0 1173 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP || 1174 * a->proto == IPPROTO_ICMP 1175 * return (1); 1176 */ 1177 return (0); 1178 } 1179 1180 1181 void 1182 skip_init(void) 1183 { 1184 struct { 1185 char *name; 1186 int skipnum; 1187 int (*func)(struct pf_rule *, struct pf_rule *); 1188 } comps[] = PF_SKIP_COMPARITORS; 1189 int skipnum, i; 1190 1191 for (skipnum = 0; skipnum < PF_SKIP_COUNT; skipnum++) { 1192 for (i = 0; i < sizeof(comps)/sizeof(*comps); i++) 1193 if (comps[i].skipnum == skipnum) { 1194 skip_comparitors[skipnum] = comps[i].func; 1195 skip_comparitors_names[skipnum] = comps[i].name; 1196 } 1197 } 1198 for (skipnum = 0; skipnum < PF_SKIP_COUNT; skipnum++) 1199 if (skip_comparitors[skipnum] == NULL) 1200 errx(1, "Need to add skip step comparitor to pfctl?!"); 1201 } 1202 1203 /* 1204 * Add a host/netmask to a table 1205 */ 1206 int 1207 add_opt_table(struct pfctl *pf, struct pf_opt_tbl **tbl, sa_family_t af, 1208 struct pf_rule_addr *addr, char *ifname) 1209 { 1210 #ifdef OPT_DEBUG 1211 char buf[128]; 1212 #endif /* OPT_DEBUG */ 1213 static int tablenum = 0; 1214 struct node_host node_host; 1215 1216 if (*tbl == NULL) { 1217 if ((*tbl = calloc(1, sizeof(**tbl))) == NULL || 1218 ((*tbl)->pt_buf = calloc(1, sizeof(*(*tbl)->pt_buf))) == 1219 NULL) 1220 err(1, "calloc"); 1221 (*tbl)->pt_buf->pfrb_type = PFRB_ADDRS; 1222 SIMPLEQ_INIT(&(*tbl)->pt_nodes); 1223 1224 /* This is just a temporary table name */ 1225 snprintf((*tbl)->pt_name, sizeof((*tbl)->pt_name), "%s%d", 1226 PF_OPT_TABLE_PREFIX, tablenum++); 1227 DEBUG("creating table <%s>", (*tbl)->pt_name); 1228 } 1229 1230 memset(&node_host, 0, sizeof(node_host)); 1231 node_host.af = af; 1232 node_host.addr = addr->addr; 1233 node_host.ifname = ifname; 1234 node_host.weight = addr->weight; 1235 1236 #ifdef OPT_DEBUG 1237 DEBUG("<%s> adding %s/%d", (*tbl)->pt_name, inet_ntop(af, 1238 &node_host.addr.v.a.addr, buf, sizeof(buf)), 1239 unmask(&node_host.addr.v.a.mask, af)); 1240 #endif /* OPT_DEBUG */ 1241 1242 if (append_addr_host((*tbl)->pt_buf, &node_host, 0, 0)) { 1243 warn("failed to add host"); 1244 return (1); 1245 } 1246 if (pf->opts & PF_OPT_VERBOSE) { 1247 struct node_tinit *ti; 1248 1249 if ((ti = calloc(1, sizeof(*ti))) == NULL) 1250 err(1, "malloc"); 1251 if ((ti->host = malloc(sizeof(*ti->host))) == NULL) 1252 err(1, "malloc"); 1253 memcpy(ti->host, &node_host, sizeof(*ti->host)); 1254 SIMPLEQ_INSERT_TAIL(&(*tbl)->pt_nodes, ti, entries); 1255 } 1256 1257 (*tbl)->pt_rulecount++; 1258 if ((*tbl)->pt_rulecount == TABLE_THRESHOLD) 1259 DEBUG("table <%s> now faster than skip steps", (*tbl)->pt_name); 1260 1261 return (0); 1262 } 1263 1264 1265 /* 1266 * Do the dirty work of choosing an unused table name and creating it. 1267 * (be careful with the table name, it might already be used in another anchor) 1268 */ 1269 int 1270 pf_opt_create_table(struct pfctl *pf, struct pf_opt_tbl *tbl) 1271 { 1272 static int tablenum; 1273 struct pfr_table *t; 1274 1275 if (table_buffer.pfrb_type == 0) { 1276 /* Initialize the list of tables */ 1277 table_buffer.pfrb_type = PFRB_TABLES; 1278 for (;;) { 1279 pfr_buf_grow(&table_buffer, table_buffer.pfrb_size); 1280 table_buffer.pfrb_size = table_buffer.pfrb_msize; 1281 if (pfr_get_tables(NULL, table_buffer.pfrb_caddr, 1282 &table_buffer.pfrb_size, PFR_FLAG_ALLRSETS)) 1283 err(1, "pfr_get_tables"); 1284 if (table_buffer.pfrb_size <= table_buffer.pfrb_msize) 1285 break; 1286 } 1287 table_identifier = arc4random(); 1288 } 1289 1290 /* XXX would be *really* nice to avoid duplicating identical tables */ 1291 1292 /* Now we have to pick a table name that isn't used */ 1293 again: 1294 DEBUG("translating temporary table <%s> to <%s%x_%d>", tbl->pt_name, 1295 PF_OPT_TABLE_PREFIX, table_identifier, tablenum); 1296 snprintf(tbl->pt_name, sizeof(tbl->pt_name), "%s%x_%d", 1297 PF_OPT_TABLE_PREFIX, table_identifier, tablenum); 1298 PFRB_FOREACH(t, &table_buffer) { 1299 if (strcasecmp(t->pfrt_name, tbl->pt_name) == 0) { 1300 /* Collision. Try again */ 1301 DEBUG("wow, table <%s> in use. trying again", 1302 tbl->pt_name); 1303 table_identifier = arc4random(); 1304 goto again; 1305 } 1306 } 1307 tablenum++; 1308 1309 if (pfctl_define_table(tbl->pt_name, PFR_TFLAG_CONST | tbl->pt_flags, 1, 1310 pf->astack[0]->name, tbl->pt_buf, pf->astack[0]->ruleset.tticket)) { 1311 warn("failed to create table %s in %s", 1312 tbl->pt_name, pf->astack[0]->name); 1313 return (1); 1314 } 1315 return (0); 1316 } 1317 1318 /* 1319 * Partition the flat ruleset into a list of distinct superblocks 1320 */ 1321 int 1322 construct_superblocks(struct pfctl *pf, struct pf_opt_queue *opt_queue, 1323 struct superblocks *superblocks) 1324 { 1325 struct superblock *block = NULL; 1326 struct pf_opt_rule *por; 1327 int i; 1328 1329 while (!TAILQ_EMPTY(opt_queue)) { 1330 por = TAILQ_FIRST(opt_queue); 1331 TAILQ_REMOVE(opt_queue, por, por_entry); 1332 if (block == NULL || !superblock_inclusive(block, por)) { 1333 if ((block = calloc(1, sizeof(*block))) == NULL) { 1334 warn("calloc"); 1335 return (1); 1336 } 1337 TAILQ_INIT(&block->sb_rules); 1338 for (i = 0; i < PF_SKIP_COUNT; i++) 1339 TAILQ_INIT(&block->sb_skipsteps[i]); 1340 TAILQ_INSERT_TAIL(superblocks, block, sb_entry); 1341 } 1342 TAILQ_INSERT_TAIL(&block->sb_rules, por, por_entry); 1343 } 1344 1345 return (0); 1346 } 1347 1348 1349 /* 1350 * Compare two rule addresses 1351 */ 1352 int 1353 addrs_equal(struct pf_rule_addr *a, struct pf_rule_addr *b) 1354 { 1355 if (a->neg != b->neg) 1356 return (0); 1357 return (memcmp(&a->addr, &b->addr, sizeof(a->addr)) == 0); 1358 } 1359 1360 1361 /* 1362 * The addresses are not equal, but can we combine them into one table? 1363 */ 1364 int 1365 addrs_combineable(struct pf_rule_addr *a, struct pf_rule_addr *b) 1366 { 1367 if (a->addr.type != PF_ADDR_ADDRMASK || 1368 b->addr.type != PF_ADDR_ADDRMASK) 1369 return (0); 1370 if (a->neg != b->neg || a->port_op != b->port_op || 1371 a->port[0] != b->port[0] || a->port[1] != b->port[1]) 1372 return (0); 1373 return (1); 1374 } 1375 1376 1377 /* 1378 * Are we allowed to combine these two rules 1379 */ 1380 int 1381 rules_combineable(struct pf_rule *p1, struct pf_rule *p2) 1382 { 1383 struct pf_rule a, b; 1384 1385 comparable_rule(&a, p1, COMBINED); 1386 comparable_rule(&b, p2, COMBINED); 1387 return (memcmp(&a, &b, sizeof(a)) == 0); 1388 } 1389 1390 1391 /* 1392 * Can a rule be included inside a superblock 1393 */ 1394 int 1395 superblock_inclusive(struct superblock *block, struct pf_opt_rule *por) 1396 { 1397 struct pf_rule a, b; 1398 int i, j; 1399 1400 /* First check for hard breaks */ 1401 for (i = 0; i < sizeof(pf_rule_desc)/sizeof(*pf_rule_desc); i++) { 1402 if (pf_rule_desc[i].prf_type == BARRIER) { 1403 for (j = 0; j < pf_rule_desc[i].prf_size; j++) 1404 if (((char *)&por->por_rule)[j + 1405 pf_rule_desc[i].prf_offset] != 0) 1406 return (0); 1407 } 1408 } 1409 1410 /* per-rule src-track is also a hard break */ 1411 if (por->por_rule.rule_flag & PFRULE_RULESRCTRACK) 1412 return (0); 1413 1414 /* 1415 * Have to handle interface groups separately. Consider the following 1416 * rules: 1417 * block on EXTIFS to any port 22 1418 * pass on em0 to any port 22 1419 * (where EXTIFS is an arbitrary interface group) 1420 * The optimizer may decide to re-order the pass rule in front of the 1421 * block rule. But what if EXTIFS includes em0??? Such a reordering 1422 * would change the meaning of the ruleset. 1423 * We can't just lookup the EXTIFS group and check if em0 is a member 1424 * because the user is allowed to add interfaces to a group during 1425 * runtime. 1426 * Ergo interface groups become a defacto superblock break :-( 1427 */ 1428 if (interface_group(por->por_rule.ifname) || 1429 interface_group(TAILQ_FIRST(&block->sb_rules)->por_rule.ifname)) { 1430 if (strcasecmp(por->por_rule.ifname, 1431 TAILQ_FIRST(&block->sb_rules)->por_rule.ifname) != 0) 1432 return (0); 1433 } 1434 1435 comparable_rule(&a, &TAILQ_FIRST(&block->sb_rules)->por_rule, NOMERGE); 1436 comparable_rule(&b, &por->por_rule, NOMERGE); 1437 if (memcmp(&a, &b, sizeof(a)) == 0) 1438 return (1); 1439 1440 #ifdef OPT_DEBUG 1441 for (i = 0; i < sizeof(por->por_rule); i++) { 1442 int closest = -1; 1443 if (((u_int8_t *)&a)[i] != ((u_int8_t *)&b)[i]) { 1444 for (j = 0; j < sizeof(pf_rule_desc) / 1445 sizeof(*pf_rule_desc); j++) { 1446 if (i >= pf_rule_desc[j].prf_offset && 1447 i < pf_rule_desc[j].prf_offset + 1448 pf_rule_desc[j].prf_size) { 1449 DEBUG("superblock break @ %d due to %s", 1450 por->por_rule.nr, 1451 pf_rule_desc[j].prf_name); 1452 return (0); 1453 } 1454 if (i > pf_rule_desc[j].prf_offset) { 1455 if (closest == -1 || 1456 i-pf_rule_desc[j].prf_offset < 1457 i-pf_rule_desc[closest].prf_offset) 1458 closest = j; 1459 } 1460 } 1461 1462 if (closest >= 0) 1463 DEBUG("superblock break @ %d on %s+%lxh", 1464 por->por_rule.nr, 1465 pf_rule_desc[closest].prf_name, 1466 i - pf_rule_desc[closest].prf_offset - 1467 pf_rule_desc[closest].prf_size); 1468 else 1469 DEBUG("superblock break @ %d on field @ %d", 1470 por->por_rule.nr, i); 1471 return (0); 1472 } 1473 } 1474 #endif /* OPT_DEBUG */ 1475 1476 return (0); 1477 } 1478 1479 1480 /* 1481 * Figure out if an interface name is an actual interface or actually a 1482 * group of interfaces. 1483 */ 1484 int 1485 interface_group(const char *ifname) 1486 { 1487 if (ifname == NULL || !ifname[0]) 1488 return (0); 1489 1490 /* Real interfaces must end in a number, interface groups do not */ 1491 if (isdigit((unsigned char)ifname[strlen(ifname) - 1])) 1492 return (0); 1493 else 1494 return (1); 1495 } 1496 1497 1498 /* 1499 * Make a rule that can directly compared by memcmp() 1500 */ 1501 void 1502 comparable_rule(struct pf_rule *dst, const struct pf_rule *src, int type) 1503 { 1504 int i; 1505 /* 1506 * To simplify the comparison, we just zero out the fields that are 1507 * allowed to be different and then do a simple memcmp() 1508 */ 1509 memcpy(dst, src, sizeof(*dst)); 1510 for (i = 0; i < sizeof(pf_rule_desc)/sizeof(*pf_rule_desc); i++) 1511 if (pf_rule_desc[i].prf_type >= type) { 1512 #ifdef OPT_DEBUG 1513 assert(pf_rule_desc[i].prf_type != NEVER || 1514 *(((char *)dst) + pf_rule_desc[i].prf_offset) == 0); 1515 #endif /* OPT_DEBUG */ 1516 memset(((char *)dst) + pf_rule_desc[i].prf_offset, 0, 1517 pf_rule_desc[i].prf_size); 1518 } 1519 } 1520 1521 1522 /* 1523 * Remove superset information from two rules so we can directly compare them 1524 * with memcmp() 1525 */ 1526 void 1527 exclude_supersets(struct pf_rule *super, struct pf_rule *sub) 1528 { 1529 if (super->ifname[0] == '\0') 1530 memset(sub->ifname, 0, sizeof(sub->ifname)); 1531 if (super->direction == PF_INOUT) 1532 sub->direction = PF_INOUT; 1533 if ((super->proto == 0 || super->proto == sub->proto) && 1534 super->flags == 0 && super->flagset == 0 && (sub->flags || 1535 sub->flagset)) { 1536 sub->flags = super->flags; 1537 sub->flagset = super->flagset; 1538 } 1539 if (super->proto == 0) 1540 sub->proto = 0; 1541 1542 if (super->src.port_op == 0) { 1543 sub->src.port_op = 0; 1544 sub->src.port[0] = 0; 1545 sub->src.port[1] = 0; 1546 } 1547 if (super->dst.port_op == 0) { 1548 sub->dst.port_op = 0; 1549 sub->dst.port[0] = 0; 1550 sub->dst.port[1] = 0; 1551 } 1552 1553 if (super->src.addr.type == PF_ADDR_ADDRMASK && !super->src.neg && 1554 !sub->src.neg && super->src.addr.v.a.mask.addr32[0] == 0 && 1555 super->src.addr.v.a.mask.addr32[1] == 0 && 1556 super->src.addr.v.a.mask.addr32[2] == 0 && 1557 super->src.addr.v.a.mask.addr32[3] == 0) 1558 memset(&sub->src.addr, 0, sizeof(sub->src.addr)); 1559 else if (super->src.addr.type == PF_ADDR_ADDRMASK && 1560 sub->src.addr.type == PF_ADDR_ADDRMASK && 1561 super->src.neg == sub->src.neg && 1562 super->af == sub->af && 1563 unmask(&super->src.addr.v.a.mask, super->af) < 1564 unmask(&sub->src.addr.v.a.mask, sub->af) && 1565 super->src.addr.v.a.addr.addr32[0] == 1566 (sub->src.addr.v.a.addr.addr32[0] & 1567 super->src.addr.v.a.mask.addr32[0]) && 1568 super->src.addr.v.a.addr.addr32[1] == 1569 (sub->src.addr.v.a.addr.addr32[1] & 1570 super->src.addr.v.a.mask.addr32[1]) && 1571 super->src.addr.v.a.addr.addr32[2] == 1572 (sub->src.addr.v.a.addr.addr32[2] & 1573 super->src.addr.v.a.mask.addr32[2]) && 1574 super->src.addr.v.a.addr.addr32[3] == 1575 (sub->src.addr.v.a.addr.addr32[3] & 1576 super->src.addr.v.a.mask.addr32[3])) { 1577 /* sub->src.addr is a subset of super->src.addr/mask */ 1578 memcpy(&sub->src.addr, &super->src.addr, sizeof(sub->src.addr)); 1579 } 1580 1581 if (super->dst.addr.type == PF_ADDR_ADDRMASK && !super->dst.neg && 1582 !sub->dst.neg && super->dst.addr.v.a.mask.addr32[0] == 0 && 1583 super->dst.addr.v.a.mask.addr32[1] == 0 && 1584 super->dst.addr.v.a.mask.addr32[2] == 0 && 1585 super->dst.addr.v.a.mask.addr32[3] == 0) 1586 memset(&sub->dst.addr, 0, sizeof(sub->dst.addr)); 1587 else if (super->dst.addr.type == PF_ADDR_ADDRMASK && 1588 sub->dst.addr.type == PF_ADDR_ADDRMASK && 1589 super->dst.neg == sub->dst.neg && 1590 super->af == sub->af && 1591 unmask(&super->dst.addr.v.a.mask, super->af) < 1592 unmask(&sub->dst.addr.v.a.mask, sub->af) && 1593 super->dst.addr.v.a.addr.addr32[0] == 1594 (sub->dst.addr.v.a.addr.addr32[0] & 1595 super->dst.addr.v.a.mask.addr32[0]) && 1596 super->dst.addr.v.a.addr.addr32[1] == 1597 (sub->dst.addr.v.a.addr.addr32[1] & 1598 super->dst.addr.v.a.mask.addr32[1]) && 1599 super->dst.addr.v.a.addr.addr32[2] == 1600 (sub->dst.addr.v.a.addr.addr32[2] & 1601 super->dst.addr.v.a.mask.addr32[2]) && 1602 super->dst.addr.v.a.addr.addr32[3] == 1603 (sub->dst.addr.v.a.addr.addr32[3] & 1604 super->dst.addr.v.a.mask.addr32[3])) { 1605 /* sub->dst.addr is a subset of super->dst.addr/mask */ 1606 memcpy(&sub->dst.addr, &super->dst.addr, sizeof(sub->dst.addr)); 1607 } 1608 1609 if (super->af == 0) 1610 sub->af = 0; 1611 } 1612 1613 1614 void 1615 superblock_free(struct pfctl *pf, struct superblock *block) 1616 { 1617 struct pf_opt_rule *por; 1618 while ((por = TAILQ_FIRST(&block->sb_rules))) { 1619 TAILQ_REMOVE(&block->sb_rules, por, por_entry); 1620 if (por->por_src_tbl) { 1621 if (por->por_src_tbl->pt_buf) { 1622 pfr_buf_clear(por->por_src_tbl->pt_buf); 1623 free(por->por_src_tbl->pt_buf); 1624 } 1625 free(por->por_src_tbl); 1626 } 1627 if (por->por_dst_tbl) { 1628 if (por->por_dst_tbl->pt_buf) { 1629 pfr_buf_clear(por->por_dst_tbl->pt_buf); 1630 free(por->por_dst_tbl->pt_buf); 1631 } 1632 free(por->por_dst_tbl); 1633 } 1634 free(por); 1635 } 1636 if (block->sb_profiled_block) 1637 superblock_free(pf, block->sb_profiled_block); 1638 free(block); 1639 } 1640 1641