Deleted Added
full compact
1/* $OpenBSD: pfctl_optimize.c,v 1.5 2005/01/03 15:18:10 frantzen Exp $ */
1/* $OpenBSD: pfctl_optimize.c,v 1.13 2006/10/31 14:17:45 mcbride 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: head/contrib/pf/pfctl/pfctl_optimize.c 145840 2005-05-03 16:55:20Z mlaier $");
20__FBSDID("$FreeBSD: head/contrib/pf/pfctl/pfctl_optimize.c 171172 2007-07-03 12:30:03Z mlaier $");
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 */
60struct 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 */
74struct 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};
80TAILQ_HEAD(superblocks, superblock);
81
82
83/*
84 * Description of the PF rule structure.
85 */
86enum {
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?!? */
93struct 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),
143 PF_RULE_FIELD(ifname, 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),
176 PF_RULE_FIELD(anchor, DC),
188 PF_RULE_FIELD(states, DC),
189 PF_RULE_FIELD(src_nodes, DC),
190 PF_RULE_FIELD(nr, DC),
191 PF_RULE_FIELD(entries, DC),
192 PF_RULE_FIELD(qid, DC),
193 PF_RULE_FIELD(pqid, DC),
194 PF_RULE_FIELD(anchor_relative, DC),
195 PF_RULE_FIELD(anchor_wildcard, DC),
196 PF_RULE_FIELD(tag, DC),
197 PF_RULE_FIELD(match_tag, DC),
198 PF_RULE_FIELD(overload_tbl, DC),
199
200 /* These fields should never be set in a PASS/BLOCK rule */
201 PF_RULE_FIELD(natpass, NEVER),
202 PF_RULE_FIELD(max_mss, NEVER),
203 PF_RULE_FIELD(min_ttl, NEVER),
204};
205
206
207
208int add_opt_table(struct pfctl *, struct pf_opt_tbl **, sa_family_t,
209 struct pf_rule_addr *);
210int addrs_combineable(struct pf_rule_addr *, struct pf_rule_addr *);
211int addrs_equal(struct pf_rule_addr *, struct pf_rule_addr *);
212int block_feedback(struct pfctl *, struct superblock *);
213int combine_rules(struct pfctl *, struct superblock *);
214void comparable_rule(struct pf_rule *, const struct pf_rule *, int);
215int construct_superblocks(struct pfctl *, struct pf_opt_queue *,
216 struct superblocks *);
217void exclude_supersets(struct pf_rule *, struct pf_rule *);
218int interface_group(const char *);
219int load_feedback_profile(struct pfctl *, struct superblocks *);
220int optimize_superblock(struct pfctl *, struct superblock *);
221int pf_opt_create_table(struct pfctl *, struct pf_opt_tbl *);
222void remove_from_skipsteps(struct skiplist *, struct superblock *,
223 struct pf_opt_rule *, struct pf_skip_step *);
224int remove_identical_rules(struct pfctl *, struct superblock *);
225int reorder_rules(struct pfctl *, struct superblock *, int);
226int rules_combineable(struct pf_rule *, struct pf_rule *);
227void skip_append(struct superblock *, int, struct pf_skip_step *,
228 struct pf_opt_rule *);
229int skip_compare(int, struct pf_skip_step *, struct pf_opt_rule *);
230void skip_init(void);
231int skip_cmp_af(struct pf_rule *, struct pf_rule *);
232int skip_cmp_dir(struct pf_rule *, struct pf_rule *);
233int skip_cmp_dst_addr(struct pf_rule *, struct pf_rule *);
234int skip_cmp_dst_port(struct pf_rule *, struct pf_rule *);
235int skip_cmp_ifp(struct pf_rule *, struct pf_rule *);
236int skip_cmp_proto(struct pf_rule *, struct pf_rule *);
237int skip_cmp_src_addr(struct pf_rule *, struct pf_rule *);
238int skip_cmp_src_port(struct pf_rule *, struct pf_rule *);
239int superblock_inclusive(struct superblock *, struct pf_opt_rule *);
240void superblock_free(struct pfctl *, struct superblock *);
241
242
243int (*skip_comparitors[PF_SKIP_COUNT])(struct pf_rule *, struct pf_rule *);
244const 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 { "af", PF_SKIP_AF, skip_cmp_af }, \
249 { "proto", PF_SKIP_PROTO, skip_cmp_proto }, \
250 { "saddr", PF_SKIP_SRC_ADDR, skip_cmp_src_addr }, \
251 { "sport", PF_SKIP_SRC_PORT, skip_cmp_src_port }, \
252 { "daddr", PF_SKIP_DST_ADDR, skip_cmp_dst_addr }, \
253 { "dport", PF_SKIP_DST_PORT, skip_cmp_dst_port } \
254}
255
256struct pfr_buffer table_buffer;
257int table_identifier;
258
259
260int
246pfctl_optimize_rules(struct pfctl *pf)
261pfctl_optimize_ruleset(struct pfctl *pf, struct pf_ruleset *rs)
262{
263 struct superblocks superblocks;
264 struct pf_opt_queue opt_queue;
265 struct superblock *block;
266 struct pf_opt_rule *por;
251 int nr;
267 struct pf_rule *r;
268 struct pf_rulequeue *old_rules;
269
270 DEBUG("optimizing ruleset");
271 memset(&table_buffer, 0, sizeof(table_buffer));
272 skip_init();
273 TAILQ_INIT(&opt_queue);
274
257 if (TAILQ_FIRST(&pf->opt_queue))
258 nr = TAILQ_FIRST(&pf->opt_queue)->por_rule.nr;
275 old_rules = rs->rules[PF_RULESET_FILTER].active.ptr;
276 rs->rules[PF_RULESET_FILTER].active.ptr =
277 rs->rules[PF_RULESET_FILTER].inactive.ptr;
278 rs->rules[PF_RULESET_FILTER].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[PF_RULESET_FILTER].inactive.ptr))
285 != NULL) {
286 TAILQ_REMOVE(rs->rules[PF_RULESET_FILTER].inactive.ptr, r,
287 entries);
288 if ((por = calloc(1, sizeof(*por))) == NULL)
289 err(1, "calloc");
290 memcpy(&por->por_rule, r, sizeof(*r));
291 if (TAILQ_FIRST(&r->rpool.list) != NULL) {
292 TAILQ_INIT(&por->por_rule.rpool.list);
293 pfctl_move_pool(&r->rpool, &por->por_rule.rpool);
294 } else
295 bzero(&por->por_rule.rpool,
296 sizeof(por->por_rule.rpool));
297
298
299 TAILQ_INSERT_TAIL(&opt_queue, por, por_entry);
300 }
301
302 TAILQ_INIT(&superblocks);
261 if (construct_superblocks(pf, &pf->opt_queue, &superblocks))
303 if (construct_superblocks(pf, &opt_queue, &superblocks))
304 goto error;
305
264 if (pf->opts & PF_OPT_OPTIMIZE_PROFILE) {
306 if (pf->optimize & PF_OPTIMIZE_PROFILE) {
307 if (load_feedback_profile(pf, &superblocks))
308 goto error;
309 }
310
311 TAILQ_FOREACH(block, &superblocks, sb_entry) {
312 if (optimize_superblock(pf, block))
313 goto error;
314 }
315
274
275 /*
276 * Optimizations are done so we turn off the optimization flag and
277 * put the rules right back into the regular codepath.
278 */
279 pf->opts &= ~PF_OPT_OPTIMIZE;
280
316 rs->anchor->refcnt = 0;
317 while ((block = TAILQ_FIRST(&superblocks))) {
318 TAILQ_REMOVE(&superblocks, block, sb_entry);
319
320 while ((por = TAILQ_FIRST(&block->sb_rules))) {
321 TAILQ_REMOVE(&block->sb_rules, por, por_entry);
286 por->por_rule.nr = nr++;
287 if (pfctl_add_rule(pf, &por->por_rule,
288 por->por_anchor)) {
289 free(por);
290 goto error;
291 }
322 por->por_rule.nr = rs->anchor->refcnt++;
323 if ((r = calloc(1, sizeof(*r))) == NULL)
324 err(1, "calloc");
325 memcpy(r, &por->por_rule, sizeof(*r));
326 TAILQ_INIT(&r->rpool.list);
327 pfctl_move_pool(&por->por_rule.rpool, &r->rpool);
328 TAILQ_INSERT_TAIL(
329 rs->rules[PF_RULESET_FILTER].active.ptr,
330 r, entries);
331 free(por);
332 }
333 free(block);
334 }
335
336 return (0);
337
338error:
300 while ((por = TAILQ_FIRST(&pf->opt_queue))) {
301 TAILQ_REMOVE(&pf->opt_queue, por, por_entry);
339 while ((por = TAILQ_FIRST(&opt_queue))) {
340 TAILQ_REMOVE(&opt_queue, por, por_entry);
341 if (por->por_src_tbl) {
342 pfr_buf_clear(por->por_src_tbl->pt_buf);
343 free(por->por_src_tbl->pt_buf);
344 free(por->por_src_tbl);
345 }
346 if (por->por_dst_tbl) {
347 pfr_buf_clear(por->por_dst_tbl->pt_buf);
348 free(por->por_dst_tbl->pt_buf);
349 free(por->por_dst_tbl);
350 }
351 free(por);
352 }
353 while ((block = TAILQ_FIRST(&superblocks))) {
354 TAILQ_REMOVE(&superblocks, block, sb_entry);
355 superblock_free(pf, block);
356 }
357 return (1);
358}
359
360
361/*
362 * Go ahead and optimize a superblock
363 */
364int
365optimize_superblock(struct pfctl *pf, struct superblock *block)
366{
367#ifdef OPT_DEBUG
368 struct pf_opt_rule *por;
369#endif /* OPT_DEBUG */
370
371 /* We have a few optimization passes:
372 * 1) remove duplicate rules or rules that are a subset of other
373 * rules
374 * 2) combine otherwise identical rules with different IP addresses
375 * into a single rule and put the addresses in a table.
376 * 3) re-order the rules to improve kernel skip steps
377 * 4) re-order the 'quick' rules based on feedback from the
378 * active ruleset statistics
379 *
380 * XXX combine_rules() doesn't combine v4 and v6 rules. would just
381 * have to keep af in the table container, make af 'COMBINE' and
382 * twiddle the af on the merged rule
383 * XXX maybe add a weighting to the metric on skipsteps when doing
384 * reordering. sometimes two sequential tables will be better
385 * that four consecutive interfaces.
386 * XXX need to adjust the skipstep count of everything after PROTO,
387 * since they aren't actually checked on a proto mismatch in
388 * pf_test_{tcp, udp, icmp}()
389 * XXX should i treat proto=0, af=0 or dir=0 special in skepstep
390 * calculation since they are a DC?
391 * XXX keep last skiplist of last superblock to influence this
392 * superblock. '5 inet6 log' should make '3 inet6' come before '4
393 * inet' in the next superblock.
394 * XXX would be useful to add tables for ports
395 * XXX we can also re-order some mutually exclusive superblocks to
396 * try merging superblocks before any of these optimization passes.
397 * for instance a single 'log in' rule in the middle of non-logging
398 * out rules.
399 */
400
401 /* shortcut. there will be alot of 1-rule superblocks */
402 if (!TAILQ_NEXT(TAILQ_FIRST(&block->sb_rules), por_entry))
403 return (0);
404
405#ifdef OPT_DEBUG
406 printf("--- Superblock ---\n");
407 TAILQ_FOREACH(por, &block->sb_rules, por_entry) {
408 printf(" ");
370 print_rule(&por->por_rule, por->por_anchor, 1);
409 print_rule(&por->por_rule, por->por_rule.anchor ?
410 por->por_rule.anchor->name : "", 1);
411 }
412#endif /* OPT_DEBUG */
413
414
415 if (remove_identical_rules(pf, block))
416 return (1);
417 if (combine_rules(pf, block))
418 return (1);
379 if ((pf->opts & PF_OPT_OPTIMIZE_PROFILE) &&
419 if ((pf->optimize & PF_OPTIMIZE_PROFILE) &&
420 TAILQ_FIRST(&block->sb_rules)->por_rule.quick &&
421 block->sb_profiled_block) {
422 if (block_feedback(pf, block))
423 return (1);
424 } else if (reorder_rules(pf, block, 0)) {
425 return (1);
426 }
427
428 /*
429 * Don't add any optimization passes below reorder_rules(). It will
430 * have divided superblocks into smaller blocks for further refinement
431 * and doesn't put them back together again. What once was a true
432 * superblock might have been split into multiple superblocks.
433 */
434
435#ifdef OPT_DEBUG
436 printf("--- END Superblock ---\n");
437#endif /* OPT_DEBUG */
438 return (0);
439}
440
441
442/*
443 * Optimization pass #1: remove identical rules
444 */
445int
446remove_identical_rules(struct pfctl *pf, struct superblock *block)
447{
448 struct pf_opt_rule *por1, *por2, *por_next, *por2_next;
449 struct pf_rule a, a2, b, b2;
450
451 for (por1 = TAILQ_FIRST(&block->sb_rules); por1; por1 = por_next) {
452 por_next = TAILQ_NEXT(por1, por_entry);
453 for (por2 = por_next; por2; por2 = por2_next) {
454 por2_next = TAILQ_NEXT(por2, por_entry);
455 comparable_rule(&a, &por1->por_rule, DC);
456 comparable_rule(&b, &por2->por_rule, DC);
457 memcpy(&a2, &a, sizeof(a2));
458 memcpy(&b2, &b, sizeof(b2));
459
460 exclude_supersets(&a, &b);
461 exclude_supersets(&b2, &a2);
462 if (memcmp(&a, &b, sizeof(a)) == 0) {
463 DEBUG("removing identical rule nr%d = *nr%d*",
464 por1->por_rule.nr, por2->por_rule.nr);
465 TAILQ_REMOVE(&block->sb_rules, por2, por_entry);
466 if (por_next == por2)
467 por_next = TAILQ_NEXT(por1, por_entry);
468 free(por2);
469 } else if (memcmp(&a2, &b2, sizeof(a2)) == 0) {
470 DEBUG("removing identical rule *nr%d* = nr%d",
471 por1->por_rule.nr, por2->por_rule.nr);
472 TAILQ_REMOVE(&block->sb_rules, por1, por_entry);
473 free(por1);
474 break;
475 }
476 }
477 }
478
479 return (0);
480}
481
482
483/*
484 * Optimization pass #2: combine similar rules with different addresses
485 * into a single rule and a table
486 */
487int
488combine_rules(struct pfctl *pf, struct superblock *block)
489{
490 struct pf_opt_rule *p1, *p2, *por_next;
491 int src_eq, dst_eq;
492
493 if ((pf->loadopt & PFCTL_FLAG_TABLE) == 0) {
494 warnx("Must enable table loading for optimizations");
495 return (1);
496 }
497
498 /* First we make a pass to combine the rules. O(n log n) */
499 TAILQ_FOREACH(p1, &block->sb_rules, por_entry) {
500 for (p2 = TAILQ_NEXT(p1, por_entry); p2; p2 = por_next) {
501 por_next = TAILQ_NEXT(p2, por_entry);
502
503 src_eq = addrs_equal(&p1->por_rule.src,
504 &p2->por_rule.src);
505 dst_eq = addrs_equal(&p1->por_rule.dst,
506 &p2->por_rule.dst);
507
508 if (src_eq && !dst_eq && p1->por_src_tbl == NULL &&
509 p2->por_dst_tbl == NULL &&
510 p2->por_src_tbl == NULL &&
511 rules_combineable(&p1->por_rule, &p2->por_rule) &&
512 addrs_combineable(&p1->por_rule.dst,
513 &p2->por_rule.dst)) {
514 DEBUG("can combine rules nr%d = nr%d",
515 p1->por_rule.nr, p2->por_rule.nr);
516 if (p1->por_dst_tbl == NULL &&
517 add_opt_table(pf, &p1->por_dst_tbl,
518 p1->por_rule.af, &p1->por_rule.dst))
519 return (1);
520 if (add_opt_table(pf, &p1->por_dst_tbl,
521 p1->por_rule.af, &p2->por_rule.dst))
522 return (1);
523 p2->por_dst_tbl = p1->por_dst_tbl;
524 if (p1->por_dst_tbl->pt_rulecount >=
525 TABLE_THRESHOLD) {
526 TAILQ_REMOVE(&block->sb_rules, p2,
527 por_entry);
528 free(p2);
529 }
530 } else if (!src_eq && dst_eq && p1->por_dst_tbl == NULL
531 && p2->por_src_tbl == NULL &&
532 p2->por_dst_tbl == NULL &&
533 rules_combineable(&p1->por_rule, &p2->por_rule) &&
534 addrs_combineable(&p1->por_rule.src,
535 &p2->por_rule.src)) {
536 DEBUG("can combine rules nr%d = nr%d",
537 p1->por_rule.nr, p2->por_rule.nr);
538 if (p1->por_src_tbl == NULL &&
539 add_opt_table(pf, &p1->por_src_tbl,
540 p1->por_rule.af, &p1->por_rule.src))
541 return (1);
542 if (add_opt_table(pf, &p1->por_src_tbl,
543 p1->por_rule.af, &p2->por_rule.src))
544 return (1);
545 p2->por_src_tbl = p1->por_src_tbl;
546 if (p1->por_src_tbl->pt_rulecount >=
547 TABLE_THRESHOLD) {
548 TAILQ_REMOVE(&block->sb_rules, p2,
549 por_entry);
550 free(p2);
551 }
552 }
553 }
554 }
555
556
557 /*
558 * Then we make a final pass to create a valid table name and
559 * insert the name into the rules.
560 */
561 for (p1 = TAILQ_FIRST(&block->sb_rules); p1; p1 = por_next) {
562 por_next = TAILQ_NEXT(p1, por_entry);
563 assert(p1->por_src_tbl == NULL || p1->por_dst_tbl == NULL);
564
565 if (p1->por_src_tbl && p1->por_src_tbl->pt_rulecount >=
566 TABLE_THRESHOLD) {
567 if (p1->por_src_tbl->pt_generated) {
568 /* This rule is included in a table */
569 TAILQ_REMOVE(&block->sb_rules, p1, por_entry);
570 free(p1);
571 continue;
572 }
573 p1->por_src_tbl->pt_generated = 1;
574
575 if ((pf->opts & PF_OPT_NOACTION) == 0 &&
576 pf_opt_create_table(pf, p1->por_src_tbl))
577 return (1);
578
579 pf->tdirty = 1;
580
581 if (pf->opts & PF_OPT_VERBOSE)
582 print_tabledef(p1->por_src_tbl->pt_name,
583 PFR_TFLAG_CONST, 1,
584 &p1->por_src_tbl->pt_nodes);
585
586 memset(&p1->por_rule.src.addr, 0,
587 sizeof(p1->por_rule.src.addr));
588 p1->por_rule.src.addr.type = PF_ADDR_TABLE;
589 strlcpy(p1->por_rule.src.addr.v.tblname,
590 p1->por_src_tbl->pt_name,
591 sizeof(p1->por_rule.src.addr.v.tblname));
592
593 pfr_buf_clear(p1->por_src_tbl->pt_buf);
594 free(p1->por_src_tbl->pt_buf);
595 p1->por_src_tbl->pt_buf = NULL;
596 }
597 if (p1->por_dst_tbl && p1->por_dst_tbl->pt_rulecount >=
598 TABLE_THRESHOLD) {
599 if (p1->por_dst_tbl->pt_generated) {
600 /* This rule is included in a table */
601 TAILQ_REMOVE(&block->sb_rules, p1, por_entry);
602 free(p1);
603 continue;
604 }
605 p1->por_dst_tbl->pt_generated = 1;
606
607 if ((pf->opts & PF_OPT_NOACTION) == 0 &&
608 pf_opt_create_table(pf, p1->por_dst_tbl))
609 return (1);
610 pf->tdirty = 1;
611
612 if (pf->opts & PF_OPT_VERBOSE)
613 print_tabledef(p1->por_dst_tbl->pt_name,
614 PFR_TFLAG_CONST, 1,
615 &p1->por_dst_tbl->pt_nodes);
616
617 memset(&p1->por_rule.dst.addr, 0,
618 sizeof(p1->por_rule.dst.addr));
619 p1->por_rule.dst.addr.type = PF_ADDR_TABLE;
620 strlcpy(p1->por_rule.dst.addr.v.tblname,
621 p1->por_dst_tbl->pt_name,
622 sizeof(p1->por_rule.dst.addr.v.tblname));
623
624 pfr_buf_clear(p1->por_dst_tbl->pt_buf);
625 free(p1->por_dst_tbl->pt_buf);
626 p1->por_dst_tbl->pt_buf = NULL;
627 }
628 }
629
630 return (0);
631}
632
633
634/*
635 * Optimization pass #3: re-order rules to improve skip steps
636 */
637int
638reorder_rules(struct pfctl *pf, struct superblock *block, int depth)
639{
640 struct superblock *newblock;
641 struct pf_skip_step *skiplist;
642 struct pf_opt_rule *por;
643 int i, largest, largest_list, rule_count = 0;
644 TAILQ_HEAD( , pf_opt_rule) head;
645
646 /*
647 * Calculate the best-case skip steps. We put each rule in a list
648 * of other rules with common fields
649 */
650 for (i = 0; i < PF_SKIP_COUNT; i++) {
651 TAILQ_FOREACH(por, &block->sb_rules, por_entry) {
652 TAILQ_FOREACH(skiplist, &block->sb_skipsteps[i],
653 ps_entry) {
654 if (skip_compare(i, skiplist, por) == 0)
655 break;
656 }
657 if (skiplist == NULL) {
658 if ((skiplist = calloc(1, sizeof(*skiplist))) ==
659 NULL)
660 err(1, "calloc");
661 TAILQ_INIT(&skiplist->ps_rules);
662 TAILQ_INSERT_TAIL(&block->sb_skipsteps[i],
663 skiplist, ps_entry);
664 }
665 skip_append(block, i, skiplist, por);
666 }
667 }
668
669 TAILQ_FOREACH(por, &block->sb_rules, por_entry)
670 rule_count++;
671
672 /*
673 * Now we're going to ignore any fields that are identical between
674 * all of the rules in the superblock and those fields which differ
675 * between every rule in the superblock.
676 */
677 largest = 0;
678 for (i = 0; i < PF_SKIP_COUNT; i++) {
679 skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]);
680 if (skiplist->ps_count == rule_count) {
681 DEBUG("(%d) original skipstep '%s' is all rules",
682 depth, skip_comparitors_names[i]);
683 skiplist->ps_count = 0;
684 } else if (skiplist->ps_count == 1) {
685 skiplist->ps_count = 0;
686 } else {
687 DEBUG("(%d) original skipstep '%s' largest jump is %d",
688 depth, skip_comparitors_names[i],
689 skiplist->ps_count);
690 if (skiplist->ps_count > largest)
691 largest = skiplist->ps_count;
692 }
693 }
694 if (largest == 0) {
695 /* Ugh. There is NO commonality in the superblock on which
696 * optimize the skipsteps optimization.
697 */
698 goto done;
699 }
700
701 /*
702 * Now we're going to empty the superblock rule list and re-create
703 * it based on a more optimal skipstep order.
704 */
705 TAILQ_INIT(&head);
706 while ((por = TAILQ_FIRST(&block->sb_rules))) {
707 TAILQ_REMOVE(&block->sb_rules, por, por_entry);
708 TAILQ_INSERT_TAIL(&head, por, por_entry);
709 }
710
711
712 while (!TAILQ_EMPTY(&head)) {
713 largest = 1;
714
715 /*
716 * Find the most useful skip steps remaining
717 */
718 for (i = 0; i < PF_SKIP_COUNT; i++) {
719 skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]);
720 if (skiplist->ps_count > largest) {
721 largest = skiplist->ps_count;
722 largest_list = i;
723 }
724 }
725
726 if (largest <= 1) {
727 /*
728 * Nothing useful left. Leave remaining rules in order.
729 */
730 DEBUG("(%d) no more commonality for skip steps", depth);
731 while ((por = TAILQ_FIRST(&head))) {
732 TAILQ_REMOVE(&head, por, por_entry);
733 TAILQ_INSERT_TAIL(&block->sb_rules, por,
734 por_entry);
735 }
736 } else {
737 /*
738 * There is commonality. Extract those common rules
739 * and place them in the ruleset adjacent to each
740 * other.
741 */
742 skiplist = TAILQ_FIRST(&block->sb_skipsteps[
743 largest_list]);
744 DEBUG("(%d) skipstep '%s' largest jump is %d @ #%d",
745 depth, skip_comparitors_names[largest_list],
746 largest, TAILQ_FIRST(&TAILQ_FIRST(&block->
747 sb_skipsteps [largest_list])->ps_rules)->
748 por_rule.nr);
749 TAILQ_REMOVE(&block->sb_skipsteps[largest_list],
750 skiplist, ps_entry);
751
752
753 /*
754 * There may be further commonality inside these
755 * rules. So we'll split them off into they're own
756 * superblock and pass it back into the optimizer.
757 */
758 if (skiplist->ps_count > 2) {
759 if ((newblock = calloc(1, sizeof(*newblock)))
760 == NULL) {
761 warn("calloc");
762 return (1);
763 }
764 TAILQ_INIT(&newblock->sb_rules);
765 for (i = 0; i < PF_SKIP_COUNT; i++)
766 TAILQ_INIT(&newblock->sb_skipsteps[i]);
767 TAILQ_INSERT_BEFORE(block, newblock, sb_entry);
768 DEBUG("(%d) splitting off %d rules from superblock @ #%d",
769 depth, skiplist->ps_count,
770 TAILQ_FIRST(&skiplist->ps_rules)->
771 por_rule.nr);
772 } else {
773 newblock = block;
774 }
775
776 while ((por = TAILQ_FIRST(&skiplist->ps_rules))) {
777 TAILQ_REMOVE(&head, por, por_entry);
778 TAILQ_REMOVE(&skiplist->ps_rules, por,
779 por_skip_entry[largest_list]);
780 TAILQ_INSERT_TAIL(&newblock->sb_rules, por,
781 por_entry);
782
783 /* Remove this rule from all other skiplists */
784 remove_from_skipsteps(&block->sb_skipsteps[
785 largest_list], block, por, skiplist);
786 }
787 free(skiplist);
788 if (newblock != block)
789 if (reorder_rules(pf, newblock, depth + 1))
790 return (1);
791 }
792 }
793
794done:
795 for (i = 0; i < PF_SKIP_COUNT; i++) {
796 while ((skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]))) {
797 TAILQ_REMOVE(&block->sb_skipsteps[i], skiplist,
798 ps_entry);
799 free(skiplist);
800 }
801 }
802
803 return (0);
804}
805
806
807/*
808 * Optimization pass #4: re-order 'quick' rules based on feedback from the
809 * currently running ruleset
810 */
811int
812block_feedback(struct pfctl *pf, struct superblock *block)
813{
814 TAILQ_HEAD( , pf_opt_rule) queue;
815 struct pf_opt_rule *por1, *por2;
816 u_int64_t total_count = 0;
817 struct pf_rule a, b;
818
819
820 /*
821 * Walk through all of the profiled superblock's rules and copy
822 * the counters onto our rules.
823 */
824 TAILQ_FOREACH(por1, &block->sb_profiled_block->sb_rules, por_entry) {
825 comparable_rule(&a, &por1->por_rule, DC);
786 total_count += por1->por_rule.packets;
826 total_count += por1->por_rule.packets[0] +
827 por1->por_rule.packets[1];
828 TAILQ_FOREACH(por2, &block->sb_rules, por_entry) {
829 if (por2->por_profile_count)
830 continue;
831 comparable_rule(&b, &por2->por_rule, DC);
832 if (memcmp(&a, &b, sizeof(a)) == 0) {
833 por2->por_profile_count =
793 por1->por_rule.packets;
834 por1->por_rule.packets[0] +
835 por1->por_rule.packets[1];
836 break;
837 }
838 }
839 }
840 superblock_free(pf, block->sb_profiled_block);
841 block->sb_profiled_block = NULL;
842
843 /*
844 * Now we pull all of the rules off the superblock and re-insert them
845 * in sorted order.
846 */
847
848 TAILQ_INIT(&queue);
849 while ((por1 = TAILQ_FIRST(&block->sb_rules)) != NULL) {
850 TAILQ_REMOVE(&block->sb_rules, por1, por_entry);
851 TAILQ_INSERT_TAIL(&queue, por1, por_entry);
852 }
853
854 while ((por1 = TAILQ_FIRST(&queue)) != NULL) {
855 TAILQ_REMOVE(&queue, por1, por_entry);
856/* XXX I should sort all of the unused rules based on skip steps */
857 TAILQ_FOREACH(por2, &block->sb_rules, por_entry) {
858 if (por1->por_profile_count > por2->por_profile_count) {
859 TAILQ_INSERT_BEFORE(por2, por1, por_entry);
860 break;
861 }
862 }
863#ifdef __FreeBSD__
864 if (por2 == NULL)
865#else
866 if (por2 == TAILQ_END(&block->sb_rules))
867#endif
868 TAILQ_INSERT_TAIL(&block->sb_rules, por1, por_entry);
869 }
870
871 return (0);
872}
873
874
875/*
876 * Load the current ruleset from the kernel and try to associate them with
877 * the ruleset we're optimizing.
878 */
879int
880load_feedback_profile(struct pfctl *pf, struct superblocks *superblocks)
881{
882 struct superblock *block, *blockcur;
883 struct superblocks prof_superblocks;
884 struct pf_opt_rule *por;
885 struct pf_opt_queue queue;
886 struct pfioc_rule pr;
887 struct pf_rule a, b;
888 int nr, mnr;
889
890 TAILQ_INIT(&queue);
891 TAILQ_INIT(&prof_superblocks);
892
893 memset(&pr, 0, sizeof(pr));
894 pr.rule.action = PF_PASS;
895 if (ioctl(pf->dev, DIOCGETRULES, &pr)) {
896 warn("DIOCGETRULES");
897 return (1);
898 }
899 mnr = pr.nr;
900
901 DEBUG("Loading %d active rules for a feedback profile", mnr);
902 for (nr = 0; nr < mnr; ++nr) {
903 struct pf_ruleset *rs;
904 if ((por = calloc(1, sizeof(*por))) == NULL) {
905 warn("calloc");
906 return (1);
907 }
908 pr.nr = nr;
909 if (ioctl(pf->dev, DIOCGETRULE, &pr)) {
910 warn("DIOCGETRULES");
911 return (1);
912 }
913 memcpy(&por->por_rule, &pr.rule, sizeof(por->por_rule));
871 strlcpy(por->por_anchor, pr.anchor_call,
872 sizeof(por->por_anchor));
914 rs = pf_find_or_create_ruleset(pr.anchor_call);
915 por->por_rule.anchor = rs->anchor;
916 if (TAILQ_EMPTY(&por->por_rule.rpool.list))
917 memset(&por->por_rule.rpool, 0,
918 sizeof(por->por_rule.rpool));
919 TAILQ_INSERT_TAIL(&queue, por, por_entry);
920
921 /* XXX pfctl_get_pool(pf->dev, &pr.rule.rpool, nr, pr.ticket,
922 * PF_PASS, pf->anchor) ???
923 * ... pfctl_clear_pool(&pr.rule.rpool)
924 */
925 }
926
927 if (construct_superblocks(pf, &queue, &prof_superblocks))
928 return (1);
929
930
931 /*
932 * Now we try to associate the active ruleset's superblocks with
933 * the superblocks we're compiling.
934 */
935 block = TAILQ_FIRST(superblocks);
936 blockcur = TAILQ_FIRST(&prof_superblocks);
937 while (block && blockcur) {
938 comparable_rule(&a, &TAILQ_FIRST(&block->sb_rules)->por_rule,
939 BREAK);
940 comparable_rule(&b, &TAILQ_FIRST(&blockcur->sb_rules)->por_rule,
941 BREAK);
942 if (memcmp(&a, &b, sizeof(a)) == 0) {
943 /* The two superblocks lined up */
944 block->sb_profiled_block = blockcur;
945 } else {
946 DEBUG("superblocks don't line up between #%d and #%d",
947 TAILQ_FIRST(&block->sb_rules)->por_rule.nr,
948 TAILQ_FIRST(&blockcur->sb_rules)->por_rule.nr);
949 break;
950 }
951 block = TAILQ_NEXT(block, sb_entry);
952 blockcur = TAILQ_NEXT(blockcur, sb_entry);
953 }
954
955
956
957 /* Free any superblocks we couldn't link */
958 while (blockcur) {
959 block = TAILQ_NEXT(blockcur, sb_entry);
960 superblock_free(pf, blockcur);
961 blockcur = block;
962 }
963 return (0);
964}
965
966
967/*
968 * Compare a rule to a skiplist to see if the rule is a member
969 */
970int
971skip_compare(int skipnum, struct pf_skip_step *skiplist,
972 struct pf_opt_rule *por)
973{
974 struct pf_rule *a, *b;
975 if (skipnum >= PF_SKIP_COUNT || skipnum < 0)
976 errx(1, "skip_compare() out of bounds");
977 a = &por->por_rule;
978 b = &TAILQ_FIRST(&skiplist->ps_rules)->por_rule;
979
980 return ((skip_comparitors[skipnum])(a, b));
981}
982
983
984/*
985 * Add a rule to a skiplist
986 */
987void
988skip_append(struct superblock *superblock, int skipnum,
989 struct pf_skip_step *skiplist, struct pf_opt_rule *por)
990{
991 struct pf_skip_step *prev;
992
993 skiplist->ps_count++;
994 TAILQ_INSERT_TAIL(&skiplist->ps_rules, por, por_skip_entry[skipnum]);
995
996 /* Keep the list of skiplists sorted by whichever is larger */
997 while ((prev = TAILQ_PREV(skiplist, skiplist, ps_entry)) &&
998 prev->ps_count < skiplist->ps_count) {
999 TAILQ_REMOVE(&superblock->sb_skipsteps[skipnum],
1000 skiplist, ps_entry);
1001 TAILQ_INSERT_BEFORE(prev, skiplist, ps_entry);
1002 }
1003}
1004
1005
1006/*
1007 * Remove a rule from the other skiplist calculations.
1008 */
1009void
1010remove_from_skipsteps(struct skiplist *head, struct superblock *block,
1011 struct pf_opt_rule *por, struct pf_skip_step *active_list)
1012{
1013 struct pf_skip_step *sk, *next;
1014 struct pf_opt_rule *p2;
1015 int i, found;
1016
1017 for (i = 0; i < PF_SKIP_COUNT; i++) {
1018 sk = TAILQ_FIRST(&block->sb_skipsteps[i]);
1019 if (sk == NULL || sk == active_list || sk->ps_count <= 1)
1020 continue;
1021 found = 0;
1022 do {
1023 TAILQ_FOREACH(p2, &sk->ps_rules, por_skip_entry[i])
1024 if (p2 == por) {
1025 TAILQ_REMOVE(&sk->ps_rules, p2,
1026 por_skip_entry[i]);
1027 found = 1;
1028 sk->ps_count--;
1029 break;
1030 }
1031 } while (!found && (sk = TAILQ_NEXT(sk, ps_entry)));
1032 if (found && sk) {
1033 /* Does this change the sorting order? */
1034 while ((next = TAILQ_NEXT(sk, ps_entry)) &&
1035 next->ps_count > sk->ps_count) {
1036 TAILQ_REMOVE(head, sk, ps_entry);
1037 TAILQ_INSERT_AFTER(head, next, sk, ps_entry);
1038 }
1039#ifdef OPT_DEBUG
1040 next = TAILQ_NEXT(sk, ps_entry);
1041 assert(next == NULL || next->ps_count <= sk->ps_count);
1042#endif /* OPT_DEBUG */
1043 }
1044 }
1045}
1046
1047
1048/* Compare two rules AF field for skiplist construction */
1049int
1050skip_cmp_af(struct pf_rule *a, struct pf_rule *b)
1051{
1052 if (a->af != b->af || a->af == 0)
1053 return (1);
1054 return (0);
1055}
1056
1057/* Compare two rules DIRECTION field for skiplist construction */
1058int
1059skip_cmp_dir(struct pf_rule *a, struct pf_rule *b)
1060{
1061 if (a->direction == 0 || a->direction != b->direction)
1062 return (1);
1063 return (0);
1064}
1065
1066/* Compare two rules DST Address field for skiplist construction */
1067int
1068skip_cmp_dst_addr(struct pf_rule *a, struct pf_rule *b)
1069{
1070 if (a->dst.neg != b->dst.neg ||
1071 a->dst.addr.type != b->dst.addr.type)
1072 return (1);
1073 /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0
1074 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP ||
1075 * a->proto == IPPROTO_ICMP
1076 * return (1);
1077 */
1078 switch (a->dst.addr.type) {
1079 case PF_ADDR_ADDRMASK:
1080 if (memcmp(&a->dst.addr.v.a.addr, &b->dst.addr.v.a.addr,
1081 sizeof(a->dst.addr.v.a.addr)) ||
1082 memcmp(&a->dst.addr.v.a.mask, &b->dst.addr.v.a.mask,
1083 sizeof(a->dst.addr.v.a.mask)) ||
1084 (a->dst.addr.v.a.addr.addr32[0] == 0 &&
1085 a->dst.addr.v.a.addr.addr32[1] == 0 &&
1086 a->dst.addr.v.a.addr.addr32[2] == 0 &&
1087 a->dst.addr.v.a.addr.addr32[3] == 0))
1088 return (1);
1089 return (0);
1090 case PF_ADDR_DYNIFTL:
1091 if (strcmp(a->dst.addr.v.ifname, b->dst.addr.v.ifname) != 0 ||
1092 a->dst.addr.iflags != a->dst.addr.iflags ||
1093 memcmp(&a->dst.addr.v.a.mask, &b->dst.addr.v.a.mask,
1094 sizeof(a->dst.addr.v.a.mask)))
1095 return (1);
1096 return (0);
1097 case PF_ADDR_NOROUTE:
1098 case PF_ADDR_URPFFAILED:
1099 return (0);
1100 case PF_ADDR_TABLE:
1101 return (strcmp(a->dst.addr.v.tblname, b->dst.addr.v.tblname));
1102 }
1103 return (1);
1104}
1105
1106/* Compare two rules DST port field for skiplist construction */
1107int
1108skip_cmp_dst_port(struct pf_rule *a, struct pf_rule *b)
1109{
1110 /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0
1111 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP ||
1112 * a->proto == IPPROTO_ICMP
1113 * return (1);
1114 */
1115 if (a->dst.port_op == PF_OP_NONE || a->dst.port_op != b->dst.port_op ||
1116 a->dst.port[0] != b->dst.port[0] ||
1117 a->dst.port[1] != b->dst.port[1])
1118 return (1);
1119 return (0);
1120}
1121
1122/* Compare two rules IFP field for skiplist construction */
1123int
1124skip_cmp_ifp(struct pf_rule *a, struct pf_rule *b)
1125{
1126 if (strcmp(a->ifname, b->ifname) || a->ifname[0] == '\0')
1127 return (1);
1128 return (a->ifnot != b->ifnot);
1129}
1130
1131/* Compare two rules PROTO field for skiplist construction */
1132int
1133skip_cmp_proto(struct pf_rule *a, struct pf_rule *b)
1134{
1135 return (a->proto != b->proto || a->proto == 0);
1136}
1137
1138/* Compare two rules SRC addr field for skiplist construction */
1139int
1140skip_cmp_src_addr(struct pf_rule *a, struct pf_rule *b)
1141{
1142 if (a->src.neg != b->src.neg ||
1143 a->src.addr.type != b->src.addr.type)
1144 return (1);
1145 /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0
1146 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP ||
1147 * a->proto == IPPROTO_ICMP
1148 * return (1);
1149 */
1150 switch (a->src.addr.type) {
1151 case PF_ADDR_ADDRMASK:
1152 if (memcmp(&a->src.addr.v.a.addr, &b->src.addr.v.a.addr,
1153 sizeof(a->src.addr.v.a.addr)) ||
1154 memcmp(&a->src.addr.v.a.mask, &b->src.addr.v.a.mask,
1155 sizeof(a->src.addr.v.a.mask)) ||
1156 (a->src.addr.v.a.addr.addr32[0] == 0 &&
1157 a->src.addr.v.a.addr.addr32[1] == 0 &&
1158 a->src.addr.v.a.addr.addr32[2] == 0 &&
1159 a->src.addr.v.a.addr.addr32[3] == 0))
1160 return (1);
1161 return (0);
1162 case PF_ADDR_DYNIFTL:
1163 if (strcmp(a->src.addr.v.ifname, b->src.addr.v.ifname) != 0 ||
1164 a->src.addr.iflags != a->src.addr.iflags ||
1165 memcmp(&a->src.addr.v.a.mask, &b->src.addr.v.a.mask,
1166 sizeof(a->src.addr.v.a.mask)))
1167 return (1);
1168 return (0);
1169 case PF_ADDR_NOROUTE:
1170 case PF_ADDR_URPFFAILED:
1171 return (0);
1172 case PF_ADDR_TABLE:
1173 return (strcmp(a->src.addr.v.tblname, b->src.addr.v.tblname));
1174 }
1175 return (1);
1176}
1177
1178/* Compare two rules SRC port field for skiplist construction */
1179int
1180skip_cmp_src_port(struct pf_rule *a, struct pf_rule *b)
1181{
1182 if (a->src.port_op == PF_OP_NONE || a->src.port_op != b->src.port_op ||
1183 a->src.port[0] != b->src.port[0] ||
1184 a->src.port[1] != b->src.port[1])
1185 return (1);
1186 /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0
1187 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP ||
1188 * a->proto == IPPROTO_ICMP
1189 * return (1);
1190 */
1191 return (0);
1192}
1193
1194
1195void
1196skip_init(void)
1197{
1198 struct {
1199 char *name;
1200 int skipnum;
1201 int (*func)(struct pf_rule *, struct pf_rule *);
1202 } comps[] = PF_SKIP_COMPARITORS;
1203 int skipnum, i;
1204
1205 for (skipnum = 0; skipnum < PF_SKIP_COUNT; skipnum++) {
1206 for (i = 0; i < sizeof(comps)/sizeof(*comps); i++)
1207 if (comps[i].skipnum == skipnum) {
1208 skip_comparitors[skipnum] = comps[i].func;
1209 skip_comparitors_names[skipnum] = comps[i].name;
1210 }
1211 }
1212 for (skipnum = 0; skipnum < PF_SKIP_COUNT; skipnum++)
1213 if (skip_comparitors[skipnum] == NULL)
1214 errx(1, "Need to add skip step comparitor to pfctl?!");
1215}
1216
1217/*
1218 * Add a host/netmask to a table
1219 */
1220int
1221add_opt_table(struct pfctl *pf, struct pf_opt_tbl **tbl, sa_family_t af,
1222 struct pf_rule_addr *addr)
1223{
1224#ifdef OPT_DEBUG
1225 char buf[128];
1226#endif /* OPT_DEBUG */
1227 static int tablenum = 0;
1228 struct node_host node_host;
1229
1230 if (*tbl == NULL) {
1231 if ((*tbl = calloc(1, sizeof(**tbl))) == NULL ||
1232 ((*tbl)->pt_buf = calloc(1, sizeof(*(*tbl)->pt_buf))) ==
1233 NULL)
1234 err(1, "calloc");
1235 (*tbl)->pt_buf->pfrb_type = PFRB_ADDRS;
1236 SIMPLEQ_INIT(&(*tbl)->pt_nodes);
1237
1238 /* This is just a temporary table name */
1239 snprintf((*tbl)->pt_name, sizeof((*tbl)->pt_name), "%s%d",
1240 PF_OPT_TABLE_PREFIX, tablenum++);
1241 DEBUG("creating table <%s>", (*tbl)->pt_name);
1242 }
1243
1244 memset(&node_host, 0, sizeof(node_host));
1245 node_host.af = af;
1246 node_host.addr = addr->addr;
1247
1248#ifdef OPT_DEBUG
1249 DEBUG("<%s> adding %s/%d", (*tbl)->pt_name, inet_ntop(af,
1250 &node_host.addr.v.a.addr, buf, sizeof(buf)),
1251 unmask(&node_host.addr.v.a.mask, af));
1252#endif /* OPT_DEBUG */
1253
1254 if (append_addr_host((*tbl)->pt_buf, &node_host, 0, 0)) {
1255 warn("failed to add host");
1256 return (1);
1257 }
1258 if (pf->opts & PF_OPT_VERBOSE) {
1259 struct node_tinit *ti;
1260
1261 if ((ti = calloc(1, sizeof(*ti))) == NULL)
1262 err(1, "malloc");
1263 if ((ti->host = malloc(sizeof(*ti->host))) == NULL)
1264 err(1, "malloc");
1265 memcpy(ti->host, &node_host, sizeof(*ti->host));
1266 SIMPLEQ_INSERT_TAIL(&(*tbl)->pt_nodes, ti, entries);
1267 }
1268
1269 (*tbl)->pt_rulecount++;
1270 if ((*tbl)->pt_rulecount == TABLE_THRESHOLD)
1271 DEBUG("table <%s> now faster than skip steps", (*tbl)->pt_name);
1272
1273 return (0);
1274}
1275
1276
1277/*
1278 * Do the dirty work of choosing an unused table name and creating it.
1279 * (be careful with the table name, it might already be used in another anchor)
1280 */
1281int
1282pf_opt_create_table(struct pfctl *pf, struct pf_opt_tbl *tbl)
1283{
1284 static int tablenum;
1285 struct pfr_table *t;
1286
1287 if (table_buffer.pfrb_type == 0) {
1288 /* Initialize the list of tables */
1289 table_buffer.pfrb_type = PFRB_TABLES;
1290 for (;;) {
1291 pfr_buf_grow(&table_buffer, table_buffer.pfrb_size);
1292 table_buffer.pfrb_size = table_buffer.pfrb_msize;
1293 if (pfr_get_tables(NULL, table_buffer.pfrb_caddr,
1294 &table_buffer.pfrb_size, PFR_FLAG_ALLRSETS))
1295 err(1, "pfr_get_tables");
1296 if (table_buffer.pfrb_size <= table_buffer.pfrb_msize)
1297 break;
1298 }
1299 table_identifier = arc4random();
1300 }
1301
1302 /* XXX would be *really* nice to avoid duplicating identical tables */
1303
1304 /* Now we have to pick a table name that isn't used */
1305again:
1306 DEBUG("translating temporary table <%s> to <%s%x_%d>", tbl->pt_name,
1307 PF_OPT_TABLE_PREFIX, table_identifier, tablenum);
1308 snprintf(tbl->pt_name, sizeof(tbl->pt_name), "%s%x_%d",
1309 PF_OPT_TABLE_PREFIX, table_identifier, tablenum);
1310 PFRB_FOREACH(t, &table_buffer) {
1311 if (strcasecmp(t->pfrt_name, tbl->pt_name) == 0) {
1312 /* Collision. Try again */
1313 DEBUG("wow, table <%s> in use. trying again",
1314 tbl->pt_name);
1315 table_identifier = arc4random();
1316 goto again;
1317 }
1318 }
1319 tablenum++;
1320
1321
1277 if (pfctl_define_table(tbl->pt_name, PFR_TFLAG_CONST, 1, pf->anchor,
1278 tbl->pt_buf, pf->tticket)) {
1322 if (pfctl_define_table(tbl->pt_name, PFR_TFLAG_CONST, 1,
1323 pf->anchor->name, tbl->pt_buf, pf->anchor->ruleset.tticket)) {
1324 warn("failed to create table %s", tbl->pt_name);
1325 return (1);
1326 }
1327 return (0);
1328}
1329
1330/*
1331 * Partition the flat ruleset into a list of distinct superblocks
1332 */
1333int
1334construct_superblocks(struct pfctl *pf, struct pf_opt_queue *opt_queue,
1335 struct superblocks *superblocks)
1336{
1337 struct superblock *block = NULL;
1338 struct pf_opt_rule *por;
1339 int i;
1340
1341 while (!TAILQ_EMPTY(opt_queue)) {
1342 por = TAILQ_FIRST(opt_queue);
1343 TAILQ_REMOVE(opt_queue, por, por_entry);
1344 if (block == NULL || !superblock_inclusive(block, por)) {
1345 if ((block = calloc(1, sizeof(*block))) == NULL) {
1346 warn("calloc");
1347 return (1);
1348 }
1349 TAILQ_INIT(&block->sb_rules);
1350 for (i = 0; i < PF_SKIP_COUNT; i++)
1351 TAILQ_INIT(&block->sb_skipsteps[i]);
1352 TAILQ_INSERT_TAIL(superblocks, block, sb_entry);
1353 }
1354 TAILQ_INSERT_TAIL(&block->sb_rules, por, por_entry);
1355 }
1356
1357 return (0);
1358}
1359
1360
1361/*
1362 * Compare two rule addresses
1363 */
1364int
1365addrs_equal(struct pf_rule_addr *a, struct pf_rule_addr *b)
1366{
1367 if (a->neg != b->neg)
1368 return (0);
1369 return (memcmp(&a->addr, &b->addr, sizeof(a->addr)) == 0);
1370}
1371
1372
1373/*
1374 * The addresses are not equal, but can we combine them into one table?
1375 */
1376int
1377addrs_combineable(struct pf_rule_addr *a, struct pf_rule_addr *b)
1378{
1379 if (a->addr.type != PF_ADDR_ADDRMASK ||
1380 b->addr.type != PF_ADDR_ADDRMASK)
1381 return (0);
1382 if (a->neg != b->neg || a->port_op != b->port_op ||
1383 a->port[0] != b->port[0] || a->port[1] != b->port[1])
1384 return (0);
1385 return (1);
1386}
1387
1388
1389/*
1390 * Are we allowed to combine these two rules
1391 */
1392int
1393rules_combineable(struct pf_rule *p1, struct pf_rule *p2)
1394{
1395 struct pf_rule a, b;
1396
1397 comparable_rule(&a, p1, COMBINED);
1398 comparable_rule(&b, p2, COMBINED);
1399 return (memcmp(&a, &b, sizeof(a)) == 0);
1400}
1401
1402
1403/*
1404 * Can a rule be included inside a superblock
1405 */
1406int
1407superblock_inclusive(struct superblock *block, struct pf_opt_rule *por)
1408{
1409 struct pf_rule a, b;
1410 int i, j;
1411
1412 /* First check for hard breaks */
1413 for (i = 0; i < sizeof(pf_rule_desc)/sizeof(*pf_rule_desc); i++) {
1414 if (pf_rule_desc[i].prf_type == BARRIER) {
1415 for (j = 0; j < pf_rule_desc[i].prf_size; j++)
1416 if (((char *)&por->por_rule)[j +
1417 pf_rule_desc[i].prf_offset] != 0)
1418 return (0);
1419 }
1420 }
1421
1377 /* 'anchor' heads and per-rule src-track are also hard breaks */
1378 if (por->por_anchor[0] != '\0' ||
1379 (por->por_rule.rule_flag & PFRULE_RULESRCTRACK))
1422 /* per-rule src-track is also a hard break */
1423 if (por->por_rule.rule_flag & PFRULE_RULESRCTRACK)
1424 return (0);
1425
1426 /*
1427 * Have to handle interface groups seperately. Consider the following
1428 * rules:
1429 * block on EXTIFS to any port 22
1430 * pass on em0 to any port 22
1431 * (where EXTIFS is an arbitrary interface group)
1432 * The optimizer may decide to re-order the pass rule in front of the
1433 * block rule. But what if EXTIFS includes em0??? Such a reordering
1434 * would change the meaning of the ruleset.
1435 * We can't just lookup the EXTIFS group and check if em0 is a member
1436 * because the user is allowed to add interfaces to a group during
1437 * runtime.
1438 * Ergo interface groups become a defacto superblock break :-(
1439 */
1440 if (interface_group(por->por_rule.ifname) ||
1441 interface_group(TAILQ_FIRST(&block->sb_rules)->por_rule.ifname)) {
1442 if (strcasecmp(por->por_rule.ifname,
1443 TAILQ_FIRST(&block->sb_rules)->por_rule.ifname) != 0)
1444 return (0);
1445 }
1446
1447 comparable_rule(&a, &TAILQ_FIRST(&block->sb_rules)->por_rule, NOMERGE);
1448 comparable_rule(&b, &por->por_rule, NOMERGE);
1384 if (strcmp(TAILQ_FIRST(&block->sb_rules)->por_anchor,
1385 por->por_anchor) == 0 && memcmp(&a, &b, sizeof(a)) == 0)
1449 if (memcmp(&a, &b, sizeof(a)) == 0)
1450 return (1);
1451
1452#ifdef OPT_DEBUG
1453 for (i = 0; i < sizeof(por->por_rule); i++) {
1454 int closest = -1;
1455 if (((u_int8_t *)&a)[i] != ((u_int8_t *)&b)[i]) {
1456 for (j = 0; j < sizeof(pf_rule_desc) /
1457 sizeof(*pf_rule_desc); j++) {
1458 if (i >= pf_rule_desc[j].prf_offset &&
1459 i < pf_rule_desc[j].prf_offset +
1460 pf_rule_desc[j].prf_size) {
1461 DEBUG("superblock break @ %d due to %s",
1462 por->por_rule.nr,
1463 pf_rule_desc[j].prf_name);
1464 return (0);
1465 }
1466 if (i > pf_rule_desc[j].prf_offset) {
1467 if (closest == -1 ||
1468 i-pf_rule_desc[j].prf_offset <
1469 i-pf_rule_desc[closest].prf_offset)
1470 closest = j;
1471 }
1472 }
1473
1474 if (closest >= 0)
1475 DEBUG("superblock break @ %d on %s+%xh",
1476 por->por_rule.nr,
1477 pf_rule_desc[closest].prf_name,
1478 i - pf_rule_desc[closest].prf_offset -
1479 pf_rule_desc[closest].prf_size);
1480 else
1481 DEBUG("superblock break @ %d on field @ %d",
1482 por->por_rule.nr, i);
1483 return (0);
1484 }
1485 }
1486#endif /* OPT_DEBUG */
1487
1488 return (0);
1489}
1490
1491
1492/*
1493 * Figure out if an interface name is an actual interface or actually a
1494 * group of interfaces.
1495 */
1496int
1497interface_group(const char *ifname)
1498{
1499 if (ifname == NULL || !ifname[0])
1500 return (0);
1501
1502 /* Real interfaces must end in a number, interface groups do not */
1503 if (isdigit(ifname[strlen(ifname) - 1]))
1504 return (0);
1505 else
1506 return (1);
1507}
1508
1509
1510/*
1511 * Make a rule that can directly compared by memcmp()
1512 */
1513void
1514comparable_rule(struct pf_rule *dst, const struct pf_rule *src, int type)
1515{
1516 int i;
1517 /*
1518 * To simplify the comparison, we just zero out the fields that are
1519 * allowed to be different and then do a simple memcmp()
1520 */
1521 memcpy(dst, src, sizeof(*dst));
1522 for (i = 0; i < sizeof(pf_rule_desc)/sizeof(*pf_rule_desc); i++)
1523 if (pf_rule_desc[i].prf_type >= type) {
1524#ifdef OPT_DEBUG
1525 assert(pf_rule_desc[i].prf_type != NEVER ||
1526 *(((char *)dst) + pf_rule_desc[i].prf_offset) == 0);
1527#endif /* OPT_DEBUG */
1528 memset(((char *)dst) + pf_rule_desc[i].prf_offset, 0,
1529 pf_rule_desc[i].prf_size);
1530 }
1531}
1532
1533
1534/*
1535 * Remove superset information from two rules so we can directly compare them
1536 * with memcmp()
1537 */
1538void
1539exclude_supersets(struct pf_rule *super, struct pf_rule *sub)
1540{
1541 if (super->ifname[0] == '\0')
1542 memset(sub->ifname, 0, sizeof(sub->ifname));
1543 if (super->direction == PF_INOUT)
1544 sub->direction = PF_INOUT;
1545 if ((super->proto == 0 || super->proto == sub->proto) &&
1546 super->flags == 0 && super->flagset == 0 && (sub->flags ||
1547 sub->flagset)) {
1548 sub->flags = super->flags;
1549 sub->flagset = super->flagset;
1550 }
1551 if (super->proto == 0)
1552 sub->proto = 0;
1553
1554 if (super->src.port_op == 0) {
1555 sub->src.port_op = 0;
1556 sub->src.port[0] = 0;
1557 sub->src.port[1] = 0;
1558 }
1559 if (super->dst.port_op == 0) {
1560 sub->dst.port_op = 0;
1561 sub->dst.port[0] = 0;
1562 sub->dst.port[1] = 0;
1563 }
1564
1565 if (super->src.addr.type == PF_ADDR_ADDRMASK && !super->src.neg &&
1566 !sub->src.neg && super->src.addr.v.a.mask.addr32[0] == 0 &&
1567 super->src.addr.v.a.mask.addr32[1] == 0 &&
1568 super->src.addr.v.a.mask.addr32[2] == 0 &&
1569 super->src.addr.v.a.mask.addr32[3] == 0)
1570 memset(&sub->src.addr, 0, sizeof(sub->src.addr));
1571 else if (super->src.addr.type == PF_ADDR_ADDRMASK &&
1572 sub->src.addr.type == PF_ADDR_ADDRMASK &&
1573 super->src.neg == sub->src.neg &&
1574 super->af == sub->af &&
1575 unmask(&super->src.addr.v.a.mask, super->af) <
1576 unmask(&sub->src.addr.v.a.mask, sub->af) &&
1577 super->src.addr.v.a.addr.addr32[0] ==
1578 (sub->src.addr.v.a.addr.addr32[0] &
1579 super->src.addr.v.a.mask.addr32[0]) &&
1580 super->src.addr.v.a.addr.addr32[1] ==
1581 (sub->src.addr.v.a.addr.addr32[1] &
1582 super->src.addr.v.a.mask.addr32[1]) &&
1583 super->src.addr.v.a.addr.addr32[2] ==
1584 (sub->src.addr.v.a.addr.addr32[2] &
1585 super->src.addr.v.a.mask.addr32[2]) &&
1586 super->src.addr.v.a.addr.addr32[3] ==
1587 (sub->src.addr.v.a.addr.addr32[3] &
1588 super->src.addr.v.a.mask.addr32[3])) {
1589 /* sub->src.addr is a subset of super->src.addr/mask */
1590 memcpy(&sub->src.addr, &super->src.addr, sizeof(sub->src.addr));
1591 }
1592
1593 if (super->dst.addr.type == PF_ADDR_ADDRMASK && !super->dst.neg &&
1594 !sub->dst.neg && super->dst.addr.v.a.mask.addr32[0] == 0 &&
1595 super->dst.addr.v.a.mask.addr32[1] == 0 &&
1596 super->dst.addr.v.a.mask.addr32[2] == 0 &&
1597 super->dst.addr.v.a.mask.addr32[3] == 0)
1598 memset(&sub->dst.addr, 0, sizeof(sub->dst.addr));
1599 else if (super->dst.addr.type == PF_ADDR_ADDRMASK &&
1600 sub->dst.addr.type == PF_ADDR_ADDRMASK &&
1601 super->dst.neg == sub->dst.neg &&
1602 super->af == sub->af &&
1603 unmask(&super->dst.addr.v.a.mask, super->af) <
1604 unmask(&sub->dst.addr.v.a.mask, sub->af) &&
1605 super->dst.addr.v.a.addr.addr32[0] ==
1606 (sub->dst.addr.v.a.addr.addr32[0] &
1607 super->dst.addr.v.a.mask.addr32[0]) &&
1608 super->dst.addr.v.a.addr.addr32[1] ==
1609 (sub->dst.addr.v.a.addr.addr32[1] &
1610 super->dst.addr.v.a.mask.addr32[1]) &&
1611 super->dst.addr.v.a.addr.addr32[2] ==
1612 (sub->dst.addr.v.a.addr.addr32[2] &
1613 super->dst.addr.v.a.mask.addr32[2]) &&
1614 super->dst.addr.v.a.addr.addr32[3] ==
1615 (sub->dst.addr.v.a.addr.addr32[3] &
1616 super->dst.addr.v.a.mask.addr32[3])) {
1617 /* sub->dst.addr is a subset of super->dst.addr/mask */
1618 memcpy(&sub->dst.addr, &super->dst.addr, sizeof(sub->dst.addr));
1619 }
1620
1621 if (super->af == 0)
1622 sub->af = 0;
1623}
1624
1625
1626void
1627superblock_free(struct pfctl *pf, struct superblock *block)
1628{
1629 struct pf_opt_rule *por;
1630 while ((por = TAILQ_FIRST(&block->sb_rules))) {
1631 TAILQ_REMOVE(&block->sb_rules, por, por_entry);
1632 if (por->por_src_tbl) {
1633 if (por->por_src_tbl->pt_buf) {
1634 pfr_buf_clear(por->por_src_tbl->pt_buf);
1635 free(por->por_src_tbl->pt_buf);
1636 }
1637 free(por->por_src_tbl);
1638 }
1639 if (por->por_dst_tbl) {
1640 if (por->por_dst_tbl->pt_buf) {
1641 pfr_buf_clear(por->por_dst_tbl->pt_buf);
1642 free(por->por_dst_tbl->pt_buf);
1643 }
1644 free(por->por_dst_tbl);
1645 }
1646 free(por);
1647 }
1648 if (block->sb_profiled_block)
1649 superblock_free(pf, block->sb_profiled_block);
1650 free(block);
1651}
1652