xref: /freebsd/sys/netpfil/ipfw/ip_fw_table.c (revision a0ee8cc636cd5c2374ec44ca71226564ea0bca95)
1 /*-
2  * Copyright (c) 2004 Ruslan Ermilov and Vsevolod Lobko.
3  * Copyright (c) 2014 Yandex LLC
4  * Copyright (c) 2014 Alexander V. Chernikov
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25  * SUCH DAMAGE.
26  */
27 
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
30 
31 /*
32  * Lookup table support for ipfw.
33  *
34  * This file contains handlers for all generic tables' operations:
35  * add/del/flush entries, list/dump tables etc..
36  *
37  * Table data modification is protected by both UH and runtime lock
38  * while reading configuration/data is protected by UH lock.
39  *
40  * Lookup algorithms for all table types are located in ip_fw_table_algo.c
41  */
42 
43 #include "opt_ipfw.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/malloc.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/rwlock.h>
51 #include <sys/rmlock.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/queue.h>
55 #include <net/if.h>	/* ip_fw.h requires IFNAMSIZ */
56 
57 #include <netinet/in.h>
58 #include <netinet/ip_var.h>	/* struct ipfw_rule_ref */
59 #include <netinet/ip_fw.h>
60 
61 #include <netpfil/ipfw/ip_fw_private.h>
62 #include <netpfil/ipfw/ip_fw_table.h>
63 
64  /*
65  * Table has the following `type` concepts:
66  *
67  * `no.type` represents lookup key type (addr, ifp, uid, etc..)
68  * vmask represents bitmask of table values which are present at the moment.
69  * Special IPFW_VTYPE_LEGACY ( (uint32_t)-1 ) represents old
70  * single-value-for-all approach.
71  */
72 struct table_config {
73 	struct named_object	no;
74 	uint8_t		tflags;		/* type flags */
75 	uint8_t		locked;		/* 1 if locked from changes */
76 	uint8_t		linked;		/* 1 if already linked */
77 	uint8_t		ochanged;	/* used by set swapping */
78 	uint8_t		vshared;	/* 1 if using shared value array */
79 	uint8_t		spare[3];
80 	uint32_t	count;		/* Number of records */
81 	uint32_t	limit;		/* Max number of records */
82 	uint32_t	vmask;		/* bitmask with supported values */
83 	uint32_t	ocount;		/* used by set swapping */
84 	uint64_t	gencnt;		/* generation count */
85 	char		tablename[64];	/* table name */
86 	struct table_algo	*ta;	/* Callbacks for given algo */
87 	void		*astate;	/* algorithm state */
88 	struct table_info	ti_copy;	/* data to put to table_info */
89 	struct namedobj_instance	*vi;
90 };
91 
92 static int find_table_err(struct namedobj_instance *ni, struct tid_info *ti,
93     struct table_config **tc);
94 static struct table_config *find_table(struct namedobj_instance *ni,
95     struct tid_info *ti);
96 static struct table_config *alloc_table_config(struct ip_fw_chain *ch,
97     struct tid_info *ti, struct table_algo *ta, char *adata, uint8_t tflags);
98 static void free_table_config(struct namedobj_instance *ni,
99     struct table_config *tc);
100 static int create_table_internal(struct ip_fw_chain *ch, struct tid_info *ti,
101     char *aname, ipfw_xtable_info *i, uint16_t *pkidx, int ref);
102 static void link_table(struct ip_fw_chain *ch, struct table_config *tc);
103 static void unlink_table(struct ip_fw_chain *ch, struct table_config *tc);
104 static int find_ref_table(struct ip_fw_chain *ch, struct tid_info *ti,
105     struct tentry_info *tei, uint32_t count, int op, struct table_config **ptc);
106 #define	OP_ADD	1
107 #define	OP_DEL	0
108 static int export_tables(struct ip_fw_chain *ch, ipfw_obj_lheader *olh,
109     struct sockopt_data *sd);
110 static void export_table_info(struct ip_fw_chain *ch, struct table_config *tc,
111     ipfw_xtable_info *i);
112 static int dump_table_tentry(void *e, void *arg);
113 static int dump_table_xentry(void *e, void *arg);
114 
115 static int swap_tables(struct ip_fw_chain *ch, struct tid_info *a,
116     struct tid_info *b);
117 
118 static int check_table_name(const char *name);
119 static int check_table_space(struct ip_fw_chain *ch, struct tableop_state *ts,
120     struct table_config *tc, struct table_info *ti, uint32_t count);
121 static int destroy_table(struct ip_fw_chain *ch, struct tid_info *ti);
122 
123 static struct table_algo *find_table_algo(struct tables_config *tableconf,
124     struct tid_info *ti, char *name);
125 
126 static void objheader_to_ti(struct _ipfw_obj_header *oh, struct tid_info *ti);
127 static void ntlv_to_ti(struct _ipfw_obj_ntlv *ntlv, struct tid_info *ti);
128 
129 #define	CHAIN_TO_NI(chain)	(CHAIN_TO_TCFG(chain)->namehash)
130 #define	KIDX_TO_TI(ch, k)	(&(((struct table_info *)(ch)->tablestate)[k]))
131 
132 #define	TA_BUF_SZ	128	/* On-stack buffer for add/delete state */
133 
134 void
135 rollback_toperation_state(struct ip_fw_chain *ch, void *object)
136 {
137 	struct tables_config *tcfg;
138 	struct op_state *os;
139 
140 	tcfg = CHAIN_TO_TCFG(ch);
141 	TAILQ_FOREACH(os, &tcfg->state_list, next)
142 		os->func(object, os);
143 }
144 
145 void
146 add_toperation_state(struct ip_fw_chain *ch, struct tableop_state *ts)
147 {
148 	struct tables_config *tcfg;
149 
150 	tcfg = CHAIN_TO_TCFG(ch);
151 	TAILQ_INSERT_HEAD(&tcfg->state_list, &ts->opstate, next);
152 }
153 
154 void
155 del_toperation_state(struct ip_fw_chain *ch, struct tableop_state *ts)
156 {
157 	struct tables_config *tcfg;
158 
159 	tcfg = CHAIN_TO_TCFG(ch);
160 	TAILQ_REMOVE(&tcfg->state_list, &ts->opstate, next);
161 }
162 
163 void
164 tc_ref(struct table_config *tc)
165 {
166 
167 	tc->no.refcnt++;
168 }
169 
170 void
171 tc_unref(struct table_config *tc)
172 {
173 
174 	tc->no.refcnt--;
175 }
176 
177 static struct table_value *
178 get_table_value(struct ip_fw_chain *ch, struct table_config *tc, uint32_t kidx)
179 {
180 	struct table_value *pval;
181 
182 	pval = (struct table_value *)ch->valuestate;
183 
184 	return (&pval[kidx]);
185 }
186 
187 
188 /*
189  * Checks if we're able to insert/update entry @tei into table
190  * w.r.t @tc limits.
191  * May alter @tei to indicate insertion error / insert
192  * options.
193  *
194  * Returns 0 if operation can be performed/
195  */
196 static int
197 check_table_limit(struct table_config *tc, struct tentry_info *tei)
198 {
199 
200 	if (tc->limit == 0 || tc->count < tc->limit)
201 		return (0);
202 
203 	if ((tei->flags & TEI_FLAGS_UPDATE) == 0) {
204 		/* Notify userland on error cause */
205 		tei->flags |= TEI_FLAGS_LIMIT;
206 		return (EFBIG);
207 	}
208 
209 	/*
210 	 * We have UPDATE flag set.
211 	 * Permit updating record (if found),
212 	 * but restrict adding new one since we've
213 	 * already hit the limit.
214 	 */
215 	tei->flags |= TEI_FLAGS_DONTADD;
216 
217 	return (0);
218 }
219 
220 /*
221  * Convert algorithm callback return code into
222  * one of pre-defined states known by userland.
223  */
224 static void
225 store_tei_result(struct tentry_info *tei, int op, int error, uint32_t num)
226 {
227 	int flag;
228 
229 	flag = 0;
230 
231 	switch (error) {
232 	case 0:
233 		if (op == OP_ADD && num != 0)
234 			flag = TEI_FLAGS_ADDED;
235 		if (op == OP_DEL)
236 			flag = TEI_FLAGS_DELETED;
237 		break;
238 	case ENOENT:
239 		flag = TEI_FLAGS_NOTFOUND;
240 		break;
241 	case EEXIST:
242 		flag = TEI_FLAGS_EXISTS;
243 		break;
244 	default:
245 		flag = TEI_FLAGS_ERROR;
246 	}
247 
248 	tei->flags |= flag;
249 }
250 
251 /*
252  * Creates and references table with default parameters.
253  * Saves table config, algo and allocated kidx info @ptc, @pta and
254  * @pkidx if non-zero.
255  * Used for table auto-creation to support old binaries.
256  *
257  * Returns 0 on success.
258  */
259 static int
260 create_table_compat(struct ip_fw_chain *ch, struct tid_info *ti,
261     uint16_t *pkidx)
262 {
263 	ipfw_xtable_info xi;
264 	int error;
265 
266 	memset(&xi, 0, sizeof(xi));
267 	/* Set default value mask for legacy clients */
268 	xi.vmask = IPFW_VTYPE_LEGACY;
269 
270 	error = create_table_internal(ch, ti, NULL, &xi, pkidx, 1);
271 	if (error != 0)
272 		return (error);
273 
274 	return (0);
275 }
276 
277 /*
278  * Find and reference existing table optionally
279  * creating new one.
280  *
281  * Saves found table config into @ptc.
282  * Note function may drop/acquire UH_WLOCK.
283  * Returns 0 if table was found/created and referenced
284  * or non-zero return code.
285  */
286 static int
287 find_ref_table(struct ip_fw_chain *ch, struct tid_info *ti,
288     struct tentry_info *tei, uint32_t count, int op,
289     struct table_config **ptc)
290 {
291 	struct namedobj_instance *ni;
292 	struct table_config *tc;
293 	uint16_t kidx;
294 	int error;
295 
296 	IPFW_UH_WLOCK_ASSERT(ch);
297 
298 	ni = CHAIN_TO_NI(ch);
299 	tc = NULL;
300 	if ((tc = find_table(ni, ti)) != NULL) {
301 		/* check table type */
302 		if (tc->no.subtype != ti->type)
303 			return (EINVAL);
304 
305 		if (tc->locked != 0)
306 			return (EACCES);
307 
308 		/* Try to exit early on limit hit */
309 		if (op == OP_ADD && count == 1 &&
310 		    check_table_limit(tc, tei) != 0)
311 			return (EFBIG);
312 
313 		/* Reference and return */
314 		tc->no.refcnt++;
315 		*ptc = tc;
316 		return (0);
317 	}
318 
319 	if (op == OP_DEL)
320 		return (ESRCH);
321 
322 	/* Compability mode: create new table for old clients */
323 	if ((tei->flags & TEI_FLAGS_COMPAT) == 0)
324 		return (ESRCH);
325 
326 	IPFW_UH_WUNLOCK(ch);
327 	error = create_table_compat(ch, ti, &kidx);
328 	IPFW_UH_WLOCK(ch);
329 
330 	if (error != 0)
331 		return (error);
332 
333 	tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, kidx);
334 	KASSERT(tc != NULL, ("create_table_compat returned bad idx %d", kidx));
335 
336 	/* OK, now we've got referenced table. */
337 	*ptc = tc;
338 	return (0);
339 }
340 
341 /*
342  * Rolls back already @added to @tc entries using state array @ta_buf_m.
343  * Assume the following layout:
344  * 1) ADD state (ta_buf_m[0] ... t_buf_m[added - 1]) for handling update cases
345  * 2) DEL state (ta_buf_m[count[ ... t_buf_m[count + added - 1])
346  *   for storing deleted state
347  */
348 static void
349 rollback_added_entries(struct ip_fw_chain *ch, struct table_config *tc,
350     struct table_info *tinfo, struct tentry_info *tei, caddr_t ta_buf_m,
351     uint32_t count, uint32_t added)
352 {
353 	struct table_algo *ta;
354 	struct tentry_info *ptei;
355 	caddr_t v, vv;
356 	size_t ta_buf_sz;
357 	int error, i;
358 	uint32_t num;
359 
360 	IPFW_UH_WLOCK_ASSERT(ch);
361 
362 	ta = tc->ta;
363 	ta_buf_sz = ta->ta_buf_size;
364 	v = ta_buf_m;
365 	vv = v + count * ta_buf_sz;
366 	for (i = 0; i < added; i++, v += ta_buf_sz, vv += ta_buf_sz) {
367 		ptei = &tei[i];
368 		if ((ptei->flags & TEI_FLAGS_UPDATED) != 0) {
369 
370 			/*
371 			 * We have old value stored by previous
372 			 * call in @ptei->value. Do add once again
373 			 * to restore it.
374 			 */
375 			error = ta->add(tc->astate, tinfo, ptei, v, &num);
376 			KASSERT(error == 0, ("rollback UPDATE fail"));
377 			KASSERT(num == 0, ("rollback UPDATE fail2"));
378 			continue;
379 		}
380 
381 		error = ta->prepare_del(ch, ptei, vv);
382 		KASSERT(error == 0, ("pre-rollback INSERT failed"));
383 		error = ta->del(tc->astate, tinfo, ptei, vv, &num);
384 		KASSERT(error == 0, ("rollback INSERT failed"));
385 		tc->count -= num;
386 	}
387 }
388 
389 /*
390  * Prepares add/del state for all @count entries in @tei.
391  * Uses either stack buffer (@ta_buf) or allocates a new one.
392  * Stores pointer to allocated buffer back to @ta_buf.
393  *
394  * Returns 0 on success.
395  */
396 static int
397 prepare_batch_buffer(struct ip_fw_chain *ch, struct table_algo *ta,
398     struct tentry_info *tei, uint32_t count, int op, caddr_t *ta_buf)
399 {
400 	caddr_t ta_buf_m, v;
401 	size_t ta_buf_sz, sz;
402 	struct tentry_info *ptei;
403 	int error, i;
404 
405 	error = 0;
406 	ta_buf_sz = ta->ta_buf_size;
407 	if (count == 1) {
408 		/* Sigle add/delete, use on-stack buffer */
409 		memset(*ta_buf, 0, TA_BUF_SZ);
410 		ta_buf_m = *ta_buf;
411 	} else {
412 
413 		/*
414 		 * Multiple adds/deletes, allocate larger buffer
415 		 *
416 		 * Note we need 2xcount buffer for add case:
417 		 * we have hold both ADD state
418 		 * and DELETE state (this may be needed
419 		 * if we need to rollback all changes)
420 		 */
421 		sz = count * ta_buf_sz;
422 		ta_buf_m = malloc((op == OP_ADD) ? sz * 2 : sz, M_TEMP,
423 		    M_WAITOK | M_ZERO);
424 	}
425 
426 	v = ta_buf_m;
427 	for (i = 0; i < count; i++, v += ta_buf_sz) {
428 		ptei = &tei[i];
429 		error = (op == OP_ADD) ?
430 		    ta->prepare_add(ch, ptei, v) : ta->prepare_del(ch, ptei, v);
431 
432 		/*
433 		 * Some syntax error (incorrect mask, or address, or
434 		 * anything). Return error regardless of atomicity
435 		 * settings.
436 		 */
437 		if (error != 0)
438 			break;
439 	}
440 
441 	*ta_buf = ta_buf_m;
442 	return (error);
443 }
444 
445 /*
446  * Flushes allocated state for each @count entries in @tei.
447  * Frees @ta_buf_m if differs from stack buffer @ta_buf.
448  */
449 static void
450 flush_batch_buffer(struct ip_fw_chain *ch, struct table_algo *ta,
451     struct tentry_info *tei, uint32_t count, int rollback,
452     caddr_t ta_buf_m, caddr_t ta_buf)
453 {
454 	caddr_t v;
455 	struct tentry_info *ptei;
456 	size_t ta_buf_sz;
457 	int i;
458 
459 	ta_buf_sz = ta->ta_buf_size;
460 
461 	/* Run cleaning callback anyway */
462 	v = ta_buf_m;
463 	for (i = 0; i < count; i++, v += ta_buf_sz) {
464 		ptei = &tei[i];
465 		ta->flush_entry(ch, ptei, v);
466 		if (ptei->ptv != NULL) {
467 			free(ptei->ptv, M_IPFW);
468 			ptei->ptv = NULL;
469 		}
470 	}
471 
472 	/* Clean up "deleted" state in case of rollback */
473 	if (rollback != 0) {
474 		v = ta_buf_m + count * ta_buf_sz;
475 		for (i = 0; i < count; i++, v += ta_buf_sz)
476 			ta->flush_entry(ch, &tei[i], v);
477 	}
478 
479 	if (ta_buf_m != ta_buf)
480 		free(ta_buf_m, M_TEMP);
481 }
482 
483 
484 static void
485 rollback_add_entry(void *object, struct op_state *_state)
486 {
487 	struct ip_fw_chain *ch;
488 	struct tableop_state *ts;
489 
490 	ts = (struct tableop_state *)_state;
491 
492 	if (ts->tc != object && ts->ch != object)
493 		return;
494 
495 	ch = ts->ch;
496 
497 	IPFW_UH_WLOCK_ASSERT(ch);
498 
499 	/* Call specifid unlockers */
500 	rollback_table_values(ts);
501 
502 	/* Indicate we've called */
503 	ts->modified = 1;
504 }
505 
506 /*
507  * Adds/updates one or more entries in table @ti.
508  *
509  * Function may drop/reacquire UH wlock multiple times due to
510  * items alloc, algorithm callbacks (check_space), value linkage
511  * (new values, value storage realloc), etc..
512  * Other processes like other adds (which may involve storage resize),
513  * table swaps (which changes table data and may change algo type),
514  * table modify (which may change value mask) may be executed
515  * simultaneously so we need to deal with it.
516  *
517  * The following approach was implemented:
518  * we have per-chain linked list, protected with UH lock.
519  * add_table_entry prepares special on-stack structure wthich is passed
520  * to its descendants. Users add this structure to this list before unlock.
521  * After performing needed operations and acquiring UH lock back, each user
522  * checks if structure has changed. If true, it rolls local state back and
523  * returns without error to the caller.
524  * add_table_entry() on its own checks if structure has changed and restarts
525  * its operation from the beginning (goto restart).
526  *
527  * Functions which are modifying fields of interest (currently
528  *   resize_shared_value_storage() and swap_tables() )
529  * traverses given list while holding UH lock immediately before
530  * performing their operations calling function provided be list entry
531  * ( currently rollback_add_entry  ) which performs rollback for all necessary
532  * state and sets appropriate values in structure indicating rollback
533  * has happened.
534  *
535  * Algo interaction:
536  * Function references @ti first to ensure table won't
537  * disappear or change its type.
538  * After that, prepare_add callback is called for each @tei entry.
539  * Next, we try to add each entry under UH+WHLOCK
540  * using add() callback.
541  * Finally, we free all state by calling flush_entry callback
542  * for each @tei.
543  *
544  * Returns 0 on success.
545  */
546 int
547 add_table_entry(struct ip_fw_chain *ch, struct tid_info *ti,
548     struct tentry_info *tei, uint8_t flags, uint32_t count)
549 {
550 	struct table_config *tc;
551 	struct table_algo *ta;
552 	uint16_t kidx;
553 	int error, first_error, i, rollback;
554 	uint32_t num, numadd;
555 	struct tentry_info *ptei;
556 	struct tableop_state ts;
557 	char ta_buf[TA_BUF_SZ];
558 	caddr_t ta_buf_m, v;
559 
560 	memset(&ts, 0, sizeof(ts));
561 	ta = NULL;
562 	IPFW_UH_WLOCK(ch);
563 
564 	/*
565 	 * Find and reference existing table.
566 	 */
567 restart:
568 	if (ts.modified != 0) {
569 		IPFW_UH_WUNLOCK(ch);
570 		flush_batch_buffer(ch, ta, tei, count, rollback,
571 		    ta_buf_m, ta_buf);
572 		memset(&ts, 0, sizeof(ts));
573 		ta = NULL;
574 		IPFW_UH_WLOCK(ch);
575 	}
576 
577 	error = find_ref_table(ch, ti, tei, count, OP_ADD, &tc);
578 	if (error != 0) {
579 		IPFW_UH_WUNLOCK(ch);
580 		return (error);
581 	}
582 	ta = tc->ta;
583 
584 	/* Fill in tablestate */
585 	ts.ch = ch;
586 	ts.opstate.func = rollback_add_entry;
587 	ts.tc = tc;
588 	ts.vshared = tc->vshared;
589 	ts.vmask = tc->vmask;
590 	ts.ta = ta;
591 	ts.tei = tei;
592 	ts.count = count;
593 	rollback = 0;
594 	add_toperation_state(ch, &ts);
595 	IPFW_UH_WUNLOCK(ch);
596 
597 	/* Allocate memory and prepare record(s) */
598 	/* Pass stack buffer by default */
599 	ta_buf_m = ta_buf;
600 	error = prepare_batch_buffer(ch, ta, tei, count, OP_ADD, &ta_buf_m);
601 
602 	IPFW_UH_WLOCK(ch);
603 	del_toperation_state(ch, &ts);
604 	/* Drop reference we've used in first search */
605 	tc->no.refcnt--;
606 
607 	/* Check prepare_batch_buffer() error */
608 	if (error != 0)
609 		goto cleanup;
610 
611 	/*
612 	 * Check if table swap has happened.
613 	 * (so table algo might be changed).
614 	 * Restart operation to achieve consistent behavior.
615 	 */
616 	if (ts.modified != 0)
617 		goto restart;
618 
619 	/*
620 	 * Link all values values to shared/per-table value array.
621 	 *
622 	 * May release/reacquire UH_WLOCK.
623 	 */
624 	error = ipfw_link_table_values(ch, &ts);
625 	if (error != 0)
626 		goto cleanup;
627 	if (ts.modified != 0)
628 		goto restart;
629 
630 	/*
631 	 * Ensure we are able to add all entries without additional
632 	 * memory allocations. May release/reacquire UH_WLOCK.
633 	 */
634 	kidx = tc->no.kidx;
635 	error = check_table_space(ch, &ts, tc, KIDX_TO_TI(ch, kidx), count);
636 	if (error != 0)
637 		goto cleanup;
638 	if (ts.modified != 0)
639 		goto restart;
640 
641 	/* We've got valid table in @tc. Let's try to add data */
642 	kidx = tc->no.kidx;
643 	ta = tc->ta;
644 	numadd = 0;
645 	first_error = 0;
646 
647 	IPFW_WLOCK(ch);
648 
649 	v = ta_buf_m;
650 	for (i = 0; i < count; i++, v += ta->ta_buf_size) {
651 		ptei = &tei[i];
652 		num = 0;
653 		/* check limit before adding */
654 		if ((error = check_table_limit(tc, ptei)) == 0) {
655 			error = ta->add(tc->astate, KIDX_TO_TI(ch, kidx),
656 			    ptei, v, &num);
657 			/* Set status flag to inform userland */
658 			store_tei_result(ptei, OP_ADD, error, num);
659 		}
660 		if (error == 0) {
661 			/* Update number of records to ease limit checking */
662 			tc->count += num;
663 			numadd += num;
664 			continue;
665 		}
666 
667 		if (first_error == 0)
668 			first_error = error;
669 
670 		/*
671 		 * Some error have happened. Check our atomicity
672 		 * settings: continue if atomicity is not required,
673 		 * rollback changes otherwise.
674 		 */
675 		if ((flags & IPFW_CTF_ATOMIC) == 0)
676 			continue;
677 
678 		rollback_added_entries(ch, tc, KIDX_TO_TI(ch, kidx),
679 		    tei, ta_buf_m, count, i);
680 
681 		rollback = 1;
682 		break;
683 	}
684 
685 	IPFW_WUNLOCK(ch);
686 
687 	ipfw_garbage_table_values(ch, tc, tei, count, rollback);
688 
689 	/* Permit post-add algorithm grow/rehash. */
690 	if (numadd != 0)
691 		check_table_space(ch, NULL, tc, KIDX_TO_TI(ch, kidx), 0);
692 
693 	/* Return first error to user, if any */
694 	error = first_error;
695 
696 cleanup:
697 	IPFW_UH_WUNLOCK(ch);
698 
699 	flush_batch_buffer(ch, ta, tei, count, rollback, ta_buf_m, ta_buf);
700 
701 	return (error);
702 }
703 
704 /*
705  * Deletes one or more entries in table @ti.
706  *
707  * Returns 0 on success.
708  */
709 int
710 del_table_entry(struct ip_fw_chain *ch, struct tid_info *ti,
711     struct tentry_info *tei, uint8_t flags, uint32_t count)
712 {
713 	struct table_config *tc;
714 	struct table_algo *ta;
715 	struct tentry_info *ptei;
716 	uint16_t kidx;
717 	int error, first_error, i;
718 	uint32_t num, numdel;
719 	char ta_buf[TA_BUF_SZ];
720 	caddr_t ta_buf_m, v;
721 
722 	/*
723 	 * Find and reference existing table.
724 	 */
725 	IPFW_UH_WLOCK(ch);
726 	error = find_ref_table(ch, ti, tei, count, OP_DEL, &tc);
727 	if (error != 0) {
728 		IPFW_UH_WUNLOCK(ch);
729 		return (error);
730 	}
731 	ta = tc->ta;
732 	IPFW_UH_WUNLOCK(ch);
733 
734 	/* Allocate memory and prepare record(s) */
735 	/* Pass stack buffer by default */
736 	ta_buf_m = ta_buf;
737 	error = prepare_batch_buffer(ch, ta, tei, count, OP_DEL, &ta_buf_m);
738 	if (error != 0)
739 		goto cleanup;
740 
741 	IPFW_UH_WLOCK(ch);
742 
743 	/* Drop reference we've used in first search */
744 	tc->no.refcnt--;
745 
746 	/*
747 	 * Check if table algo is still the same.
748 	 * (changed ta may be the result of table swap).
749 	 */
750 	if (ta != tc->ta) {
751 		IPFW_UH_WUNLOCK(ch);
752 		error = EINVAL;
753 		goto cleanup;
754 	}
755 
756 	kidx = tc->no.kidx;
757 	numdel = 0;
758 	first_error = 0;
759 
760 	IPFW_WLOCK(ch);
761 	v = ta_buf_m;
762 	for (i = 0; i < count; i++, v += ta->ta_buf_size) {
763 		ptei = &tei[i];
764 		num = 0;
765 		error = ta->del(tc->astate, KIDX_TO_TI(ch, kidx), ptei, v,
766 		    &num);
767 		/* Save state for userland */
768 		store_tei_result(ptei, OP_DEL, error, num);
769 		if (error != 0 && first_error == 0)
770 			first_error = error;
771 		tc->count -= num;
772 		numdel += num;
773 	}
774 	IPFW_WUNLOCK(ch);
775 
776 	/* Unlink non-used values */
777 	ipfw_garbage_table_values(ch, tc, tei, count, 0);
778 
779 	if (numdel != 0) {
780 		/* Run post-del hook to permit shrinking */
781 		check_table_space(ch, NULL, tc, KIDX_TO_TI(ch, kidx), 0);
782 	}
783 
784 	IPFW_UH_WUNLOCK(ch);
785 
786 	/* Return first error to user, if any */
787 	error = first_error;
788 
789 cleanup:
790 	flush_batch_buffer(ch, ta, tei, count, 0, ta_buf_m, ta_buf);
791 
792 	return (error);
793 }
794 
795 /*
796  * Ensure that table @tc has enough space to add @count entries without
797  * need for reallocation.
798  *
799  * Callbacks order:
800  * 0) need_modify() (UH_WLOCK) - checks if @count items can be added w/o resize.
801  *
802  * 1) alloc_modify (no locks, M_WAITOK) - alloc new state based on @pflags.
803  * 2) prepare_modifyt (UH_WLOCK) - copy old data into new storage
804  * 3) modify (UH_WLOCK + WLOCK) - switch pointers
805  * 4) flush_modify (UH_WLOCK) - free state, if needed
806  *
807  * Returns 0 on success.
808  */
809 static int
810 check_table_space(struct ip_fw_chain *ch, struct tableop_state *ts,
811     struct table_config *tc, struct table_info *ti, uint32_t count)
812 {
813 	struct table_algo *ta;
814 	uint64_t pflags;
815 	char ta_buf[TA_BUF_SZ];
816 	int error;
817 
818 	IPFW_UH_WLOCK_ASSERT(ch);
819 
820 	error = 0;
821 	ta = tc->ta;
822 	if (ta->need_modify == NULL)
823 		return (0);
824 
825 	/* Acquire reference not to loose @tc between locks/unlocks */
826 	tc->no.refcnt++;
827 
828 	/*
829 	 * TODO: think about avoiding race between large add/large delete
830 	 * operation on algorithm which implements shrinking along with
831 	 * growing.
832 	 */
833 	while (true) {
834 		pflags = 0;
835 		if (ta->need_modify(tc->astate, ti, count, &pflags) == 0) {
836 			error = 0;
837 			break;
838 		}
839 
840 		/* We have to shrink/grow table */
841 		if (ts != NULL)
842 			add_toperation_state(ch, ts);
843 		IPFW_UH_WUNLOCK(ch);
844 
845 		memset(&ta_buf, 0, sizeof(ta_buf));
846 		error = ta->prepare_mod(ta_buf, &pflags);
847 
848 		IPFW_UH_WLOCK(ch);
849 		if (ts != NULL)
850 			del_toperation_state(ch, ts);
851 
852 		if (error != 0)
853 			break;
854 
855 		if (ts != NULL && ts->modified != 0) {
856 
857 			/*
858 			 * Swap operation has happened
859 			 * so we're currently operating on other
860 			 * table data. Stop doing this.
861 			 */
862 			ta->flush_mod(ta_buf);
863 			break;
864 		}
865 
866 		/* Check if we still need to alter table */
867 		ti = KIDX_TO_TI(ch, tc->no.kidx);
868 		if (ta->need_modify(tc->astate, ti, count, &pflags) == 0) {
869 			IPFW_UH_WUNLOCK(ch);
870 
871 			/*
872 			 * Other thread has already performed resize.
873 			 * Flush our state and return.
874 			 */
875 			ta->flush_mod(ta_buf);
876 			break;
877 		}
878 
879 		error = ta->fill_mod(tc->astate, ti, ta_buf, &pflags);
880 		if (error == 0) {
881 			/* Do actual modification */
882 			IPFW_WLOCK(ch);
883 			ta->modify(tc->astate, ti, ta_buf, pflags);
884 			IPFW_WUNLOCK(ch);
885 		}
886 
887 		/* Anyway, flush data and retry */
888 		ta->flush_mod(ta_buf);
889 	}
890 
891 	tc->no.refcnt--;
892 	return (error);
893 }
894 
895 /*
896  * Adds or deletes record in table.
897  * Data layout (v0):
898  * Request: [ ip_fw3_opheader ipfw_table_xentry ]
899  *
900  * Returns 0 on success
901  */
902 static int
903 manage_table_ent_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
904     struct sockopt_data *sd)
905 {
906 	ipfw_table_xentry *xent;
907 	struct tentry_info tei;
908 	struct tid_info ti;
909 	struct table_value v;
910 	int error, hdrlen, read;
911 
912 	hdrlen = offsetof(ipfw_table_xentry, k);
913 
914 	/* Check minimum header size */
915 	if (sd->valsize < (sizeof(*op3) + hdrlen))
916 		return (EINVAL);
917 
918 	read = sizeof(ip_fw3_opheader);
919 
920 	/* Check if xentry len field is valid */
921 	xent = (ipfw_table_xentry *)(op3 + 1);
922 	if (xent->len < hdrlen || xent->len + read > sd->valsize)
923 		return (EINVAL);
924 
925 	memset(&tei, 0, sizeof(tei));
926 	tei.paddr = &xent->k;
927 	tei.masklen = xent->masklen;
928 	ipfw_import_table_value_legacy(xent->value, &v);
929 	tei.pvalue = &v;
930 	/* Old requests compability */
931 	tei.flags = TEI_FLAGS_COMPAT;
932 	if (xent->type == IPFW_TABLE_ADDR) {
933 		if (xent->len - hdrlen == sizeof(in_addr_t))
934 			tei.subtype = AF_INET;
935 		else
936 			tei.subtype = AF_INET6;
937 	}
938 
939 	memset(&ti, 0, sizeof(ti));
940 	ti.uidx = xent->tbl;
941 	ti.type = xent->type;
942 
943 	error = (op3->opcode == IP_FW_TABLE_XADD) ?
944 	    add_table_entry(ch, &ti, &tei, 0, 1) :
945 	    del_table_entry(ch, &ti, &tei, 0, 1);
946 
947 	return (error);
948 }
949 
950 /*
951  * Adds or deletes record in table.
952  * Data layout (v1)(current):
953  * Request: [ ipfw_obj_header
954  *   ipfw_obj_ctlv(IPFW_TLV_TBLENT_LIST) [ ipfw_obj_tentry x N ]
955  * ]
956  *
957  * Returns 0 on success
958  */
959 static int
960 manage_table_ent_v1(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
961     struct sockopt_data *sd)
962 {
963 	ipfw_obj_tentry *tent, *ptent;
964 	ipfw_obj_ctlv *ctlv;
965 	ipfw_obj_header *oh;
966 	struct tentry_info *ptei, tei, *tei_buf;
967 	struct tid_info ti;
968 	int error, i, kidx, read;
969 
970 	/* Check minimum header size */
971 	if (sd->valsize < (sizeof(*oh) + sizeof(*ctlv)))
972 		return (EINVAL);
973 
974 	/* Check if passed data is too long */
975 	if (sd->valsize != sd->kavail)
976 		return (EINVAL);
977 
978 	oh = (ipfw_obj_header *)sd->kbuf;
979 
980 	/* Basic length checks for TLVs */
981 	if (oh->ntlv.head.length != sizeof(oh->ntlv))
982 		return (EINVAL);
983 
984 	read = sizeof(*oh);
985 
986 	ctlv = (ipfw_obj_ctlv *)(oh + 1);
987 	if (ctlv->head.length + read != sd->valsize)
988 		return (EINVAL);
989 
990 	read += sizeof(*ctlv);
991 	tent = (ipfw_obj_tentry *)(ctlv + 1);
992 	if (ctlv->count * sizeof(*tent) + read != sd->valsize)
993 		return (EINVAL);
994 
995 	if (ctlv->count == 0)
996 		return (0);
997 
998 	/*
999 	 * Mark entire buffer as "read".
1000 	 * This instructs sopt api write it back
1001 	 * after function return.
1002 	 */
1003 	ipfw_get_sopt_header(sd, sd->valsize);
1004 
1005 	/* Perform basic checks for each entry */
1006 	ptent = tent;
1007 	kidx = tent->idx;
1008 	for (i = 0; i < ctlv->count; i++, ptent++) {
1009 		if (ptent->head.length != sizeof(*ptent))
1010 			return (EINVAL);
1011 		if (ptent->idx != kidx)
1012 			return (ENOTSUP);
1013 	}
1014 
1015 	/* Convert data into kernel request objects */
1016 	objheader_to_ti(oh, &ti);
1017 	ti.type = oh->ntlv.type;
1018 	ti.uidx = kidx;
1019 
1020 	/* Use on-stack buffer for single add/del */
1021 	if (ctlv->count == 1) {
1022 		memset(&tei, 0, sizeof(tei));
1023 		tei_buf = &tei;
1024 	} else
1025 		tei_buf = malloc(ctlv->count * sizeof(tei), M_TEMP,
1026 		    M_WAITOK | M_ZERO);
1027 
1028 	ptei = tei_buf;
1029 	ptent = tent;
1030 	for (i = 0; i < ctlv->count; i++, ptent++, ptei++) {
1031 		ptei->paddr = &ptent->k;
1032 		ptei->subtype = ptent->subtype;
1033 		ptei->masklen = ptent->masklen;
1034 		if (ptent->head.flags & IPFW_TF_UPDATE)
1035 			ptei->flags |= TEI_FLAGS_UPDATE;
1036 
1037 		ipfw_import_table_value_v1(&ptent->v.value);
1038 		ptei->pvalue = (struct table_value *)&ptent->v.value;
1039 	}
1040 
1041 	error = (oh->opheader.opcode == IP_FW_TABLE_XADD) ?
1042 	    add_table_entry(ch, &ti, tei_buf, ctlv->flags, ctlv->count) :
1043 	    del_table_entry(ch, &ti, tei_buf, ctlv->flags, ctlv->count);
1044 
1045 	/* Translate result back to userland */
1046 	ptei = tei_buf;
1047 	ptent = tent;
1048 	for (i = 0; i < ctlv->count; i++, ptent++, ptei++) {
1049 		if (ptei->flags & TEI_FLAGS_ADDED)
1050 			ptent->result = IPFW_TR_ADDED;
1051 		else if (ptei->flags & TEI_FLAGS_DELETED)
1052 			ptent->result = IPFW_TR_DELETED;
1053 		else if (ptei->flags & TEI_FLAGS_UPDATED)
1054 			ptent->result = IPFW_TR_UPDATED;
1055 		else if (ptei->flags & TEI_FLAGS_LIMIT)
1056 			ptent->result = IPFW_TR_LIMIT;
1057 		else if (ptei->flags & TEI_FLAGS_ERROR)
1058 			ptent->result = IPFW_TR_ERROR;
1059 		else if (ptei->flags & TEI_FLAGS_NOTFOUND)
1060 			ptent->result = IPFW_TR_NOTFOUND;
1061 		else if (ptei->flags & TEI_FLAGS_EXISTS)
1062 			ptent->result = IPFW_TR_EXISTS;
1063 		ipfw_export_table_value_v1(ptei->pvalue, &ptent->v.value);
1064 	}
1065 
1066 	if (tei_buf != &tei)
1067 		free(tei_buf, M_TEMP);
1068 
1069 	return (error);
1070 }
1071 
1072 /*
1073  * Looks up an entry in given table.
1074  * Data layout (v0)(current):
1075  * Request: [ ipfw_obj_header ipfw_obj_tentry ]
1076  * Reply: [ ipfw_obj_header ipfw_obj_tentry ]
1077  *
1078  * Returns 0 on success
1079  */
1080 static int
1081 find_table_entry(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1082     struct sockopt_data *sd)
1083 {
1084 	ipfw_obj_tentry *tent;
1085 	ipfw_obj_header *oh;
1086 	struct tid_info ti;
1087 	struct table_config *tc;
1088 	struct table_algo *ta;
1089 	struct table_info *kti;
1090 	struct namedobj_instance *ni;
1091 	int error;
1092 	size_t sz;
1093 
1094 	/* Check minimum header size */
1095 	sz = sizeof(*oh) + sizeof(*tent);
1096 	if (sd->valsize != sz)
1097 		return (EINVAL);
1098 
1099 	oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz);
1100 	tent = (ipfw_obj_tentry *)(oh + 1);
1101 
1102 	/* Basic length checks for TLVs */
1103 	if (oh->ntlv.head.length != sizeof(oh->ntlv))
1104 		return (EINVAL);
1105 
1106 	objheader_to_ti(oh, &ti);
1107 	ti.type = oh->ntlv.type;
1108 	ti.uidx = tent->idx;
1109 
1110 	IPFW_UH_RLOCK(ch);
1111 	ni = CHAIN_TO_NI(ch);
1112 
1113 	/*
1114 	 * Find existing table and check its type .
1115 	 */
1116 	ta = NULL;
1117 	if ((tc = find_table(ni, &ti)) == NULL) {
1118 		IPFW_UH_RUNLOCK(ch);
1119 		return (ESRCH);
1120 	}
1121 
1122 	/* check table type */
1123 	if (tc->no.subtype != ti.type) {
1124 		IPFW_UH_RUNLOCK(ch);
1125 		return (EINVAL);
1126 	}
1127 
1128 	kti = KIDX_TO_TI(ch, tc->no.kidx);
1129 	ta = tc->ta;
1130 
1131 	if (ta->find_tentry == NULL)
1132 		return (ENOTSUP);
1133 
1134 	error = ta->find_tentry(tc->astate, kti, tent);
1135 
1136 	IPFW_UH_RUNLOCK(ch);
1137 
1138 	return (error);
1139 }
1140 
1141 /*
1142  * Flushes all entries or destroys given table.
1143  * Data layout (v0)(current):
1144  * Request: [ ipfw_obj_header ]
1145  *
1146  * Returns 0 on success
1147  */
1148 static int
1149 flush_table_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1150     struct sockopt_data *sd)
1151 {
1152 	int error;
1153 	struct _ipfw_obj_header *oh;
1154 	struct tid_info ti;
1155 
1156 	if (sd->valsize != sizeof(*oh))
1157 		return (EINVAL);
1158 
1159 	oh = (struct _ipfw_obj_header *)op3;
1160 	objheader_to_ti(oh, &ti);
1161 
1162 	if (op3->opcode == IP_FW_TABLE_XDESTROY)
1163 		error = destroy_table(ch, &ti);
1164 	else if (op3->opcode == IP_FW_TABLE_XFLUSH)
1165 		error = flush_table(ch, &ti);
1166 	else
1167 		return (ENOTSUP);
1168 
1169 	return (error);
1170 }
1171 
1172 static void
1173 restart_flush(void *object, struct op_state *_state)
1174 {
1175 	struct tableop_state *ts;
1176 
1177 	ts = (struct tableop_state *)_state;
1178 
1179 	if (ts->tc != object)
1180 		return;
1181 
1182 	/* Indicate we've called */
1183 	ts->modified = 1;
1184 }
1185 
1186 /*
1187  * Flushes given table.
1188  *
1189  * Function create new table instance with the same
1190  * parameters, swaps it with old one and
1191  * flushes state without holding runtime WLOCK.
1192  *
1193  * Returns 0 on success.
1194  */
1195 int
1196 flush_table(struct ip_fw_chain *ch, struct tid_info *ti)
1197 {
1198 	struct namedobj_instance *ni;
1199 	struct table_config *tc;
1200 	struct table_algo *ta;
1201 	struct table_info ti_old, ti_new, *tablestate;
1202 	void *astate_old, *astate_new;
1203 	char algostate[64], *pstate;
1204 	struct tableop_state ts;
1205 	int error, need_gc;
1206 	uint16_t kidx;
1207 	uint8_t tflags;
1208 
1209 	/*
1210 	 * Stage 1: save table algoritm.
1211 	 * Reference found table to ensure it won't disappear.
1212 	 */
1213 	IPFW_UH_WLOCK(ch);
1214 	ni = CHAIN_TO_NI(ch);
1215 	if ((tc = find_table(ni, ti)) == NULL) {
1216 		IPFW_UH_WUNLOCK(ch);
1217 		return (ESRCH);
1218 	}
1219 	need_gc = 0;
1220 	astate_new = NULL;
1221 	memset(&ti_new, 0, sizeof(ti_new));
1222 restart:
1223 	/* Set up swap handler */
1224 	memset(&ts, 0, sizeof(ts));
1225 	ts.opstate.func = restart_flush;
1226 	ts.tc = tc;
1227 
1228 	ta = tc->ta;
1229 	/* Do not flush readonly tables */
1230 	if ((ta->flags & TA_FLAG_READONLY) != 0) {
1231 		IPFW_UH_WUNLOCK(ch);
1232 		return (EACCES);
1233 	}
1234 	/* Save startup algo parameters */
1235 	if (ta->print_config != NULL) {
1236 		ta->print_config(tc->astate, KIDX_TO_TI(ch, tc->no.kidx),
1237 		    algostate, sizeof(algostate));
1238 		pstate = algostate;
1239 	} else
1240 		pstate = NULL;
1241 	tflags = tc->tflags;
1242 	tc->no.refcnt++;
1243 	add_toperation_state(ch, &ts);
1244 	IPFW_UH_WUNLOCK(ch);
1245 
1246 	/*
1247 	 * Stage 1.5: if this is not the first attempt, destroy previous state
1248 	 */
1249 	if (need_gc != 0) {
1250 		ta->destroy(astate_new, &ti_new);
1251 		need_gc = 0;
1252 	}
1253 
1254 	/*
1255 	 * Stage 2: allocate new table instance using same algo.
1256 	 */
1257 	memset(&ti_new, 0, sizeof(struct table_info));
1258 	error = ta->init(ch, &astate_new, &ti_new, pstate, tflags);
1259 
1260 	/*
1261 	 * Stage 3: swap old state pointers with newly-allocated ones.
1262 	 * Decrease refcount.
1263 	 */
1264 	IPFW_UH_WLOCK(ch);
1265 	tc->no.refcnt--;
1266 	del_toperation_state(ch, &ts);
1267 
1268 	if (error != 0) {
1269 		IPFW_UH_WUNLOCK(ch);
1270 		return (error);
1271 	}
1272 
1273 	/*
1274 	 * Restart operation if table swap has happened:
1275 	 * even if algo may be the same, algo init parameters
1276 	 * may change. Restart operation instead of doing
1277 	 * complex checks.
1278 	 */
1279 	if (ts.modified != 0) {
1280 		/* Delay destroying data since we're holding UH lock */
1281 		need_gc = 1;
1282 		goto restart;
1283 	}
1284 
1285 	ni = CHAIN_TO_NI(ch);
1286 	kidx = tc->no.kidx;
1287 	tablestate = (struct table_info *)ch->tablestate;
1288 
1289 	IPFW_WLOCK(ch);
1290 	ti_old = tablestate[kidx];
1291 	tablestate[kidx] = ti_new;
1292 	IPFW_WUNLOCK(ch);
1293 
1294 	astate_old = tc->astate;
1295 	tc->astate = astate_new;
1296 	tc->ti_copy = ti_new;
1297 	tc->count = 0;
1298 
1299 	/* Notify algo on real @ti address */
1300 	if (ta->change_ti != NULL)
1301 		ta->change_ti(tc->astate, &tablestate[kidx]);
1302 
1303 	/*
1304 	 * Stage 4: unref values.
1305 	 */
1306 	ipfw_unref_table_values(ch, tc, ta, astate_old, &ti_old);
1307 	IPFW_UH_WUNLOCK(ch);
1308 
1309 	/*
1310 	 * Stage 5: perform real flush/destroy.
1311 	 */
1312 	ta->destroy(astate_old, &ti_old);
1313 
1314 	return (0);
1315 }
1316 
1317 /*
1318  * Swaps two tables.
1319  * Data layout (v0)(current):
1320  * Request: [ ipfw_obj_header ipfw_obj_ntlv ]
1321  *
1322  * Returns 0 on success
1323  */
1324 static int
1325 swap_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1326     struct sockopt_data *sd)
1327 {
1328 	int error;
1329 	struct _ipfw_obj_header *oh;
1330 	struct tid_info ti_a, ti_b;
1331 
1332 	if (sd->valsize != sizeof(*oh) + sizeof(ipfw_obj_ntlv))
1333 		return (EINVAL);
1334 
1335 	oh = (struct _ipfw_obj_header *)op3;
1336 	ntlv_to_ti(&oh->ntlv, &ti_a);
1337 	ntlv_to_ti((ipfw_obj_ntlv *)(oh + 1), &ti_b);
1338 
1339 	error = swap_tables(ch, &ti_a, &ti_b);
1340 
1341 	return (error);
1342 }
1343 
1344 /*
1345  * Swaps two tables of the same type/valtype.
1346  *
1347  * Checks if tables are compatible and limits
1348  * permits swap, than actually perform swap.
1349  *
1350  * Each table consists of 2 different parts:
1351  * config:
1352  *   @tc (with name, set, kidx) and rule bindings, which is "stable".
1353  *   number of items
1354  *   table algo
1355  * runtime:
1356  *   runtime data @ti (ch->tablestate)
1357  *   runtime cache in @tc
1358  *   algo-specific data (@tc->astate)
1359  *
1360  * So we switch:
1361  *  all runtime data
1362  *   number of items
1363  *   table algo
1364  *
1365  * After that we call @ti change handler for each table.
1366  *
1367  * Note that referencing @tc won't protect tc->ta from change.
1368  * XXX: Do we need to restrict swap between locked tables?
1369  * XXX: Do we need to exchange ftype?
1370  *
1371  * Returns 0 on success.
1372  */
1373 static int
1374 swap_tables(struct ip_fw_chain *ch, struct tid_info *a,
1375     struct tid_info *b)
1376 {
1377 	struct namedobj_instance *ni;
1378 	struct table_config *tc_a, *tc_b;
1379 	struct table_algo *ta;
1380 	struct table_info ti, *tablestate;
1381 	void *astate;
1382 	uint32_t count;
1383 
1384 	/*
1385 	 * Stage 1: find both tables and ensure they are of
1386 	 * the same type.
1387 	 */
1388 	IPFW_UH_WLOCK(ch);
1389 	ni = CHAIN_TO_NI(ch);
1390 	if ((tc_a = find_table(ni, a)) == NULL) {
1391 		IPFW_UH_WUNLOCK(ch);
1392 		return (ESRCH);
1393 	}
1394 	if ((tc_b = find_table(ni, b)) == NULL) {
1395 		IPFW_UH_WUNLOCK(ch);
1396 		return (ESRCH);
1397 	}
1398 
1399 	/* It is very easy to swap between the same table */
1400 	if (tc_a == tc_b) {
1401 		IPFW_UH_WUNLOCK(ch);
1402 		return (0);
1403 	}
1404 
1405 	/* Check type and value are the same */
1406 	if (tc_a->no.subtype!=tc_b->no.subtype || tc_a->tflags!=tc_b->tflags) {
1407 		IPFW_UH_WUNLOCK(ch);
1408 		return (EINVAL);
1409 	}
1410 
1411 	/* Check limits before swap */
1412 	if ((tc_a->limit != 0 && tc_b->count > tc_a->limit) ||
1413 	    (tc_b->limit != 0 && tc_a->count > tc_b->limit)) {
1414 		IPFW_UH_WUNLOCK(ch);
1415 		return (EFBIG);
1416 	}
1417 
1418 	/* Check if one of the tables is readonly */
1419 	if (((tc_a->ta->flags | tc_b->ta->flags) & TA_FLAG_READONLY) != 0) {
1420 		IPFW_UH_WUNLOCK(ch);
1421 		return (EACCES);
1422 	}
1423 
1424 	/* Notify we're going to swap */
1425 	rollback_toperation_state(ch, tc_a);
1426 	rollback_toperation_state(ch, tc_b);
1427 
1428 	/* Everything is fine, prepare to swap */
1429 	tablestate = (struct table_info *)ch->tablestate;
1430 	ti = tablestate[tc_a->no.kidx];
1431 	ta = tc_a->ta;
1432 	astate = tc_a->astate;
1433 	count = tc_a->count;
1434 
1435 	IPFW_WLOCK(ch);
1436 	/* a <- b */
1437 	tablestate[tc_a->no.kidx] = tablestate[tc_b->no.kidx];
1438 	tc_a->ta = tc_b->ta;
1439 	tc_a->astate = tc_b->astate;
1440 	tc_a->count = tc_b->count;
1441 	/* b <- a */
1442 	tablestate[tc_b->no.kidx] = ti;
1443 	tc_b->ta = ta;
1444 	tc_b->astate = astate;
1445 	tc_b->count = count;
1446 	IPFW_WUNLOCK(ch);
1447 
1448 	/* Ensure tc.ti copies are in sync */
1449 	tc_a->ti_copy = tablestate[tc_a->no.kidx];
1450 	tc_b->ti_copy = tablestate[tc_b->no.kidx];
1451 
1452 	/* Notify both tables on @ti change */
1453 	if (tc_a->ta->change_ti != NULL)
1454 		tc_a->ta->change_ti(tc_a->astate, &tablestate[tc_a->no.kidx]);
1455 	if (tc_b->ta->change_ti != NULL)
1456 		tc_b->ta->change_ti(tc_b->astate, &tablestate[tc_b->no.kidx]);
1457 
1458 	IPFW_UH_WUNLOCK(ch);
1459 
1460 	return (0);
1461 }
1462 
1463 /*
1464  * Destroys table specified by @ti.
1465  * Data layout (v0)(current):
1466  * Request: [ ip_fw3_opheader ]
1467  *
1468  * Returns 0 on success
1469  */
1470 static int
1471 destroy_table(struct ip_fw_chain *ch, struct tid_info *ti)
1472 {
1473 	struct namedobj_instance *ni;
1474 	struct table_config *tc;
1475 
1476 	IPFW_UH_WLOCK(ch);
1477 
1478 	ni = CHAIN_TO_NI(ch);
1479 	if ((tc = find_table(ni, ti)) == NULL) {
1480 		IPFW_UH_WUNLOCK(ch);
1481 		return (ESRCH);
1482 	}
1483 
1484 	/* Do not permit destroying referenced tables */
1485 	if (tc->no.refcnt > 0) {
1486 		IPFW_UH_WUNLOCK(ch);
1487 		return (EBUSY);
1488 	}
1489 
1490 	IPFW_WLOCK(ch);
1491 	unlink_table(ch, tc);
1492 	IPFW_WUNLOCK(ch);
1493 
1494 	/* Free obj index */
1495 	if (ipfw_objhash_free_idx(ni, tc->no.kidx) != 0)
1496 		printf("Error unlinking kidx %d from table %s\n",
1497 		    tc->no.kidx, tc->tablename);
1498 
1499 	/* Unref values used in tables while holding UH lock */
1500 	ipfw_unref_table_values(ch, tc, tc->ta, tc->astate, &tc->ti_copy);
1501 	IPFW_UH_WUNLOCK(ch);
1502 
1503 	free_table_config(ni, tc);
1504 
1505 	return (0);
1506 }
1507 
1508 static uint32_t
1509 roundup2p(uint32_t v)
1510 {
1511 
1512 	v--;
1513 	v |= v >> 1;
1514 	v |= v >> 2;
1515 	v |= v >> 4;
1516 	v |= v >> 8;
1517 	v |= v >> 16;
1518 	v++;
1519 
1520 	return (v);
1521 }
1522 
1523 /*
1524  * Grow tables index.
1525  *
1526  * Returns 0 on success.
1527  */
1528 int
1529 ipfw_resize_tables(struct ip_fw_chain *ch, unsigned int ntables)
1530 {
1531 	unsigned int ntables_old, tbl;
1532 	struct namedobj_instance *ni;
1533 	void *new_idx, *old_tablestate, *tablestate;
1534 	struct table_info *ti;
1535 	struct table_config *tc;
1536 	int i, new_blocks;
1537 
1538 	/* Check new value for validity */
1539 	if (ntables == 0)
1540 		return (EINVAL);
1541 	if (ntables > IPFW_TABLES_MAX)
1542 		ntables = IPFW_TABLES_MAX;
1543 	/* Alight to nearest power of 2 */
1544 	ntables = (unsigned int)roundup2p(ntables);
1545 
1546 	/* Allocate new pointers */
1547 	tablestate = malloc(ntables * sizeof(struct table_info),
1548 	    M_IPFW, M_WAITOK | M_ZERO);
1549 
1550 	ipfw_objhash_bitmap_alloc(ntables, (void *)&new_idx, &new_blocks);
1551 
1552 	IPFW_UH_WLOCK(ch);
1553 
1554 	tbl = (ntables >= V_fw_tables_max) ? V_fw_tables_max : ntables;
1555 	ni = CHAIN_TO_NI(ch);
1556 
1557 	/* Temporary restrict decreasing max_tables */
1558 	if (ntables < V_fw_tables_max) {
1559 
1560 		/*
1561 		 * FIXME: Check if we really can shrink
1562 		 */
1563 		IPFW_UH_WUNLOCK(ch);
1564 		return (EINVAL);
1565 	}
1566 
1567 	/* Copy table info/indices */
1568 	memcpy(tablestate, ch->tablestate, sizeof(struct table_info) * tbl);
1569 	ipfw_objhash_bitmap_merge(ni, &new_idx, &new_blocks);
1570 
1571 	IPFW_WLOCK(ch);
1572 
1573 	/* Change pointers */
1574 	old_tablestate = ch->tablestate;
1575 	ch->tablestate = tablestate;
1576 	ipfw_objhash_bitmap_swap(ni, &new_idx, &new_blocks);
1577 
1578 	ntables_old = V_fw_tables_max;
1579 	V_fw_tables_max = ntables;
1580 
1581 	IPFW_WUNLOCK(ch);
1582 
1583 	/* Notify all consumers that their @ti pointer has changed */
1584 	ti = (struct table_info *)ch->tablestate;
1585 	for (i = 0; i < tbl; i++, ti++) {
1586 		if (ti->lookup == NULL)
1587 			continue;
1588 		tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, i);
1589 		if (tc == NULL || tc->ta->change_ti == NULL)
1590 			continue;
1591 
1592 		tc->ta->change_ti(tc->astate, ti);
1593 	}
1594 
1595 	IPFW_UH_WUNLOCK(ch);
1596 
1597 	/* Free old pointers */
1598 	free(old_tablestate, M_IPFW);
1599 	ipfw_objhash_bitmap_free(new_idx, new_blocks);
1600 
1601 	return (0);
1602 }
1603 
1604 /*
1605  * Switch between "set 0" and "rule's set" table binding,
1606  * Check all ruleset bindings and permits changing
1607  * IFF each binding has both rule AND table in default set (set 0).
1608  *
1609  * Returns 0 on success.
1610  */
1611 int
1612 ipfw_switch_tables_namespace(struct ip_fw_chain *ch, unsigned int sets)
1613 {
1614 	struct namedobj_instance *ni;
1615 	struct named_object *no;
1616 	struct ip_fw *rule;
1617 	ipfw_insn *cmd;
1618 	int cmdlen, i, l;
1619 	uint16_t kidx;
1620 
1621 	IPFW_UH_WLOCK(ch);
1622 
1623 	if (V_fw_tables_sets == sets) {
1624 		IPFW_UH_WUNLOCK(ch);
1625 		return (0);
1626 	}
1627 
1628 	ni = CHAIN_TO_NI(ch);
1629 
1630 	/*
1631 	 * Scan all rules and examine tables opcodes.
1632 	 */
1633 	for (i = 0; i < ch->n_rules; i++) {
1634 		rule = ch->map[i];
1635 
1636 		l = rule->cmd_len;
1637 		cmd = rule->cmd;
1638 		cmdlen = 0;
1639 		for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
1640 			cmdlen = F_LEN(cmd);
1641 
1642 			if (classify_opcode_kidx(cmd, &kidx) != 0)
1643 				continue;
1644 
1645 			no = ipfw_objhash_lookup_kidx(ni, kidx);
1646 
1647 			/* Check if both table object and rule has the set 0 */
1648 			if (no->set != 0 || rule->set != 0) {
1649 				IPFW_UH_WUNLOCK(ch);
1650 				return (EBUSY);
1651 			}
1652 
1653 		}
1654 	}
1655 	V_fw_tables_sets = sets;
1656 
1657 	IPFW_UH_WUNLOCK(ch);
1658 
1659 	return (0);
1660 }
1661 
1662 /*
1663  * Lookup an IP @addr in table @tbl.
1664  * Stores found value in @val.
1665  *
1666  * Returns 1 if @addr was found.
1667  */
1668 int
1669 ipfw_lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1670     uint32_t *val)
1671 {
1672 	struct table_info *ti;
1673 
1674 	ti = KIDX_TO_TI(ch, tbl);
1675 
1676 	return (ti->lookup(ti, &addr, sizeof(in_addr_t), val));
1677 }
1678 
1679 /*
1680  * Lookup an arbtrary key @paddr of legth @plen in table @tbl.
1681  * Stores found value in @val.
1682  *
1683  * Returns 1 if key was found.
1684  */
1685 int
1686 ipfw_lookup_table_extended(struct ip_fw_chain *ch, uint16_t tbl, uint16_t plen,
1687     void *paddr, uint32_t *val)
1688 {
1689 	struct table_info *ti;
1690 
1691 	ti = KIDX_TO_TI(ch, tbl);
1692 
1693 	return (ti->lookup(ti, paddr, plen, val));
1694 }
1695 
1696 /*
1697  * Info/List/dump support for tables.
1698  *
1699  */
1700 
1701 /*
1702  * High-level 'get' cmds sysctl handlers
1703  */
1704 
1705 /*
1706  * Lists all tables currently available in kernel.
1707  * Data layout (v0)(current):
1708  * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size
1709  * Reply: [ ipfw_obj_lheader ipfw_xtable_info x N ]
1710  *
1711  * Returns 0 on success
1712  */
1713 static int
1714 list_tables(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1715     struct sockopt_data *sd)
1716 {
1717 	struct _ipfw_obj_lheader *olh;
1718 	int error;
1719 
1720 	olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh));
1721 	if (olh == NULL)
1722 		return (EINVAL);
1723 	if (sd->valsize < olh->size)
1724 		return (EINVAL);
1725 
1726 	IPFW_UH_RLOCK(ch);
1727 	error = export_tables(ch, olh, sd);
1728 	IPFW_UH_RUNLOCK(ch);
1729 
1730 	return (error);
1731 }
1732 
1733 /*
1734  * Store table info to buffer provided by @sd.
1735  * Data layout (v0)(current):
1736  * Request: [ ipfw_obj_header ipfw_xtable_info(empty)]
1737  * Reply: [ ipfw_obj_header ipfw_xtable_info ]
1738  *
1739  * Returns 0 on success.
1740  */
1741 static int
1742 describe_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1743     struct sockopt_data *sd)
1744 {
1745 	struct _ipfw_obj_header *oh;
1746 	struct table_config *tc;
1747 	struct tid_info ti;
1748 	size_t sz;
1749 
1750 	sz = sizeof(*oh) + sizeof(ipfw_xtable_info);
1751 	oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz);
1752 	if (oh == NULL)
1753 		return (EINVAL);
1754 
1755 	objheader_to_ti(oh, &ti);
1756 
1757 	IPFW_UH_RLOCK(ch);
1758 	if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) {
1759 		IPFW_UH_RUNLOCK(ch);
1760 		return (ESRCH);
1761 	}
1762 
1763 	export_table_info(ch, tc, (ipfw_xtable_info *)(oh + 1));
1764 	IPFW_UH_RUNLOCK(ch);
1765 
1766 	return (0);
1767 }
1768 
1769 /*
1770  * Modifies existing table.
1771  * Data layout (v0)(current):
1772  * Request: [ ipfw_obj_header ipfw_xtable_info ]
1773  *
1774  * Returns 0 on success
1775  */
1776 static int
1777 modify_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1778     struct sockopt_data *sd)
1779 {
1780 	struct _ipfw_obj_header *oh;
1781 	ipfw_xtable_info *i;
1782 	char *tname;
1783 	struct tid_info ti;
1784 	struct namedobj_instance *ni;
1785 	struct table_config *tc;
1786 
1787 	if (sd->valsize != sizeof(*oh) + sizeof(ipfw_xtable_info))
1788 		return (EINVAL);
1789 
1790 	oh = (struct _ipfw_obj_header *)sd->kbuf;
1791 	i = (ipfw_xtable_info *)(oh + 1);
1792 
1793 	/*
1794 	 * Verify user-supplied strings.
1795 	 * Check for null-terminated/zero-length strings/
1796 	 */
1797 	tname = oh->ntlv.name;
1798 	if (check_table_name(tname) != 0)
1799 		return (EINVAL);
1800 
1801 	objheader_to_ti(oh, &ti);
1802 	ti.type = i->type;
1803 
1804 	IPFW_UH_WLOCK(ch);
1805 	ni = CHAIN_TO_NI(ch);
1806 	if ((tc = find_table(ni, &ti)) == NULL) {
1807 		IPFW_UH_WUNLOCK(ch);
1808 		return (ESRCH);
1809 	}
1810 
1811 	/* Do not support any modifications for readonly tables */
1812 	if ((tc->ta->flags & TA_FLAG_READONLY) != 0) {
1813 		IPFW_UH_WUNLOCK(ch);
1814 		return (EACCES);
1815 	}
1816 
1817 	if ((i->mflags & IPFW_TMFLAGS_LIMIT) != 0)
1818 		tc->limit = i->limit;
1819 	if ((i->mflags & IPFW_TMFLAGS_LOCK) != 0)
1820 		tc->locked = ((i->flags & IPFW_TGFLAGS_LOCKED) != 0);
1821 	IPFW_UH_WUNLOCK(ch);
1822 
1823 	return (0);
1824 }
1825 
1826 /*
1827  * Creates new table.
1828  * Data layout (v0)(current):
1829  * Request: [ ipfw_obj_header ipfw_xtable_info ]
1830  *
1831  * Returns 0 on success
1832  */
1833 static int
1834 create_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1835     struct sockopt_data *sd)
1836 {
1837 	struct _ipfw_obj_header *oh;
1838 	ipfw_xtable_info *i;
1839 	char *tname, *aname;
1840 	struct tid_info ti;
1841 	struct namedobj_instance *ni;
1842 
1843 	if (sd->valsize != sizeof(*oh) + sizeof(ipfw_xtable_info))
1844 		return (EINVAL);
1845 
1846 	oh = (struct _ipfw_obj_header *)sd->kbuf;
1847 	i = (ipfw_xtable_info *)(oh + 1);
1848 
1849 	/*
1850 	 * Verify user-supplied strings.
1851 	 * Check for null-terminated/zero-length strings/
1852 	 */
1853 	tname = oh->ntlv.name;
1854 	aname = i->algoname;
1855 	if (check_table_name(tname) != 0 ||
1856 	    strnlen(aname, sizeof(i->algoname)) == sizeof(i->algoname))
1857 		return (EINVAL);
1858 
1859 	if (aname[0] == '\0') {
1860 		/* Use default algorithm */
1861 		aname = NULL;
1862 	}
1863 
1864 	objheader_to_ti(oh, &ti);
1865 	ti.type = i->type;
1866 
1867 	ni = CHAIN_TO_NI(ch);
1868 
1869 	IPFW_UH_RLOCK(ch);
1870 	if (find_table(ni, &ti) != NULL) {
1871 		IPFW_UH_RUNLOCK(ch);
1872 		return (EEXIST);
1873 	}
1874 	IPFW_UH_RUNLOCK(ch);
1875 
1876 	return (create_table_internal(ch, &ti, aname, i, NULL, 0));
1877 }
1878 
1879 /*
1880  * Creates new table based on @ti and @aname.
1881  *
1882  * Relies on table name checking inside find_name_tlv()
1883  * Assume @aname to be checked and valid.
1884  * Stores allocated table kidx inside @pkidx (if non-NULL).
1885  * Reference created table if @compat is non-zero.
1886  *
1887  * Returns 0 on success.
1888  */
1889 static int
1890 create_table_internal(struct ip_fw_chain *ch, struct tid_info *ti,
1891     char *aname, ipfw_xtable_info *i, uint16_t *pkidx, int compat)
1892 {
1893 	struct namedobj_instance *ni;
1894 	struct table_config *tc, *tc_new, *tmp;
1895 	struct table_algo *ta;
1896 	uint16_t kidx;
1897 
1898 	ni = CHAIN_TO_NI(ch);
1899 
1900 	ta = find_table_algo(CHAIN_TO_TCFG(ch), ti, aname);
1901 	if (ta == NULL)
1902 		return (ENOTSUP);
1903 
1904 	tc = alloc_table_config(ch, ti, ta, aname, i->tflags);
1905 	if (tc == NULL)
1906 		return (ENOMEM);
1907 
1908 	tc->vmask = i->vmask;
1909 	tc->limit = i->limit;
1910 	if (ta->flags & TA_FLAG_READONLY)
1911 		tc->locked = 1;
1912 	else
1913 		tc->locked = (i->flags & IPFW_TGFLAGS_LOCKED) != 0;
1914 
1915 	IPFW_UH_WLOCK(ch);
1916 
1917 	/* Check if table has been already created */
1918 	tc_new = find_table(ni, ti);
1919 	if (tc_new != NULL) {
1920 
1921 		/*
1922 		 * Compat: do not fail if we're
1923 		 * requesting to create existing table
1924 		 * which has the same type
1925 		 */
1926 		if (compat == 0 || tc_new->no.subtype != tc->no.subtype) {
1927 			IPFW_UH_WUNLOCK(ch);
1928 			free_table_config(ni, tc);
1929 			return (EEXIST);
1930 		}
1931 
1932 		/* Exchange tc and tc_new for proper refcounting & freeing */
1933 		tmp = tc;
1934 		tc = tc_new;
1935 		tc_new = tmp;
1936 	} else {
1937 		/* New table */
1938 		if (ipfw_objhash_alloc_idx(ni, &kidx) != 0) {
1939 			IPFW_UH_WUNLOCK(ch);
1940 			printf("Unable to allocate table index."
1941 			    " Consider increasing net.inet.ip.fw.tables_max");
1942 			free_table_config(ni, tc);
1943 			return (EBUSY);
1944 		}
1945 		tc->no.kidx = kidx;
1946 		tc->no.etlv = IPFW_TLV_TBL_NAME;
1947 
1948 		IPFW_WLOCK(ch);
1949 		link_table(ch, tc);
1950 		IPFW_WUNLOCK(ch);
1951 	}
1952 
1953 	if (compat != 0)
1954 		tc->no.refcnt++;
1955 	if (pkidx != NULL)
1956 		*pkidx = tc->no.kidx;
1957 
1958 	IPFW_UH_WUNLOCK(ch);
1959 
1960 	if (tc_new != NULL)
1961 		free_table_config(ni, tc_new);
1962 
1963 	return (0);
1964 }
1965 
1966 static void
1967 ntlv_to_ti(ipfw_obj_ntlv *ntlv, struct tid_info *ti)
1968 {
1969 
1970 	memset(ti, 0, sizeof(struct tid_info));
1971 	ti->set = ntlv->set;
1972 	ti->uidx = ntlv->idx;
1973 	ti->tlvs = ntlv;
1974 	ti->tlen = ntlv->head.length;
1975 }
1976 
1977 static void
1978 objheader_to_ti(struct _ipfw_obj_header *oh, struct tid_info *ti)
1979 {
1980 
1981 	ntlv_to_ti(&oh->ntlv, ti);
1982 }
1983 
1984 struct namedobj_instance *
1985 ipfw_get_table_objhash(struct ip_fw_chain *ch)
1986 {
1987 
1988 	return (CHAIN_TO_NI(ch));
1989 }
1990 
1991 /*
1992  * Exports basic table info as name TLV.
1993  * Used inside dump_static_rules() to provide info
1994  * about all tables referenced by current ruleset.
1995  *
1996  * Returns 0 on success.
1997  */
1998 int
1999 ipfw_export_table_ntlv(struct ip_fw_chain *ch, uint16_t kidx,
2000     struct sockopt_data *sd)
2001 {
2002 	struct namedobj_instance *ni;
2003 	struct named_object *no;
2004 	ipfw_obj_ntlv *ntlv;
2005 
2006 	ni = CHAIN_TO_NI(ch);
2007 
2008 	no = ipfw_objhash_lookup_kidx(ni, kidx);
2009 	KASSERT(no != NULL, ("invalid table kidx passed"));
2010 
2011 	ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv));
2012 	if (ntlv == NULL)
2013 		return (ENOMEM);
2014 
2015 	ntlv->head.type = IPFW_TLV_TBL_NAME;
2016 	ntlv->head.length = sizeof(*ntlv);
2017 	ntlv->idx = no->kidx;
2018 	strlcpy(ntlv->name, no->name, sizeof(ntlv->name));
2019 
2020 	return (0);
2021 }
2022 
2023 struct dump_args {
2024 	struct ip_fw_chain *ch;
2025 	struct table_info *ti;
2026 	struct table_config *tc;
2027 	struct sockopt_data *sd;
2028 	uint32_t cnt;
2029 	uint16_t uidx;
2030 	int error;
2031 	uint32_t size;
2032 	ipfw_table_entry *ent;
2033 	ta_foreach_f *f;
2034 	void *farg;
2035 	ipfw_obj_tentry tent;
2036 };
2037 
2038 static int
2039 count_ext_entries(void *e, void *arg)
2040 {
2041 	struct dump_args *da;
2042 
2043 	da = (struct dump_args *)arg;
2044 	da->cnt++;
2045 
2046 	return (0);
2047 }
2048 
2049 /*
2050  * Gets number of items from table either using
2051  * internal counter or calling algo callback for
2052  * externally-managed tables.
2053  *
2054  * Returns number of records.
2055  */
2056 static uint32_t
2057 table_get_count(struct ip_fw_chain *ch, struct table_config *tc)
2058 {
2059 	struct table_info *ti;
2060 	struct table_algo *ta;
2061 	struct dump_args da;
2062 
2063 	ti = KIDX_TO_TI(ch, tc->no.kidx);
2064 	ta = tc->ta;
2065 
2066 	/* Use internal counter for self-managed tables */
2067 	if ((ta->flags & TA_FLAG_READONLY) == 0)
2068 		return (tc->count);
2069 
2070 	/* Use callback to quickly get number of items */
2071 	if ((ta->flags & TA_FLAG_EXTCOUNTER) != 0)
2072 		return (ta->get_count(tc->astate, ti));
2073 
2074 	/* Count number of iterms ourselves */
2075 	memset(&da, 0, sizeof(da));
2076 	ta->foreach(tc->astate, ti, count_ext_entries, &da);
2077 
2078 	return (da.cnt);
2079 }
2080 
2081 /*
2082  * Exports table @tc info into standard ipfw_xtable_info format.
2083  */
2084 static void
2085 export_table_info(struct ip_fw_chain *ch, struct table_config *tc,
2086     ipfw_xtable_info *i)
2087 {
2088 	struct table_info *ti;
2089 	struct table_algo *ta;
2090 
2091 	i->type = tc->no.subtype;
2092 	i->tflags = tc->tflags;
2093 	i->vmask = tc->vmask;
2094 	i->set = tc->no.set;
2095 	i->kidx = tc->no.kidx;
2096 	i->refcnt = tc->no.refcnt;
2097 	i->count = table_get_count(ch, tc);
2098 	i->limit = tc->limit;
2099 	i->flags |= (tc->locked != 0) ? IPFW_TGFLAGS_LOCKED : 0;
2100 	i->size = i->count * sizeof(ipfw_obj_tentry);
2101 	i->size += sizeof(ipfw_obj_header) + sizeof(ipfw_xtable_info);
2102 	strlcpy(i->tablename, tc->tablename, sizeof(i->tablename));
2103 	ti = KIDX_TO_TI(ch, tc->no.kidx);
2104 	ta = tc->ta;
2105 	if (ta->print_config != NULL) {
2106 		/* Use algo function to print table config to string */
2107 		ta->print_config(tc->astate, ti, i->algoname,
2108 		    sizeof(i->algoname));
2109 	} else
2110 		strlcpy(i->algoname, ta->name, sizeof(i->algoname));
2111 	/* Dump algo-specific data, if possible */
2112 	if (ta->dump_tinfo != NULL) {
2113 		ta->dump_tinfo(tc->astate, ti, &i->ta_info);
2114 		i->ta_info.flags |= IPFW_TATFLAGS_DATA;
2115 	}
2116 }
2117 
2118 struct dump_table_args {
2119 	struct ip_fw_chain *ch;
2120 	struct sockopt_data *sd;
2121 };
2122 
2123 static void
2124 export_table_internal(struct namedobj_instance *ni, struct named_object *no,
2125     void *arg)
2126 {
2127 	ipfw_xtable_info *i;
2128 	struct dump_table_args *dta;
2129 
2130 	dta = (struct dump_table_args *)arg;
2131 
2132 	i = (ipfw_xtable_info *)ipfw_get_sopt_space(dta->sd, sizeof(*i));
2133 	KASSERT(i != 0, ("previously checked buffer is not enough"));
2134 
2135 	export_table_info(dta->ch, (struct table_config *)no, i);
2136 }
2137 
2138 /*
2139  * Export all tables as ipfw_xtable_info structures to
2140  * storage provided by @sd.
2141  *
2142  * If supplied buffer is too small, fills in required size
2143  * and returns ENOMEM.
2144  * Returns 0 on success.
2145  */
2146 static int
2147 export_tables(struct ip_fw_chain *ch, ipfw_obj_lheader *olh,
2148     struct sockopt_data *sd)
2149 {
2150 	uint32_t size;
2151 	uint32_t count;
2152 	struct dump_table_args dta;
2153 
2154 	count = ipfw_objhash_count(CHAIN_TO_NI(ch));
2155 	size = count * sizeof(ipfw_xtable_info) + sizeof(ipfw_obj_lheader);
2156 
2157 	/* Fill in header regadless of buffer size */
2158 	olh->count = count;
2159 	olh->objsize = sizeof(ipfw_xtable_info);
2160 
2161 	if (size > olh->size) {
2162 		olh->size = size;
2163 		return (ENOMEM);
2164 	}
2165 
2166 	olh->size = size;
2167 
2168 	dta.ch = ch;
2169 	dta.sd = sd;
2170 
2171 	ipfw_objhash_foreach(CHAIN_TO_NI(ch), export_table_internal, &dta);
2172 
2173 	return (0);
2174 }
2175 
2176 /*
2177  * Dumps all table data
2178  * Data layout (v1)(current):
2179  * Request: [ ipfw_obj_header ], size = ipfw_xtable_info.size
2180  * Reply: [ ipfw_obj_header ipfw_xtable_info ipfw_obj_tentry x N ]
2181  *
2182  * Returns 0 on success
2183  */
2184 static int
2185 dump_table_v1(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
2186     struct sockopt_data *sd)
2187 {
2188 	struct _ipfw_obj_header *oh;
2189 	ipfw_xtable_info *i;
2190 	struct tid_info ti;
2191 	struct table_config *tc;
2192 	struct table_algo *ta;
2193 	struct dump_args da;
2194 	uint32_t sz;
2195 
2196 	sz = sizeof(ipfw_obj_header) + sizeof(ipfw_xtable_info);
2197 	oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz);
2198 	if (oh == NULL)
2199 		return (EINVAL);
2200 
2201 	i = (ipfw_xtable_info *)(oh + 1);
2202 	objheader_to_ti(oh, &ti);
2203 
2204 	IPFW_UH_RLOCK(ch);
2205 	if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) {
2206 		IPFW_UH_RUNLOCK(ch);
2207 		return (ESRCH);
2208 	}
2209 	export_table_info(ch, tc, i);
2210 
2211 	if (sd->valsize < i->size) {
2212 
2213 		/*
2214 		 * Submitted buffer size is not enough.
2215 		 * WE've already filled in @i structure with
2216 		 * relevant table info including size, so we
2217 		 * can return. Buffer will be flushed automatically.
2218 		 */
2219 		IPFW_UH_RUNLOCK(ch);
2220 		return (ENOMEM);
2221 	}
2222 
2223 	/*
2224 	 * Do the actual dump in eXtended format
2225 	 */
2226 	memset(&da, 0, sizeof(da));
2227 	da.ch = ch;
2228 	da.ti = KIDX_TO_TI(ch, tc->no.kidx);
2229 	da.tc = tc;
2230 	da.sd = sd;
2231 
2232 	ta = tc->ta;
2233 
2234 	ta->foreach(tc->astate, da.ti, dump_table_tentry, &da);
2235 	IPFW_UH_RUNLOCK(ch);
2236 
2237 	return (da.error);
2238 }
2239 
2240 /*
2241  * Dumps all table data
2242  * Data layout (version 0)(legacy):
2243  * Request: [ ipfw_xtable ], size = IP_FW_TABLE_XGETSIZE()
2244  * Reply: [ ipfw_xtable ipfw_table_xentry x N ]
2245  *
2246  * Returns 0 on success
2247  */
2248 static int
2249 dump_table_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
2250     struct sockopt_data *sd)
2251 {
2252 	ipfw_xtable *xtbl;
2253 	struct tid_info ti;
2254 	struct table_config *tc;
2255 	struct table_algo *ta;
2256 	struct dump_args da;
2257 	size_t sz, count;
2258 
2259 	xtbl = (ipfw_xtable *)ipfw_get_sopt_header(sd, sizeof(ipfw_xtable));
2260 	if (xtbl == NULL)
2261 		return (EINVAL);
2262 
2263 	memset(&ti, 0, sizeof(ti));
2264 	ti.uidx = xtbl->tbl;
2265 
2266 	IPFW_UH_RLOCK(ch);
2267 	if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) {
2268 		IPFW_UH_RUNLOCK(ch);
2269 		return (0);
2270 	}
2271 	count = table_get_count(ch, tc);
2272 	sz = count * sizeof(ipfw_table_xentry) + sizeof(ipfw_xtable);
2273 
2274 	xtbl->cnt = count;
2275 	xtbl->size = sz;
2276 	xtbl->type = tc->no.subtype;
2277 	xtbl->tbl = ti.uidx;
2278 
2279 	if (sd->valsize < sz) {
2280 
2281 		/*
2282 		 * Submitted buffer size is not enough.
2283 		 * WE've already filled in @i structure with
2284 		 * relevant table info including size, so we
2285 		 * can return. Buffer will be flushed automatically.
2286 		 */
2287 		IPFW_UH_RUNLOCK(ch);
2288 		return (ENOMEM);
2289 	}
2290 
2291 	/* Do the actual dump in eXtended format */
2292 	memset(&da, 0, sizeof(da));
2293 	da.ch = ch;
2294 	da.ti = KIDX_TO_TI(ch, tc->no.kidx);
2295 	da.tc = tc;
2296 	da.sd = sd;
2297 
2298 	ta = tc->ta;
2299 
2300 	ta->foreach(tc->astate, da.ti, dump_table_xentry, &da);
2301 	IPFW_UH_RUNLOCK(ch);
2302 
2303 	return (0);
2304 }
2305 
2306 /*
2307  * Legacy function to retrieve number of items in table.
2308  */
2309 static int
2310 get_table_size(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
2311     struct sockopt_data *sd)
2312 {
2313 	uint32_t *tbl;
2314 	struct tid_info ti;
2315 	size_t sz;
2316 	int error;
2317 
2318 	sz = sizeof(*op3) + sizeof(uint32_t);
2319 	op3 = (ip_fw3_opheader *)ipfw_get_sopt_header(sd, sz);
2320 	if (op3 == NULL)
2321 		return (EINVAL);
2322 
2323 	tbl = (uint32_t *)(op3 + 1);
2324 	memset(&ti, 0, sizeof(ti));
2325 	ti.uidx = *tbl;
2326 	IPFW_UH_RLOCK(ch);
2327 	error = ipfw_count_xtable(ch, &ti, tbl);
2328 	IPFW_UH_RUNLOCK(ch);
2329 	return (error);
2330 }
2331 
2332 /*
2333  * Legacy IP_FW_TABLE_GETSIZE handler
2334  */
2335 int
2336 ipfw_count_table(struct ip_fw_chain *ch, struct tid_info *ti, uint32_t *cnt)
2337 {
2338 	struct table_config *tc;
2339 
2340 	if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL)
2341 		return (ESRCH);
2342 	*cnt = table_get_count(ch, tc);
2343 	return (0);
2344 }
2345 
2346 /*
2347  * Legacy IP_FW_TABLE_XGETSIZE handler
2348  */
2349 int
2350 ipfw_count_xtable(struct ip_fw_chain *ch, struct tid_info *ti, uint32_t *cnt)
2351 {
2352 	struct table_config *tc;
2353 	uint32_t count;
2354 
2355 	if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL) {
2356 		*cnt = 0;
2357 		return (0); /* 'table all list' requires success */
2358 	}
2359 
2360 	count = table_get_count(ch, tc);
2361 	*cnt = count * sizeof(ipfw_table_xentry);
2362 	if (count > 0)
2363 		*cnt += sizeof(ipfw_xtable);
2364 	return (0);
2365 }
2366 
2367 static int
2368 dump_table_entry(void *e, void *arg)
2369 {
2370 	struct dump_args *da;
2371 	struct table_config *tc;
2372 	struct table_algo *ta;
2373 	ipfw_table_entry *ent;
2374 	struct table_value *pval;
2375 	int error;
2376 
2377 	da = (struct dump_args *)arg;
2378 
2379 	tc = da->tc;
2380 	ta = tc->ta;
2381 
2382 	/* Out of memory, returning */
2383 	if (da->cnt == da->size)
2384 		return (1);
2385 	ent = da->ent++;
2386 	ent->tbl = da->uidx;
2387 	da->cnt++;
2388 
2389 	error = ta->dump_tentry(tc->astate, da->ti, e, &da->tent);
2390 	if (error != 0)
2391 		return (error);
2392 
2393 	ent->addr = da->tent.k.addr.s_addr;
2394 	ent->masklen = da->tent.masklen;
2395 	pval = get_table_value(da->ch, da->tc, da->tent.v.kidx);
2396 	ent->value = ipfw_export_table_value_legacy(pval);
2397 
2398 	return (0);
2399 }
2400 
2401 /*
2402  * Dumps table in pre-8.1 legacy format.
2403  */
2404 int
2405 ipfw_dump_table_legacy(struct ip_fw_chain *ch, struct tid_info *ti,
2406     ipfw_table *tbl)
2407 {
2408 	struct table_config *tc;
2409 	struct table_algo *ta;
2410 	struct dump_args da;
2411 
2412 	tbl->cnt = 0;
2413 
2414 	if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL)
2415 		return (0);	/* XXX: We should return ESRCH */
2416 
2417 	ta = tc->ta;
2418 
2419 	/* This dump format supports IPv4 only */
2420 	if (tc->no.subtype != IPFW_TABLE_ADDR)
2421 		return (0);
2422 
2423 	memset(&da, 0, sizeof(da));
2424 	da.ch = ch;
2425 	da.ti = KIDX_TO_TI(ch, tc->no.kidx);
2426 	da.tc = tc;
2427 	da.ent = &tbl->ent[0];
2428 	da.size = tbl->size;
2429 
2430 	tbl->cnt = 0;
2431 	ta->foreach(tc->astate, da.ti, dump_table_entry, &da);
2432 	tbl->cnt = da.cnt;
2433 
2434 	return (0);
2435 }
2436 
2437 /*
2438  * Dumps table entry in eXtended format (v1)(current).
2439  */
2440 static int
2441 dump_table_tentry(void *e, void *arg)
2442 {
2443 	struct dump_args *da;
2444 	struct table_config *tc;
2445 	struct table_algo *ta;
2446 	struct table_value *pval;
2447 	ipfw_obj_tentry *tent;
2448 	int error;
2449 
2450 	da = (struct dump_args *)arg;
2451 
2452 	tc = da->tc;
2453 	ta = tc->ta;
2454 
2455 	tent = (ipfw_obj_tentry *)ipfw_get_sopt_space(da->sd, sizeof(*tent));
2456 	/* Out of memory, returning */
2457 	if (tent == NULL) {
2458 		da->error = ENOMEM;
2459 		return (1);
2460 	}
2461 	tent->head.length = sizeof(ipfw_obj_tentry);
2462 	tent->idx = da->uidx;
2463 
2464 	error = ta->dump_tentry(tc->astate, da->ti, e, tent);
2465 	if (error != 0)
2466 		return (error);
2467 
2468 	pval = get_table_value(da->ch, da->tc, tent->v.kidx);
2469 	ipfw_export_table_value_v1(pval, &tent->v.value);
2470 
2471 	return (0);
2472 }
2473 
2474 /*
2475  * Dumps table entry in eXtended format (v0).
2476  */
2477 static int
2478 dump_table_xentry(void *e, void *arg)
2479 {
2480 	struct dump_args *da;
2481 	struct table_config *tc;
2482 	struct table_algo *ta;
2483 	ipfw_table_xentry *xent;
2484 	ipfw_obj_tentry *tent;
2485 	struct table_value *pval;
2486 	int error;
2487 
2488 	da = (struct dump_args *)arg;
2489 
2490 	tc = da->tc;
2491 	ta = tc->ta;
2492 
2493 	xent = (ipfw_table_xentry *)ipfw_get_sopt_space(da->sd, sizeof(*xent));
2494 	/* Out of memory, returning */
2495 	if (xent == NULL)
2496 		return (1);
2497 	xent->len = sizeof(ipfw_table_xentry);
2498 	xent->tbl = da->uidx;
2499 
2500 	memset(&da->tent, 0, sizeof(da->tent));
2501 	tent = &da->tent;
2502 	error = ta->dump_tentry(tc->astate, da->ti, e, tent);
2503 	if (error != 0)
2504 		return (error);
2505 
2506 	/* Convert current format to previous one */
2507 	xent->masklen = tent->masklen;
2508 	pval = get_table_value(da->ch, da->tc, da->tent.v.kidx);
2509 	xent->value = ipfw_export_table_value_legacy(pval);
2510 	/* Apply some hacks */
2511 	if (tc->no.subtype == IPFW_TABLE_ADDR && tent->subtype == AF_INET) {
2512 		xent->k.addr6.s6_addr32[3] = tent->k.addr.s_addr;
2513 		xent->flags = IPFW_TCF_INET;
2514 	} else
2515 		memcpy(&xent->k, &tent->k, sizeof(xent->k));
2516 
2517 	return (0);
2518 }
2519 
2520 /*
2521  * Helper function to export table algo data
2522  * to tentry format before calling user function.
2523  *
2524  * Returns 0 on success.
2525  */
2526 static int
2527 prepare_table_tentry(void *e, void *arg)
2528 {
2529 	struct dump_args *da;
2530 	struct table_config *tc;
2531 	struct table_algo *ta;
2532 	int error;
2533 
2534 	da = (struct dump_args *)arg;
2535 
2536 	tc = da->tc;
2537 	ta = tc->ta;
2538 
2539 	error = ta->dump_tentry(tc->astate, da->ti, e, &da->tent);
2540 	if (error != 0)
2541 		return (error);
2542 
2543 	da->f(&da->tent, da->farg);
2544 
2545 	return (0);
2546 }
2547 
2548 /*
2549  * Allow external consumers to read table entries in standard format.
2550  */
2551 int
2552 ipfw_foreach_table_tentry(struct ip_fw_chain *ch, uint16_t kidx,
2553     ta_foreach_f *f, void *arg)
2554 {
2555 	struct namedobj_instance *ni;
2556 	struct table_config *tc;
2557 	struct table_algo *ta;
2558 	struct dump_args da;
2559 
2560 	ni = CHAIN_TO_NI(ch);
2561 
2562 	tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, kidx);
2563 	if (tc == NULL)
2564 		return (ESRCH);
2565 
2566 	ta = tc->ta;
2567 
2568 	memset(&da, 0, sizeof(da));
2569 	da.ch = ch;
2570 	da.ti = KIDX_TO_TI(ch, tc->no.kidx);
2571 	da.tc = tc;
2572 	da.f = f;
2573 	da.farg = arg;
2574 
2575 	ta->foreach(tc->astate, da.ti, prepare_table_tentry, &da);
2576 
2577 	return (0);
2578 }
2579 
2580 /*
2581  * Table algorithms
2582  */
2583 
2584 /*
2585  * Finds algoritm by index, table type or supplied name.
2586  *
2587  * Returns pointer to algo or NULL.
2588  */
2589 static struct table_algo *
2590 find_table_algo(struct tables_config *tcfg, struct tid_info *ti, char *name)
2591 {
2592 	int i, l;
2593 	struct table_algo *ta;
2594 
2595 	if (ti->type > IPFW_TABLE_MAXTYPE)
2596 		return (NULL);
2597 
2598 	/* Search by index */
2599 	if (ti->atype != 0) {
2600 		if (ti->atype > tcfg->algo_count)
2601 			return (NULL);
2602 		return (tcfg->algo[ti->atype]);
2603 	}
2604 
2605 	if (name == NULL) {
2606 		/* Return default algorithm for given type if set */
2607 		return (tcfg->def_algo[ti->type]);
2608 	}
2609 
2610 	/* Search by name */
2611 	/* TODO: better search */
2612 	for (i = 1; i <= tcfg->algo_count; i++) {
2613 		ta = tcfg->algo[i];
2614 
2615 		/*
2616 		 * One can supply additional algorithm
2617 		 * parameters so we compare only the first word
2618 		 * of supplied name:
2619 		 * 'addr:chash hsize=32'
2620 		 * '^^^^^^^^^'
2621 		 *
2622 		 */
2623 		l = strlen(ta->name);
2624 		if (strncmp(name, ta->name, l) != 0)
2625 			continue;
2626 		if (name[l] != '\0' && name[l] != ' ')
2627 			continue;
2628 		/* Check if we're requesting proper table type */
2629 		if (ti->type != 0 && ti->type != ta->type)
2630 			return (NULL);
2631 		return (ta);
2632 	}
2633 
2634 	return (NULL);
2635 }
2636 
2637 /*
2638  * Register new table algo @ta.
2639  * Stores algo id inside @idx.
2640  *
2641  * Returns 0 on success.
2642  */
2643 int
2644 ipfw_add_table_algo(struct ip_fw_chain *ch, struct table_algo *ta, size_t size,
2645     int *idx)
2646 {
2647 	struct tables_config *tcfg;
2648 	struct table_algo *ta_new;
2649 	size_t sz;
2650 
2651 	if (size > sizeof(struct table_algo))
2652 		return (EINVAL);
2653 
2654 	/* Check for the required on-stack size for add/del */
2655 	sz = roundup2(ta->ta_buf_size, sizeof(void *));
2656 	if (sz > TA_BUF_SZ)
2657 		return (EINVAL);
2658 
2659 	KASSERT(ta->type <= IPFW_TABLE_MAXTYPE,("Increase IPFW_TABLE_MAXTYPE"));
2660 
2661 	/* Copy algorithm data to stable storage. */
2662 	ta_new = malloc(sizeof(struct table_algo), M_IPFW, M_WAITOK | M_ZERO);
2663 	memcpy(ta_new, ta, size);
2664 
2665 	tcfg = CHAIN_TO_TCFG(ch);
2666 
2667 	KASSERT(tcfg->algo_count < 255, ("Increase algo array size"));
2668 
2669 	tcfg->algo[++tcfg->algo_count] = ta_new;
2670 	ta_new->idx = tcfg->algo_count;
2671 
2672 	/* Set algorithm as default one for given type */
2673 	if ((ta_new->flags & TA_FLAG_DEFAULT) != 0 &&
2674 	    tcfg->def_algo[ta_new->type] == NULL)
2675 		tcfg->def_algo[ta_new->type] = ta_new;
2676 
2677 	*idx = ta_new->idx;
2678 
2679 	return (0);
2680 }
2681 
2682 /*
2683  * Unregisters table algo using @idx as id.
2684  * XXX: It is NOT safe to call this function in any place
2685  * other than ipfw instance destroy handler.
2686  */
2687 void
2688 ipfw_del_table_algo(struct ip_fw_chain *ch, int idx)
2689 {
2690 	struct tables_config *tcfg;
2691 	struct table_algo *ta;
2692 
2693 	tcfg = CHAIN_TO_TCFG(ch);
2694 
2695 	KASSERT(idx <= tcfg->algo_count, ("algo idx %d out of range 1..%d",
2696 	    idx, tcfg->algo_count));
2697 
2698 	ta = tcfg->algo[idx];
2699 	KASSERT(ta != NULL, ("algo idx %d is NULL", idx));
2700 
2701 	if (tcfg->def_algo[ta->type] == ta)
2702 		tcfg->def_algo[ta->type] = NULL;
2703 
2704 	free(ta, M_IPFW);
2705 }
2706 
2707 /*
2708  * Lists all table algorithms currently available.
2709  * Data layout (v0)(current):
2710  * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size
2711  * Reply: [ ipfw_obj_lheader ipfw_ta_info x N ]
2712  *
2713  * Returns 0 on success
2714  */
2715 static int
2716 list_table_algo(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
2717     struct sockopt_data *sd)
2718 {
2719 	struct _ipfw_obj_lheader *olh;
2720 	struct tables_config *tcfg;
2721 	ipfw_ta_info *i;
2722 	struct table_algo *ta;
2723 	uint32_t count, n, size;
2724 
2725 	olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh));
2726 	if (olh == NULL)
2727 		return (EINVAL);
2728 	if (sd->valsize < olh->size)
2729 		return (EINVAL);
2730 
2731 	IPFW_UH_RLOCK(ch);
2732 	tcfg = CHAIN_TO_TCFG(ch);
2733 	count = tcfg->algo_count;
2734 	size = count * sizeof(ipfw_ta_info) + sizeof(ipfw_obj_lheader);
2735 
2736 	/* Fill in header regadless of buffer size */
2737 	olh->count = count;
2738 	olh->objsize = sizeof(ipfw_ta_info);
2739 
2740 	if (size > olh->size) {
2741 		olh->size = size;
2742 		IPFW_UH_RUNLOCK(ch);
2743 		return (ENOMEM);
2744 	}
2745 	olh->size = size;
2746 
2747 	for (n = 1; n <= count; n++) {
2748 		i = (ipfw_ta_info *)ipfw_get_sopt_space(sd, sizeof(*i));
2749 		KASSERT(i != 0, ("previously checked buffer is not enough"));
2750 		ta = tcfg->algo[n];
2751 		strlcpy(i->algoname, ta->name, sizeof(i->algoname));
2752 		i->type = ta->type;
2753 		i->refcnt = ta->refcnt;
2754 	}
2755 
2756 	IPFW_UH_RUNLOCK(ch);
2757 
2758 	return (0);
2759 }
2760 
2761 static int
2762 classify_srcdst(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype)
2763 {
2764 	/* Basic IPv4/IPv6 or u32 lookups */
2765 	*puidx = cmd->arg1;
2766 	/* Assume ADDR by default */
2767 	*ptype = IPFW_TABLE_ADDR;
2768 	int v;
2769 
2770 	if (F_LEN(cmd) > F_INSN_SIZE(ipfw_insn_u32)) {
2771 		/*
2772 		 * generic lookup. The key must be
2773 		 * in 32bit big-endian format.
2774 		 */
2775 		v = ((ipfw_insn_u32 *)cmd)->d[1];
2776 		switch (v) {
2777 		case 0:
2778 		case 1:
2779 			/* IPv4 src/dst */
2780 			break;
2781 		case 2:
2782 		case 3:
2783 			/* src/dst port */
2784 			*ptype = IPFW_TABLE_NUMBER;
2785 			break;
2786 		case 4:
2787 			/* uid/gid */
2788 			*ptype = IPFW_TABLE_NUMBER;
2789 			break;
2790 		case 5:
2791 			/* jid */
2792 			*ptype = IPFW_TABLE_NUMBER;
2793 			break;
2794 		case 6:
2795 			/* dscp */
2796 			*ptype = IPFW_TABLE_NUMBER;
2797 			break;
2798 		}
2799 	}
2800 
2801 	return (0);
2802 }
2803 
2804 static int
2805 classify_via(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype)
2806 {
2807 	ipfw_insn_if *cmdif;
2808 
2809 	/* Interface table, possibly */
2810 	cmdif = (ipfw_insn_if *)cmd;
2811 	if (cmdif->name[0] != '\1')
2812 		return (1);
2813 
2814 	*ptype = IPFW_TABLE_INTERFACE;
2815 	*puidx = cmdif->p.kidx;
2816 
2817 	return (0);
2818 }
2819 
2820 static int
2821 classify_flow(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype)
2822 {
2823 
2824 	*puidx = cmd->arg1;
2825 	*ptype = IPFW_TABLE_FLOW;
2826 
2827 	return (0);
2828 }
2829 
2830 static void
2831 update_arg1(ipfw_insn *cmd, uint16_t idx)
2832 {
2833 
2834 	cmd->arg1 = idx;
2835 }
2836 
2837 static void
2838 update_via(ipfw_insn *cmd, uint16_t idx)
2839 {
2840 	ipfw_insn_if *cmdif;
2841 
2842 	cmdif = (ipfw_insn_if *)cmd;
2843 	cmdif->p.kidx = idx;
2844 }
2845 
2846 static int
2847 table_findbyname(struct ip_fw_chain *ch, struct tid_info *ti,
2848     struct named_object **pno)
2849 {
2850 	struct table_config *tc;
2851 	int error;
2852 
2853 	IPFW_UH_WLOCK_ASSERT(ch);
2854 
2855 	error = find_table_err(CHAIN_TO_NI(ch), ti, &tc);
2856 	if (error != 0)
2857 		return (error);
2858 
2859 	*pno = &tc->no;
2860 	return (0);
2861 }
2862 
2863 /* XXX: sets-sets! */
2864 static struct named_object *
2865 table_findbykidx(struct ip_fw_chain *ch, uint16_t idx)
2866 {
2867 	struct namedobj_instance *ni;
2868 	struct table_config *tc;
2869 
2870 	IPFW_UH_WLOCK_ASSERT(ch);
2871 	ni = CHAIN_TO_NI(ch);
2872 	tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, idx);
2873 	KASSERT(tc != NULL, ("Table with index %d not found", idx));
2874 
2875 	return (&tc->no);
2876 }
2877 
2878 static struct opcode_obj_rewrite opcodes[] = {
2879 	{
2880 		O_IP_SRC_LOOKUP, IPFW_TLV_TBL_NAME,
2881 		classify_srcdst, update_arg1,
2882 		table_findbyname, table_findbykidx, create_table_compat
2883 	},
2884 	{
2885 		O_IP_DST_LOOKUP, IPFW_TLV_TBL_NAME,
2886 		classify_srcdst, update_arg1,
2887 		table_findbyname, table_findbykidx, create_table_compat
2888 	},
2889 	{
2890 		O_IP_FLOW_LOOKUP, IPFW_TLV_TBL_NAME,
2891 		classify_flow, update_arg1,
2892 		table_findbyname, table_findbykidx, create_table_compat
2893 	},
2894 	{
2895 		O_XMIT, IPFW_TLV_TBL_NAME,
2896 		classify_via, update_via,
2897 		table_findbyname, table_findbykidx, create_table_compat
2898 	},
2899 	{
2900 		O_RECV, IPFW_TLV_TBL_NAME,
2901 		classify_via, update_via,
2902 		table_findbyname, table_findbykidx, create_table_compat
2903 	},
2904 	{
2905 		O_VIA, IPFW_TLV_TBL_NAME,
2906 		classify_via, update_via,
2907 		table_findbyname, table_findbykidx, create_table_compat
2908 	},
2909 };
2910 
2911 
2912 /*
2913  * Checks table name for validity.
2914  * Enforce basic length checks, the rest
2915  * should be done in userland.
2916  *
2917  * Returns 0 if name is considered valid.
2918  */
2919 static int
2920 check_table_name(const char *name)
2921 {
2922 
2923 	/*
2924 	 * TODO: do some more complicated checks
2925 	 */
2926 	return (ipfw_check_object_name_generic(name));
2927 }
2928 
2929 /*
2930  * Find tablename TLV by @uid.
2931  * Check @tlvs for valid data inside.
2932  *
2933  * Returns pointer to found TLV or NULL.
2934  */
2935 static ipfw_obj_ntlv *
2936 find_name_tlv(void *tlvs, int len, uint16_t uidx)
2937 {
2938 	ipfw_obj_ntlv *ntlv;
2939 	uintptr_t pa, pe;
2940 	int l;
2941 
2942 	pa = (uintptr_t)tlvs;
2943 	pe = pa + len;
2944 	l = 0;
2945 	for (; pa < pe; pa += l) {
2946 		ntlv = (ipfw_obj_ntlv *)pa;
2947 		l = ntlv->head.length;
2948 
2949 		if (l != sizeof(*ntlv))
2950 			return (NULL);
2951 
2952 		if (ntlv->head.type != IPFW_TLV_TBL_NAME)
2953 			continue;
2954 
2955 		if (ntlv->idx != uidx)
2956 			continue;
2957 
2958 		if (check_table_name(ntlv->name) != 0)
2959 			return (NULL);
2960 
2961 		return (ntlv);
2962 	}
2963 
2964 	return (NULL);
2965 }
2966 
2967 /*
2968  * Finds table config based on either legacy index
2969  * or name in ntlv.
2970  * Note @ti structure contains unchecked data from userland.
2971  *
2972  * Returns 0 in success and fills in @tc with found config
2973  */
2974 static int
2975 find_table_err(struct namedobj_instance *ni, struct tid_info *ti,
2976     struct table_config **tc)
2977 {
2978 	char *name, bname[16];
2979 	struct named_object *no;
2980 	ipfw_obj_ntlv *ntlv;
2981 	uint32_t set;
2982 
2983 	if (ti->tlvs != NULL) {
2984 		ntlv = find_name_tlv(ti->tlvs, ti->tlen, ti->uidx);
2985 		if (ntlv == NULL)
2986 			return (EINVAL);
2987 		name = ntlv->name;
2988 
2989 		/*
2990 		 * Use set provided by @ti instead of @ntlv one.
2991 		 * This is needed due to different sets behavior
2992 		 * controlled by V_fw_tables_sets.
2993 		 */
2994 		set = ti->set;
2995 	} else {
2996 		snprintf(bname, sizeof(bname), "%d", ti->uidx);
2997 		name = bname;
2998 		set = 0;
2999 	}
3000 
3001 	no = ipfw_objhash_lookup_name(ni, set, name);
3002 	*tc = (struct table_config *)no;
3003 
3004 	return (0);
3005 }
3006 
3007 /*
3008  * Finds table config based on either legacy index
3009  * or name in ntlv.
3010  * Note @ti structure contains unchecked data from userland.
3011  *
3012  * Returns pointer to table_config or NULL.
3013  */
3014 static struct table_config *
3015 find_table(struct namedobj_instance *ni, struct tid_info *ti)
3016 {
3017 	struct table_config *tc;
3018 
3019 	if (find_table_err(ni, ti, &tc) != 0)
3020 		return (NULL);
3021 
3022 	return (tc);
3023 }
3024 
3025 /*
3026  * Allocate new table config structure using
3027  * specified @algo and @aname.
3028  *
3029  * Returns pointer to config or NULL.
3030  */
3031 static struct table_config *
3032 alloc_table_config(struct ip_fw_chain *ch, struct tid_info *ti,
3033     struct table_algo *ta, char *aname, uint8_t tflags)
3034 {
3035 	char *name, bname[16];
3036 	struct table_config *tc;
3037 	int error;
3038 	ipfw_obj_ntlv *ntlv;
3039 	uint32_t set;
3040 
3041 	if (ti->tlvs != NULL) {
3042 		ntlv = find_name_tlv(ti->tlvs, ti->tlen, ti->uidx);
3043 		if (ntlv == NULL)
3044 			return (NULL);
3045 		name = ntlv->name;
3046 		set = ntlv->set;
3047 	} else {
3048 		/* Compat part: convert number to string representation */
3049 		snprintf(bname, sizeof(bname), "%d", ti->uidx);
3050 		name = bname;
3051 		set = 0;
3052 	}
3053 
3054 	tc = malloc(sizeof(struct table_config), M_IPFW, M_WAITOK | M_ZERO);
3055 	tc->no.name = tc->tablename;
3056 	tc->no.subtype = ta->type;
3057 	tc->no.set = set;
3058 	tc->tflags = tflags;
3059 	tc->ta = ta;
3060 	strlcpy(tc->tablename, name, sizeof(tc->tablename));
3061 	/* Set "shared" value type by default */
3062 	tc->vshared = 1;
3063 
3064 	/* Preallocate data structures for new tables */
3065 	error = ta->init(ch, &tc->astate, &tc->ti_copy, aname, tflags);
3066 	if (error != 0) {
3067 		free(tc, M_IPFW);
3068 		return (NULL);
3069 	}
3070 
3071 	return (tc);
3072 }
3073 
3074 /*
3075  * Destroys table state and config.
3076  */
3077 static void
3078 free_table_config(struct namedobj_instance *ni, struct table_config *tc)
3079 {
3080 
3081 	KASSERT(tc->linked == 0, ("free() on linked config"));
3082 	/* UH lock MUST NOT be held */
3083 
3084 	/*
3085 	 * We're using ta without any locking/referencing.
3086 	 * TODO: fix this if we're going to use unloadable algos.
3087 	 */
3088 	tc->ta->destroy(tc->astate, &tc->ti_copy);
3089 	free(tc, M_IPFW);
3090 }
3091 
3092 /*
3093  * Links @tc to @chain table named instance.
3094  * Sets appropriate type/states in @chain table info.
3095  */
3096 static void
3097 link_table(struct ip_fw_chain *ch, struct table_config *tc)
3098 {
3099 	struct namedobj_instance *ni;
3100 	struct table_info *ti;
3101 	uint16_t kidx;
3102 
3103 	IPFW_UH_WLOCK_ASSERT(ch);
3104 	IPFW_WLOCK_ASSERT(ch);
3105 
3106 	ni = CHAIN_TO_NI(ch);
3107 	kidx = tc->no.kidx;
3108 
3109 	ipfw_objhash_add(ni, &tc->no);
3110 
3111 	ti = KIDX_TO_TI(ch, kidx);
3112 	*ti = tc->ti_copy;
3113 
3114 	/* Notify algo on real @ti address */
3115 	if (tc->ta->change_ti != NULL)
3116 		tc->ta->change_ti(tc->astate, ti);
3117 
3118 	tc->linked = 1;
3119 	tc->ta->refcnt++;
3120 }
3121 
3122 /*
3123  * Unlinks @tc from @chain table named instance.
3124  * Zeroes states in @chain and stores them in @tc.
3125  */
3126 static void
3127 unlink_table(struct ip_fw_chain *ch, struct table_config *tc)
3128 {
3129 	struct namedobj_instance *ni;
3130 	struct table_info *ti;
3131 	uint16_t kidx;
3132 
3133 	IPFW_UH_WLOCK_ASSERT(ch);
3134 	IPFW_WLOCK_ASSERT(ch);
3135 
3136 	ni = CHAIN_TO_NI(ch);
3137 	kidx = tc->no.kidx;
3138 
3139 	/* Clear state. @ti copy is already saved inside @tc */
3140 	ipfw_objhash_del(ni, &tc->no);
3141 	ti = KIDX_TO_TI(ch, kidx);
3142 	memset(ti, 0, sizeof(struct table_info));
3143 	tc->linked = 0;
3144 	tc->ta->refcnt--;
3145 
3146 	/* Notify algo on real @ti address */
3147 	if (tc->ta->change_ti != NULL)
3148 		tc->ta->change_ti(tc->astate, NULL);
3149 }
3150 
3151 struct swap_table_args {
3152 	int set;
3153 	int new_set;
3154 	int mv;
3155 };
3156 
3157 /*
3158  * Change set for each matching table.
3159  *
3160  * Ensure we dispatch each table once by setting/checking ochange
3161  * fields.
3162  */
3163 static void
3164 swap_table_set(struct namedobj_instance *ni, struct named_object *no,
3165     void *arg)
3166 {
3167 	struct table_config *tc;
3168 	struct swap_table_args *sta;
3169 
3170 	tc = (struct table_config *)no;
3171 	sta = (struct swap_table_args *)arg;
3172 
3173 	if (no->set != sta->set && (no->set != sta->new_set || sta->mv != 0))
3174 		return;
3175 
3176 	if (tc->ochanged != 0)
3177 		return;
3178 
3179 	tc->ochanged = 1;
3180 	ipfw_objhash_del(ni, no);
3181 	if (no->set == sta->set)
3182 		no->set = sta->new_set;
3183 	else
3184 		no->set = sta->set;
3185 	ipfw_objhash_add(ni, no);
3186 }
3187 
3188 /*
3189  * Cleans up ochange field for all tables.
3190  */
3191 static void
3192 clean_table_set_data(struct namedobj_instance *ni, struct named_object *no,
3193     void *arg)
3194 {
3195 	struct table_config *tc;
3196 	struct swap_table_args *sta;
3197 
3198 	tc = (struct table_config *)no;
3199 	sta = (struct swap_table_args *)arg;
3200 
3201 	tc->ochanged = 0;
3202 }
3203 
3204 /*
3205  * Swaps tables within two sets.
3206  */
3207 void
3208 ipfw_swap_tables_sets(struct ip_fw_chain *ch, uint32_t set,
3209     uint32_t new_set, int mv)
3210 {
3211 	struct swap_table_args sta;
3212 
3213 	IPFW_UH_WLOCK_ASSERT(ch);
3214 
3215 	sta.set = set;
3216 	sta.new_set = new_set;
3217 	sta.mv = mv;
3218 
3219 	ipfw_objhash_foreach(CHAIN_TO_NI(ch), swap_table_set, &sta);
3220 	ipfw_objhash_foreach(CHAIN_TO_NI(ch), clean_table_set_data, &sta);
3221 }
3222 
3223 /*
3224  * Move all tables which are reference by rules in @rr to set @new_set.
3225  * Makes sure that all relevant tables are referenced ONLLY by given rules.
3226  *
3227  * Retuns 0 on success,
3228  */
3229 int
3230 ipfw_move_tables_sets(struct ip_fw_chain *ch, ipfw_range_tlv *rt,
3231     uint32_t new_set)
3232 {
3233 	struct ip_fw *rule;
3234 	struct table_config *tc;
3235 	struct named_object *no;
3236 	struct namedobj_instance *ni;
3237 	int bad, i, l, cmdlen;
3238 	uint16_t kidx;
3239 	ipfw_insn *cmd;
3240 
3241 	IPFW_UH_WLOCK_ASSERT(ch);
3242 
3243 	ni = CHAIN_TO_NI(ch);
3244 
3245 	/* Stage 1: count number of references by given rules */
3246 	for (i = 0; i < ch->n_rules - 1; i++) {
3247 		rule = ch->map[i];
3248 		if (ipfw_match_range(rule, rt) == 0)
3249 			continue;
3250 
3251 		l = rule->cmd_len;
3252 		cmd = rule->cmd;
3253 		cmdlen = 0;
3254 		for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
3255 			cmdlen = F_LEN(cmd);
3256 			if (classify_opcode_kidx(cmd, &kidx) != 0)
3257 				continue;
3258 			no = ipfw_objhash_lookup_kidx(ni, kidx);
3259 			KASSERT(no != NULL,
3260 			    ("objhash lookup failed on index %d", kidx));
3261 			tc = (struct table_config *)no;
3262 			tc->ocount++;
3263 		}
3264 
3265 	}
3266 
3267 	/* Stage 2: verify "ownership" */
3268 	bad = 0;
3269 	for (i = 0; i < ch->n_rules - 1; i++) {
3270 		rule = ch->map[i];
3271 		if (ipfw_match_range(rule, rt) == 0)
3272 			continue;
3273 
3274 		l = rule->cmd_len;
3275 		cmd = rule->cmd;
3276 		cmdlen = 0;
3277 		for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
3278 			cmdlen = F_LEN(cmd);
3279 			if (classify_opcode_kidx(cmd, &kidx) != 0)
3280 				continue;
3281 			no = ipfw_objhash_lookup_kidx(ni, kidx);
3282 			KASSERT(no != NULL,
3283 			    ("objhash lookup failed on index %d", kidx));
3284 			tc = (struct table_config *)no;
3285 			if (tc->no.refcnt != tc->ocount) {
3286 
3287 				/*
3288 				 * Number of references differ:
3289 				 * Other rule(s) are holding reference to given
3290 				 * table, so it is not possible to change its set.
3291 				 *
3292 				 * Note that refcnt may account
3293 				 * references to some going-to-be-added rules.
3294 				 * Since we don't know their numbers (and event
3295 				 * if they will be added) it is perfectly OK
3296 				 * to return error here.
3297 				 */
3298 				bad = 1;
3299 				break;
3300 			}
3301 		}
3302 
3303 		if (bad != 0)
3304 			break;
3305 	}
3306 
3307 	/* Stage 3: change set or cleanup */
3308 	for (i = 0; i < ch->n_rules - 1; i++) {
3309 		rule = ch->map[i];
3310 		if (ipfw_match_range(rule, rt) == 0)
3311 			continue;
3312 
3313 		l = rule->cmd_len;
3314 		cmd = rule->cmd;
3315 		cmdlen = 0;
3316 		for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
3317 			cmdlen = F_LEN(cmd);
3318 			if (classify_opcode_kidx(cmd, &kidx) != 0)
3319 				continue;
3320 			no = ipfw_objhash_lookup_kidx(ni, kidx);
3321 			KASSERT(no != NULL,
3322 			    ("objhash lookup failed on index %d", kidx));
3323 			tc = (struct table_config *)no;
3324 
3325 			tc->ocount = 0;
3326 			if (bad != 0)
3327 				continue;
3328 
3329 			/* Actually change set. */
3330 			ipfw_objhash_del(ni, no);
3331 			no->set = new_set;
3332 			ipfw_objhash_add(ni, no);
3333 		}
3334 	}
3335 
3336 	return (bad);
3337 }
3338 
3339 /*
3340  * Finds and bumps refcount for objects referenced by given @rule.
3341  * Auto-creates non-existing tables.
3342  * Fills in @oib array with userland/kernel indexes.
3343  *
3344  * Returns 0 on success.
3345  */
3346 static int
3347 ref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule,
3348     struct rule_check_info *ci, struct obj_idx *oib, struct tid_info *ti)
3349 {
3350 	int cmdlen, error, l, numnew;
3351 	ipfw_insn *cmd;
3352 	struct obj_idx *pidx;
3353 	int found, unresolved;
3354 
3355 	pidx = oib;
3356 	l = rule->cmd_len;
3357 	cmd = rule->cmd;
3358 	cmdlen = 0;
3359 	error = 0;
3360 	numnew = 0;
3361 	found = 0;
3362 	unresolved = 0;
3363 
3364 	IPFW_UH_WLOCK(ch);
3365 
3366 	/* Increase refcount on each existing referenced table. */
3367 	for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
3368 		cmdlen = F_LEN(cmd);
3369 
3370 		error = ref_opcode_object(ch, cmd, ti, pidx, &found, &unresolved);
3371 		if (error != 0)
3372 			break;
3373 		if (found || unresolved) {
3374 			pidx->off = rule->cmd_len - l;
3375 			pidx++;
3376 		}
3377 		/*
3378 		 * Compability stuff for old clients:
3379 		 * prepare to manually create non-existing objects.
3380 		 */
3381 		if (unresolved)
3382 			numnew++;
3383 	}
3384 
3385 	if (error != 0) {
3386 		/* Unref everything we have already done */
3387 		unref_oib_objects(ch, rule->cmd, oib, pidx);
3388 		IPFW_UH_WUNLOCK(ch);
3389 		return (error);
3390 	}
3391 	IPFW_UH_WUNLOCK(ch);
3392 
3393 	/* Perform auto-creation for non-existing objects */
3394 	if (numnew != 0)
3395 		error = create_objects_compat(ch, rule->cmd, oib, pidx, ti);
3396 
3397 	/* Calculate real number of dynamic objects */
3398 	ci->object_opcodes = (uint16_t)(pidx - oib);
3399 
3400 	return (error);
3401 }
3402 
3403 /*
3404  * Checks is opcode is referencing table of appropriate type.
3405  * Adds reference count for found table if true.
3406  * Rewrites user-supplied opcode values with kernel ones.
3407  *
3408  * Returns 0 on success and appropriate error code otherwise.
3409  */
3410 int
3411 ipfw_rewrite_rule_uidx(struct ip_fw_chain *chain,
3412     struct rule_check_info *ci)
3413 {
3414 	int error;
3415 	ipfw_insn *cmd;
3416 	uint8_t type;
3417 	struct obj_idx *p, *pidx_first, *pidx_last;
3418 	struct tid_info ti;
3419 
3420 	/*
3421 	 * Prepare an array for storing opcode indices.
3422 	 * Use stack allocation by default.
3423 	 */
3424 	if (ci->object_opcodes <= (sizeof(ci->obuf)/sizeof(ci->obuf[0]))) {
3425 		/* Stack */
3426 		pidx_first = ci->obuf;
3427 	} else
3428 		pidx_first = malloc(ci->object_opcodes * sizeof(struct obj_idx),
3429 		    M_IPFW, M_WAITOK | M_ZERO);
3430 
3431 	error = 0;
3432 	type = 0;
3433 	memset(&ti, 0, sizeof(ti));
3434 
3435 	/*
3436 	 * Use default set for looking up tables (old way) or
3437 	 * use set rule is assigned to (new way).
3438 	 */
3439 	ti.set = (V_fw_tables_sets != 0) ? ci->krule->set : 0;
3440 	if (ci->ctlv != NULL) {
3441 		ti.tlvs = (void *)(ci->ctlv + 1);
3442 		ti.tlen = ci->ctlv->head.length - sizeof(ipfw_obj_ctlv);
3443 	}
3444 
3445 	/* Reference all used tables and other objects */
3446 	error = ref_rule_objects(chain, ci->krule, ci, pidx_first, &ti);
3447 	if (error != 0)
3448 		goto free;
3449 	/*
3450 	 * Note that ref_rule_objects() might have updated ci->object_opcodes
3451 	 * to reflect actual number of object opcodes.
3452 	 */
3453 
3454 	/* Perform rule rewrite */
3455 	p = pidx_first;
3456 	pidx_last = pidx_first + ci->object_opcodes;
3457 	for (p = pidx_first; p < pidx_last; p++) {
3458 		cmd = ci->krule->cmd + p->off;
3459 		update_opcode_kidx(cmd, p->kidx);
3460 	}
3461 
3462 free:
3463 	if (pidx_first != ci->obuf)
3464 		free(pidx_first, M_IPFW);
3465 
3466 	return (error);
3467 }
3468 
3469 static struct ipfw_sopt_handler	scodes[] = {
3470 	{ IP_FW_TABLE_XCREATE,	0,	HDIR_SET,	create_table },
3471 	{ IP_FW_TABLE_XDESTROY,	0,	HDIR_SET,	flush_table_v0 },
3472 	{ IP_FW_TABLE_XFLUSH,	0,	HDIR_SET,	flush_table_v0 },
3473 	{ IP_FW_TABLE_XMODIFY,	0,	HDIR_BOTH,	modify_table },
3474 	{ IP_FW_TABLE_XINFO,	0,	HDIR_GET,	describe_table },
3475 	{ IP_FW_TABLES_XLIST,	0,	HDIR_GET,	list_tables },
3476 	{ IP_FW_TABLE_XLIST,	0,	HDIR_GET,	dump_table_v0 },
3477 	{ IP_FW_TABLE_XLIST,	1,	HDIR_GET,	dump_table_v1 },
3478 	{ IP_FW_TABLE_XADD,	0,	HDIR_BOTH,	manage_table_ent_v0 },
3479 	{ IP_FW_TABLE_XADD,	1,	HDIR_BOTH,	manage_table_ent_v1 },
3480 	{ IP_FW_TABLE_XDEL,	0,	HDIR_BOTH,	manage_table_ent_v0 },
3481 	{ IP_FW_TABLE_XDEL,	1,	HDIR_BOTH,	manage_table_ent_v1 },
3482 	{ IP_FW_TABLE_XFIND,	0,	HDIR_GET,	find_table_entry },
3483 	{ IP_FW_TABLE_XSWAP,	0,	HDIR_SET,	swap_table },
3484 	{ IP_FW_TABLES_ALIST,	0,	HDIR_GET,	list_table_algo },
3485 	{ IP_FW_TABLE_XGETSIZE,	0,	HDIR_GET,	get_table_size },
3486 };
3487 
3488 static void
3489 destroy_table_locked(struct namedobj_instance *ni, struct named_object *no,
3490     void *arg)
3491 {
3492 
3493 	unlink_table((struct ip_fw_chain *)arg, (struct table_config *)no);
3494 	if (ipfw_objhash_free_idx(ni, no->kidx) != 0)
3495 		printf("Error unlinking kidx %d from table %s\n",
3496 		    no->kidx, no->name);
3497 	free_table_config(ni, (struct table_config *)no);
3498 }
3499 
3500 /*
3501  * Shuts tables module down.
3502  */
3503 void
3504 ipfw_destroy_tables(struct ip_fw_chain *ch, int last)
3505 {
3506 
3507 	IPFW_DEL_SOPT_HANDLER(last, scodes);
3508 	IPFW_DEL_OBJ_REWRITER(last, opcodes);
3509 
3510 	/* Remove all tables from working set */
3511 	IPFW_UH_WLOCK(ch);
3512 	IPFW_WLOCK(ch);
3513 	ipfw_objhash_foreach(CHAIN_TO_NI(ch), destroy_table_locked, ch);
3514 	IPFW_WUNLOCK(ch);
3515 	IPFW_UH_WUNLOCK(ch);
3516 
3517 	/* Free pointers itself */
3518 	free(ch->tablestate, M_IPFW);
3519 
3520 	ipfw_table_value_destroy(ch, last);
3521 	ipfw_table_algo_destroy(ch);
3522 
3523 	ipfw_objhash_destroy(CHAIN_TO_NI(ch));
3524 	free(CHAIN_TO_TCFG(ch), M_IPFW);
3525 }
3526 
3527 /*
3528  * Starts tables module.
3529  */
3530 int
3531 ipfw_init_tables(struct ip_fw_chain *ch, int first)
3532 {
3533 	struct tables_config *tcfg;
3534 
3535 	/* Allocate pointers */
3536 	ch->tablestate = malloc(V_fw_tables_max * sizeof(struct table_info),
3537 	    M_IPFW, M_WAITOK | M_ZERO);
3538 
3539 	tcfg = malloc(sizeof(struct tables_config), M_IPFW, M_WAITOK | M_ZERO);
3540 	tcfg->namehash = ipfw_objhash_create(V_fw_tables_max);
3541 	ch->tblcfg = tcfg;
3542 
3543 	ipfw_table_value_init(ch, first);
3544 	ipfw_table_algo_init(ch);
3545 
3546 	IPFW_ADD_OBJ_REWRITER(first, opcodes);
3547 	IPFW_ADD_SOPT_HANDLER(first, scodes);
3548 	return (0);
3549 }
3550 
3551 
3552 
3553