2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2004 Ruslan Ermilov and Vsevolod Lobko.
5 * Copyright (c) 2014 Yandex LLC
6 * Copyright (c) 2014 Alexander V. Chernikov
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
34 * Lookup table support for ipfw.
36 * This file contains handlers for all generic tables' operations:
37 * add/del/flush entries, list/dump tables etc..
39 * Table data modification is protected by both UH and runtime lock
40 * while reading configuration/data is protected by UH lock.
42 * Lookup algorithms for all table types are located in ip_fw_table_algo.c
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/malloc.h>
50 #include <sys/kernel.h>
52 #include <sys/rwlock.h>
53 #include <sys/rmlock.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/queue.h>
57 #include <net/if.h> /* ip_fw.h requires IFNAMSIZ */
59 #include <netinet/in.h>
60 #include <netinet/ip_var.h> /* struct ipfw_rule_ref */
61 #include <netinet/ip_fw.h>
63 #include <netpfil/ipfw/ip_fw_private.h>
64 #include <netpfil/ipfw/ip_fw_table.h>
67 * Table has the following `type` concepts:
69 * `no.type` represents lookup key type (addr, ifp, uid, etc..)
70 * vmask represents bitmask of table values which are present at the moment.
71 * Special IPFW_VTYPE_LEGACY ( (uint32_t)-1 ) represents old
72 * single-value-for-all approach.
75 struct named_object no;
76 uint8_t tflags; /* type flags */
77 uint8_t locked; /* 1 if locked from changes */
78 uint8_t linked; /* 1 if already linked */
79 uint8_t ochanged; /* used by set swapping */
80 uint8_t vshared; /* 1 if using shared value array */
82 uint32_t count; /* Number of records */
83 uint32_t limit; /* Max number of records */
84 uint32_t vmask; /* bitmask with supported values */
85 uint32_t ocount; /* used by set swapping */
86 uint64_t gencnt; /* generation count */
87 char tablename[64]; /* table name */
88 struct table_algo *ta; /* Callbacks for given algo */
89 void *astate; /* algorithm state */
90 struct table_info ti_copy; /* data to put to table_info */
91 struct namedobj_instance *vi;
94 static int find_table_err(struct namedobj_instance *ni, struct tid_info *ti,
95 struct table_config **tc);
96 static struct table_config *find_table(struct namedobj_instance *ni,
98 static struct table_config *alloc_table_config(struct ip_fw_chain *ch,
99 struct tid_info *ti, struct table_algo *ta, char *adata, uint8_t tflags);
100 static void free_table_config(struct namedobj_instance *ni,
101 struct table_config *tc);
102 static int create_table_internal(struct ip_fw_chain *ch, struct tid_info *ti,
103 char *aname, ipfw_xtable_info *i, uint16_t *pkidx, int ref);
104 static void link_table(struct ip_fw_chain *ch, struct table_config *tc);
105 static void unlink_table(struct ip_fw_chain *ch, struct table_config *tc);
106 static int find_ref_table(struct ip_fw_chain *ch, struct tid_info *ti,
107 struct tentry_info *tei, uint32_t count, int op, struct table_config **ptc);
110 static int export_tables(struct ip_fw_chain *ch, ipfw_obj_lheader *olh,
111 struct sockopt_data *sd);
112 static void export_table_info(struct ip_fw_chain *ch, struct table_config *tc,
113 ipfw_xtable_info *i);
114 static int dump_table_tentry(void *e, void *arg);
115 static int dump_table_xentry(void *e, void *arg);
117 static int swap_tables(struct ip_fw_chain *ch, struct tid_info *a,
120 static int check_table_name(const char *name);
121 static int check_table_space(struct ip_fw_chain *ch, struct tableop_state *ts,
122 struct table_config *tc, struct table_info *ti, uint32_t count);
123 static int destroy_table(struct ip_fw_chain *ch, struct tid_info *ti);
125 static struct table_algo *find_table_algo(struct tables_config *tableconf,
126 struct tid_info *ti, char *name);
128 static void objheader_to_ti(struct _ipfw_obj_header *oh, struct tid_info *ti);
129 static void ntlv_to_ti(struct _ipfw_obj_ntlv *ntlv, struct tid_info *ti);
131 #define CHAIN_TO_NI(chain) (CHAIN_TO_TCFG(chain)->namehash)
132 #define KIDX_TO_TI(ch, k) (&(((struct table_info *)(ch)->tablestate)[k]))
134 #define TA_BUF_SZ 128 /* On-stack buffer for add/delete state */
137 rollback_toperation_state(struct ip_fw_chain *ch, void *object)
139 struct tables_config *tcfg;
142 tcfg = CHAIN_TO_TCFG(ch);
143 TAILQ_FOREACH(os, &tcfg->state_list, next)
144 os->func(object, os);
148 add_toperation_state(struct ip_fw_chain *ch, struct tableop_state *ts)
150 struct tables_config *tcfg;
152 tcfg = CHAIN_TO_TCFG(ch);
153 TAILQ_INSERT_HEAD(&tcfg->state_list, &ts->opstate, next);
157 del_toperation_state(struct ip_fw_chain *ch, struct tableop_state *ts)
159 struct tables_config *tcfg;
161 tcfg = CHAIN_TO_TCFG(ch);
162 TAILQ_REMOVE(&tcfg->state_list, &ts->opstate, next);
166 tc_ref(struct table_config *tc)
173 tc_unref(struct table_config *tc)
179 static struct table_value *
180 get_table_value(struct ip_fw_chain *ch, struct table_config *tc, uint32_t kidx)
182 struct table_value *pval;
184 pval = (struct table_value *)ch->valuestate;
186 return (&pval[kidx]);
191 * Checks if we're able to insert/update entry @tei into table
193 * May alter @tei to indicate insertion error / insert
196 * Returns 0 if operation can be performed/
199 check_table_limit(struct table_config *tc, struct tentry_info *tei)
202 if (tc->limit == 0 || tc->count < tc->limit)
205 if ((tei->flags & TEI_FLAGS_UPDATE) == 0) {
206 /* Notify userland on error cause */
207 tei->flags |= TEI_FLAGS_LIMIT;
212 * We have UPDATE flag set.
213 * Permit updating record (if found),
214 * but restrict adding new one since we've
215 * already hit the limit.
217 tei->flags |= TEI_FLAGS_DONTADD;
223 * Convert algorithm callback return code into
224 * one of pre-defined states known by userland.
227 store_tei_result(struct tentry_info *tei, int op, int error, uint32_t num)
235 if (op == OP_ADD && num != 0)
236 flag = TEI_FLAGS_ADDED;
238 flag = TEI_FLAGS_DELETED;
241 flag = TEI_FLAGS_NOTFOUND;
244 flag = TEI_FLAGS_EXISTS;
247 flag = TEI_FLAGS_ERROR;
254 * Creates and references table with default parameters.
255 * Saves table config, algo and allocated kidx info @ptc, @pta and
256 * @pkidx if non-zero.
257 * Used for table auto-creation to support old binaries.
259 * Returns 0 on success.
262 create_table_compat(struct ip_fw_chain *ch, struct tid_info *ti,
268 memset(&xi, 0, sizeof(xi));
269 /* Set default value mask for legacy clients */
270 xi.vmask = IPFW_VTYPE_LEGACY;
272 error = create_table_internal(ch, ti, NULL, &xi, pkidx, 1);
280 * Find and reference existing table optionally
283 * Saves found table config into @ptc.
284 * Note function may drop/acquire UH_WLOCK.
285 * Returns 0 if table was found/created and referenced
286 * or non-zero return code.
289 find_ref_table(struct ip_fw_chain *ch, struct tid_info *ti,
290 struct tentry_info *tei, uint32_t count, int op,
291 struct table_config **ptc)
293 struct namedobj_instance *ni;
294 struct table_config *tc;
298 IPFW_UH_WLOCK_ASSERT(ch);
300 ni = CHAIN_TO_NI(ch);
302 if ((tc = find_table(ni, ti)) != NULL) {
303 /* check table type */
304 if (tc->no.subtype != ti->type)
310 /* Try to exit early on limit hit */
311 if (op == OP_ADD && count == 1 &&
312 check_table_limit(tc, tei) != 0)
315 /* Reference and return */
324 /* Compatibility mode: create new table for old clients */
325 if ((tei->flags & TEI_FLAGS_COMPAT) == 0)
329 error = create_table_compat(ch, ti, &kidx);
335 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, kidx);
336 KASSERT(tc != NULL, ("create_table_compat returned bad idx %d", kidx));
338 /* OK, now we've got referenced table. */
344 * Rolls back already @added to @tc entries using state array @ta_buf_m.
345 * Assume the following layout:
346 * 1) ADD state (ta_buf_m[0] ... t_buf_m[added - 1]) for handling update cases
347 * 2) DEL state (ta_buf_m[count[ ... t_buf_m[count + added - 1])
348 * for storing deleted state
351 rollback_added_entries(struct ip_fw_chain *ch, struct table_config *tc,
352 struct table_info *tinfo, struct tentry_info *tei, caddr_t ta_buf_m,
353 uint32_t count, uint32_t added)
355 struct table_algo *ta;
356 struct tentry_info *ptei;
362 IPFW_UH_WLOCK_ASSERT(ch);
365 ta_buf_sz = ta->ta_buf_size;
367 vv = v + count * ta_buf_sz;
368 for (i = 0; i < added; i++, v += ta_buf_sz, vv += ta_buf_sz) {
370 if ((ptei->flags & TEI_FLAGS_UPDATED) != 0) {
373 * We have old value stored by previous
374 * call in @ptei->value. Do add once again
377 error = ta->add(tc->astate, tinfo, ptei, v, &num);
378 KASSERT(error == 0, ("rollback UPDATE fail"));
379 KASSERT(num == 0, ("rollback UPDATE fail2"));
383 error = ta->prepare_del(ch, ptei, vv);
384 KASSERT(error == 0, ("pre-rollback INSERT failed"));
385 error = ta->del(tc->astate, tinfo, ptei, vv, &num);
386 KASSERT(error == 0, ("rollback INSERT failed"));
392 * Prepares add/del state for all @count entries in @tei.
393 * Uses either stack buffer (@ta_buf) or allocates a new one.
394 * Stores pointer to allocated buffer back to @ta_buf.
396 * Returns 0 on success.
399 prepare_batch_buffer(struct ip_fw_chain *ch, struct table_algo *ta,
400 struct tentry_info *tei, uint32_t count, int op, caddr_t *ta_buf)
403 size_t ta_buf_sz, sz;
404 struct tentry_info *ptei;
408 ta_buf_sz = ta->ta_buf_size;
410 /* Single add/delete, use on-stack buffer */
411 memset(*ta_buf, 0, TA_BUF_SZ);
416 * Multiple adds/deletes, allocate larger buffer
418 * Note we need 2xcount buffer for add case:
419 * we have hold both ADD state
420 * and DELETE state (this may be needed
421 * if we need to rollback all changes)
423 sz = count * ta_buf_sz;
424 ta_buf_m = malloc((op == OP_ADD) ? sz * 2 : sz, M_TEMP,
429 for (i = 0; i < count; i++, v += ta_buf_sz) {
431 error = (op == OP_ADD) ?
432 ta->prepare_add(ch, ptei, v) : ta->prepare_del(ch, ptei, v);
435 * Some syntax error (incorrect mask, or address, or
436 * anything). Return error regardless of atomicity
448 * Flushes allocated state for each @count entries in @tei.
449 * Frees @ta_buf_m if differs from stack buffer @ta_buf.
452 flush_batch_buffer(struct ip_fw_chain *ch, struct table_algo *ta,
453 struct tentry_info *tei, uint32_t count, int rollback,
454 caddr_t ta_buf_m, caddr_t ta_buf)
457 struct tentry_info *ptei;
461 ta_buf_sz = ta->ta_buf_size;
463 /* Run cleaning callback anyway */
465 for (i = 0; i < count; i++, v += ta_buf_sz) {
467 ta->flush_entry(ch, ptei, v);
468 if (ptei->ptv != NULL) {
469 free(ptei->ptv, M_IPFW);
474 /* Clean up "deleted" state in case of rollback */
476 v = ta_buf_m + count * ta_buf_sz;
477 for (i = 0; i < count; i++, v += ta_buf_sz)
478 ta->flush_entry(ch, &tei[i], v);
481 if (ta_buf_m != ta_buf)
482 free(ta_buf_m, M_TEMP);
487 rollback_add_entry(void *object, struct op_state *_state)
489 struct ip_fw_chain *ch;
490 struct tableop_state *ts;
492 ts = (struct tableop_state *)_state;
494 if (ts->tc != object && ts->ch != object)
499 IPFW_UH_WLOCK_ASSERT(ch);
501 /* Call specifid unlockers */
502 rollback_table_values(ts);
504 /* Indicate we've called */
509 * Adds/updates one or more entries in table @ti.
511 * Function may drop/reacquire UH wlock multiple times due to
512 * items alloc, algorithm callbacks (check_space), value linkage
513 * (new values, value storage realloc), etc..
514 * Other processes like other adds (which may involve storage resize),
515 * table swaps (which changes table data and may change algo type),
516 * table modify (which may change value mask) may be executed
517 * simultaneously so we need to deal with it.
519 * The following approach was implemented:
520 * we have per-chain linked list, protected with UH lock.
521 * add_table_entry prepares special on-stack structure wthich is passed
522 * to its descendants. Users add this structure to this list before unlock.
523 * After performing needed operations and acquiring UH lock back, each user
524 * checks if structure has changed. If true, it rolls local state back and
525 * returns without error to the caller.
526 * add_table_entry() on its own checks if structure has changed and restarts
527 * its operation from the beginning (goto restart).
529 * Functions which are modifying fields of interest (currently
530 * resize_shared_value_storage() and swap_tables() )
531 * traverses given list while holding UH lock immediately before
532 * performing their operations calling function provided be list entry
533 * ( currently rollback_add_entry ) which performs rollback for all necessary
534 * state and sets appropriate values in structure indicating rollback
538 * Function references @ti first to ensure table won't
539 * disappear or change its type.
540 * After that, prepare_add callback is called for each @tei entry.
541 * Next, we try to add each entry under UH+WHLOCK
542 * using add() callback.
543 * Finally, we free all state by calling flush_entry callback
546 * Returns 0 on success.
549 add_table_entry(struct ip_fw_chain *ch, struct tid_info *ti,
550 struct tentry_info *tei, uint8_t flags, uint32_t count)
552 struct table_config *tc;
553 struct table_algo *ta;
555 int error, first_error, i, rollback;
556 uint32_t num, numadd;
557 struct tentry_info *ptei;
558 struct tableop_state ts;
559 char ta_buf[TA_BUF_SZ];
562 memset(&ts, 0, sizeof(ts));
567 * Find and reference existing table.
570 if (ts.modified != 0) {
572 flush_batch_buffer(ch, ta, tei, count, rollback,
574 memset(&ts, 0, sizeof(ts));
579 error = find_ref_table(ch, ti, tei, count, OP_ADD, &tc);
586 /* Fill in tablestate */
588 ts.opstate.func = rollback_add_entry;
590 ts.vshared = tc->vshared;
591 ts.vmask = tc->vmask;
596 add_toperation_state(ch, &ts);
599 /* Allocate memory and prepare record(s) */
600 /* Pass stack buffer by default */
602 error = prepare_batch_buffer(ch, ta, tei, count, OP_ADD, &ta_buf_m);
605 del_toperation_state(ch, &ts);
606 /* Drop reference we've used in first search */
609 /* Check prepare_batch_buffer() error */
614 * Check if table swap has happened.
615 * (so table algo might be changed).
616 * Restart operation to achieve consistent behavior.
618 if (ts.modified != 0)
622 * Link all values values to shared/per-table value array.
624 * May release/reacquire UH_WLOCK.
626 error = ipfw_link_table_values(ch, &ts);
629 if (ts.modified != 0)
633 * Ensure we are able to add all entries without additional
634 * memory allocations. May release/reacquire UH_WLOCK.
637 error = check_table_space(ch, &ts, tc, KIDX_TO_TI(ch, kidx), count);
640 if (ts.modified != 0)
643 /* We've got valid table in @tc. Let's try to add data */
652 for (i = 0; i < count; i++, v += ta->ta_buf_size) {
655 /* check limit before adding */
656 if ((error = check_table_limit(tc, ptei)) == 0) {
657 error = ta->add(tc->astate, KIDX_TO_TI(ch, kidx),
659 /* Set status flag to inform userland */
660 store_tei_result(ptei, OP_ADD, error, num);
663 /* Update number of records to ease limit checking */
669 if (first_error == 0)
673 * Some error have happened. Check our atomicity
674 * settings: continue if atomicity is not required,
675 * rollback changes otherwise.
677 if ((flags & IPFW_CTF_ATOMIC) == 0)
680 rollback_added_entries(ch, tc, KIDX_TO_TI(ch, kidx),
681 tei, ta_buf_m, count, i);
689 ipfw_garbage_table_values(ch, tc, tei, count, rollback);
691 /* Permit post-add algorithm grow/rehash. */
693 check_table_space(ch, NULL, tc, KIDX_TO_TI(ch, kidx), 0);
695 /* Return first error to user, if any */
701 flush_batch_buffer(ch, ta, tei, count, rollback, ta_buf_m, ta_buf);
707 * Deletes one or more entries in table @ti.
709 * Returns 0 on success.
712 del_table_entry(struct ip_fw_chain *ch, struct tid_info *ti,
713 struct tentry_info *tei, uint8_t flags, uint32_t count)
715 struct table_config *tc;
716 struct table_algo *ta;
717 struct tentry_info *ptei;
719 int error, first_error, i;
720 uint32_t num, numdel;
721 char ta_buf[TA_BUF_SZ];
725 * Find and reference existing table.
728 error = find_ref_table(ch, ti, tei, count, OP_DEL, &tc);
736 /* Allocate memory and prepare record(s) */
737 /* Pass stack buffer by default */
739 error = prepare_batch_buffer(ch, ta, tei, count, OP_DEL, &ta_buf_m);
745 /* Drop reference we've used in first search */
749 * Check if table algo is still the same.
750 * (changed ta may be the result of table swap).
764 for (i = 0; i < count; i++, v += ta->ta_buf_size) {
767 error = ta->del(tc->astate, KIDX_TO_TI(ch, kidx), ptei, v,
769 /* Save state for userland */
770 store_tei_result(ptei, OP_DEL, error, num);
771 if (error != 0 && first_error == 0)
778 /* Unlink non-used values */
779 ipfw_garbage_table_values(ch, tc, tei, count, 0);
782 /* Run post-del hook to permit shrinking */
783 check_table_space(ch, NULL, tc, KIDX_TO_TI(ch, kidx), 0);
788 /* Return first error to user, if any */
792 flush_batch_buffer(ch, ta, tei, count, 0, ta_buf_m, ta_buf);
798 * Ensure that table @tc has enough space to add @count entries without
799 * need for reallocation.
802 * 0) need_modify() (UH_WLOCK) - checks if @count items can be added w/o resize.
804 * 1) alloc_modify (no locks, M_WAITOK) - alloc new state based on @pflags.
805 * 2) prepare_modifyt (UH_WLOCK) - copy old data into new storage
806 * 3) modify (UH_WLOCK + WLOCK) - switch pointers
807 * 4) flush_modify (UH_WLOCK) - free state, if needed
809 * Returns 0 on success.
812 check_table_space(struct ip_fw_chain *ch, struct tableop_state *ts,
813 struct table_config *tc, struct table_info *ti, uint32_t count)
815 struct table_algo *ta;
817 char ta_buf[TA_BUF_SZ];
820 IPFW_UH_WLOCK_ASSERT(ch);
824 if (ta->need_modify == NULL)
827 /* Acquire reference not to loose @tc between locks/unlocks */
831 * TODO: think about avoiding race between large add/large delete
832 * operation on algorithm which implements shrinking along with
837 if (ta->need_modify(tc->astate, ti, count, &pflags) == 0) {
842 /* We have to shrink/grow table */
844 add_toperation_state(ch, ts);
847 memset(&ta_buf, 0, sizeof(ta_buf));
848 error = ta->prepare_mod(ta_buf, &pflags);
852 del_toperation_state(ch, ts);
857 if (ts != NULL && ts->modified != 0) {
860 * Swap operation has happened
861 * so we're currently operating on other
862 * table data. Stop doing this.
864 ta->flush_mod(ta_buf);
868 /* Check if we still need to alter table */
869 ti = KIDX_TO_TI(ch, tc->no.kidx);
870 if (ta->need_modify(tc->astate, ti, count, &pflags) == 0) {
874 * Other thread has already performed resize.
875 * Flush our state and return.
877 ta->flush_mod(ta_buf);
881 error = ta->fill_mod(tc->astate, ti, ta_buf, &pflags);
883 /* Do actual modification */
885 ta->modify(tc->astate, ti, ta_buf, pflags);
889 /* Anyway, flush data and retry */
890 ta->flush_mod(ta_buf);
898 * Adds or deletes record in table.
900 * Request: [ ip_fw3_opheader ipfw_table_xentry ]
902 * Returns 0 on success
905 manage_table_ent_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
906 struct sockopt_data *sd)
908 ipfw_table_xentry *xent;
909 struct tentry_info tei;
911 struct table_value v;
912 int error, hdrlen, read;
914 hdrlen = offsetof(ipfw_table_xentry, k);
916 /* Check minimum header size */
917 if (sd->valsize < (sizeof(*op3) + hdrlen))
920 read = sizeof(ip_fw3_opheader);
922 /* Check if xentry len field is valid */
923 xent = (ipfw_table_xentry *)(op3 + 1);
924 if (xent->len < hdrlen || xent->len + read > sd->valsize)
927 memset(&tei, 0, sizeof(tei));
928 tei.paddr = &xent->k;
929 tei.masklen = xent->masklen;
930 ipfw_import_table_value_legacy(xent->value, &v);
932 /* Old requests compatibility */
933 tei.flags = TEI_FLAGS_COMPAT;
934 if (xent->type == IPFW_TABLE_ADDR) {
935 if (xent->len - hdrlen == sizeof(in_addr_t))
936 tei.subtype = AF_INET;
938 tei.subtype = AF_INET6;
941 memset(&ti, 0, sizeof(ti));
943 ti.type = xent->type;
945 error = (op3->opcode == IP_FW_TABLE_XADD) ?
946 add_table_entry(ch, &ti, &tei, 0, 1) :
947 del_table_entry(ch, &ti, &tei, 0, 1);
953 * Adds or deletes record in table.
954 * Data layout (v1)(current):
955 * Request: [ ipfw_obj_header
956 * ipfw_obj_ctlv(IPFW_TLV_TBLENT_LIST) [ ipfw_obj_tentry x N ]
959 * Returns 0 on success
962 manage_table_ent_v1(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
963 struct sockopt_data *sd)
965 ipfw_obj_tentry *tent, *ptent;
968 struct tentry_info *ptei, tei, *tei_buf;
970 int error, i, kidx, read;
972 /* Check minimum header size */
973 if (sd->valsize < (sizeof(*oh) + sizeof(*ctlv)))
976 /* Check if passed data is too long */
977 if (sd->valsize != sd->kavail)
980 oh = (ipfw_obj_header *)sd->kbuf;
982 /* Basic length checks for TLVs */
983 if (oh->ntlv.head.length != sizeof(oh->ntlv))
988 ctlv = (ipfw_obj_ctlv *)(oh + 1);
989 if (ctlv->head.length + read != sd->valsize)
992 read += sizeof(*ctlv);
993 tent = (ipfw_obj_tentry *)(ctlv + 1);
994 if (ctlv->count * sizeof(*tent) + read != sd->valsize)
997 if (ctlv->count == 0)
1001 * Mark entire buffer as "read".
1002 * This instructs sopt api write it back
1003 * after function return.
1005 ipfw_get_sopt_header(sd, sd->valsize);
1007 /* Perform basic checks for each entry */
1010 for (i = 0; i < ctlv->count; i++, ptent++) {
1011 if (ptent->head.length != sizeof(*ptent))
1013 if (ptent->idx != kidx)
1017 /* Convert data into kernel request objects */
1018 objheader_to_ti(oh, &ti);
1019 ti.type = oh->ntlv.type;
1022 /* Use on-stack buffer for single add/del */
1023 if (ctlv->count == 1) {
1024 memset(&tei, 0, sizeof(tei));
1027 tei_buf = malloc(ctlv->count * sizeof(tei), M_TEMP,
1032 for (i = 0; i < ctlv->count; i++, ptent++, ptei++) {
1033 ptei->paddr = &ptent->k;
1034 ptei->subtype = ptent->subtype;
1035 ptei->masklen = ptent->masklen;
1036 if (ptent->head.flags & IPFW_TF_UPDATE)
1037 ptei->flags |= TEI_FLAGS_UPDATE;
1039 ipfw_import_table_value_v1(&ptent->v.value);
1040 ptei->pvalue = (struct table_value *)&ptent->v.value;
1043 error = (oh->opheader.opcode == IP_FW_TABLE_XADD) ?
1044 add_table_entry(ch, &ti, tei_buf, ctlv->flags, ctlv->count) :
1045 del_table_entry(ch, &ti, tei_buf, ctlv->flags, ctlv->count);
1047 /* Translate result back to userland */
1050 for (i = 0; i < ctlv->count; i++, ptent++, ptei++) {
1051 if (ptei->flags & TEI_FLAGS_ADDED)
1052 ptent->result = IPFW_TR_ADDED;
1053 else if (ptei->flags & TEI_FLAGS_DELETED)
1054 ptent->result = IPFW_TR_DELETED;
1055 else if (ptei->flags & TEI_FLAGS_UPDATED)
1056 ptent->result = IPFW_TR_UPDATED;
1057 else if (ptei->flags & TEI_FLAGS_LIMIT)
1058 ptent->result = IPFW_TR_LIMIT;
1059 else if (ptei->flags & TEI_FLAGS_ERROR)
1060 ptent->result = IPFW_TR_ERROR;
1061 else if (ptei->flags & TEI_FLAGS_NOTFOUND)
1062 ptent->result = IPFW_TR_NOTFOUND;
1063 else if (ptei->flags & TEI_FLAGS_EXISTS)
1064 ptent->result = IPFW_TR_EXISTS;
1065 ipfw_export_table_value_v1(ptei->pvalue, &ptent->v.value);
1068 if (tei_buf != &tei)
1069 free(tei_buf, M_TEMP);
1075 * Looks up an entry in given table.
1076 * Data layout (v0)(current):
1077 * Request: [ ipfw_obj_header ipfw_obj_tentry ]
1078 * Reply: [ ipfw_obj_header ipfw_obj_tentry ]
1080 * Returns 0 on success
1083 find_table_entry(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1084 struct sockopt_data *sd)
1086 ipfw_obj_tentry *tent;
1087 ipfw_obj_header *oh;
1089 struct table_config *tc;
1090 struct table_algo *ta;
1091 struct table_info *kti;
1092 struct table_value *pval;
1093 struct namedobj_instance *ni;
1097 /* Check minimum header size */
1098 sz = sizeof(*oh) + sizeof(*tent);
1099 if (sd->valsize != sz)
1102 oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz);
1103 tent = (ipfw_obj_tentry *)(oh + 1);
1105 /* Basic length checks for TLVs */
1106 if (oh->ntlv.head.length != sizeof(oh->ntlv))
1109 objheader_to_ti(oh, &ti);
1110 ti.type = oh->ntlv.type;
1111 ti.uidx = tent->idx;
1114 ni = CHAIN_TO_NI(ch);
1117 * Find existing table and check its type .
1120 if ((tc = find_table(ni, &ti)) == NULL) {
1121 IPFW_UH_RUNLOCK(ch);
1125 /* check table type */
1126 if (tc->no.subtype != ti.type) {
1127 IPFW_UH_RUNLOCK(ch);
1131 kti = KIDX_TO_TI(ch, tc->no.kidx);
1134 if (ta->find_tentry == NULL)
1137 error = ta->find_tentry(tc->astate, kti, tent);
1139 pval = get_table_value(ch, tc, tent->v.kidx);
1140 ipfw_export_table_value_v1(pval, &tent->v.value);
1142 IPFW_UH_RUNLOCK(ch);
1148 * Flushes all entries or destroys given table.
1149 * Data layout (v0)(current):
1150 * Request: [ ipfw_obj_header ]
1152 * Returns 0 on success
1155 flush_table_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1156 struct sockopt_data *sd)
1159 struct _ipfw_obj_header *oh;
1162 if (sd->valsize != sizeof(*oh))
1165 oh = (struct _ipfw_obj_header *)op3;
1166 objheader_to_ti(oh, &ti);
1168 if (op3->opcode == IP_FW_TABLE_XDESTROY)
1169 error = destroy_table(ch, &ti);
1170 else if (op3->opcode == IP_FW_TABLE_XFLUSH)
1171 error = flush_table(ch, &ti);
1179 restart_flush(void *object, struct op_state *_state)
1181 struct tableop_state *ts;
1183 ts = (struct tableop_state *)_state;
1185 if (ts->tc != object)
1188 /* Indicate we've called */
1193 * Flushes given table.
1195 * Function create new table instance with the same
1196 * parameters, swaps it with old one and
1197 * flushes state without holding runtime WLOCK.
1199 * Returns 0 on success.
1202 flush_table(struct ip_fw_chain *ch, struct tid_info *ti)
1204 struct namedobj_instance *ni;
1205 struct table_config *tc;
1206 struct table_algo *ta;
1207 struct table_info ti_old, ti_new, *tablestate;
1208 void *astate_old, *astate_new;
1209 char algostate[64], *pstate;
1210 struct tableop_state ts;
1216 * Stage 1: save table algorithm.
1217 * Reference found table to ensure it won't disappear.
1220 ni = CHAIN_TO_NI(ch);
1221 if ((tc = find_table(ni, ti)) == NULL) {
1222 IPFW_UH_WUNLOCK(ch);
1227 memset(&ti_new, 0, sizeof(ti_new));
1229 /* Set up swap handler */
1230 memset(&ts, 0, sizeof(ts));
1231 ts.opstate.func = restart_flush;
1235 /* Do not flush readonly tables */
1236 if ((ta->flags & TA_FLAG_READONLY) != 0) {
1237 IPFW_UH_WUNLOCK(ch);
1240 /* Save startup algo parameters */
1241 if (ta->print_config != NULL) {
1242 ta->print_config(tc->astate, KIDX_TO_TI(ch, tc->no.kidx),
1243 algostate, sizeof(algostate));
1247 tflags = tc->tflags;
1249 add_toperation_state(ch, &ts);
1250 IPFW_UH_WUNLOCK(ch);
1253 * Stage 1.5: if this is not the first attempt, destroy previous state
1256 ta->destroy(astate_new, &ti_new);
1261 * Stage 2: allocate new table instance using same algo.
1263 memset(&ti_new, 0, sizeof(struct table_info));
1264 error = ta->init(ch, &astate_new, &ti_new, pstate, tflags);
1267 * Stage 3: swap old state pointers with newly-allocated ones.
1268 * Decrease refcount.
1272 del_toperation_state(ch, &ts);
1275 IPFW_UH_WUNLOCK(ch);
1280 * Restart operation if table swap has happened:
1281 * even if algo may be the same, algo init parameters
1282 * may change. Restart operation instead of doing
1285 if (ts.modified != 0) {
1286 /* Delay destroying data since we're holding UH lock */
1291 ni = CHAIN_TO_NI(ch);
1293 tablestate = (struct table_info *)ch->tablestate;
1296 ti_old = tablestate[kidx];
1297 tablestate[kidx] = ti_new;
1300 astate_old = tc->astate;
1301 tc->astate = astate_new;
1302 tc->ti_copy = ti_new;
1305 /* Notify algo on real @ti address */
1306 if (ta->change_ti != NULL)
1307 ta->change_ti(tc->astate, &tablestate[kidx]);
1310 * Stage 4: unref values.
1312 ipfw_unref_table_values(ch, tc, ta, astate_old, &ti_old);
1313 IPFW_UH_WUNLOCK(ch);
1316 * Stage 5: perform real flush/destroy.
1318 ta->destroy(astate_old, &ti_old);
1325 * Data layout (v0)(current):
1326 * Request: [ ipfw_obj_header ipfw_obj_ntlv ]
1328 * Returns 0 on success
1331 swap_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1332 struct sockopt_data *sd)
1335 struct _ipfw_obj_header *oh;
1336 struct tid_info ti_a, ti_b;
1338 if (sd->valsize != sizeof(*oh) + sizeof(ipfw_obj_ntlv))
1341 oh = (struct _ipfw_obj_header *)op3;
1342 ntlv_to_ti(&oh->ntlv, &ti_a);
1343 ntlv_to_ti((ipfw_obj_ntlv *)(oh + 1), &ti_b);
1345 error = swap_tables(ch, &ti_a, &ti_b);
1351 * Swaps two tables of the same type/valtype.
1353 * Checks if tables are compatible and limits
1354 * permits swap, than actually perform swap.
1356 * Each table consists of 2 different parts:
1358 * @tc (with name, set, kidx) and rule bindings, which is "stable".
1362 * runtime data @ti (ch->tablestate)
1363 * runtime cache in @tc
1364 * algo-specific data (@tc->astate)
1371 * After that we call @ti change handler for each table.
1373 * Note that referencing @tc won't protect tc->ta from change.
1374 * XXX: Do we need to restrict swap between locked tables?
1375 * XXX: Do we need to exchange ftype?
1377 * Returns 0 on success.
1380 swap_tables(struct ip_fw_chain *ch, struct tid_info *a,
1383 struct namedobj_instance *ni;
1384 struct table_config *tc_a, *tc_b;
1385 struct table_algo *ta;
1386 struct table_info ti, *tablestate;
1391 * Stage 1: find both tables and ensure they are of
1395 ni = CHAIN_TO_NI(ch);
1396 if ((tc_a = find_table(ni, a)) == NULL) {
1397 IPFW_UH_WUNLOCK(ch);
1400 if ((tc_b = find_table(ni, b)) == NULL) {
1401 IPFW_UH_WUNLOCK(ch);
1405 /* It is very easy to swap between the same table */
1407 IPFW_UH_WUNLOCK(ch);
1411 /* Check type and value are the same */
1412 if (tc_a->no.subtype!=tc_b->no.subtype || tc_a->tflags!=tc_b->tflags) {
1413 IPFW_UH_WUNLOCK(ch);
1417 /* Check limits before swap */
1418 if ((tc_a->limit != 0 && tc_b->count > tc_a->limit) ||
1419 (tc_b->limit != 0 && tc_a->count > tc_b->limit)) {
1420 IPFW_UH_WUNLOCK(ch);
1424 /* Check if one of the tables is readonly */
1425 if (((tc_a->ta->flags | tc_b->ta->flags) & TA_FLAG_READONLY) != 0) {
1426 IPFW_UH_WUNLOCK(ch);
1430 /* Notify we're going to swap */
1431 rollback_toperation_state(ch, tc_a);
1432 rollback_toperation_state(ch, tc_b);
1434 /* Everything is fine, prepare to swap */
1435 tablestate = (struct table_info *)ch->tablestate;
1436 ti = tablestate[tc_a->no.kidx];
1438 astate = tc_a->astate;
1439 count = tc_a->count;
1443 tablestate[tc_a->no.kidx] = tablestate[tc_b->no.kidx];
1444 tc_a->ta = tc_b->ta;
1445 tc_a->astate = tc_b->astate;
1446 tc_a->count = tc_b->count;
1448 tablestate[tc_b->no.kidx] = ti;
1450 tc_b->astate = astate;
1451 tc_b->count = count;
1454 /* Ensure tc.ti copies are in sync */
1455 tc_a->ti_copy = tablestate[tc_a->no.kidx];
1456 tc_b->ti_copy = tablestate[tc_b->no.kidx];
1458 /* Notify both tables on @ti change */
1459 if (tc_a->ta->change_ti != NULL)
1460 tc_a->ta->change_ti(tc_a->astate, &tablestate[tc_a->no.kidx]);
1461 if (tc_b->ta->change_ti != NULL)
1462 tc_b->ta->change_ti(tc_b->astate, &tablestate[tc_b->no.kidx]);
1464 IPFW_UH_WUNLOCK(ch);
1470 * Destroys table specified by @ti.
1471 * Data layout (v0)(current):
1472 * Request: [ ip_fw3_opheader ]
1474 * Returns 0 on success
1477 destroy_table(struct ip_fw_chain *ch, struct tid_info *ti)
1479 struct namedobj_instance *ni;
1480 struct table_config *tc;
1484 ni = CHAIN_TO_NI(ch);
1485 if ((tc = find_table(ni, ti)) == NULL) {
1486 IPFW_UH_WUNLOCK(ch);
1490 /* Do not permit destroying referenced tables */
1491 if (tc->no.refcnt > 0) {
1492 IPFW_UH_WUNLOCK(ch);
1497 unlink_table(ch, tc);
1500 /* Free obj index */
1501 if (ipfw_objhash_free_idx(ni, tc->no.kidx) != 0)
1502 printf("Error unlinking kidx %d from table %s\n",
1503 tc->no.kidx, tc->tablename);
1505 /* Unref values used in tables while holding UH lock */
1506 ipfw_unref_table_values(ch, tc, tc->ta, tc->astate, &tc->ti_copy);
1507 IPFW_UH_WUNLOCK(ch);
1509 free_table_config(ni, tc);
1515 roundup2p(uint32_t v)
1530 * Grow tables index.
1532 * Returns 0 on success.
1535 ipfw_resize_tables(struct ip_fw_chain *ch, unsigned int ntables)
1537 unsigned int ntables_old, tbl;
1538 struct namedobj_instance *ni;
1539 void *new_idx, *old_tablestate, *tablestate;
1540 struct table_info *ti;
1541 struct table_config *tc;
1544 /* Check new value for validity */
1547 if (ntables > IPFW_TABLES_MAX)
1548 ntables = IPFW_TABLES_MAX;
1549 /* Alight to nearest power of 2 */
1550 ntables = (unsigned int)roundup2p(ntables);
1552 /* Allocate new pointers */
1553 tablestate = malloc(ntables * sizeof(struct table_info),
1554 M_IPFW, M_WAITOK | M_ZERO);
1556 ipfw_objhash_bitmap_alloc(ntables, (void *)&new_idx, &new_blocks);
1560 tbl = (ntables >= V_fw_tables_max) ? V_fw_tables_max : ntables;
1561 ni = CHAIN_TO_NI(ch);
1563 /* Temporary restrict decreasing max_tables */
1564 if (ntables < V_fw_tables_max) {
1567 * FIXME: Check if we really can shrink
1569 IPFW_UH_WUNLOCK(ch);
1573 /* Copy table info/indices */
1574 memcpy(tablestate, ch->tablestate, sizeof(struct table_info) * tbl);
1575 ipfw_objhash_bitmap_merge(ni, &new_idx, &new_blocks);
1579 /* Change pointers */
1580 old_tablestate = ch->tablestate;
1581 ch->tablestate = tablestate;
1582 ipfw_objhash_bitmap_swap(ni, &new_idx, &new_blocks);
1584 ntables_old = V_fw_tables_max;
1585 V_fw_tables_max = ntables;
1589 /* Notify all consumers that their @ti pointer has changed */
1590 ti = (struct table_info *)ch->tablestate;
1591 for (i = 0; i < tbl; i++, ti++) {
1592 if (ti->lookup == NULL)
1594 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, i);
1595 if (tc == NULL || tc->ta->change_ti == NULL)
1598 tc->ta->change_ti(tc->astate, ti);
1601 IPFW_UH_WUNLOCK(ch);
1603 /* Free old pointers */
1604 free(old_tablestate, M_IPFW);
1605 ipfw_objhash_bitmap_free(new_idx, new_blocks);
1611 * Lookup table's named object by its @kidx.
1613 struct named_object *
1614 ipfw_objhash_lookup_table_kidx(struct ip_fw_chain *ch, uint16_t kidx)
1617 return (ipfw_objhash_lookup_kidx(CHAIN_TO_NI(ch), kidx));
1621 * Take reference to table specified in @ntlv.
1622 * On success return its @kidx.
1625 ipfw_ref_table(struct ip_fw_chain *ch, ipfw_obj_ntlv *ntlv, uint16_t *kidx)
1628 struct table_config *tc;
1631 IPFW_UH_WLOCK_ASSERT(ch);
1633 ntlv_to_ti(ntlv, &ti);
1634 error = find_table_err(CHAIN_TO_NI(ch), &ti, &tc);
1642 *kidx = tc->no.kidx;
1648 ipfw_unref_table(struct ip_fw_chain *ch, uint16_t kidx)
1651 struct namedobj_instance *ni;
1652 struct named_object *no;
1654 IPFW_UH_WLOCK_ASSERT(ch);
1655 ni = CHAIN_TO_NI(ch);
1656 no = ipfw_objhash_lookup_kidx(ni, kidx);
1657 KASSERT(no != NULL, ("Table with index %d not found", kidx));
1662 * Lookup an arbitrary key @paddr of length @plen in table @tbl.
1663 * Stores found value in @val.
1665 * Returns 1 if key was found.
1668 ipfw_lookup_table(struct ip_fw_chain *ch, uint16_t tbl, uint16_t plen,
1669 void *paddr, uint32_t *val)
1671 struct table_info *ti;
1673 ti = KIDX_TO_TI(ch, tbl);
1675 return (ti->lookup(ti, paddr, plen, val));
1679 * Info/List/dump support for tables.
1684 * High-level 'get' cmds sysctl handlers
1688 * Lists all tables currently available in kernel.
1689 * Data layout (v0)(current):
1690 * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size
1691 * Reply: [ ipfw_obj_lheader ipfw_xtable_info x N ]
1693 * Returns 0 on success
1696 list_tables(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1697 struct sockopt_data *sd)
1699 struct _ipfw_obj_lheader *olh;
1702 olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh));
1705 if (sd->valsize < olh->size)
1709 error = export_tables(ch, olh, sd);
1710 IPFW_UH_RUNLOCK(ch);
1716 * Store table info to buffer provided by @sd.
1717 * Data layout (v0)(current):
1718 * Request: [ ipfw_obj_header ipfw_xtable_info(empty)]
1719 * Reply: [ ipfw_obj_header ipfw_xtable_info ]
1721 * Returns 0 on success.
1724 describe_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1725 struct sockopt_data *sd)
1727 struct _ipfw_obj_header *oh;
1728 struct table_config *tc;
1732 sz = sizeof(*oh) + sizeof(ipfw_xtable_info);
1733 oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz);
1737 objheader_to_ti(oh, &ti);
1740 if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) {
1741 IPFW_UH_RUNLOCK(ch);
1745 export_table_info(ch, tc, (ipfw_xtable_info *)(oh + 1));
1746 IPFW_UH_RUNLOCK(ch);
1752 * Modifies existing table.
1753 * Data layout (v0)(current):
1754 * Request: [ ipfw_obj_header ipfw_xtable_info ]
1756 * Returns 0 on success
1759 modify_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1760 struct sockopt_data *sd)
1762 struct _ipfw_obj_header *oh;
1763 ipfw_xtable_info *i;
1766 struct namedobj_instance *ni;
1767 struct table_config *tc;
1769 if (sd->valsize != sizeof(*oh) + sizeof(ipfw_xtable_info))
1772 oh = (struct _ipfw_obj_header *)sd->kbuf;
1773 i = (ipfw_xtable_info *)(oh + 1);
1776 * Verify user-supplied strings.
1777 * Check for null-terminated/zero-length strings/
1779 tname = oh->ntlv.name;
1780 if (check_table_name(tname) != 0)
1783 objheader_to_ti(oh, &ti);
1787 ni = CHAIN_TO_NI(ch);
1788 if ((tc = find_table(ni, &ti)) == NULL) {
1789 IPFW_UH_WUNLOCK(ch);
1793 /* Do not support any modifications for readonly tables */
1794 if ((tc->ta->flags & TA_FLAG_READONLY) != 0) {
1795 IPFW_UH_WUNLOCK(ch);
1799 if ((i->mflags & IPFW_TMFLAGS_LIMIT) != 0)
1800 tc->limit = i->limit;
1801 if ((i->mflags & IPFW_TMFLAGS_LOCK) != 0)
1802 tc->locked = ((i->flags & IPFW_TGFLAGS_LOCKED) != 0);
1803 IPFW_UH_WUNLOCK(ch);
1809 * Creates new table.
1810 * Data layout (v0)(current):
1811 * Request: [ ipfw_obj_header ipfw_xtable_info ]
1813 * Returns 0 on success
1816 create_table(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
1817 struct sockopt_data *sd)
1819 struct _ipfw_obj_header *oh;
1820 ipfw_xtable_info *i;
1821 char *tname, *aname;
1823 struct namedobj_instance *ni;
1825 if (sd->valsize != sizeof(*oh) + sizeof(ipfw_xtable_info))
1828 oh = (struct _ipfw_obj_header *)sd->kbuf;
1829 i = (ipfw_xtable_info *)(oh + 1);
1832 * Verify user-supplied strings.
1833 * Check for null-terminated/zero-length strings/
1835 tname = oh->ntlv.name;
1836 aname = i->algoname;
1837 if (check_table_name(tname) != 0 ||
1838 strnlen(aname, sizeof(i->algoname)) == sizeof(i->algoname))
1841 if (aname[0] == '\0') {
1842 /* Use default algorithm */
1846 objheader_to_ti(oh, &ti);
1849 ni = CHAIN_TO_NI(ch);
1852 if (find_table(ni, &ti) != NULL) {
1853 IPFW_UH_RUNLOCK(ch);
1856 IPFW_UH_RUNLOCK(ch);
1858 return (create_table_internal(ch, &ti, aname, i, NULL, 0));
1862 * Creates new table based on @ti and @aname.
1864 * Assume @aname to be checked and valid.
1865 * Stores allocated table kidx inside @pkidx (if non-NULL).
1866 * Reference created table if @compat is non-zero.
1868 * Returns 0 on success.
1871 create_table_internal(struct ip_fw_chain *ch, struct tid_info *ti,
1872 char *aname, ipfw_xtable_info *i, uint16_t *pkidx, int compat)
1874 struct namedobj_instance *ni;
1875 struct table_config *tc, *tc_new, *tmp;
1876 struct table_algo *ta;
1879 ni = CHAIN_TO_NI(ch);
1881 ta = find_table_algo(CHAIN_TO_TCFG(ch), ti, aname);
1885 tc = alloc_table_config(ch, ti, ta, aname, i->tflags);
1889 tc->vmask = i->vmask;
1890 tc->limit = i->limit;
1891 if (ta->flags & TA_FLAG_READONLY)
1894 tc->locked = (i->flags & IPFW_TGFLAGS_LOCKED) != 0;
1898 /* Check if table has been already created */
1899 tc_new = find_table(ni, ti);
1900 if (tc_new != NULL) {
1903 * Compat: do not fail if we're
1904 * requesting to create existing table
1905 * which has the same type
1907 if (compat == 0 || tc_new->no.subtype != tc->no.subtype) {
1908 IPFW_UH_WUNLOCK(ch);
1909 free_table_config(ni, tc);
1913 /* Exchange tc and tc_new for proper refcounting & freeing */
1919 if (ipfw_objhash_alloc_idx(ni, &kidx) != 0) {
1920 IPFW_UH_WUNLOCK(ch);
1921 printf("Unable to allocate table index."
1922 " Consider increasing net.inet.ip.fw.tables_max");
1923 free_table_config(ni, tc);
1927 tc->no.etlv = IPFW_TLV_TBL_NAME;
1935 *pkidx = tc->no.kidx;
1937 IPFW_UH_WUNLOCK(ch);
1940 free_table_config(ni, tc_new);
1946 ntlv_to_ti(ipfw_obj_ntlv *ntlv, struct tid_info *ti)
1949 memset(ti, 0, sizeof(struct tid_info));
1950 ti->set = ntlv->set;
1951 ti->uidx = ntlv->idx;
1953 ti->tlen = ntlv->head.length;
1957 objheader_to_ti(struct _ipfw_obj_header *oh, struct tid_info *ti)
1960 ntlv_to_ti(&oh->ntlv, ti);
1963 struct namedobj_instance *
1964 ipfw_get_table_objhash(struct ip_fw_chain *ch)
1967 return (CHAIN_TO_NI(ch));
1971 * Exports basic table info as name TLV.
1972 * Used inside dump_static_rules() to provide info
1973 * about all tables referenced by current ruleset.
1975 * Returns 0 on success.
1978 ipfw_export_table_ntlv(struct ip_fw_chain *ch, uint16_t kidx,
1979 struct sockopt_data *sd)
1981 struct namedobj_instance *ni;
1982 struct named_object *no;
1983 ipfw_obj_ntlv *ntlv;
1985 ni = CHAIN_TO_NI(ch);
1987 no = ipfw_objhash_lookup_kidx(ni, kidx);
1988 KASSERT(no != NULL, ("invalid table kidx passed"));
1990 ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv));
1994 ntlv->head.type = IPFW_TLV_TBL_NAME;
1995 ntlv->head.length = sizeof(*ntlv);
1996 ntlv->idx = no->kidx;
1997 strlcpy(ntlv->name, no->name, sizeof(ntlv->name));
2003 struct ip_fw_chain *ch;
2004 struct table_info *ti;
2005 struct table_config *tc;
2006 struct sockopt_data *sd;
2011 ipfw_table_entry *ent;
2014 ipfw_obj_tentry tent;
2018 count_ext_entries(void *e, void *arg)
2020 struct dump_args *da;
2022 da = (struct dump_args *)arg;
2029 * Gets number of items from table either using
2030 * internal counter or calling algo callback for
2031 * externally-managed tables.
2033 * Returns number of records.
2036 table_get_count(struct ip_fw_chain *ch, struct table_config *tc)
2038 struct table_info *ti;
2039 struct table_algo *ta;
2040 struct dump_args da;
2042 ti = KIDX_TO_TI(ch, tc->no.kidx);
2045 /* Use internal counter for self-managed tables */
2046 if ((ta->flags & TA_FLAG_READONLY) == 0)
2049 /* Use callback to quickly get number of items */
2050 if ((ta->flags & TA_FLAG_EXTCOUNTER) != 0)
2051 return (ta->get_count(tc->astate, ti));
2053 /* Count number of iterms ourselves */
2054 memset(&da, 0, sizeof(da));
2055 ta->foreach(tc->astate, ti, count_ext_entries, &da);
2061 * Exports table @tc info into standard ipfw_xtable_info format.
2064 export_table_info(struct ip_fw_chain *ch, struct table_config *tc,
2065 ipfw_xtable_info *i)
2067 struct table_info *ti;
2068 struct table_algo *ta;
2070 i->type = tc->no.subtype;
2071 i->tflags = tc->tflags;
2072 i->vmask = tc->vmask;
2073 i->set = tc->no.set;
2074 i->kidx = tc->no.kidx;
2075 i->refcnt = tc->no.refcnt;
2076 i->count = table_get_count(ch, tc);
2077 i->limit = tc->limit;
2078 i->flags |= (tc->locked != 0) ? IPFW_TGFLAGS_LOCKED : 0;
2079 i->size = i->count * sizeof(ipfw_obj_tentry);
2080 i->size += sizeof(ipfw_obj_header) + sizeof(ipfw_xtable_info);
2081 strlcpy(i->tablename, tc->tablename, sizeof(i->tablename));
2082 ti = KIDX_TO_TI(ch, tc->no.kidx);
2084 if (ta->print_config != NULL) {
2085 /* Use algo function to print table config to string */
2086 ta->print_config(tc->astate, ti, i->algoname,
2087 sizeof(i->algoname));
2089 strlcpy(i->algoname, ta->name, sizeof(i->algoname));
2090 /* Dump algo-specific data, if possible */
2091 if (ta->dump_tinfo != NULL) {
2092 ta->dump_tinfo(tc->astate, ti, &i->ta_info);
2093 i->ta_info.flags |= IPFW_TATFLAGS_DATA;
2097 struct dump_table_args {
2098 struct ip_fw_chain *ch;
2099 struct sockopt_data *sd;
2103 export_table_internal(struct namedobj_instance *ni, struct named_object *no,
2106 ipfw_xtable_info *i;
2107 struct dump_table_args *dta;
2109 dta = (struct dump_table_args *)arg;
2111 i = (ipfw_xtable_info *)ipfw_get_sopt_space(dta->sd, sizeof(*i));
2112 KASSERT(i != NULL, ("previously checked buffer is not enough"));
2114 export_table_info(dta->ch, (struct table_config *)no, i);
2119 * Export all tables as ipfw_xtable_info structures to
2120 * storage provided by @sd.
2122 * If supplied buffer is too small, fills in required size
2123 * and returns ENOMEM.
2124 * Returns 0 on success.
2127 export_tables(struct ip_fw_chain *ch, ipfw_obj_lheader *olh,
2128 struct sockopt_data *sd)
2132 struct dump_table_args dta;
2134 count = ipfw_objhash_count(CHAIN_TO_NI(ch));
2135 size = count * sizeof(ipfw_xtable_info) + sizeof(ipfw_obj_lheader);
2137 /* Fill in header regadless of buffer size */
2139 olh->objsize = sizeof(ipfw_xtable_info);
2141 if (size > olh->size) {
2151 ipfw_objhash_foreach(CHAIN_TO_NI(ch), export_table_internal, &dta);
2157 * Dumps all table data
2158 * Data layout (v1)(current):
2159 * Request: [ ipfw_obj_header ], size = ipfw_xtable_info.size
2160 * Reply: [ ipfw_obj_header ipfw_xtable_info ipfw_obj_tentry x N ]
2162 * Returns 0 on success
2165 dump_table_v1(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
2166 struct sockopt_data *sd)
2168 struct _ipfw_obj_header *oh;
2169 ipfw_xtable_info *i;
2171 struct table_config *tc;
2172 struct table_algo *ta;
2173 struct dump_args da;
2176 sz = sizeof(ipfw_obj_header) + sizeof(ipfw_xtable_info);
2177 oh = (struct _ipfw_obj_header *)ipfw_get_sopt_header(sd, sz);
2181 i = (ipfw_xtable_info *)(oh + 1);
2182 objheader_to_ti(oh, &ti);
2185 if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) {
2186 IPFW_UH_RUNLOCK(ch);
2189 export_table_info(ch, tc, i);
2191 if (sd->valsize < i->size) {
2194 * Submitted buffer size is not enough.
2195 * WE've already filled in @i structure with
2196 * relevant table info including size, so we
2197 * can return. Buffer will be flushed automatically.
2199 IPFW_UH_RUNLOCK(ch);
2204 * Do the actual dump in eXtended format
2206 memset(&da, 0, sizeof(da));
2208 da.ti = KIDX_TO_TI(ch, tc->no.kidx);
2214 ta->foreach(tc->astate, da.ti, dump_table_tentry, &da);
2215 IPFW_UH_RUNLOCK(ch);
2221 * Dumps all table data
2222 * Data layout (version 0)(legacy):
2223 * Request: [ ipfw_xtable ], size = IP_FW_TABLE_XGETSIZE()
2224 * Reply: [ ipfw_xtable ipfw_table_xentry x N ]
2226 * Returns 0 on success
2229 dump_table_v0(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
2230 struct sockopt_data *sd)
2234 struct table_config *tc;
2235 struct table_algo *ta;
2236 struct dump_args da;
2239 xtbl = (ipfw_xtable *)ipfw_get_sopt_header(sd, sizeof(ipfw_xtable));
2243 memset(&ti, 0, sizeof(ti));
2244 ti.uidx = xtbl->tbl;
2247 if ((tc = find_table(CHAIN_TO_NI(ch), &ti)) == NULL) {
2248 IPFW_UH_RUNLOCK(ch);
2251 count = table_get_count(ch, tc);
2252 sz = count * sizeof(ipfw_table_xentry) + sizeof(ipfw_xtable);
2256 xtbl->type = tc->no.subtype;
2257 xtbl->tbl = ti.uidx;
2259 if (sd->valsize < sz) {
2262 * Submitted buffer size is not enough.
2263 * WE've already filled in @i structure with
2264 * relevant table info including size, so we
2265 * can return. Buffer will be flushed automatically.
2267 IPFW_UH_RUNLOCK(ch);
2271 /* Do the actual dump in eXtended format */
2272 memset(&da, 0, sizeof(da));
2274 da.ti = KIDX_TO_TI(ch, tc->no.kidx);
2280 ta->foreach(tc->astate, da.ti, dump_table_xentry, &da);
2281 IPFW_UH_RUNLOCK(ch);
2287 * Legacy function to retrieve number of items in table.
2290 get_table_size(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
2291 struct sockopt_data *sd)
2298 sz = sizeof(*op3) + sizeof(uint32_t);
2299 op3 = (ip_fw3_opheader *)ipfw_get_sopt_header(sd, sz);
2303 tbl = (uint32_t *)(op3 + 1);
2304 memset(&ti, 0, sizeof(ti));
2307 error = ipfw_count_xtable(ch, &ti, tbl);
2308 IPFW_UH_RUNLOCK(ch);
2313 * Legacy IP_FW_TABLE_GETSIZE handler
2316 ipfw_count_table(struct ip_fw_chain *ch, struct tid_info *ti, uint32_t *cnt)
2318 struct table_config *tc;
2320 if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL)
2322 *cnt = table_get_count(ch, tc);
2327 * Legacy IP_FW_TABLE_XGETSIZE handler
2330 ipfw_count_xtable(struct ip_fw_chain *ch, struct tid_info *ti, uint32_t *cnt)
2332 struct table_config *tc;
2335 if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL) {
2337 return (0); /* 'table all list' requires success */
2340 count = table_get_count(ch, tc);
2341 *cnt = count * sizeof(ipfw_table_xentry);
2343 *cnt += sizeof(ipfw_xtable);
2348 dump_table_entry(void *e, void *arg)
2350 struct dump_args *da;
2351 struct table_config *tc;
2352 struct table_algo *ta;
2353 ipfw_table_entry *ent;
2354 struct table_value *pval;
2357 da = (struct dump_args *)arg;
2362 /* Out of memory, returning */
2363 if (da->cnt == da->size)
2366 ent->tbl = da->uidx;
2369 error = ta->dump_tentry(tc->astate, da->ti, e, &da->tent);
2373 ent->addr = da->tent.k.addr.s_addr;
2374 ent->masklen = da->tent.masklen;
2375 pval = get_table_value(da->ch, da->tc, da->tent.v.kidx);
2376 ent->value = ipfw_export_table_value_legacy(pval);
2382 * Dumps table in pre-8.1 legacy format.
2385 ipfw_dump_table_legacy(struct ip_fw_chain *ch, struct tid_info *ti,
2388 struct table_config *tc;
2389 struct table_algo *ta;
2390 struct dump_args da;
2394 if ((tc = find_table(CHAIN_TO_NI(ch), ti)) == NULL)
2395 return (0); /* XXX: We should return ESRCH */
2399 /* This dump format supports IPv4 only */
2400 if (tc->no.subtype != IPFW_TABLE_ADDR)
2403 memset(&da, 0, sizeof(da));
2405 da.ti = KIDX_TO_TI(ch, tc->no.kidx);
2407 da.ent = &tbl->ent[0];
2408 da.size = tbl->size;
2411 ta->foreach(tc->astate, da.ti, dump_table_entry, &da);
2418 * Dumps table entry in eXtended format (v1)(current).
2421 dump_table_tentry(void *e, void *arg)
2423 struct dump_args *da;
2424 struct table_config *tc;
2425 struct table_algo *ta;
2426 struct table_value *pval;
2427 ipfw_obj_tentry *tent;
2430 da = (struct dump_args *)arg;
2435 tent = (ipfw_obj_tentry *)ipfw_get_sopt_space(da->sd, sizeof(*tent));
2436 /* Out of memory, returning */
2441 tent->head.length = sizeof(ipfw_obj_tentry);
2442 tent->idx = da->uidx;
2444 error = ta->dump_tentry(tc->astate, da->ti, e, tent);
2448 pval = get_table_value(da->ch, da->tc, tent->v.kidx);
2449 ipfw_export_table_value_v1(pval, &tent->v.value);
2455 * Dumps table entry in eXtended format (v0).
2458 dump_table_xentry(void *e, void *arg)
2460 struct dump_args *da;
2461 struct table_config *tc;
2462 struct table_algo *ta;
2463 ipfw_table_xentry *xent;
2464 ipfw_obj_tentry *tent;
2465 struct table_value *pval;
2468 da = (struct dump_args *)arg;
2473 xent = (ipfw_table_xentry *)ipfw_get_sopt_space(da->sd, sizeof(*xent));
2474 /* Out of memory, returning */
2477 xent->len = sizeof(ipfw_table_xentry);
2478 xent->tbl = da->uidx;
2480 memset(&da->tent, 0, sizeof(da->tent));
2482 error = ta->dump_tentry(tc->astate, da->ti, e, tent);
2486 /* Convert current format to previous one */
2487 xent->masklen = tent->masklen;
2488 pval = get_table_value(da->ch, da->tc, da->tent.v.kidx);
2489 xent->value = ipfw_export_table_value_legacy(pval);
2490 /* Apply some hacks */
2491 if (tc->no.subtype == IPFW_TABLE_ADDR && tent->subtype == AF_INET) {
2492 xent->k.addr6.s6_addr32[3] = tent->k.addr.s_addr;
2493 xent->flags = IPFW_TCF_INET;
2495 memcpy(&xent->k, &tent->k, sizeof(xent->k));
2501 * Helper function to export table algo data
2502 * to tentry format before calling user function.
2504 * Returns 0 on success.
2507 prepare_table_tentry(void *e, void *arg)
2509 struct dump_args *da;
2510 struct table_config *tc;
2511 struct table_algo *ta;
2514 da = (struct dump_args *)arg;
2519 error = ta->dump_tentry(tc->astate, da->ti, e, &da->tent);
2523 da->f(&da->tent, da->farg);
2529 * Allow external consumers to read table entries in standard format.
2532 ipfw_foreach_table_tentry(struct ip_fw_chain *ch, uint16_t kidx,
2533 ta_foreach_f *f, void *arg)
2535 struct namedobj_instance *ni;
2536 struct table_config *tc;
2537 struct table_algo *ta;
2538 struct dump_args da;
2540 ni = CHAIN_TO_NI(ch);
2542 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, kidx);
2548 memset(&da, 0, sizeof(da));
2550 da.ti = KIDX_TO_TI(ch, tc->no.kidx);
2555 ta->foreach(tc->astate, da.ti, prepare_table_tentry, &da);
2565 * Finds algorithm by index, table type or supplied name.
2567 * Returns pointer to algo or NULL.
2569 static struct table_algo *
2570 find_table_algo(struct tables_config *tcfg, struct tid_info *ti, char *name)
2573 struct table_algo *ta;
2575 if (ti->type > IPFW_TABLE_MAXTYPE)
2578 /* Search by index */
2579 if (ti->atype != 0) {
2580 if (ti->atype > tcfg->algo_count)
2582 return (tcfg->algo[ti->atype]);
2586 /* Return default algorithm for given type if set */
2587 return (tcfg->def_algo[ti->type]);
2590 /* Search by name */
2591 /* TODO: better search */
2592 for (i = 1; i <= tcfg->algo_count; i++) {
2596 * One can supply additional algorithm
2597 * parameters so we compare only the first word
2599 * 'addr:chash hsize=32'
2603 l = strlen(ta->name);
2604 if (strncmp(name, ta->name, l) != 0)
2606 if (name[l] != '\0' && name[l] != ' ')
2608 /* Check if we're requesting proper table type */
2609 if (ti->type != 0 && ti->type != ta->type)
2618 * Register new table algo @ta.
2619 * Stores algo id inside @idx.
2621 * Returns 0 on success.
2624 ipfw_add_table_algo(struct ip_fw_chain *ch, struct table_algo *ta, size_t size,
2627 struct tables_config *tcfg;
2628 struct table_algo *ta_new;
2631 if (size > sizeof(struct table_algo))
2634 /* Check for the required on-stack size for add/del */
2635 sz = roundup2(ta->ta_buf_size, sizeof(void *));
2639 KASSERT(ta->type <= IPFW_TABLE_MAXTYPE,("Increase IPFW_TABLE_MAXTYPE"));
2641 /* Copy algorithm data to stable storage. */
2642 ta_new = malloc(sizeof(struct table_algo), M_IPFW, M_WAITOK | M_ZERO);
2643 memcpy(ta_new, ta, size);
2645 tcfg = CHAIN_TO_TCFG(ch);
2647 KASSERT(tcfg->algo_count < 255, ("Increase algo array size"));
2649 tcfg->algo[++tcfg->algo_count] = ta_new;
2650 ta_new->idx = tcfg->algo_count;
2652 /* Set algorithm as default one for given type */
2653 if ((ta_new->flags & TA_FLAG_DEFAULT) != 0 &&
2654 tcfg->def_algo[ta_new->type] == NULL)
2655 tcfg->def_algo[ta_new->type] = ta_new;
2663 * Unregisters table algo using @idx as id.
2664 * XXX: It is NOT safe to call this function in any place
2665 * other than ipfw instance destroy handler.
2668 ipfw_del_table_algo(struct ip_fw_chain *ch, int idx)
2670 struct tables_config *tcfg;
2671 struct table_algo *ta;
2673 tcfg = CHAIN_TO_TCFG(ch);
2675 KASSERT(idx <= tcfg->algo_count, ("algo idx %d out of range 1..%d",
2676 idx, tcfg->algo_count));
2678 ta = tcfg->algo[idx];
2679 KASSERT(ta != NULL, ("algo idx %d is NULL", idx));
2681 if (tcfg->def_algo[ta->type] == ta)
2682 tcfg->def_algo[ta->type] = NULL;
2688 * Lists all table algorithms currently available.
2689 * Data layout (v0)(current):
2690 * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size
2691 * Reply: [ ipfw_obj_lheader ipfw_ta_info x N ]
2693 * Returns 0 on success
2696 list_table_algo(struct ip_fw_chain *ch, ip_fw3_opheader *op3,
2697 struct sockopt_data *sd)
2699 struct _ipfw_obj_lheader *olh;
2700 struct tables_config *tcfg;
2702 struct table_algo *ta;
2703 uint32_t count, n, size;
2705 olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh));
2708 if (sd->valsize < olh->size)
2712 tcfg = CHAIN_TO_TCFG(ch);
2713 count = tcfg->algo_count;
2714 size = count * sizeof(ipfw_ta_info) + sizeof(ipfw_obj_lheader);
2716 /* Fill in header regadless of buffer size */
2718 olh->objsize = sizeof(ipfw_ta_info);
2720 if (size > olh->size) {
2722 IPFW_UH_RUNLOCK(ch);
2727 for (n = 1; n <= count; n++) {
2728 i = (ipfw_ta_info *)ipfw_get_sopt_space(sd, sizeof(*i));
2729 KASSERT(i != NULL, ("previously checked buffer is not enough"));
2731 strlcpy(i->algoname, ta->name, sizeof(i->algoname));
2733 i->refcnt = ta->refcnt;
2736 IPFW_UH_RUNLOCK(ch);
2742 classify_srcdst(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype)
2744 /* Basic IPv4/IPv6 or u32 lookups */
2746 /* Assume ADDR by default */
2747 *ptype = IPFW_TABLE_ADDR;
2750 if (F_LEN(cmd) > F_INSN_SIZE(ipfw_insn_u32)) {
2752 * generic lookup. The key must be
2753 * in 32bit big-endian format.
2755 v = ((ipfw_insn_u32 *)cmd)->d[1];
2764 *ptype = IPFW_TABLE_NUMBER;
2768 *ptype = IPFW_TABLE_NUMBER;
2772 *ptype = IPFW_TABLE_NUMBER;
2776 *ptype = IPFW_TABLE_NUMBER;
2785 classify_via(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype)
2787 ipfw_insn_if *cmdif;
2789 /* Interface table, possibly */
2790 cmdif = (ipfw_insn_if *)cmd;
2791 if (cmdif->name[0] != '\1')
2794 *ptype = IPFW_TABLE_INTERFACE;
2795 *puidx = cmdif->p.kidx;
2801 classify_flow(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype)
2805 *ptype = IPFW_TABLE_FLOW;
2811 update_arg1(ipfw_insn *cmd, uint16_t idx)
2818 update_via(ipfw_insn *cmd, uint16_t idx)
2820 ipfw_insn_if *cmdif;
2822 cmdif = (ipfw_insn_if *)cmd;
2823 cmdif->p.kidx = idx;
2827 table_findbyname(struct ip_fw_chain *ch, struct tid_info *ti,
2828 struct named_object **pno)
2830 struct table_config *tc;
2833 IPFW_UH_WLOCK_ASSERT(ch);
2835 error = find_table_err(CHAIN_TO_NI(ch), ti, &tc);
2843 /* XXX: sets-sets! */
2844 static struct named_object *
2845 table_findbykidx(struct ip_fw_chain *ch, uint16_t idx)
2847 struct namedobj_instance *ni;
2848 struct table_config *tc;
2850 IPFW_UH_WLOCK_ASSERT(ch);
2851 ni = CHAIN_TO_NI(ch);
2852 tc = (struct table_config *)ipfw_objhash_lookup_kidx(ni, idx);
2853 KASSERT(tc != NULL, ("Table with index %d not found", idx));
2859 table_manage_sets(struct ip_fw_chain *ch, uint16_t set, uint8_t new_set,
2860 enum ipfw_sets_cmd cmd)
2868 * Always return success, the real action and decision
2869 * should make table_manage_sets_all().
2875 * NOTE: we need to use ipfw_objhash_del/ipfw_objhash_add
2876 * if set number will be used in hash function. Currently
2877 * we can just use generic handler that replaces set value.
2879 if (V_fw_tables_sets == 0)
2884 * Return EOPNOTSUPP for COUNT_ONE when per-set sysctl is
2885 * disabled. This allow skip table's opcodes from additional
2886 * checks when specific rules moved to another set.
2888 if (V_fw_tables_sets == 0)
2889 return (EOPNOTSUPP);
2891 /* Use generic sets handler when per-set sysctl is enabled. */
2892 return (ipfw_obj_manage_sets(CHAIN_TO_NI(ch), IPFW_TLV_TBL_NAME,
2893 set, new_set, cmd));
2897 * We register several opcode rewriters for lookup tables.
2898 * All tables opcodes have the same ETLV type, but different subtype.
2899 * To avoid invoking sets handler several times for XXX_ALL commands,
2900 * we use separate manage_sets handler. O_RECV has the lowest value,
2901 * so it should be called first.
2904 table_manage_sets_all(struct ip_fw_chain *ch, uint16_t set, uint8_t new_set,
2905 enum ipfw_sets_cmd cmd)
2912 * Return success for TEST_ALL, since nothing prevents
2913 * move rules from one set to another. All tables are
2914 * accessible from all sets when per-set tables sysctl
2918 if (V_fw_tables_sets == 0)
2922 return (table_manage_sets(ch, set, new_set, cmd));
2924 /* Use generic sets handler when per-set sysctl is enabled. */
2925 return (ipfw_obj_manage_sets(CHAIN_TO_NI(ch), IPFW_TLV_TBL_NAME,
2926 set, new_set, cmd));
2929 static struct opcode_obj_rewrite opcodes[] = {
2931 .opcode = O_IP_SRC_LOOKUP,
2932 .etlv = IPFW_TLV_TBL_NAME,
2933 .classifier = classify_srcdst,
2934 .update = update_arg1,
2935 .find_byname = table_findbyname,
2936 .find_bykidx = table_findbykidx,
2937 .create_object = create_table_compat,
2938 .manage_sets = table_manage_sets,
2941 .opcode = O_IP_DST_LOOKUP,
2942 .etlv = IPFW_TLV_TBL_NAME,
2943 .classifier = classify_srcdst,
2944 .update = update_arg1,
2945 .find_byname = table_findbyname,
2946 .find_bykidx = table_findbykidx,
2947 .create_object = create_table_compat,
2948 .manage_sets = table_manage_sets,
2951 .opcode = O_IP_FLOW_LOOKUP,
2952 .etlv = IPFW_TLV_TBL_NAME,
2953 .classifier = classify_flow,
2954 .update = update_arg1,
2955 .find_byname = table_findbyname,
2956 .find_bykidx = table_findbykidx,
2957 .create_object = create_table_compat,
2958 .manage_sets = table_manage_sets,
2962 .etlv = IPFW_TLV_TBL_NAME,
2963 .classifier = classify_via,
2964 .update = update_via,
2965 .find_byname = table_findbyname,
2966 .find_bykidx = table_findbykidx,
2967 .create_object = create_table_compat,
2968 .manage_sets = table_manage_sets,
2972 .etlv = IPFW_TLV_TBL_NAME,
2973 .classifier = classify_via,
2974 .update = update_via,
2975 .find_byname = table_findbyname,
2976 .find_bykidx = table_findbykidx,
2977 .create_object = create_table_compat,
2978 .manage_sets = table_manage_sets_all,
2982 .etlv = IPFW_TLV_TBL_NAME,
2983 .classifier = classify_via,
2984 .update = update_via,
2985 .find_byname = table_findbyname,
2986 .find_bykidx = table_findbykidx,
2987 .create_object = create_table_compat,
2988 .manage_sets = table_manage_sets,
2993 test_sets_cb(struct namedobj_instance *ni __unused, struct named_object *no,
2997 /* Check that there aren't any tables in not default set */
3004 * Switch between "set 0" and "rule's set" table binding,
3005 * Check all ruleset bindings and permits changing
3006 * IFF each binding has both rule AND table in default set (set 0).
3008 * Returns 0 on success.
3011 ipfw_switch_tables_namespace(struct ip_fw_chain *ch, unsigned int sets)
3013 struct opcode_obj_rewrite *rw;
3014 struct namedobj_instance *ni;
3015 struct named_object *no;
3024 if (V_fw_tables_sets == sets) {
3025 IPFW_UH_WUNLOCK(ch);
3028 ni = CHAIN_TO_NI(ch);
3031 * Prevent disabling sets support if we have some tables
3032 * in not default sets.
3034 if (ipfw_objhash_foreach_type(ni, test_sets_cb,
3035 NULL, IPFW_TLV_TBL_NAME) != 0) {
3036 IPFW_UH_WUNLOCK(ch);
3041 * Scan all rules and examine tables opcodes.
3043 for (i = 0; i < ch->n_rules; i++) {
3049 for ( ; l > 0 ; l -= cmdlen, cmd += cmdlen) {
3050 cmdlen = F_LEN(cmd);
3051 /* Check only tables opcodes */
3052 for (kidx = 0, rw = opcodes;
3053 rw < opcodes + nitems(opcodes); rw++) {
3054 if (rw->opcode != cmd->opcode)
3056 if (rw->classifier(cmd, &kidx, &subtype) == 0)
3061 no = ipfw_objhash_lookup_kidx(ni, kidx);
3062 /* Check if both table object and rule has the set 0 */
3063 if (no->set != 0 || rule->set != 0) {
3064 IPFW_UH_WUNLOCK(ch);
3070 V_fw_tables_sets = sets;
3071 IPFW_UH_WUNLOCK(ch);
3076 * Checks table name for validity.
3077 * Enforce basic length checks, the rest
3078 * should be done in userland.
3080 * Returns 0 if name is considered valid.
3083 check_table_name(const char *name)
3087 * TODO: do some more complicated checks
3089 return (ipfw_check_object_name_generic(name));
3093 * Finds table config based on either legacy index
3095 * Note @ti structure contains unchecked data from userland.
3097 * Returns 0 in success and fills in @tc with found config
3100 find_table_err(struct namedobj_instance *ni, struct tid_info *ti,
3101 struct table_config **tc)
3103 char *name, bname[16];
3104 struct named_object *no;
3105 ipfw_obj_ntlv *ntlv;
3108 if (ti->tlvs != NULL) {
3109 ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx,
3116 * Use set provided by @ti instead of @ntlv one.
3117 * This is needed due to different sets behavior
3118 * controlled by V_fw_tables_sets.
3120 set = (V_fw_tables_sets != 0) ? ti->set : 0;
3122 snprintf(bname, sizeof(bname), "%d", ti->uidx);
3127 no = ipfw_objhash_lookup_name(ni, set, name);
3128 *tc = (struct table_config *)no;
3134 * Finds table config based on either legacy index
3136 * Note @ti structure contains unchecked data from userland.
3138 * Returns pointer to table_config or NULL.
3140 static struct table_config *
3141 find_table(struct namedobj_instance *ni, struct tid_info *ti)
3143 struct table_config *tc;
3145 if (find_table_err(ni, ti, &tc) != 0)
3152 * Allocate new table config structure using
3153 * specified @algo and @aname.
3155 * Returns pointer to config or NULL.
3157 static struct table_config *
3158 alloc_table_config(struct ip_fw_chain *ch, struct tid_info *ti,
3159 struct table_algo *ta, char *aname, uint8_t tflags)
3161 char *name, bname[16];
3162 struct table_config *tc;
3164 ipfw_obj_ntlv *ntlv;
3167 if (ti->tlvs != NULL) {
3168 ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx,
3173 set = (V_fw_tables_sets == 0) ? 0 : ntlv->set;
3175 /* Compat part: convert number to string representation */
3176 snprintf(bname, sizeof(bname), "%d", ti->uidx);
3181 tc = malloc(sizeof(struct table_config), M_IPFW, M_WAITOK | M_ZERO);
3182 tc->no.name = tc->tablename;
3183 tc->no.subtype = ta->type;
3185 tc->tflags = tflags;
3187 strlcpy(tc->tablename, name, sizeof(tc->tablename));
3188 /* Set "shared" value type by default */
3191 /* Preallocate data structures for new tables */
3192 error = ta->init(ch, &tc->astate, &tc->ti_copy, aname, tflags);
3202 * Destroys table state and config.
3205 free_table_config(struct namedobj_instance *ni, struct table_config *tc)
3208 KASSERT(tc->linked == 0, ("free() on linked config"));
3209 /* UH lock MUST NOT be held */
3212 * We're using ta without any locking/referencing.
3213 * TODO: fix this if we're going to use unloadable algos.
3215 tc->ta->destroy(tc->astate, &tc->ti_copy);
3220 * Links @tc to @chain table named instance.
3221 * Sets appropriate type/states in @chain table info.
3224 link_table(struct ip_fw_chain *ch, struct table_config *tc)
3226 struct namedobj_instance *ni;
3227 struct table_info *ti;
3230 IPFW_UH_WLOCK_ASSERT(ch);
3232 ni = CHAIN_TO_NI(ch);
3235 ipfw_objhash_add(ni, &tc->no);
3237 ti = KIDX_TO_TI(ch, kidx);
3240 /* Notify algo on real @ti address */
3241 if (tc->ta->change_ti != NULL)
3242 tc->ta->change_ti(tc->astate, ti);
3249 * Unlinks @tc from @chain table named instance.
3250 * Zeroes states in @chain and stores them in @tc.
3253 unlink_table(struct ip_fw_chain *ch, struct table_config *tc)
3255 struct namedobj_instance *ni;
3256 struct table_info *ti;
3259 IPFW_UH_WLOCK_ASSERT(ch);
3260 IPFW_WLOCK_ASSERT(ch);
3262 ni = CHAIN_TO_NI(ch);
3265 /* Clear state. @ti copy is already saved inside @tc */
3266 ipfw_objhash_del(ni, &tc->no);
3267 ti = KIDX_TO_TI(ch, kidx);
3268 memset(ti, 0, sizeof(struct table_info));
3272 /* Notify algo on real @ti address */
3273 if (tc->ta->change_ti != NULL)
3274 tc->ta->change_ti(tc->astate, NULL);
3277 static struct ipfw_sopt_handler scodes[] = {
3278 { IP_FW_TABLE_XCREATE, 0, HDIR_SET, create_table },
3279 { IP_FW_TABLE_XDESTROY, 0, HDIR_SET, flush_table_v0 },
3280 { IP_FW_TABLE_XFLUSH, 0, HDIR_SET, flush_table_v0 },
3281 { IP_FW_TABLE_XMODIFY, 0, HDIR_BOTH, modify_table },
3282 { IP_FW_TABLE_XINFO, 0, HDIR_GET, describe_table },
3283 { IP_FW_TABLES_XLIST, 0, HDIR_GET, list_tables },
3284 { IP_FW_TABLE_XLIST, 0, HDIR_GET, dump_table_v0 },
3285 { IP_FW_TABLE_XLIST, 1, HDIR_GET, dump_table_v1 },
3286 { IP_FW_TABLE_XADD, 0, HDIR_BOTH, manage_table_ent_v0 },
3287 { IP_FW_TABLE_XADD, 1, HDIR_BOTH, manage_table_ent_v1 },
3288 { IP_FW_TABLE_XDEL, 0, HDIR_BOTH, manage_table_ent_v0 },
3289 { IP_FW_TABLE_XDEL, 1, HDIR_BOTH, manage_table_ent_v1 },
3290 { IP_FW_TABLE_XFIND, 0, HDIR_GET, find_table_entry },
3291 { IP_FW_TABLE_XSWAP, 0, HDIR_SET, swap_table },
3292 { IP_FW_TABLES_ALIST, 0, HDIR_GET, list_table_algo },
3293 { IP_FW_TABLE_XGETSIZE, 0, HDIR_GET, get_table_size },
3297 destroy_table_locked(struct namedobj_instance *ni, struct named_object *no,
3301 unlink_table((struct ip_fw_chain *)arg, (struct table_config *)no);
3302 if (ipfw_objhash_free_idx(ni, no->kidx) != 0)
3303 printf("Error unlinking kidx %d from table %s\n",
3304 no->kidx, no->name);
3305 free_table_config(ni, (struct table_config *)no);
3310 * Shuts tables module down.
3313 ipfw_destroy_tables(struct ip_fw_chain *ch, int last)
3316 IPFW_DEL_SOPT_HANDLER(last, scodes);
3317 IPFW_DEL_OBJ_REWRITER(last, opcodes);
3319 /* Remove all tables from working set */
3322 ipfw_objhash_foreach(CHAIN_TO_NI(ch), destroy_table_locked, ch);
3324 IPFW_UH_WUNLOCK(ch);
3326 /* Free pointers itself */
3327 free(ch->tablestate, M_IPFW);
3329 ipfw_table_value_destroy(ch, last);
3330 ipfw_table_algo_destroy(ch);
3332 ipfw_objhash_destroy(CHAIN_TO_NI(ch));
3333 free(CHAIN_TO_TCFG(ch), M_IPFW);
3337 * Starts tables module.
3340 ipfw_init_tables(struct ip_fw_chain *ch, int first)
3342 struct tables_config *tcfg;
3344 /* Allocate pointers */
3345 ch->tablestate = malloc(V_fw_tables_max * sizeof(struct table_info),
3346 M_IPFW, M_WAITOK | M_ZERO);
3348 tcfg = malloc(sizeof(struct tables_config), M_IPFW, M_WAITOK | M_ZERO);
3349 tcfg->namehash = ipfw_objhash_create(V_fw_tables_max);
3352 ipfw_table_value_init(ch, first);
3353 ipfw_table_algo_init(ch);
3355 IPFW_ADD_OBJ_REWRITER(first, opcodes);
3356 IPFW_ADD_SOPT_HANDLER(first, scodes);