1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2001 Daniel Hartmeier
5 * Copyright (c) 2002,2003 Henning Brauer
6 * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
13 * - Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * - Redistributions in binary form must reproduce the above
16 * copyright notice, this list of conditions and the following
17 * disclaimer in the documentation and/or other materials provided
18 * with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 * POSSIBILITY OF SUCH DAMAGE.
32 *
33 * Effort sponsored in part by the Defense Advanced Research Projects
34 * Agency (DARPA) and Air Force Research Laboratory, Air Force
35 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36 *
37 * $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38 */
39
40 #include <sys/cdefs.h>
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45
46 #include <sys/param.h>
47 #include <sys/_bitset.h>
48 #include <sys/bitset.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/hash.h>
55 #include <sys/interrupt.h>
56 #include <sys/jail.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/lock.h>
60 #include <sys/mbuf.h>
61 #include <sys/module.h>
62 #include <sys/nv.h>
63 #include <sys/proc.h>
64 #include <sys/sdt.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/if_private.h>
74 #include <net/vnet.h>
75 #include <net/route.h>
76 #include <net/pfil.h>
77 #include <net/pfvar.h>
78 #include <net/if_pfsync.h>
79 #include <net/if_pflog.h>
80
81 #include <netinet/in.h>
82 #include <netinet/ip.h>
83 #include <netinet/ip_var.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet/ip_icmp.h>
86 #include <netpfil/pf/pf_nl.h>
87 #include <netpfil/pf/pf_nv.h>
88
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101
102 static struct pf_kpool *pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 u_int32_t, u_int8_t, u_int8_t, u_int8_t, int);
104
105 static void pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void pf_empty_kpool(struct pf_kpalist *);
107 static int pfioctl(struct cdev *, u_long, caddr_t, int,
108 struct thread *);
109 static int pf_begin_eth(uint32_t *, const char *);
110 static int pf_rollback_eth(uint32_t, const char *);
111 static int pf_commit_eth(uint32_t, const char *);
112 static void pf_free_eth_rule(struct pf_keth_rule *);
113 #ifdef ALTQ
114 static int pf_begin_altq(u_int32_t *);
115 static int pf_rollback_altq(u_int32_t);
116 static int pf_commit_altq(u_int32_t);
117 static int pf_enable_altq(struct pf_altq *);
118 static int pf_disable_altq(struct pf_altq *);
119 static uint16_t pf_qname2qid(const char *);
120 static void pf_qid_unref(uint16_t);
121 #endif /* ALTQ */
122 static int pf_begin_rules(u_int32_t *, int, const char *);
123 static int pf_rollback_rules(u_int32_t, int, char *);
124 static int pf_setup_pfsync_matching(struct pf_kruleset *);
125 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
126 static void pf_hash_rule(struct pf_krule *);
127 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
128 static int pf_commit_rules(u_int32_t, int, char *);
129 static int pf_addr_setup(struct pf_kruleset *,
130 struct pf_addr_wrap *, sa_family_t);
131 static void pf_src_node_copy(const struct pf_ksrc_node *,
132 struct pf_src_node *);
133 #ifdef ALTQ
134 static int pf_export_kaltq(struct pf_altq *,
135 struct pfioc_altq_v1 *, size_t);
136 static int pf_import_kaltq(struct pfioc_altq_v1 *,
137 struct pf_altq *, size_t);
138 #endif /* ALTQ */
139
140 VNET_DEFINE(struct pf_krule, pf_default_rule);
141
142 static __inline int pf_krule_compare(struct pf_krule *,
143 struct pf_krule *);
144
145 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
146
147 #ifdef ALTQ
148 VNET_DEFINE_STATIC(int, pf_altq_running);
149 #define V_pf_altq_running VNET(pf_altq_running)
150 #endif
151
152 #define TAGID_MAX 50000
153 struct pf_tagname {
154 TAILQ_ENTRY(pf_tagname) namehash_entries;
155 TAILQ_ENTRY(pf_tagname) taghash_entries;
156 char name[PF_TAG_NAME_SIZE];
157 uint16_t tag;
158 int ref;
159 };
160
161 struct pf_tagset {
162 TAILQ_HEAD(, pf_tagname) *namehash;
163 TAILQ_HEAD(, pf_tagname) *taghash;
164 unsigned int mask;
165 uint32_t seed;
166 BITSET_DEFINE(, TAGID_MAX) avail;
167 };
168
169 VNET_DEFINE(struct pf_tagset, pf_tags);
170 #define V_pf_tags VNET(pf_tags)
171 static unsigned int pf_rule_tag_hashsize;
172 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128
173 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
174 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
175 "Size of pf(4) rule tag hashtable");
176
177 #ifdef ALTQ
178 VNET_DEFINE(struct pf_tagset, pf_qids);
179 #define V_pf_qids VNET(pf_qids)
180 static unsigned int pf_queue_tag_hashsize;
181 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128
182 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
183 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
184 "Size of pf(4) queue tag hashtable");
185 #endif
186 VNET_DEFINE(uma_zone_t, pf_tag_z);
187 #define V_pf_tag_z VNET(pf_tag_z)
188 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
189 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
190
191 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
192 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
193 #endif
194
195 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
196 #define V_pf_filter_local VNET(pf_filter_local)
197 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
198 &VNET_NAME(pf_filter_local), false,
199 "Enable filtering for packets delivered to local network stack");
200
201 #ifdef PF_DEFAULT_TO_DROP
202 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
203 #else
204 VNET_DEFINE_STATIC(bool, default_to_drop);
205 #endif
206 #define V_default_to_drop VNET(default_to_drop)
207 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
208 &VNET_NAME(default_to_drop), false,
209 "Make the default rule drop all packets.");
210
211 static void pf_init_tagset(struct pf_tagset *, unsigned int *,
212 unsigned int);
213 static void pf_cleanup_tagset(struct pf_tagset *);
214 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *);
215 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t);
216 static u_int16_t tagname2tag(struct pf_tagset *, const char *);
217 static u_int16_t pf_tagname2tag(const char *);
218 static void tag_unref(struct pf_tagset *, u_int16_t);
219
220 struct cdev *pf_dev;
221
222 /*
223 * XXX - These are new and need to be checked when moveing to a new version
224 */
225 static void pf_clear_all_states(void);
226 static int pf_killstates_row(struct pf_kstate_kill *,
227 struct pf_idhash *);
228 static int pf_killstates_nv(struct pfioc_nv *);
229 static int pf_clearstates_nv(struct pfioc_nv *);
230 static int pf_getstate(struct pfioc_nv *);
231 static int pf_getstatus(struct pfioc_nv *);
232 static int pf_clear_tables(void);
233 static void pf_kill_srcnodes(struct pfioc_src_node_kill *);
234 static int pf_keepcounters(struct pfioc_nv *);
235 static void pf_tbladdr_copyout(struct pf_addr_wrap *);
236
237 /*
238 * Wrapper functions for pfil(9) hooks
239 */
240 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
241 int flags, void *ruleset __unused, struct inpcb *inp);
242 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
243 int flags, void *ruleset __unused, struct inpcb *inp);
244 #ifdef INET
245 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
246 int flags, void *ruleset __unused, struct inpcb *inp);
247 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
248 int flags, void *ruleset __unused, struct inpcb *inp);
249 #endif
250 #ifdef INET6
251 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
252 int flags, void *ruleset __unused, struct inpcb *inp);
253 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
254 int flags, void *ruleset __unused, struct inpcb *inp);
255 #endif
256
257 static void hook_pf_eth(void);
258 static void hook_pf(void);
259 static void dehook_pf_eth(void);
260 static void dehook_pf(void);
261 static int shutdown_pf(void);
262 static int pf_load(void);
263 static void pf_unload(void);
264
265 static struct cdevsw pf_cdevsw = {
266 .d_ioctl = pfioctl,
267 .d_name = PF_NAME,
268 .d_version = D_VERSION,
269 };
270
271 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
272 #define V_pf_pfil_hooked VNET(pf_pfil_hooked)
273 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
274 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked)
275
276 /*
277 * We need a flag that is neither hooked nor running to know when
278 * the VNET is "valid". We primarily need this to control (global)
279 * external event, e.g., eventhandlers.
280 */
281 VNET_DEFINE(int, pf_vnet_active);
282 #define V_pf_vnet_active VNET(pf_vnet_active)
283
284 int pf_end_threads;
285 struct proc *pf_purge_proc;
286
287 VNET_DEFINE(struct rmlock, pf_rules_lock);
288 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
289 #define V_pf_ioctl_lock VNET(pf_ioctl_lock)
290 struct sx pf_end_lock;
291
292 /* pfsync */
293 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
294 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
295 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
296 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
297 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
298 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
299 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
300 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
301
302 /* pflog */
303 pflog_packet_t *pflog_packet_ptr = NULL;
304
305 /*
306 * Copy a user-provided string, returning an error if truncation would occur.
307 * Avoid scanning past "sz" bytes in the source string since there's no
308 * guarantee that it's nul-terminated.
309 */
310 static int
pf_user_strcpy(char * dst,const char * src,size_t sz)311 pf_user_strcpy(char *dst, const char *src, size_t sz)
312 {
313 if (strnlen(src, sz) == sz)
314 return (EINVAL);
315 (void)strlcpy(dst, src, sz);
316 return (0);
317 }
318
319 static void
pfattach_vnet(void)320 pfattach_vnet(void)
321 {
322 u_int32_t *my_timeout = V_pf_default_rule.timeout;
323
324 bzero(&V_pf_status, sizeof(V_pf_status));
325
326 pf_initialize();
327 pfr_initialize();
328 pfi_initialize_vnet();
329 pf_normalize_init();
330 pf_syncookies_init();
331
332 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
333 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
334
335 RB_INIT(&V_pf_anchors);
336 pf_init_kruleset(&pf_main_ruleset);
337
338 pf_init_keth(V_pf_keth);
339
340 /* default rule should never be garbage collected */
341 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
342 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
343 V_pf_default_rule.nr = (uint32_t)-1;
344 V_pf_default_rule.rtableid = -1;
345
346 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
347 for (int i = 0; i < 2; i++) {
348 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
349 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
350 }
351 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
352 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
353 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++)
354 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
355
356 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
357 M_WAITOK | M_ZERO);
358
359 #ifdef PF_WANT_32_TO_64_COUNTER
360 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
361 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
362 PF_RULES_WLOCK();
363 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
364 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
365 V_pf_allrulecount++;
366 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
367 PF_RULES_WUNLOCK();
368 #endif
369
370 /* initialize default timeouts */
371 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
372 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
373 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
374 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
375 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
376 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
377 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
378 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
379 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
380 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
381 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
382 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
383 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
384 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
385 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
386 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
387 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
388 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
389 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
390 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
391 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
392 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
393 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
394 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
395 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
396
397 V_pf_status.debug = PF_DEBUG_URGENT;
398 /*
399 * XXX This is different than in OpenBSD where reassembly is enabled by
400 * defult. In FreeBSD we expect people to still use scrub rules and
401 * switch to the new syntax later. Only when they switch they must
402 * explicitly enable reassemle. We could change the default once the
403 * scrub rule functionality is hopefully removed some day in future.
404 */
405 V_pf_status.reass = 0;
406
407 V_pf_pfil_hooked = false;
408 V_pf_pfil_eth_hooked = false;
409
410 /* XXX do our best to avoid a conflict */
411 V_pf_status.hostid = arc4random();
412
413 for (int i = 0; i < PFRES_MAX; i++)
414 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
415 for (int i = 0; i < KLCNT_MAX; i++)
416 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
417 for (int i = 0; i < FCNT_MAX; i++)
418 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
419 for (int i = 0; i < SCNT_MAX; i++)
420 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
421
422 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
423 INTR_MPSAFE, &V_pf_swi_cookie) != 0)
424 /* XXXGL: leaked all above. */
425 return;
426 }
427
428 static struct pf_kpool *
pf_get_kpool(const char * anchor,u_int32_t ticket,u_int8_t rule_action,u_int32_t rule_number,u_int8_t r_last,u_int8_t active,u_int8_t check_ticket,int which)429 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
430 u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
431 u_int8_t check_ticket, int which)
432 {
433 struct pf_kruleset *ruleset;
434 struct pf_krule *rule;
435 int rs_num;
436
437 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT);
438
439 ruleset = pf_find_kruleset(anchor);
440 if (ruleset == NULL)
441 return (NULL);
442 rs_num = pf_get_ruleset_number(rule_action);
443 if (rs_num >= PF_RULESET_MAX)
444 return (NULL);
445 if (active) {
446 if (check_ticket && ticket !=
447 ruleset->rules[rs_num].active.ticket)
448 return (NULL);
449 if (r_last)
450 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
451 pf_krulequeue);
452 else
453 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
454 } else {
455 if (check_ticket && ticket !=
456 ruleset->rules[rs_num].inactive.ticket)
457 return (NULL);
458 if (r_last)
459 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
460 pf_krulequeue);
461 else
462 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
463 }
464 if (!r_last) {
465 while ((rule != NULL) && (rule->nr != rule_number))
466 rule = TAILQ_NEXT(rule, entries);
467 }
468 if (rule == NULL)
469 return (NULL);
470
471 switch (which) {
472 case PF_RDR:
473 return (&rule->rdr);
474 case PF_NAT:
475 return (&rule->nat);
476 case PF_RT:
477 return (&rule->route);
478 default:
479 panic("Unknow pool type %d", which);
480 }
481 }
482
483 static void
pf_mv_kpool(struct pf_kpalist * poola,struct pf_kpalist * poolb)484 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
485 {
486 struct pf_kpooladdr *mv_pool_pa;
487
488 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
489 TAILQ_REMOVE(poola, mv_pool_pa, entries);
490 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
491 }
492 }
493
494 static void
pf_empty_kpool(struct pf_kpalist * poola)495 pf_empty_kpool(struct pf_kpalist *poola)
496 {
497 struct pf_kpooladdr *pa;
498
499 while ((pa = TAILQ_FIRST(poola)) != NULL) {
500 switch (pa->addr.type) {
501 case PF_ADDR_DYNIFTL:
502 pfi_dynaddr_remove(pa->addr.p.dyn);
503 break;
504 case PF_ADDR_TABLE:
505 /* XXX: this could be unfinished pooladdr on pabuf */
506 if (pa->addr.p.tbl != NULL)
507 pfr_detach_table(pa->addr.p.tbl);
508 break;
509 }
510 if (pa->kif)
511 pfi_kkif_unref(pa->kif);
512 TAILQ_REMOVE(poola, pa, entries);
513 free(pa, M_PFRULE);
514 }
515 }
516
517 static void
pf_unlink_rule_locked(struct pf_krulequeue * rulequeue,struct pf_krule * rule)518 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
519 {
520
521 PF_RULES_WASSERT();
522 PF_UNLNKDRULES_ASSERT();
523
524 TAILQ_REMOVE(rulequeue, rule, entries);
525
526 rule->rule_ref |= PFRULE_REFS;
527 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
528 }
529
530 static void
pf_unlink_rule(struct pf_krulequeue * rulequeue,struct pf_krule * rule)531 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
532 {
533
534 PF_RULES_WASSERT();
535
536 PF_UNLNKDRULES_LOCK();
537 pf_unlink_rule_locked(rulequeue, rule);
538 PF_UNLNKDRULES_UNLOCK();
539 }
540
541 static void
pf_free_eth_rule(struct pf_keth_rule * rule)542 pf_free_eth_rule(struct pf_keth_rule *rule)
543 {
544 PF_RULES_WASSERT();
545
546 if (rule == NULL)
547 return;
548
549 if (rule->tag)
550 tag_unref(&V_pf_tags, rule->tag);
551 if (rule->match_tag)
552 tag_unref(&V_pf_tags, rule->match_tag);
553 #ifdef ALTQ
554 pf_qid_unref(rule->qid);
555 #endif
556
557 if (rule->bridge_to)
558 pfi_kkif_unref(rule->bridge_to);
559 if (rule->kif)
560 pfi_kkif_unref(rule->kif);
561
562 if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
563 pfr_detach_table(rule->ipsrc.addr.p.tbl);
564 if (rule->ipdst.addr.type == PF_ADDR_TABLE)
565 pfr_detach_table(rule->ipdst.addr.p.tbl);
566
567 counter_u64_free(rule->evaluations);
568 for (int i = 0; i < 2; i++) {
569 counter_u64_free(rule->packets[i]);
570 counter_u64_free(rule->bytes[i]);
571 }
572 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
573 pf_keth_anchor_remove(rule);
574
575 free(rule, M_PFRULE);
576 }
577
578 void
pf_free_rule(struct pf_krule * rule)579 pf_free_rule(struct pf_krule *rule)
580 {
581
582 PF_RULES_WASSERT();
583 PF_CONFIG_ASSERT();
584
585 if (rule->tag)
586 tag_unref(&V_pf_tags, rule->tag);
587 if (rule->match_tag)
588 tag_unref(&V_pf_tags, rule->match_tag);
589 #ifdef ALTQ
590 if (rule->pqid != rule->qid)
591 pf_qid_unref(rule->pqid);
592 pf_qid_unref(rule->qid);
593 #endif
594 switch (rule->src.addr.type) {
595 case PF_ADDR_DYNIFTL:
596 pfi_dynaddr_remove(rule->src.addr.p.dyn);
597 break;
598 case PF_ADDR_TABLE:
599 pfr_detach_table(rule->src.addr.p.tbl);
600 break;
601 }
602 switch (rule->dst.addr.type) {
603 case PF_ADDR_DYNIFTL:
604 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
605 break;
606 case PF_ADDR_TABLE:
607 pfr_detach_table(rule->dst.addr.p.tbl);
608 break;
609 }
610 if (rule->overload_tbl)
611 pfr_detach_table(rule->overload_tbl);
612 if (rule->kif)
613 pfi_kkif_unref(rule->kif);
614 if (rule->rcv_kif)
615 pfi_kkif_unref(rule->rcv_kif);
616 pf_remove_kanchor(rule);
617 pf_empty_kpool(&rule->rdr.list);
618 pf_empty_kpool(&rule->nat.list);
619 pf_empty_kpool(&rule->route.list);
620
621 pf_krule_free(rule);
622 }
623
624 static void
pf_init_tagset(struct pf_tagset * ts,unsigned int * tunable_size,unsigned int default_size)625 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
626 unsigned int default_size)
627 {
628 unsigned int i;
629 unsigned int hashsize;
630
631 if (*tunable_size == 0 || !powerof2(*tunable_size))
632 *tunable_size = default_size;
633
634 hashsize = *tunable_size;
635 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
636 M_WAITOK);
637 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
638 M_WAITOK);
639 ts->mask = hashsize - 1;
640 ts->seed = arc4random();
641 for (i = 0; i < hashsize; i++) {
642 TAILQ_INIT(&ts->namehash[i]);
643 TAILQ_INIT(&ts->taghash[i]);
644 }
645 BIT_FILL(TAGID_MAX, &ts->avail);
646 }
647
648 static void
pf_cleanup_tagset(struct pf_tagset * ts)649 pf_cleanup_tagset(struct pf_tagset *ts)
650 {
651 unsigned int i;
652 unsigned int hashsize;
653 struct pf_tagname *t, *tmp;
654
655 /*
656 * Only need to clean up one of the hashes as each tag is hashed
657 * into each table.
658 */
659 hashsize = ts->mask + 1;
660 for (i = 0; i < hashsize; i++)
661 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
662 uma_zfree(V_pf_tag_z, t);
663
664 free(ts->namehash, M_PFHASH);
665 free(ts->taghash, M_PFHASH);
666 }
667
668 static uint16_t
tagname2hashindex(const struct pf_tagset * ts,const char * tagname)669 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
670 {
671 size_t len;
672
673 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
674 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
675 }
676
677 static uint16_t
tag2hashindex(const struct pf_tagset * ts,uint16_t tag)678 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
679 {
680
681 return (tag & ts->mask);
682 }
683
684 static u_int16_t
tagname2tag(struct pf_tagset * ts,const char * tagname)685 tagname2tag(struct pf_tagset *ts, const char *tagname)
686 {
687 struct pf_tagname *tag;
688 u_int32_t index;
689 u_int16_t new_tagid;
690
691 PF_RULES_WASSERT();
692
693 index = tagname2hashindex(ts, tagname);
694 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
695 if (strcmp(tagname, tag->name) == 0) {
696 tag->ref++;
697 return (tag->tag);
698 }
699
700 /*
701 * new entry
702 *
703 * to avoid fragmentation, we do a linear search from the beginning
704 * and take the first free slot we find.
705 */
706 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
707 /*
708 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
709 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
710 * set. It may also return a bit number greater than TAGID_MAX due
711 * to rounding of the number of bits in the vector up to a multiple
712 * of the vector word size at declaration/allocation time.
713 */
714 if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
715 return (0);
716
717 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */
718 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
719
720 /* allocate and fill new struct pf_tagname */
721 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
722 if (tag == NULL)
723 return (0);
724 strlcpy(tag->name, tagname, sizeof(tag->name));
725 tag->tag = new_tagid;
726 tag->ref = 1;
727
728 /* Insert into namehash */
729 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
730
731 /* Insert into taghash */
732 index = tag2hashindex(ts, new_tagid);
733 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
734
735 return (tag->tag);
736 }
737
738 static void
tag_unref(struct pf_tagset * ts,u_int16_t tag)739 tag_unref(struct pf_tagset *ts, u_int16_t tag)
740 {
741 struct pf_tagname *t;
742 uint16_t index;
743
744 PF_RULES_WASSERT();
745
746 index = tag2hashindex(ts, tag);
747 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
748 if (tag == t->tag) {
749 if (--t->ref == 0) {
750 TAILQ_REMOVE(&ts->taghash[index], t,
751 taghash_entries);
752 index = tagname2hashindex(ts, t->name);
753 TAILQ_REMOVE(&ts->namehash[index], t,
754 namehash_entries);
755 /* Bits are 0-based for BIT_SET() */
756 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
757 uma_zfree(V_pf_tag_z, t);
758 }
759 break;
760 }
761 }
762
763 static uint16_t
pf_tagname2tag(const char * tagname)764 pf_tagname2tag(const char *tagname)
765 {
766 return (tagname2tag(&V_pf_tags, tagname));
767 }
768
769 static int
pf_begin_eth(uint32_t * ticket,const char * anchor)770 pf_begin_eth(uint32_t *ticket, const char *anchor)
771 {
772 struct pf_keth_rule *rule, *tmp;
773 struct pf_keth_ruleset *rs;
774
775 PF_RULES_WASSERT();
776
777 rs = pf_find_or_create_keth_ruleset(anchor);
778 if (rs == NULL)
779 return (EINVAL);
780
781 /* Purge old inactive rules. */
782 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
783 tmp) {
784 TAILQ_REMOVE(rs->inactive.rules, rule,
785 entries);
786 pf_free_eth_rule(rule);
787 }
788
789 *ticket = ++rs->inactive.ticket;
790 rs->inactive.open = 1;
791
792 return (0);
793 }
794
795 static int
pf_rollback_eth(uint32_t ticket,const char * anchor)796 pf_rollback_eth(uint32_t ticket, const char *anchor)
797 {
798 struct pf_keth_rule *rule, *tmp;
799 struct pf_keth_ruleset *rs;
800
801 PF_RULES_WASSERT();
802
803 rs = pf_find_keth_ruleset(anchor);
804 if (rs == NULL)
805 return (EINVAL);
806
807 if (!rs->inactive.open ||
808 ticket != rs->inactive.ticket)
809 return (0);
810
811 /* Purge old inactive rules. */
812 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
813 tmp) {
814 TAILQ_REMOVE(rs->inactive.rules, rule, entries);
815 pf_free_eth_rule(rule);
816 }
817
818 rs->inactive.open = 0;
819
820 pf_remove_if_empty_keth_ruleset(rs);
821
822 return (0);
823 }
824
825 #define PF_SET_SKIP_STEPS(i) \
826 do { \
827 while (head[i] != cur) { \
828 head[i]->skip[i].ptr = cur; \
829 head[i] = TAILQ_NEXT(head[i], entries); \
830 } \
831 } while (0)
832
833 static void
pf_eth_calc_skip_steps(struct pf_keth_ruleq * rules)834 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
835 {
836 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
837 int i;
838
839 cur = TAILQ_FIRST(rules);
840 prev = cur;
841 for (i = 0; i < PFE_SKIP_COUNT; ++i)
842 head[i] = cur;
843 while (cur != NULL) {
844 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
845 PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
846 if (cur->direction != prev->direction)
847 PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
848 if (cur->proto != prev->proto)
849 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
850 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
851 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
852 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
853 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
854 if (cur->ipsrc.neg != prev->ipsrc.neg ||
855 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
856 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
857 if (cur->ipdst.neg != prev->ipdst.neg ||
858 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
859 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
860
861 prev = cur;
862 cur = TAILQ_NEXT(cur, entries);
863 }
864 for (i = 0; i < PFE_SKIP_COUNT; ++i)
865 PF_SET_SKIP_STEPS(i);
866 }
867
868 static int
pf_commit_eth(uint32_t ticket,const char * anchor)869 pf_commit_eth(uint32_t ticket, const char *anchor)
870 {
871 struct pf_keth_ruleq *rules;
872 struct pf_keth_ruleset *rs;
873
874 rs = pf_find_keth_ruleset(anchor);
875 if (rs == NULL) {
876 return (EINVAL);
877 }
878
879 if (!rs->inactive.open ||
880 ticket != rs->inactive.ticket)
881 return (EBUSY);
882
883 PF_RULES_WASSERT();
884
885 pf_eth_calc_skip_steps(rs->inactive.rules);
886
887 rules = rs->active.rules;
888 atomic_store_ptr(&rs->active.rules, rs->inactive.rules);
889 rs->inactive.rules = rules;
890 rs->inactive.ticket = rs->active.ticket;
891
892 return (pf_rollback_eth(rs->inactive.ticket,
893 rs->anchor ? rs->anchor->path : ""));
894 }
895
896 #ifdef ALTQ
897 static uint16_t
pf_qname2qid(const char * qname)898 pf_qname2qid(const char *qname)
899 {
900 return (tagname2tag(&V_pf_qids, qname));
901 }
902
903 static void
pf_qid_unref(uint16_t qid)904 pf_qid_unref(uint16_t qid)
905 {
906 tag_unref(&V_pf_qids, qid);
907 }
908
909 static int
pf_begin_altq(u_int32_t * ticket)910 pf_begin_altq(u_int32_t *ticket)
911 {
912 struct pf_altq *altq, *tmp;
913 int error = 0;
914
915 PF_RULES_WASSERT();
916
917 /* Purge the old altq lists */
918 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
919 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
920 /* detach and destroy the discipline */
921 error = altq_remove(altq);
922 }
923 free(altq, M_PFALTQ);
924 }
925 TAILQ_INIT(V_pf_altq_ifs_inactive);
926 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
927 pf_qid_unref(altq->qid);
928 free(altq, M_PFALTQ);
929 }
930 TAILQ_INIT(V_pf_altqs_inactive);
931 if (error)
932 return (error);
933 *ticket = ++V_ticket_altqs_inactive;
934 V_altqs_inactive_open = 1;
935 return (0);
936 }
937
938 static int
pf_rollback_altq(u_int32_t ticket)939 pf_rollback_altq(u_int32_t ticket)
940 {
941 struct pf_altq *altq, *tmp;
942 int error = 0;
943
944 PF_RULES_WASSERT();
945
946 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
947 return (0);
948 /* Purge the old altq lists */
949 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
950 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
951 /* detach and destroy the discipline */
952 error = altq_remove(altq);
953 }
954 free(altq, M_PFALTQ);
955 }
956 TAILQ_INIT(V_pf_altq_ifs_inactive);
957 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
958 pf_qid_unref(altq->qid);
959 free(altq, M_PFALTQ);
960 }
961 TAILQ_INIT(V_pf_altqs_inactive);
962 V_altqs_inactive_open = 0;
963 return (error);
964 }
965
966 static int
pf_commit_altq(u_int32_t ticket)967 pf_commit_altq(u_int32_t ticket)
968 {
969 struct pf_altqqueue *old_altqs, *old_altq_ifs;
970 struct pf_altq *altq, *tmp;
971 int err, error = 0;
972
973 PF_RULES_WASSERT();
974
975 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
976 return (EBUSY);
977
978 /* swap altqs, keep the old. */
979 old_altqs = V_pf_altqs_active;
980 old_altq_ifs = V_pf_altq_ifs_active;
981 V_pf_altqs_active = V_pf_altqs_inactive;
982 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
983 V_pf_altqs_inactive = old_altqs;
984 V_pf_altq_ifs_inactive = old_altq_ifs;
985 V_ticket_altqs_active = V_ticket_altqs_inactive;
986
987 /* Attach new disciplines */
988 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
989 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
990 /* attach the discipline */
991 error = altq_pfattach(altq);
992 if (error == 0 && V_pf_altq_running)
993 error = pf_enable_altq(altq);
994 if (error != 0)
995 return (error);
996 }
997 }
998
999 /* Purge the old altq lists */
1000 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1001 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1002 /* detach and destroy the discipline */
1003 if (V_pf_altq_running)
1004 error = pf_disable_altq(altq);
1005 err = altq_pfdetach(altq);
1006 if (err != 0 && error == 0)
1007 error = err;
1008 err = altq_remove(altq);
1009 if (err != 0 && error == 0)
1010 error = err;
1011 }
1012 free(altq, M_PFALTQ);
1013 }
1014 TAILQ_INIT(V_pf_altq_ifs_inactive);
1015 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1016 pf_qid_unref(altq->qid);
1017 free(altq, M_PFALTQ);
1018 }
1019 TAILQ_INIT(V_pf_altqs_inactive);
1020
1021 V_altqs_inactive_open = 0;
1022 return (error);
1023 }
1024
1025 static int
pf_enable_altq(struct pf_altq * altq)1026 pf_enable_altq(struct pf_altq *altq)
1027 {
1028 struct ifnet *ifp;
1029 struct tb_profile tb;
1030 int error = 0;
1031
1032 if ((ifp = ifunit(altq->ifname)) == NULL)
1033 return (EINVAL);
1034
1035 if (ifp->if_snd.altq_type != ALTQT_NONE)
1036 error = altq_enable(&ifp->if_snd);
1037
1038 /* set tokenbucket regulator */
1039 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1040 tb.rate = altq->ifbandwidth;
1041 tb.depth = altq->tbrsize;
1042 error = tbr_set(&ifp->if_snd, &tb);
1043 }
1044
1045 return (error);
1046 }
1047
1048 static int
pf_disable_altq(struct pf_altq * altq)1049 pf_disable_altq(struct pf_altq *altq)
1050 {
1051 struct ifnet *ifp;
1052 struct tb_profile tb;
1053 int error;
1054
1055 if ((ifp = ifunit(altq->ifname)) == NULL)
1056 return (EINVAL);
1057
1058 /*
1059 * when the discipline is no longer referenced, it was overridden
1060 * by a new one. if so, just return.
1061 */
1062 if (altq->altq_disc != ifp->if_snd.altq_disc)
1063 return (0);
1064
1065 error = altq_disable(&ifp->if_snd);
1066
1067 if (error == 0) {
1068 /* clear tokenbucket regulator */
1069 tb.rate = 0;
1070 error = tbr_set(&ifp->if_snd, &tb);
1071 }
1072
1073 return (error);
1074 }
1075
1076 static int
pf_altq_ifnet_event_add(struct ifnet * ifp,int remove,u_int32_t ticket,struct pf_altq * altq)1077 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1078 struct pf_altq *altq)
1079 {
1080 struct ifnet *ifp1;
1081 int error = 0;
1082
1083 /* Deactivate the interface in question */
1084 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1085 if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1086 (remove && ifp1 == ifp)) {
1087 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1088 } else {
1089 error = altq_add(ifp1, altq);
1090
1091 if (ticket != V_ticket_altqs_inactive)
1092 error = EBUSY;
1093
1094 if (error)
1095 free(altq, M_PFALTQ);
1096 }
1097
1098 return (error);
1099 }
1100
1101 void
pf_altq_ifnet_event(struct ifnet * ifp,int remove)1102 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1103 {
1104 struct pf_altq *a1, *a2, *a3;
1105 u_int32_t ticket;
1106 int error = 0;
1107
1108 /*
1109 * No need to re-evaluate the configuration for events on interfaces
1110 * that do not support ALTQ, as it's not possible for such
1111 * interfaces to be part of the configuration.
1112 */
1113 if (!ALTQ_IS_READY(&ifp->if_snd))
1114 return;
1115
1116 /* Interrupt userland queue modifications */
1117 if (V_altqs_inactive_open)
1118 pf_rollback_altq(V_ticket_altqs_inactive);
1119
1120 /* Start new altq ruleset */
1121 if (pf_begin_altq(&ticket))
1122 return;
1123
1124 /* Copy the current active set */
1125 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1126 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1127 if (a2 == NULL) {
1128 error = ENOMEM;
1129 break;
1130 }
1131 bcopy(a1, a2, sizeof(struct pf_altq));
1132
1133 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1134 if (error)
1135 break;
1136
1137 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1138 }
1139 if (error)
1140 goto out;
1141 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1142 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1143 if (a2 == NULL) {
1144 error = ENOMEM;
1145 break;
1146 }
1147 bcopy(a1, a2, sizeof(struct pf_altq));
1148
1149 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1150 error = EBUSY;
1151 free(a2, M_PFALTQ);
1152 break;
1153 }
1154 a2->altq_disc = NULL;
1155 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1156 if (strncmp(a3->ifname, a2->ifname,
1157 IFNAMSIZ) == 0) {
1158 a2->altq_disc = a3->altq_disc;
1159 break;
1160 }
1161 }
1162 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1163 if (error)
1164 break;
1165
1166 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1167 }
1168
1169 out:
1170 if (error != 0)
1171 pf_rollback_altq(ticket);
1172 else
1173 pf_commit_altq(ticket);
1174 }
1175 #endif /* ALTQ */
1176
1177 static struct pf_krule_global *
pf_rule_tree_alloc(int flags)1178 pf_rule_tree_alloc(int flags)
1179 {
1180 struct pf_krule_global *tree;
1181
1182 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1183 if (tree == NULL)
1184 return (NULL);
1185 RB_INIT(tree);
1186 return (tree);
1187 }
1188
1189 static void
pf_rule_tree_free(struct pf_krule_global * tree)1190 pf_rule_tree_free(struct pf_krule_global *tree)
1191 {
1192
1193 free(tree, M_TEMP);
1194 }
1195
1196 static int
pf_begin_rules(u_int32_t * ticket,int rs_num,const char * anchor)1197 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1198 {
1199 struct pf_krule_global *tree;
1200 struct pf_kruleset *rs;
1201 struct pf_krule *rule;
1202
1203 PF_RULES_WASSERT();
1204
1205 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1206 return (EINVAL);
1207 tree = pf_rule_tree_alloc(M_NOWAIT);
1208 if (tree == NULL)
1209 return (ENOMEM);
1210 rs = pf_find_or_create_kruleset(anchor);
1211 if (rs == NULL) {
1212 free(tree, M_TEMP);
1213 return (EINVAL);
1214 }
1215 pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1216 rs->rules[rs_num].inactive.tree = tree;
1217
1218 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1219 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1220 rs->rules[rs_num].inactive.rcount--;
1221 }
1222 *ticket = ++rs->rules[rs_num].inactive.ticket;
1223 rs->rules[rs_num].inactive.open = 1;
1224 return (0);
1225 }
1226
1227 static int
pf_rollback_rules(u_int32_t ticket,int rs_num,char * anchor)1228 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1229 {
1230 struct pf_kruleset *rs;
1231 struct pf_krule *rule;
1232
1233 PF_RULES_WASSERT();
1234
1235 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1236 return (EINVAL);
1237 rs = pf_find_kruleset(anchor);
1238 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1239 rs->rules[rs_num].inactive.ticket != ticket)
1240 return (0);
1241 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1242 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1243 rs->rules[rs_num].inactive.rcount--;
1244 }
1245 rs->rules[rs_num].inactive.open = 0;
1246 return (0);
1247 }
1248
1249 #define PF_MD5_UPD(st, elm) \
1250 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1251
1252 #define PF_MD5_UPD_STR(st, elm) \
1253 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1254
1255 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \
1256 (stor) = htonl((st)->elm); \
1257 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1258 } while (0)
1259
1260 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \
1261 (stor) = htons((st)->elm); \
1262 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1263 } while (0)
1264
1265 static void
pf_hash_rule_addr(MD5_CTX * ctx,struct pf_rule_addr * pfr)1266 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1267 {
1268 PF_MD5_UPD(pfr, addr.type);
1269 switch (pfr->addr.type) {
1270 case PF_ADDR_DYNIFTL:
1271 PF_MD5_UPD(pfr, addr.v.ifname);
1272 PF_MD5_UPD(pfr, addr.iflags);
1273 break;
1274 case PF_ADDR_TABLE:
1275 if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX,
1276 strlen(PF_OPTIMIZER_TABLE_PFX)))
1277 PF_MD5_UPD(pfr, addr.v.tblname);
1278 break;
1279 case PF_ADDR_ADDRMASK:
1280 /* XXX ignore af? */
1281 PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1282 PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1283 break;
1284 }
1285
1286 PF_MD5_UPD(pfr, port[0]);
1287 PF_MD5_UPD(pfr, port[1]);
1288 PF_MD5_UPD(pfr, neg);
1289 PF_MD5_UPD(pfr, port_op);
1290 }
1291
1292 static void
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)1293 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1294 {
1295 u_int16_t x;
1296 u_int32_t y;
1297
1298 pf_hash_rule_addr(ctx, &rule->src);
1299 pf_hash_rule_addr(ctx, &rule->dst);
1300 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1301 PF_MD5_UPD_STR(rule, label[i]);
1302 PF_MD5_UPD_STR(rule, ifname);
1303 PF_MD5_UPD_STR(rule, rcv_ifname);
1304 PF_MD5_UPD_STR(rule, match_tagname);
1305 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1306 PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1307 PF_MD5_UPD_HTONL(rule, prob, y);
1308 PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1309 PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1310 PF_MD5_UPD(rule, uid.op);
1311 PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1312 PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1313 PF_MD5_UPD(rule, gid.op);
1314 PF_MD5_UPD_HTONL(rule, rule_flag, y);
1315 PF_MD5_UPD(rule, action);
1316 PF_MD5_UPD(rule, direction);
1317 PF_MD5_UPD(rule, af);
1318 PF_MD5_UPD(rule, quick);
1319 PF_MD5_UPD(rule, ifnot);
1320 PF_MD5_UPD(rule, rcvifnot);
1321 PF_MD5_UPD(rule, match_tag_not);
1322 PF_MD5_UPD(rule, natpass);
1323 PF_MD5_UPD(rule, keep_state);
1324 PF_MD5_UPD(rule, proto);
1325 PF_MD5_UPD(rule, type);
1326 PF_MD5_UPD(rule, code);
1327 PF_MD5_UPD(rule, flags);
1328 PF_MD5_UPD(rule, flagset);
1329 PF_MD5_UPD(rule, allow_opts);
1330 PF_MD5_UPD(rule, rt);
1331 PF_MD5_UPD(rule, tos);
1332 PF_MD5_UPD(rule, scrub_flags);
1333 PF_MD5_UPD(rule, min_ttl);
1334 PF_MD5_UPD(rule, set_tos);
1335 if (rule->anchor != NULL)
1336 PF_MD5_UPD_STR(rule, anchor->path);
1337 }
1338
1339 static void
pf_hash_rule(struct pf_krule * rule)1340 pf_hash_rule(struct pf_krule *rule)
1341 {
1342 MD5_CTX ctx;
1343
1344 MD5Init(&ctx);
1345 pf_hash_rule_rolling(&ctx, rule);
1346 MD5Final(rule->md5sum, &ctx);
1347 }
1348
1349 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1350 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1351 {
1352
1353 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1354 }
1355
1356 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1357 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1358 {
1359 struct pf_kruleset *rs;
1360 struct pf_krule *rule, *old_rule;
1361 struct pf_krulequeue *old_rules;
1362 struct pf_krule_global *old_tree;
1363 int error;
1364 u_int32_t old_rcount;
1365
1366 PF_RULES_WASSERT();
1367
1368 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1369 return (EINVAL);
1370 rs = pf_find_kruleset(anchor);
1371 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1372 ticket != rs->rules[rs_num].inactive.ticket)
1373 return (EBUSY);
1374
1375 /* Calculate checksum for the main ruleset */
1376 if (rs == &pf_main_ruleset) {
1377 error = pf_setup_pfsync_matching(rs);
1378 if (error != 0)
1379 return (error);
1380 }
1381
1382 /* Swap rules, keep the old. */
1383 old_rules = rs->rules[rs_num].active.ptr;
1384 old_rcount = rs->rules[rs_num].active.rcount;
1385 old_tree = rs->rules[rs_num].active.tree;
1386
1387 rs->rules[rs_num].active.ptr =
1388 rs->rules[rs_num].inactive.ptr;
1389 rs->rules[rs_num].active.tree =
1390 rs->rules[rs_num].inactive.tree;
1391 rs->rules[rs_num].active.rcount =
1392 rs->rules[rs_num].inactive.rcount;
1393
1394 /* Attempt to preserve counter information. */
1395 if (V_pf_status.keep_counters && old_tree != NULL) {
1396 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1397 entries) {
1398 old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1399 if (old_rule == NULL) {
1400 continue;
1401 }
1402 pf_counter_u64_critical_enter();
1403 pf_counter_u64_rollup_protected(&rule->evaluations,
1404 pf_counter_u64_fetch(&old_rule->evaluations));
1405 pf_counter_u64_rollup_protected(&rule->packets[0],
1406 pf_counter_u64_fetch(&old_rule->packets[0]));
1407 pf_counter_u64_rollup_protected(&rule->packets[1],
1408 pf_counter_u64_fetch(&old_rule->packets[1]));
1409 pf_counter_u64_rollup_protected(&rule->bytes[0],
1410 pf_counter_u64_fetch(&old_rule->bytes[0]));
1411 pf_counter_u64_rollup_protected(&rule->bytes[1],
1412 pf_counter_u64_fetch(&old_rule->bytes[1]));
1413 pf_counter_u64_critical_exit();
1414 }
1415 }
1416
1417 rs->rules[rs_num].inactive.ptr = old_rules;
1418 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1419 rs->rules[rs_num].inactive.rcount = old_rcount;
1420
1421 rs->rules[rs_num].active.ticket =
1422 rs->rules[rs_num].inactive.ticket;
1423 pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1424
1425 /* Purge the old rule list. */
1426 PF_UNLNKDRULES_LOCK();
1427 while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1428 pf_unlink_rule_locked(old_rules, rule);
1429 PF_UNLNKDRULES_UNLOCK();
1430 rs->rules[rs_num].inactive.rcount = 0;
1431 rs->rules[rs_num].inactive.open = 0;
1432 pf_remove_if_empty_kruleset(rs);
1433 free(old_tree, M_TEMP);
1434
1435 return (0);
1436 }
1437
1438 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1439 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1440 {
1441 MD5_CTX ctx;
1442 struct pf_krule *rule;
1443 int rs_cnt;
1444 u_int8_t digest[PF_MD5_DIGEST_LENGTH];
1445
1446 MD5Init(&ctx);
1447 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1448 /* XXX PF_RULESET_SCRUB as well? */
1449 if (rs_cnt == PF_RULESET_SCRUB)
1450 continue;
1451
1452 if (rs->rules[rs_cnt].inactive.rcount) {
1453 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1454 entries) {
1455 pf_hash_rule_rolling(&ctx, rule);
1456 }
1457 }
1458 }
1459
1460 MD5Final(digest, &ctx);
1461 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1462 return (0);
1463 }
1464
1465 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1466 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1467 {
1468 int error = 0;
1469
1470 switch (addr->type) {
1471 case PF_ADDR_TABLE:
1472 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1473 if (addr->p.tbl == NULL)
1474 error = ENOMEM;
1475 break;
1476 default:
1477 error = EINVAL;
1478 }
1479
1480 return (error);
1481 }
1482
1483 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1484 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1485 sa_family_t af)
1486 {
1487 int error = 0;
1488
1489 switch (addr->type) {
1490 case PF_ADDR_TABLE:
1491 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1492 if (addr->p.tbl == NULL)
1493 error = ENOMEM;
1494 break;
1495 case PF_ADDR_DYNIFTL:
1496 error = pfi_dynaddr_setup(addr, af);
1497 break;
1498 }
1499
1500 return (error);
1501 }
1502
1503 void
pf_addr_copyout(struct pf_addr_wrap * addr)1504 pf_addr_copyout(struct pf_addr_wrap *addr)
1505 {
1506
1507 switch (addr->type) {
1508 case PF_ADDR_DYNIFTL:
1509 pfi_dynaddr_copyout(addr);
1510 break;
1511 case PF_ADDR_TABLE:
1512 pf_tbladdr_copyout(addr);
1513 break;
1514 }
1515 }
1516
1517 static void
pf_src_node_copy(const struct pf_ksrc_node * in,struct pf_src_node * out)1518 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1519 {
1520 int secs = time_uptime;
1521
1522 bzero(out, sizeof(struct pf_src_node));
1523
1524 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1525 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1526
1527 if (in->rule != NULL)
1528 out->rule.nr = in->rule->nr;
1529
1530 for (int i = 0; i < 2; i++) {
1531 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1532 out->packets[i] = counter_u64_fetch(in->packets[i]);
1533 }
1534
1535 out->states = in->states;
1536 out->conn = in->conn;
1537 out->af = in->af;
1538 out->ruletype = in->ruletype;
1539
1540 out->creation = secs - in->creation;
1541 if (out->expire > secs)
1542 out->expire -= secs;
1543 else
1544 out->expire = 0;
1545
1546 /* Adjust the connection rate estimate. */
1547 out->conn_rate.limit = in->conn_rate.limit;
1548 out->conn_rate.seconds = in->conn_rate.seconds;
1549 /* If there's no limit there's no counter_rate. */
1550 if (in->conn_rate.cr != NULL)
1551 out->conn_rate.count = counter_rate_get(in->conn_rate.cr);
1552 }
1553
1554 #ifdef ALTQ
1555 /*
1556 * Handle export of struct pf_kaltq to user binaries that may be using any
1557 * version of struct pf_altq.
1558 */
1559 static int
pf_export_kaltq(struct pf_altq * q,struct pfioc_altq_v1 * pa,size_t ioc_size)1560 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1561 {
1562 u_int32_t version;
1563
1564 if (ioc_size == sizeof(struct pfioc_altq_v0))
1565 version = 0;
1566 else
1567 version = pa->version;
1568
1569 if (version > PFIOC_ALTQ_VERSION)
1570 return (EINVAL);
1571
1572 #define ASSIGN(x) exported_q->x = q->x
1573 #define COPY(x) \
1574 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1575 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1576 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1577
1578 switch (version) {
1579 case 0: {
1580 struct pf_altq_v0 *exported_q =
1581 &((struct pfioc_altq_v0 *)pa)->altq;
1582
1583 COPY(ifname);
1584
1585 ASSIGN(scheduler);
1586 ASSIGN(tbrsize);
1587 exported_q->tbrsize = SATU16(q->tbrsize);
1588 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1589
1590 COPY(qname);
1591 COPY(parent);
1592 ASSIGN(parent_qid);
1593 exported_q->bandwidth = SATU32(q->bandwidth);
1594 ASSIGN(priority);
1595 ASSIGN(local_flags);
1596
1597 ASSIGN(qlimit);
1598 ASSIGN(flags);
1599
1600 if (q->scheduler == ALTQT_HFSC) {
1601 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1602 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1603 SATU32(q->pq_u.hfsc_opts.x)
1604
1605 ASSIGN_OPT_SATU32(rtsc_m1);
1606 ASSIGN_OPT(rtsc_d);
1607 ASSIGN_OPT_SATU32(rtsc_m2);
1608
1609 ASSIGN_OPT_SATU32(lssc_m1);
1610 ASSIGN_OPT(lssc_d);
1611 ASSIGN_OPT_SATU32(lssc_m2);
1612
1613 ASSIGN_OPT_SATU32(ulsc_m1);
1614 ASSIGN_OPT(ulsc_d);
1615 ASSIGN_OPT_SATU32(ulsc_m2);
1616
1617 ASSIGN_OPT(flags);
1618
1619 #undef ASSIGN_OPT
1620 #undef ASSIGN_OPT_SATU32
1621 } else
1622 COPY(pq_u);
1623
1624 ASSIGN(qid);
1625 break;
1626 }
1627 case 1: {
1628 struct pf_altq_v1 *exported_q =
1629 &((struct pfioc_altq_v1 *)pa)->altq;
1630
1631 COPY(ifname);
1632
1633 ASSIGN(scheduler);
1634 ASSIGN(tbrsize);
1635 ASSIGN(ifbandwidth);
1636
1637 COPY(qname);
1638 COPY(parent);
1639 ASSIGN(parent_qid);
1640 ASSIGN(bandwidth);
1641 ASSIGN(priority);
1642 ASSIGN(local_flags);
1643
1644 ASSIGN(qlimit);
1645 ASSIGN(flags);
1646 COPY(pq_u);
1647
1648 ASSIGN(qid);
1649 break;
1650 }
1651 default:
1652 panic("%s: unhandled struct pfioc_altq version", __func__);
1653 break;
1654 }
1655
1656 #undef ASSIGN
1657 #undef COPY
1658 #undef SATU16
1659 #undef SATU32
1660
1661 return (0);
1662 }
1663
1664 /*
1665 * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1666 * that may be using any version of it.
1667 */
1668 static int
pf_import_kaltq(struct pfioc_altq_v1 * pa,struct pf_altq * q,size_t ioc_size)1669 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1670 {
1671 u_int32_t version;
1672
1673 if (ioc_size == sizeof(struct pfioc_altq_v0))
1674 version = 0;
1675 else
1676 version = pa->version;
1677
1678 if (version > PFIOC_ALTQ_VERSION)
1679 return (EINVAL);
1680
1681 #define ASSIGN(x) q->x = imported_q->x
1682 #define COPY(x) \
1683 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1684
1685 switch (version) {
1686 case 0: {
1687 struct pf_altq_v0 *imported_q =
1688 &((struct pfioc_altq_v0 *)pa)->altq;
1689
1690 COPY(ifname);
1691
1692 ASSIGN(scheduler);
1693 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1694 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1695
1696 COPY(qname);
1697 COPY(parent);
1698 ASSIGN(parent_qid);
1699 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1700 ASSIGN(priority);
1701 ASSIGN(local_flags);
1702
1703 ASSIGN(qlimit);
1704 ASSIGN(flags);
1705
1706 if (imported_q->scheduler == ALTQT_HFSC) {
1707 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1708
1709 /*
1710 * The m1 and m2 parameters are being copied from
1711 * 32-bit to 64-bit.
1712 */
1713 ASSIGN_OPT(rtsc_m1);
1714 ASSIGN_OPT(rtsc_d);
1715 ASSIGN_OPT(rtsc_m2);
1716
1717 ASSIGN_OPT(lssc_m1);
1718 ASSIGN_OPT(lssc_d);
1719 ASSIGN_OPT(lssc_m2);
1720
1721 ASSIGN_OPT(ulsc_m1);
1722 ASSIGN_OPT(ulsc_d);
1723 ASSIGN_OPT(ulsc_m2);
1724
1725 ASSIGN_OPT(flags);
1726
1727 #undef ASSIGN_OPT
1728 } else
1729 COPY(pq_u);
1730
1731 ASSIGN(qid);
1732 break;
1733 }
1734 case 1: {
1735 struct pf_altq_v1 *imported_q =
1736 &((struct pfioc_altq_v1 *)pa)->altq;
1737
1738 COPY(ifname);
1739
1740 ASSIGN(scheduler);
1741 ASSIGN(tbrsize);
1742 ASSIGN(ifbandwidth);
1743
1744 COPY(qname);
1745 COPY(parent);
1746 ASSIGN(parent_qid);
1747 ASSIGN(bandwidth);
1748 ASSIGN(priority);
1749 ASSIGN(local_flags);
1750
1751 ASSIGN(qlimit);
1752 ASSIGN(flags);
1753 COPY(pq_u);
1754
1755 ASSIGN(qid);
1756 break;
1757 }
1758 default:
1759 panic("%s: unhandled struct pfioc_altq version", __func__);
1760 break;
1761 }
1762
1763 #undef ASSIGN
1764 #undef COPY
1765
1766 return (0);
1767 }
1768
1769 static struct pf_altq *
pf_altq_get_nth_active(u_int32_t n)1770 pf_altq_get_nth_active(u_int32_t n)
1771 {
1772 struct pf_altq *altq;
1773 u_int32_t nr;
1774
1775 nr = 0;
1776 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1777 if (nr == n)
1778 return (altq);
1779 nr++;
1780 }
1781
1782 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1783 if (nr == n)
1784 return (altq);
1785 nr++;
1786 }
1787
1788 return (NULL);
1789 }
1790 #endif /* ALTQ */
1791
1792 struct pf_krule *
pf_krule_alloc(void)1793 pf_krule_alloc(void)
1794 {
1795 struct pf_krule *rule;
1796
1797 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1798 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF);
1799 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF);
1800 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF);
1801 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1802 M_WAITOK | M_ZERO);
1803 return (rule);
1804 }
1805
1806 void
pf_krule_free(struct pf_krule * rule)1807 pf_krule_free(struct pf_krule *rule)
1808 {
1809 #ifdef PF_WANT_32_TO_64_COUNTER
1810 bool wowned;
1811 #endif
1812
1813 if (rule == NULL)
1814 return;
1815
1816 #ifdef PF_WANT_32_TO_64_COUNTER
1817 if (rule->allrulelinked) {
1818 wowned = PF_RULES_WOWNED();
1819 if (!wowned)
1820 PF_RULES_WLOCK();
1821 LIST_REMOVE(rule, allrulelist);
1822 V_pf_allrulecount--;
1823 if (!wowned)
1824 PF_RULES_WUNLOCK();
1825 }
1826 #endif
1827
1828 pf_counter_u64_deinit(&rule->evaluations);
1829 for (int i = 0; i < 2; i++) {
1830 pf_counter_u64_deinit(&rule->packets[i]);
1831 pf_counter_u64_deinit(&rule->bytes[i]);
1832 }
1833 counter_u64_free(rule->states_cur);
1834 counter_u64_free(rule->states_tot);
1835 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
1836 counter_u64_free(rule->src_nodes[sn_type]);
1837 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1838
1839 mtx_destroy(&rule->nat.mtx);
1840 mtx_destroy(&rule->rdr.mtx);
1841 mtx_destroy(&rule->route.mtx);
1842 free(rule, M_PFRULE);
1843 }
1844
1845 void
pf_krule_clear_counters(struct pf_krule * rule)1846 pf_krule_clear_counters(struct pf_krule *rule)
1847 {
1848 pf_counter_u64_zero(&rule->evaluations);
1849 for (int i = 0; i < 2; i++) {
1850 pf_counter_u64_zero(&rule->packets[i]);
1851 pf_counter_u64_zero(&rule->bytes[i]);
1852 }
1853 counter_u64_zero(rule->states_tot);
1854 }
1855
1856 static void
pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr * kpool,struct pf_pooladdr * pool)1857 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1858 struct pf_pooladdr *pool)
1859 {
1860
1861 bzero(pool, sizeof(*pool));
1862 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1863 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1864 }
1865
1866 static int
pf_pooladdr_to_kpooladdr(const struct pf_pooladdr * pool,struct pf_kpooladdr * kpool)1867 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1868 struct pf_kpooladdr *kpool)
1869 {
1870 int ret;
1871
1872 bzero(kpool, sizeof(*kpool));
1873 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1874 ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1875 sizeof(kpool->ifname));
1876 return (ret);
1877 }
1878
1879 static void
pf_pool_to_kpool(const struct pf_pool * pool,struct pf_kpool * kpool)1880 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1881 {
1882 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1883 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1884
1885 bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1886 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1887
1888 kpool->tblidx = pool->tblidx;
1889 kpool->proxy_port[0] = pool->proxy_port[0];
1890 kpool->proxy_port[1] = pool->proxy_port[1];
1891 kpool->opts = pool->opts;
1892 }
1893
1894 static int
pf_rule_to_krule(const struct pf_rule * rule,struct pf_krule * krule)1895 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1896 {
1897 int ret;
1898
1899 #ifndef INET
1900 if (rule->af == AF_INET) {
1901 return (EAFNOSUPPORT);
1902 }
1903 #endif /* INET */
1904 #ifndef INET6
1905 if (rule->af == AF_INET6) {
1906 return (EAFNOSUPPORT);
1907 }
1908 #endif /* INET6 */
1909
1910 ret = pf_check_rule_addr(&rule->src);
1911 if (ret != 0)
1912 return (ret);
1913 ret = pf_check_rule_addr(&rule->dst);
1914 if (ret != 0)
1915 return (ret);
1916
1917 bcopy(&rule->src, &krule->src, sizeof(rule->src));
1918 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1919
1920 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1921 if (ret != 0)
1922 return (ret);
1923 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1924 if (ret != 0)
1925 return (ret);
1926 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1927 if (ret != 0)
1928 return (ret);
1929 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1930 if (ret != 0)
1931 return (ret);
1932 ret = pf_user_strcpy(krule->tagname, rule->tagname,
1933 sizeof(rule->tagname));
1934 if (ret != 0)
1935 return (ret);
1936 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1937 sizeof(rule->match_tagname));
1938 if (ret != 0)
1939 return (ret);
1940 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1941 sizeof(rule->overload_tblname));
1942 if (ret != 0)
1943 return (ret);
1944
1945 pf_pool_to_kpool(&rule->rpool, &krule->rdr);
1946
1947 /* Don't allow userspace to set evaluations, packets or bytes. */
1948 /* kif, anchor, overload_tbl are not copied over. */
1949
1950 krule->os_fingerprint = rule->os_fingerprint;
1951
1952 krule->rtableid = rule->rtableid;
1953 /* pf_rule->timeout is smaller than pf_krule->timeout */
1954 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1955 krule->max_states = rule->max_states;
1956 krule->max_src_nodes = rule->max_src_nodes;
1957 krule->max_src_states = rule->max_src_states;
1958 krule->max_src_conn = rule->max_src_conn;
1959 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1960 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1961 krule->qid = rule->qid;
1962 krule->pqid = rule->pqid;
1963 krule->nr = rule->nr;
1964 krule->prob = rule->prob;
1965 krule->cuid = rule->cuid;
1966 krule->cpid = rule->cpid;
1967
1968 krule->return_icmp = rule->return_icmp;
1969 krule->return_icmp6 = rule->return_icmp6;
1970 krule->max_mss = rule->max_mss;
1971 krule->tag = rule->tag;
1972 krule->match_tag = rule->match_tag;
1973 krule->scrub_flags = rule->scrub_flags;
1974
1975 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1976 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1977
1978 krule->rule_flag = rule->rule_flag;
1979 krule->action = rule->action;
1980 krule->direction = rule->direction;
1981 krule->log = rule->log;
1982 krule->logif = rule->logif;
1983 krule->quick = rule->quick;
1984 krule->ifnot = rule->ifnot;
1985 krule->match_tag_not = rule->match_tag_not;
1986 krule->natpass = rule->natpass;
1987
1988 krule->keep_state = rule->keep_state;
1989 krule->af = rule->af;
1990 krule->proto = rule->proto;
1991 krule->type = rule->type;
1992 krule->code = rule->code;
1993 krule->flags = rule->flags;
1994 krule->flagset = rule->flagset;
1995 krule->min_ttl = rule->min_ttl;
1996 krule->allow_opts = rule->allow_opts;
1997 krule->rt = rule->rt;
1998 krule->return_ttl = rule->return_ttl;
1999 krule->tos = rule->tos;
2000 krule->set_tos = rule->set_tos;
2001
2002 krule->flush = rule->flush;
2003 krule->prio = rule->prio;
2004 krule->set_prio[0] = rule->set_prio[0];
2005 krule->set_prio[1] = rule->set_prio[1];
2006
2007 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2008
2009 return (0);
2010 }
2011
2012 int
pf_ioctl_getrules(struct pfioc_rule * pr)2013 pf_ioctl_getrules(struct pfioc_rule *pr)
2014 {
2015 struct pf_kruleset *ruleset;
2016 struct pf_krule *tail;
2017 int rs_num;
2018
2019 PF_RULES_WLOCK();
2020 ruleset = pf_find_kruleset(pr->anchor);
2021 if (ruleset == NULL) {
2022 PF_RULES_WUNLOCK();
2023 return (EINVAL);
2024 }
2025 rs_num = pf_get_ruleset_number(pr->rule.action);
2026 if (rs_num >= PF_RULESET_MAX) {
2027 PF_RULES_WUNLOCK();
2028 return (EINVAL);
2029 }
2030 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2031 pf_krulequeue);
2032 if (tail)
2033 pr->nr = tail->nr + 1;
2034 else
2035 pr->nr = 0;
2036 pr->ticket = ruleset->rules[rs_num].active.ticket;
2037 PF_RULES_WUNLOCK();
2038
2039 return (0);
2040 }
2041
2042 static int
pf_rule_checkaf(struct pf_krule * r)2043 pf_rule_checkaf(struct pf_krule *r)
2044 {
2045 switch (r->af) {
2046 case 0:
2047 if (r->rule_flag & PFRULE_AFTO)
2048 return (EPFNOSUPPORT);
2049 break;
2050 case AF_INET:
2051 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET6)
2052 return (EPFNOSUPPORT);
2053 break;
2054 #ifdef INET6
2055 case AF_INET6:
2056 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET)
2057 return (EPFNOSUPPORT);
2058 break;
2059 #endif /* INET6 */
2060 default:
2061 return (EPFNOSUPPORT);
2062 }
2063
2064 if ((r->rule_flag & PFRULE_AFTO) == 0 && r->naf != 0)
2065 return (EPFNOSUPPORT);
2066
2067 return (0);
2068 }
2069
2070 static int
pf_validate_range(uint8_t op,uint16_t port[2])2071 pf_validate_range(uint8_t op, uint16_t port[2])
2072 {
2073 uint16_t a = ntohs(port[0]);
2074 uint16_t b = ntohs(port[1]);
2075
2076 if ((op == PF_OP_RRG && a > b) || /* 34:12, i.e. none */
2077 (op == PF_OP_IRG && a >= b) || /* 34><12, i.e. none */
2078 (op == PF_OP_XRG && a > b)) /* 34<>22, i.e. all */
2079 return 1;
2080 return 0;
2081 }
2082
2083 int
pf_ioctl_addrule(struct pf_krule * rule,uint32_t ticket,uint32_t pool_ticket,const char * anchor,const char * anchor_call,uid_t uid,pid_t pid)2084 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2085 uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2086 uid_t uid, pid_t pid)
2087 {
2088 struct pf_kruleset *ruleset;
2089 struct pf_krule *tail;
2090 struct pf_kpooladdr *pa;
2091 struct pfi_kkif *kif = NULL, *rcv_kif = NULL;
2092 int rs_num;
2093 int error = 0;
2094
2095 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
2096 #define ERROUT_UNLOCKED(x) ERROUT_FUNCTION(errout_unlocked, x)
2097
2098 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE)
2099 ERROUT_UNLOCKED(EINVAL);
2100
2101 if ((error = pf_rule_checkaf(rule)))
2102 ERROUT_UNLOCKED(error);
2103 if (pf_validate_range(rule->src.port_op, rule->src.port))
2104 ERROUT_UNLOCKED(EINVAL);
2105 if (pf_validate_range(rule->dst.port_op, rule->dst.port))
2106 ERROUT_UNLOCKED(EINVAL);
2107
2108 if (rule->ifname[0])
2109 kif = pf_kkif_create(M_WAITOK);
2110 if (rule->rcv_ifname[0])
2111 rcv_kif = pf_kkif_create(M_WAITOK);
2112 pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2113 for (int i = 0; i < 2; i++) {
2114 pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2115 pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2116 }
2117 rule->states_cur = counter_u64_alloc(M_WAITOK);
2118 rule->states_tot = counter_u64_alloc(M_WAITOK);
2119 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
2120 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
2121 rule->cuid = uid;
2122 rule->cpid = pid;
2123 TAILQ_INIT(&rule->rdr.list);
2124 TAILQ_INIT(&rule->nat.list);
2125 TAILQ_INIT(&rule->route.list);
2126
2127 PF_CONFIG_LOCK();
2128 PF_RULES_WLOCK();
2129 #ifdef PF_WANT_32_TO_64_COUNTER
2130 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2131 MPASS(!rule->allrulelinked);
2132 rule->allrulelinked = true;
2133 V_pf_allrulecount++;
2134 #endif
2135 ruleset = pf_find_kruleset(anchor);
2136 if (ruleset == NULL)
2137 ERROUT(EINVAL);
2138 rs_num = pf_get_ruleset_number(rule->action);
2139 if (rs_num >= PF_RULESET_MAX)
2140 ERROUT(EINVAL);
2141 if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2142 DPFPRINTF(PF_DEBUG_MISC,
2143 "ticket: %d != [%d]%d", ticket, rs_num,
2144 ruleset->rules[rs_num].inactive.ticket);
2145 ERROUT(EBUSY);
2146 }
2147 if (pool_ticket != V_ticket_pabuf) {
2148 DPFPRINTF(PF_DEBUG_MISC,
2149 "pool_ticket: %d != %d", pool_ticket,
2150 V_ticket_pabuf);
2151 ERROUT(EBUSY);
2152 }
2153 /*
2154 * XXXMJG hack: there is no mechanism to ensure they started the
2155 * transaction. Ticket checked above may happen to match by accident,
2156 * even if nobody called DIOCXBEGIN, let alone this process.
2157 * Partially work around it by checking if the RB tree got allocated,
2158 * see pf_begin_rules.
2159 */
2160 if (ruleset->rules[rs_num].inactive.tree == NULL) {
2161 ERROUT(EINVAL);
2162 }
2163
2164 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2165 pf_krulequeue);
2166 if (tail)
2167 rule->nr = tail->nr + 1;
2168 else
2169 rule->nr = 0;
2170 if (rule->ifname[0]) {
2171 rule->kif = pfi_kkif_attach(kif, rule->ifname);
2172 kif = NULL;
2173 pfi_kkif_ref(rule->kif);
2174 } else
2175 rule->kif = NULL;
2176
2177 if (rule->rcv_ifname[0]) {
2178 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname);
2179 rcv_kif = NULL;
2180 pfi_kkif_ref(rule->rcv_kif);
2181 } else
2182 rule->rcv_kif = NULL;
2183
2184 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2185 ERROUT(EBUSY);
2186 #ifdef ALTQ
2187 /* set queue IDs */
2188 if (rule->qname[0] != 0) {
2189 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2190 ERROUT(EBUSY);
2191 else if (rule->pqname[0] != 0) {
2192 if ((rule->pqid =
2193 pf_qname2qid(rule->pqname)) == 0)
2194 ERROUT(EBUSY);
2195 } else
2196 rule->pqid = rule->qid;
2197 }
2198 #endif
2199 if (rule->tagname[0])
2200 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2201 ERROUT(EBUSY);
2202 if (rule->match_tagname[0])
2203 if ((rule->match_tag =
2204 pf_tagname2tag(rule->match_tagname)) == 0)
2205 ERROUT(EBUSY);
2206 if (rule->rt && !rule->direction)
2207 ERROUT(EINVAL);
2208 if (!rule->log)
2209 rule->logif = 0;
2210 if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit,
2211 rule->pktrate.seconds))
2212 ERROUT(ENOMEM);
2213 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2214 ERROUT(ENOMEM);
2215 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2216 ERROUT(ENOMEM);
2217 if (pf_kanchor_setup(rule, ruleset, anchor_call))
2218 ERROUT(EINVAL);
2219 if (rule->scrub_flags & PFSTATE_SETPRIO &&
2220 (rule->set_prio[0] > PF_PRIO_MAX ||
2221 rule->set_prio[1] > PF_PRIO_MAX))
2222 ERROUT(EINVAL);
2223 for (int i = 0; i < 3; i++) {
2224 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
2225 if (pa->addr.type == PF_ADDR_TABLE) {
2226 pa->addr.p.tbl = pfr_attach_table(ruleset,
2227 pa->addr.v.tblname);
2228 if (pa->addr.p.tbl == NULL)
2229 ERROUT(ENOMEM);
2230 }
2231 }
2232
2233 rule->overload_tbl = NULL;
2234 if (rule->overload_tblname[0]) {
2235 if ((rule->overload_tbl = pfr_attach_table(ruleset,
2236 rule->overload_tblname)) == NULL)
2237 ERROUT(EINVAL);
2238 else
2239 rule->overload_tbl->pfrkt_flags |=
2240 PFR_TFLAG_ACTIVE;
2241 }
2242
2243 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list);
2244
2245 /*
2246 * Old version of pfctl provide route redirection pools in single
2247 * common redirection pool rdr. New versions use rdr only for
2248 * rdr-to rules.
2249 */
2250 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) {
2251 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list);
2252 } else {
2253 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list);
2254 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list);
2255 }
2256
2257 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2258 (rule->action == PF_BINAT)) && rule->anchor == NULL &&
2259 TAILQ_FIRST(&rule->rdr.list) == NULL) {
2260 ERROUT(EINVAL);
2261 }
2262
2263 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) {
2264 ERROUT(EINVAL);
2265 }
2266
2267 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR ||
2268 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) {
2269 ERROUT(EINVAL);
2270 }
2271
2272 MPASS(error == 0);
2273
2274 rule->nat.cur = TAILQ_FIRST(&rule->nat.list);
2275 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list);
2276 rule->route.cur = TAILQ_FIRST(&rule->route.list);
2277 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2278 rule, entries);
2279 ruleset->rules[rs_num].inactive.rcount++;
2280
2281 PF_RULES_WUNLOCK();
2282 pf_hash_rule(rule);
2283 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2284 PF_RULES_WLOCK();
2285 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2286 ruleset->rules[rs_num].inactive.rcount--;
2287 pf_free_rule(rule);
2288 rule = NULL;
2289 ERROUT(EEXIST);
2290 }
2291 PF_CONFIG_UNLOCK();
2292
2293 return (0);
2294
2295 #undef ERROUT
2296 #undef ERROUT_UNLOCKED
2297 errout:
2298 PF_RULES_WUNLOCK();
2299 PF_CONFIG_UNLOCK();
2300 errout_unlocked:
2301 pf_kkif_free(rcv_kif);
2302 pf_kkif_free(kif);
2303 pf_krule_free(rule);
2304 return (error);
2305 }
2306
2307 static bool
pf_label_match(const struct pf_krule * rule,const char * label)2308 pf_label_match(const struct pf_krule *rule, const char *label)
2309 {
2310 int i = 0;
2311
2312 while (*rule->label[i]) {
2313 if (strcmp(rule->label[i], label) == 0)
2314 return (true);
2315 i++;
2316 }
2317
2318 return (false);
2319 }
2320
2321 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)2322 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2323 {
2324 struct pf_kstate *s;
2325 int more = 0;
2326
2327 s = pf_find_state_all(key, dir, &more);
2328 if (s == NULL)
2329 return (0);
2330
2331 if (more) {
2332 PF_STATE_UNLOCK(s);
2333 return (0);
2334 }
2335
2336 pf_remove_state(s);
2337 return (1);
2338 }
2339
2340 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)2341 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2342 {
2343 struct pf_kstate *s;
2344 struct pf_state_key *sk;
2345 struct pf_addr *srcaddr, *dstaddr;
2346 struct pf_state_key_cmp match_key;
2347 int idx, killed = 0;
2348 unsigned int dir;
2349 u_int16_t srcport, dstport;
2350 struct pfi_kkif *kif;
2351
2352 relock_DIOCKILLSTATES:
2353 PF_HASHROW_LOCK(ih);
2354 LIST_FOREACH(s, &ih->states, entry) {
2355 /* For floating states look at the original kif. */
2356 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2357
2358 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2359 if (s->direction == PF_OUT) {
2360 srcaddr = &sk->addr[1];
2361 dstaddr = &sk->addr[0];
2362 srcport = sk->port[1];
2363 dstport = sk->port[0];
2364 } else {
2365 srcaddr = &sk->addr[0];
2366 dstaddr = &sk->addr[1];
2367 srcport = sk->port[0];
2368 dstport = sk->port[1];
2369 }
2370
2371 if (psk->psk_af && sk->af != psk->psk_af)
2372 continue;
2373
2374 if (psk->psk_proto && psk->psk_proto != sk->proto)
2375 continue;
2376
2377 if (! pf_match_addr(psk->psk_src.neg,
2378 &psk->psk_src.addr.v.a.addr,
2379 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2380 continue;
2381
2382 if (! pf_match_addr(psk->psk_dst.neg,
2383 &psk->psk_dst.addr.v.a.addr,
2384 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2385 continue;
2386
2387 if (! pf_match_addr(psk->psk_rt_addr.neg,
2388 &psk->psk_rt_addr.addr.v.a.addr,
2389 &psk->psk_rt_addr.addr.v.a.mask,
2390 &s->act.rt_addr, sk->af))
2391 continue;
2392
2393 if (psk->psk_src.port_op != 0 &&
2394 ! pf_match_port(psk->psk_src.port_op,
2395 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2396 continue;
2397
2398 if (psk->psk_dst.port_op != 0 &&
2399 ! pf_match_port(psk->psk_dst.port_op,
2400 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2401 continue;
2402
2403 if (psk->psk_label[0] &&
2404 ! pf_label_match(s->rule, psk->psk_label))
2405 continue;
2406
2407 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2408 kif->pfik_name))
2409 continue;
2410
2411 if (psk->psk_kill_match) {
2412 /* Create the key to find matching states, with lock
2413 * held. */
2414
2415 bzero(&match_key, sizeof(match_key));
2416
2417 if (s->direction == PF_OUT) {
2418 dir = PF_IN;
2419 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2420 } else {
2421 dir = PF_OUT;
2422 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2423 }
2424
2425 match_key.af = s->key[idx]->af;
2426 match_key.proto = s->key[idx]->proto;
2427 pf_addrcpy(&match_key.addr[0],
2428 &s->key[idx]->addr[1], match_key.af);
2429 match_key.port[0] = s->key[idx]->port[1];
2430 pf_addrcpy(&match_key.addr[1],
2431 &s->key[idx]->addr[0], match_key.af);
2432 match_key.port[1] = s->key[idx]->port[0];
2433 }
2434
2435 pf_remove_state(s);
2436 killed++;
2437
2438 if (psk->psk_kill_match)
2439 killed += pf_kill_matching_state(&match_key, dir);
2440
2441 goto relock_DIOCKILLSTATES;
2442 }
2443 PF_HASHROW_UNLOCK(ih);
2444
2445 return (killed);
2446 }
2447
2448 void
unhandled_af(int af)2449 unhandled_af(int af)
2450 {
2451 panic("unhandled af %d", af);
2452 }
2453
2454 int
pf_start(void)2455 pf_start(void)
2456 {
2457 int error = 0;
2458
2459 sx_xlock(&V_pf_ioctl_lock);
2460 if (V_pf_status.running)
2461 error = EEXIST;
2462 else {
2463 hook_pf();
2464 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2465 hook_pf_eth();
2466 V_pf_status.running = 1;
2467 V_pf_status.since = time_uptime;
2468 new_unrhdr64(&V_pf_stateid, time_second);
2469
2470 DPFPRINTF(PF_DEBUG_MISC, "pf: started");
2471 }
2472 sx_xunlock(&V_pf_ioctl_lock);
2473
2474 return (error);
2475 }
2476
2477 int
pf_stop(void)2478 pf_stop(void)
2479 {
2480 int error = 0;
2481
2482 sx_xlock(&V_pf_ioctl_lock);
2483 if (!V_pf_status.running)
2484 error = ENOENT;
2485 else {
2486 V_pf_status.running = 0;
2487 dehook_pf();
2488 dehook_pf_eth();
2489 V_pf_status.since = time_uptime;
2490 DPFPRINTF(PF_DEBUG_MISC, "pf: stopped");
2491 }
2492 sx_xunlock(&V_pf_ioctl_lock);
2493
2494 return (error);
2495 }
2496
2497 void
pf_ioctl_clear_status(void)2498 pf_ioctl_clear_status(void)
2499 {
2500 PF_RULES_WLOCK();
2501 for (int i = 0; i < PFRES_MAX; i++)
2502 counter_u64_zero(V_pf_status.counters[i]);
2503 for (int i = 0; i < FCNT_MAX; i++)
2504 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2505 for (int i = 0; i < SCNT_MAX; i++)
2506 counter_u64_zero(V_pf_status.scounters[i]);
2507 for (int i = 0; i < KLCNT_MAX; i++)
2508 counter_u64_zero(V_pf_status.lcounters[i]);
2509 V_pf_status.since = time_uptime;
2510 if (*V_pf_status.ifname)
2511 pfi_update_status(V_pf_status.ifname, NULL);
2512 PF_RULES_WUNLOCK();
2513 }
2514
2515 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2516 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2517 {
2518 uint32_t old;
2519
2520 if (timeout < 0 || timeout >= PFTM_MAX ||
2521 seconds < 0)
2522 return (EINVAL);
2523
2524 PF_RULES_WLOCK();
2525 old = V_pf_default_rule.timeout[timeout];
2526 if (timeout == PFTM_INTERVAL && seconds == 0)
2527 seconds = 1;
2528 V_pf_default_rule.timeout[timeout] = seconds;
2529 if (timeout == PFTM_INTERVAL && seconds < old)
2530 wakeup(pf_purge_thread);
2531
2532 if (prev_seconds != NULL)
2533 *prev_seconds = old;
2534
2535 PF_RULES_WUNLOCK();
2536
2537 return (0);
2538 }
2539
2540 int
pf_ioctl_get_timeout(int timeout,int * seconds)2541 pf_ioctl_get_timeout(int timeout, int *seconds)
2542 {
2543 PF_RULES_RLOCK_TRACKER;
2544
2545 if (timeout < 0 || timeout >= PFTM_MAX)
2546 return (EINVAL);
2547
2548 PF_RULES_RLOCK();
2549 *seconds = V_pf_default_rule.timeout[timeout];
2550 PF_RULES_RUNLOCK();
2551
2552 return (0);
2553 }
2554
2555 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2556 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2557 {
2558
2559 PF_RULES_WLOCK();
2560 if (index < 0 || index >= PF_LIMIT_MAX ||
2561 V_pf_limits[index].zone == NULL) {
2562 PF_RULES_WUNLOCK();
2563 return (EINVAL);
2564 }
2565 uma_zone_set_max(V_pf_limits[index].zone,
2566 limit == 0 ? INT_MAX : limit);
2567 if (old_limit != NULL)
2568 *old_limit = V_pf_limits[index].limit;
2569 V_pf_limits[index].limit = limit;
2570 PF_RULES_WUNLOCK();
2571
2572 return (0);
2573 }
2574
2575 int
pf_ioctl_get_limit(int index,unsigned int * limit)2576 pf_ioctl_get_limit(int index, unsigned int *limit)
2577 {
2578 PF_RULES_RLOCK_TRACKER;
2579
2580 if (index < 0 || index >= PF_LIMIT_MAX)
2581 return (EINVAL);
2582
2583 PF_RULES_RLOCK();
2584 *limit = V_pf_limits[index].limit;
2585 PF_RULES_RUNLOCK();
2586
2587 return (0);
2588 }
2589
2590 int
pf_ioctl_begin_addrs(uint32_t * ticket)2591 pf_ioctl_begin_addrs(uint32_t *ticket)
2592 {
2593 PF_RULES_WLOCK();
2594 pf_empty_kpool(&V_pf_pabuf[0]);
2595 pf_empty_kpool(&V_pf_pabuf[1]);
2596 pf_empty_kpool(&V_pf_pabuf[2]);
2597 *ticket = ++V_ticket_pabuf;
2598 PF_RULES_WUNLOCK();
2599
2600 return (0);
2601 }
2602
2603 int
pf_ioctl_add_addr(struct pf_nl_pooladdr * pp)2604 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
2605 {
2606 struct pf_kpooladdr *pa = NULL;
2607 struct pfi_kkif *kif = NULL;
2608 int error;
2609
2610 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2611 pp->which != PF_RT)
2612 return (EINVAL);
2613
2614 switch (pp->af) {
2615 #ifdef INET
2616 case AF_INET:
2617 /* FALLTHROUGH */
2618 #endif /* INET */
2619 #ifdef INET6
2620 case AF_INET6:
2621 /* FALLTHROUGH */
2622 #endif /* INET6 */
2623 case AF_UNSPEC:
2624 break;
2625 default:
2626 return (EAFNOSUPPORT);
2627 }
2628
2629 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2630 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2631 pp->addr.addr.type != PF_ADDR_TABLE)
2632 return (EINVAL);
2633
2634 if (pp->addr.addr.p.dyn != NULL)
2635 return (EINVAL);
2636
2637 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2638 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2639 if (error != 0)
2640 goto out;
2641 if (pa->ifname[0])
2642 kif = pf_kkif_create(M_WAITOK);
2643 PF_RULES_WLOCK();
2644 if (pp->ticket != V_ticket_pabuf) {
2645 PF_RULES_WUNLOCK();
2646 if (pa->ifname[0])
2647 pf_kkif_free(kif);
2648 error = EBUSY;
2649 goto out;
2650 }
2651 if (pa->ifname[0]) {
2652 pa->kif = pfi_kkif_attach(kif, pa->ifname);
2653 kif = NULL;
2654 pfi_kkif_ref(pa->kif);
2655 } else
2656 pa->kif = NULL;
2657 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2658 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2659 if (pa->ifname[0])
2660 pfi_kkif_unref(pa->kif);
2661 PF_RULES_WUNLOCK();
2662 goto out;
2663 }
2664 pa->af = pp->af;
2665 switch (pp->which) {
2666 case PF_NAT:
2667 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries);
2668 break;
2669 case PF_RDR:
2670 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries);
2671 break;
2672 case PF_RT:
2673 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries);
2674 break;
2675 }
2676 PF_RULES_WUNLOCK();
2677
2678 return (0);
2679
2680 out:
2681 free(pa, M_PFRULE);
2682 return (error);
2683 }
2684
2685 int
pf_ioctl_get_addrs(struct pf_nl_pooladdr * pp)2686 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
2687 {
2688 struct pf_kpool *pool;
2689 struct pf_kpooladdr *pa;
2690
2691 PF_RULES_RLOCK_TRACKER;
2692
2693 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2694 pp->which != PF_RT)
2695 return (EINVAL);
2696
2697 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2698 pp->nr = 0;
2699
2700 PF_RULES_RLOCK();
2701 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2702 pp->r_num, 0, 1, 0, pp->which);
2703 if (pool == NULL) {
2704 PF_RULES_RUNLOCK();
2705 return (EBUSY);
2706 }
2707 TAILQ_FOREACH(pa, &pool->list, entries)
2708 pp->nr++;
2709 PF_RULES_RUNLOCK();
2710
2711 return (0);
2712 }
2713
2714 int
pf_ioctl_get_addr(struct pf_nl_pooladdr * pp)2715 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
2716 {
2717 struct pf_kpool *pool;
2718 struct pf_kpooladdr *pa;
2719 u_int32_t nr = 0;
2720
2721 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2722 pp->which != PF_RT)
2723 return (EINVAL);
2724
2725 PF_RULES_RLOCK_TRACKER;
2726
2727 pp->anchor[sizeof(pp->anchor) - 1] = '\0';
2728
2729 PF_RULES_RLOCK();
2730 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2731 pp->r_num, 0, 1, 1, pp->which);
2732 if (pool == NULL) {
2733 PF_RULES_RUNLOCK();
2734 return (EBUSY);
2735 }
2736 pa = TAILQ_FIRST(&pool->list);
2737 while ((pa != NULL) && (nr < pp->nr)) {
2738 pa = TAILQ_NEXT(pa, entries);
2739 nr++;
2740 }
2741 if (pa == NULL) {
2742 PF_RULES_RUNLOCK();
2743 return (EBUSY);
2744 }
2745 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2746 pp->af = pa->af;
2747 pf_addr_copyout(&pp->addr.addr);
2748 PF_RULES_RUNLOCK();
2749
2750 return (0);
2751 }
2752
2753 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)2754 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
2755 {
2756 struct pf_kruleset *ruleset;
2757 struct pf_kanchor *anchor;
2758
2759 PF_RULES_RLOCK_TRACKER;
2760
2761 pr->path[sizeof(pr->path) - 1] = '\0';
2762
2763 PF_RULES_RLOCK();
2764 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2765 PF_RULES_RUNLOCK();
2766 return (ENOENT);
2767 }
2768 pr->nr = 0;
2769 if (ruleset == &pf_main_ruleset) {
2770 /* XXX kludge for pf_main_ruleset */
2771 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2772 if (anchor->parent == NULL)
2773 pr->nr++;
2774 } else {
2775 RB_FOREACH(anchor, pf_kanchor_node,
2776 &ruleset->anchor->children)
2777 pr->nr++;
2778 }
2779 PF_RULES_RUNLOCK();
2780
2781 return (0);
2782 }
2783
2784 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)2785 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
2786 {
2787 struct pf_kruleset *ruleset;
2788 struct pf_kanchor *anchor;
2789 u_int32_t nr = 0;
2790 int error = 0;
2791
2792 PF_RULES_RLOCK_TRACKER;
2793
2794 PF_RULES_RLOCK();
2795 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2796 PF_RULES_RUNLOCK();
2797 return (ENOENT);
2798 }
2799
2800 pr->name[0] = '\0';
2801 if (ruleset == &pf_main_ruleset) {
2802 /* XXX kludge for pf_main_ruleset */
2803 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2804 if (anchor->parent == NULL && nr++ == pr->nr) {
2805 strlcpy(pr->name, anchor->name,
2806 sizeof(pr->name));
2807 break;
2808 }
2809 } else {
2810 RB_FOREACH(anchor, pf_kanchor_node,
2811 &ruleset->anchor->children)
2812 if (nr++ == pr->nr) {
2813 strlcpy(pr->name, anchor->name,
2814 sizeof(pr->name));
2815 break;
2816 }
2817 }
2818 if (!pr->name[0])
2819 error = EBUSY;
2820 PF_RULES_RUNLOCK();
2821
2822 return (error);
2823 }
2824
2825 int
pf_ioctl_natlook(struct pfioc_natlook * pnl)2826 pf_ioctl_natlook(struct pfioc_natlook *pnl)
2827 {
2828 struct pf_state_key *sk;
2829 struct pf_kstate *state;
2830 struct pf_state_key_cmp key;
2831 int m = 0, direction = pnl->direction;
2832 int sidx, didx;
2833
2834 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
2835 sidx = (direction == PF_IN) ? 1 : 0;
2836 didx = (direction == PF_IN) ? 0 : 1;
2837
2838 if (!pnl->proto ||
2839 PF_AZERO(&pnl->saddr, pnl->af) ||
2840 PF_AZERO(&pnl->daddr, pnl->af) ||
2841 ((pnl->proto == IPPROTO_TCP ||
2842 pnl->proto == IPPROTO_UDP) &&
2843 (!pnl->dport || !pnl->sport)))
2844 return (EINVAL);
2845
2846 switch (pnl->direction) {
2847 case PF_IN:
2848 case PF_OUT:
2849 case PF_INOUT:
2850 break;
2851 default:
2852 return (EINVAL);
2853 }
2854
2855 switch (pnl->af) {
2856 #ifdef INET
2857 case AF_INET:
2858 break;
2859 #endif /* INET */
2860 #ifdef INET6
2861 case AF_INET6:
2862 break;
2863 #endif /* INET6 */
2864 default:
2865 return (EAFNOSUPPORT);
2866 }
2867
2868 bzero(&key, sizeof(key));
2869 key.af = pnl->af;
2870 key.proto = pnl->proto;
2871 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af);
2872 key.port[sidx] = pnl->sport;
2873 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af);
2874 key.port[didx] = pnl->dport;
2875
2876 state = pf_find_state_all(&key, direction, &m);
2877 if (state == NULL)
2878 return (ENOENT);
2879
2880 if (m > 1) {
2881 PF_STATE_UNLOCK(state);
2882 return (E2BIG); /* more than one state */
2883 }
2884
2885 sk = state->key[sidx];
2886 pf_addrcpy(&pnl->rsaddr,
2887 &sk->addr[sidx], sk->af);
2888 pnl->rsport = sk->port[sidx];
2889 pf_addrcpy(&pnl->rdaddr,
2890 &sk->addr[didx], sk->af);
2891 pnl->rdport = sk->port[didx];
2892 PF_STATE_UNLOCK(state);
2893
2894 return (0);
2895 }
2896
2897 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2898 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2899 {
2900 int error = 0;
2901 PF_RULES_RLOCK_TRACKER;
2902
2903 #define ERROUT_IOCTL(target, x) \
2904 do { \
2905 error = (x); \
2906 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
2907 goto target; \
2908 } while (0)
2909
2910
2911 /* XXX keep in sync with switch() below */
2912 if (securelevel_gt(td->td_ucred, 2))
2913 switch (cmd) {
2914 case DIOCGETRULES:
2915 case DIOCGETRULENV:
2916 case DIOCGETADDRS:
2917 case DIOCGETADDR:
2918 case DIOCGETSTATE:
2919 case DIOCGETSTATENV:
2920 case DIOCSETSTATUSIF:
2921 case DIOCGETSTATUSNV:
2922 case DIOCCLRSTATUS:
2923 case DIOCNATLOOK:
2924 case DIOCSETDEBUG:
2925 #ifdef COMPAT_FREEBSD14
2926 case DIOCGETSTATES:
2927 case DIOCGETSTATESV2:
2928 #endif
2929 case DIOCGETTIMEOUT:
2930 case DIOCCLRRULECTRS:
2931 case DIOCGETLIMIT:
2932 case DIOCGETALTQSV0:
2933 case DIOCGETALTQSV1:
2934 case DIOCGETALTQV0:
2935 case DIOCGETALTQV1:
2936 case DIOCGETQSTATSV0:
2937 case DIOCGETQSTATSV1:
2938 case DIOCGETRULESETS:
2939 case DIOCGETRULESET:
2940 case DIOCRGETTABLES:
2941 case DIOCRGETTSTATS:
2942 case DIOCRCLRTSTATS:
2943 case DIOCRCLRADDRS:
2944 case DIOCRADDADDRS:
2945 case DIOCRDELADDRS:
2946 case DIOCRSETADDRS:
2947 case DIOCRGETADDRS:
2948 case DIOCRGETASTATS:
2949 case DIOCRCLRASTATS:
2950 case DIOCRTSTADDRS:
2951 case DIOCOSFPGET:
2952 case DIOCGETSRCNODES:
2953 case DIOCCLRSRCNODES:
2954 case DIOCGETSYNCOOKIES:
2955 case DIOCIGETIFACES:
2956 case DIOCGIFSPEEDV0:
2957 case DIOCGIFSPEEDV1:
2958 case DIOCSETIFFLAG:
2959 case DIOCCLRIFFLAG:
2960 case DIOCGETETHRULES:
2961 case DIOCGETETHRULE:
2962 case DIOCGETETHRULESETS:
2963 case DIOCGETETHRULESET:
2964 break;
2965 case DIOCRCLRTABLES:
2966 case DIOCRADDTABLES:
2967 case DIOCRDELTABLES:
2968 case DIOCRSETTFLAGS:
2969 if (((struct pfioc_table *)addr)->pfrio_flags &
2970 PFR_FLAG_DUMMY)
2971 break; /* dummy operation ok */
2972 return (EPERM);
2973 default:
2974 return (EPERM);
2975 }
2976
2977 if (!(flags & FWRITE))
2978 switch (cmd) {
2979 case DIOCGETRULES:
2980 case DIOCGETADDRS:
2981 case DIOCGETADDR:
2982 case DIOCGETSTATE:
2983 case DIOCGETSTATENV:
2984 case DIOCGETSTATUSNV:
2985 #ifdef COMPAT_FREEBSD14
2986 case DIOCGETSTATES:
2987 case DIOCGETSTATESV2:
2988 #endif
2989 case DIOCGETTIMEOUT:
2990 case DIOCGETLIMIT:
2991 case DIOCGETALTQSV0:
2992 case DIOCGETALTQSV1:
2993 case DIOCGETALTQV0:
2994 case DIOCGETALTQV1:
2995 case DIOCGETQSTATSV0:
2996 case DIOCGETQSTATSV1:
2997 case DIOCGETRULESETS:
2998 case DIOCGETRULESET:
2999 case DIOCNATLOOK:
3000 case DIOCRGETTABLES:
3001 case DIOCRGETTSTATS:
3002 case DIOCRGETADDRS:
3003 case DIOCRGETASTATS:
3004 case DIOCRTSTADDRS:
3005 case DIOCOSFPGET:
3006 case DIOCGETSRCNODES:
3007 case DIOCGETSYNCOOKIES:
3008 case DIOCIGETIFACES:
3009 case DIOCGIFSPEEDV1:
3010 case DIOCGIFSPEEDV0:
3011 case DIOCGETRULENV:
3012 case DIOCGETETHRULES:
3013 case DIOCGETETHRULE:
3014 case DIOCGETETHRULESETS:
3015 case DIOCGETETHRULESET:
3016 break;
3017 case DIOCRCLRTABLES:
3018 case DIOCRADDTABLES:
3019 case DIOCRDELTABLES:
3020 case DIOCRCLRTSTATS:
3021 case DIOCRCLRADDRS:
3022 case DIOCRADDADDRS:
3023 case DIOCRDELADDRS:
3024 case DIOCRSETADDRS:
3025 case DIOCRSETTFLAGS:
3026 if (((struct pfioc_table *)addr)->pfrio_flags &
3027 PFR_FLAG_DUMMY) {
3028 flags |= FWRITE; /* need write lock for dummy */
3029 break; /* dummy operation ok */
3030 }
3031 return (EACCES);
3032 default:
3033 return (EACCES);
3034 }
3035
3036 CURVNET_SET(TD_TO_VNET(td));
3037
3038 switch (cmd) {
3039 #ifdef COMPAT_FREEBSD14
3040 case DIOCSTART:
3041 error = pf_start();
3042 break;
3043
3044 case DIOCSTOP:
3045 error = pf_stop();
3046 break;
3047 #endif
3048
3049 case DIOCGETETHRULES: {
3050 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3051 nvlist_t *nvl;
3052 void *packed;
3053 struct pf_keth_rule *tail;
3054 struct pf_keth_ruleset *rs;
3055 u_int32_t ticket, nr;
3056 const char *anchor = "";
3057
3058 nvl = NULL;
3059 packed = NULL;
3060
3061 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
3062
3063 if (nv->len > pf_ioctl_maxcount)
3064 ERROUT(ENOMEM);
3065
3066 /* Copy the request in */
3067 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
3068 error = copyin(nv->data, packed, nv->len);
3069 if (error)
3070 ERROUT(error);
3071
3072 nvl = nvlist_unpack(packed, nv->len, 0);
3073 if (nvl == NULL)
3074 ERROUT(EBADMSG);
3075
3076 if (! nvlist_exists_string(nvl, "anchor"))
3077 ERROUT(EBADMSG);
3078
3079 anchor = nvlist_get_string(nvl, "anchor");
3080
3081 rs = pf_find_keth_ruleset(anchor);
3082
3083 nvlist_destroy(nvl);
3084 nvl = NULL;
3085 free(packed, M_NVLIST);
3086 packed = NULL;
3087
3088 if (rs == NULL)
3089 ERROUT(ENOENT);
3090
3091 /* Reply */
3092 nvl = nvlist_create(0);
3093 if (nvl == NULL)
3094 ERROUT(ENOMEM);
3095
3096 PF_RULES_RLOCK();
3097
3098 ticket = rs->active.ticket;
3099 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
3100 if (tail)
3101 nr = tail->nr + 1;
3102 else
3103 nr = 0;
3104
3105 PF_RULES_RUNLOCK();
3106
3107 nvlist_add_number(nvl, "ticket", ticket);
3108 nvlist_add_number(nvl, "nr", nr);
3109
3110 packed = nvlist_pack(nvl, &nv->len);
3111 if (packed == NULL)
3112 ERROUT(ENOMEM);
3113
3114 if (nv->size == 0)
3115 ERROUT(0);
3116 else if (nv->size < nv->len)
3117 ERROUT(ENOSPC);
3118
3119 error = copyout(packed, nv->data, nv->len);
3120
3121 #undef ERROUT
3122 DIOCGETETHRULES_error:
3123 free(packed, M_NVLIST);
3124 nvlist_destroy(nvl);
3125 break;
3126 }
3127
3128 case DIOCGETETHRULE: {
3129 struct epoch_tracker et;
3130 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3131 nvlist_t *nvl = NULL;
3132 void *nvlpacked = NULL;
3133 struct pf_keth_rule *rule = NULL;
3134 struct pf_keth_ruleset *rs;
3135 u_int32_t ticket, nr;
3136 bool clear = false;
3137 const char *anchor;
3138
3139 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
3140
3141 if (nv->len > pf_ioctl_maxcount)
3142 ERROUT(ENOMEM);
3143
3144 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3145 error = copyin(nv->data, nvlpacked, nv->len);
3146 if (error)
3147 ERROUT(error);
3148
3149 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3150 if (nvl == NULL)
3151 ERROUT(EBADMSG);
3152 if (! nvlist_exists_number(nvl, "ticket"))
3153 ERROUT(EBADMSG);
3154 ticket = nvlist_get_number(nvl, "ticket");
3155 if (! nvlist_exists_string(nvl, "anchor"))
3156 ERROUT(EBADMSG);
3157 anchor = nvlist_get_string(nvl, "anchor");
3158
3159 if (nvlist_exists_bool(nvl, "clear"))
3160 clear = nvlist_get_bool(nvl, "clear");
3161
3162 if (clear && !(flags & FWRITE))
3163 ERROUT(EACCES);
3164
3165 if (! nvlist_exists_number(nvl, "nr"))
3166 ERROUT(EBADMSG);
3167 nr = nvlist_get_number(nvl, "nr");
3168
3169 PF_RULES_RLOCK();
3170 rs = pf_find_keth_ruleset(anchor);
3171 if (rs == NULL) {
3172 PF_RULES_RUNLOCK();
3173 ERROUT(ENOENT);
3174 }
3175 if (ticket != rs->active.ticket) {
3176 PF_RULES_RUNLOCK();
3177 ERROUT(EBUSY);
3178 }
3179
3180 nvlist_destroy(nvl);
3181 nvl = NULL;
3182 free(nvlpacked, M_NVLIST);
3183 nvlpacked = NULL;
3184
3185 rule = TAILQ_FIRST(rs->active.rules);
3186 while ((rule != NULL) && (rule->nr != nr))
3187 rule = TAILQ_NEXT(rule, entries);
3188 if (rule == NULL) {
3189 PF_RULES_RUNLOCK();
3190 ERROUT(ENOENT);
3191 }
3192 /* Make sure rule can't go away. */
3193 NET_EPOCH_ENTER(et);
3194 PF_RULES_RUNLOCK();
3195 nvl = pf_keth_rule_to_nveth_rule(rule);
3196 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
3197 NET_EPOCH_EXIT(et);
3198 ERROUT(EBUSY);
3199 }
3200 NET_EPOCH_EXIT(et);
3201 if (nvl == NULL)
3202 ERROUT(ENOMEM);
3203
3204 nvlpacked = nvlist_pack(nvl, &nv->len);
3205 if (nvlpacked == NULL)
3206 ERROUT(ENOMEM);
3207
3208 if (nv->size == 0)
3209 ERROUT(0);
3210 else if (nv->size < nv->len)
3211 ERROUT(ENOSPC);
3212
3213 error = copyout(nvlpacked, nv->data, nv->len);
3214 if (error == 0 && clear) {
3215 counter_u64_zero(rule->evaluations);
3216 for (int i = 0; i < 2; i++) {
3217 counter_u64_zero(rule->packets[i]);
3218 counter_u64_zero(rule->bytes[i]);
3219 }
3220 }
3221
3222 #undef ERROUT
3223 DIOCGETETHRULE_error:
3224 free(nvlpacked, M_NVLIST);
3225 nvlist_destroy(nvl);
3226 break;
3227 }
3228
3229 case DIOCADDETHRULE: {
3230 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3231 nvlist_t *nvl = NULL;
3232 void *nvlpacked = NULL;
3233 struct pf_keth_rule *rule = NULL, *tail = NULL;
3234 struct pf_keth_ruleset *ruleset = NULL;
3235 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
3236 const char *anchor = "", *anchor_call = "";
3237
3238 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3239
3240 if (nv->len > pf_ioctl_maxcount)
3241 ERROUT(ENOMEM);
3242
3243 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3244 error = copyin(nv->data, nvlpacked, nv->len);
3245 if (error)
3246 ERROUT(error);
3247
3248 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3249 if (nvl == NULL)
3250 ERROUT(EBADMSG);
3251
3252 if (! nvlist_exists_number(nvl, "ticket"))
3253 ERROUT(EBADMSG);
3254
3255 if (nvlist_exists_string(nvl, "anchor"))
3256 anchor = nvlist_get_string(nvl, "anchor");
3257 if (nvlist_exists_string(nvl, "anchor_call"))
3258 anchor_call = nvlist_get_string(nvl, "anchor_call");
3259
3260 ruleset = pf_find_keth_ruleset(anchor);
3261 if (ruleset == NULL)
3262 ERROUT(EINVAL);
3263
3264 if (nvlist_get_number(nvl, "ticket") !=
3265 ruleset->inactive.ticket) {
3266 DPFPRINTF(PF_DEBUG_MISC,
3267 "ticket: %d != %d",
3268 (u_int32_t)nvlist_get_number(nvl, "ticket"),
3269 ruleset->inactive.ticket);
3270 ERROUT(EBUSY);
3271 }
3272
3273 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3274 rule->timestamp = NULL;
3275
3276 error = pf_nveth_rule_to_keth_rule(nvl, rule);
3277 if (error != 0)
3278 ERROUT(error);
3279
3280 if (rule->ifname[0])
3281 kif = pf_kkif_create(M_WAITOK);
3282 if (rule->bridge_to_name[0])
3283 bridge_to_kif = pf_kkif_create(M_WAITOK);
3284 rule->evaluations = counter_u64_alloc(M_WAITOK);
3285 for (int i = 0; i < 2; i++) {
3286 rule->packets[i] = counter_u64_alloc(M_WAITOK);
3287 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3288 }
3289 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3290 M_WAITOK | M_ZERO);
3291
3292 PF_RULES_WLOCK();
3293
3294 if (rule->ifname[0]) {
3295 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3296 pfi_kkif_ref(rule->kif);
3297 } else
3298 rule->kif = NULL;
3299 if (rule->bridge_to_name[0]) {
3300 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3301 rule->bridge_to_name);
3302 pfi_kkif_ref(rule->bridge_to);
3303 } else
3304 rule->bridge_to = NULL;
3305
3306 #ifdef ALTQ
3307 /* set queue IDs */
3308 if (rule->qname[0] != 0) {
3309 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3310 error = EBUSY;
3311 else
3312 rule->qid = rule->qid;
3313 }
3314 #endif
3315 if (rule->tagname[0])
3316 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3317 error = EBUSY;
3318 if (rule->match_tagname[0])
3319 if ((rule->match_tag = pf_tagname2tag(
3320 rule->match_tagname)) == 0)
3321 error = EBUSY;
3322
3323 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3324 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3325 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3326 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3327
3328 if (error) {
3329 pf_free_eth_rule(rule);
3330 PF_RULES_WUNLOCK();
3331 ERROUT(error);
3332 }
3333
3334 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3335 pf_free_eth_rule(rule);
3336 PF_RULES_WUNLOCK();
3337 ERROUT(EINVAL);
3338 }
3339
3340 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3341 if (tail)
3342 rule->nr = tail->nr + 1;
3343 else
3344 rule->nr = 0;
3345
3346 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3347
3348 PF_RULES_WUNLOCK();
3349
3350 #undef ERROUT
3351 DIOCADDETHRULE_error:
3352 nvlist_destroy(nvl);
3353 free(nvlpacked, M_NVLIST);
3354 break;
3355 }
3356
3357 case DIOCGETETHRULESETS: {
3358 struct epoch_tracker et;
3359 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3360 nvlist_t *nvl = NULL;
3361 void *nvlpacked = NULL;
3362 struct pf_keth_ruleset *ruleset;
3363 struct pf_keth_anchor *anchor;
3364 int nr = 0;
3365
3366 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3367
3368 if (nv->len > pf_ioctl_maxcount)
3369 ERROUT(ENOMEM);
3370
3371 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3372 error = copyin(nv->data, nvlpacked, nv->len);
3373 if (error)
3374 ERROUT(error);
3375
3376 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3377 if (nvl == NULL)
3378 ERROUT(EBADMSG);
3379 if (! nvlist_exists_string(nvl, "path"))
3380 ERROUT(EBADMSG);
3381
3382 NET_EPOCH_ENTER(et);
3383
3384 if ((ruleset = pf_find_keth_ruleset(
3385 nvlist_get_string(nvl, "path"))) == NULL) {
3386 NET_EPOCH_EXIT(et);
3387 ERROUT(ENOENT);
3388 }
3389
3390 if (ruleset->anchor == NULL) {
3391 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3392 if (anchor->parent == NULL)
3393 nr++;
3394 } else {
3395 RB_FOREACH(anchor, pf_keth_anchor_node,
3396 &ruleset->anchor->children)
3397 nr++;
3398 }
3399
3400 NET_EPOCH_EXIT(et);
3401
3402 nvlist_destroy(nvl);
3403 nvl = NULL;
3404 free(nvlpacked, M_NVLIST);
3405 nvlpacked = NULL;
3406
3407 nvl = nvlist_create(0);
3408 if (nvl == NULL)
3409 ERROUT(ENOMEM);
3410
3411 nvlist_add_number(nvl, "nr", nr);
3412
3413 nvlpacked = nvlist_pack(nvl, &nv->len);
3414 if (nvlpacked == NULL)
3415 ERROUT(ENOMEM);
3416
3417 if (nv->size == 0)
3418 ERROUT(0);
3419 else if (nv->size < nv->len)
3420 ERROUT(ENOSPC);
3421
3422 error = copyout(nvlpacked, nv->data, nv->len);
3423
3424 #undef ERROUT
3425 DIOCGETETHRULESETS_error:
3426 free(nvlpacked, M_NVLIST);
3427 nvlist_destroy(nvl);
3428 break;
3429 }
3430
3431 case DIOCGETETHRULESET: {
3432 struct epoch_tracker et;
3433 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3434 nvlist_t *nvl = NULL;
3435 void *nvlpacked = NULL;
3436 struct pf_keth_ruleset *ruleset;
3437 struct pf_keth_anchor *anchor;
3438 int nr = 0, req_nr = 0;
3439 bool found = false;
3440
3441 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3442
3443 if (nv->len > pf_ioctl_maxcount)
3444 ERROUT(ENOMEM);
3445
3446 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3447 error = copyin(nv->data, nvlpacked, nv->len);
3448 if (error)
3449 ERROUT(error);
3450
3451 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3452 if (nvl == NULL)
3453 ERROUT(EBADMSG);
3454 if (! nvlist_exists_string(nvl, "path"))
3455 ERROUT(EBADMSG);
3456 if (! nvlist_exists_number(nvl, "nr"))
3457 ERROUT(EBADMSG);
3458
3459 req_nr = nvlist_get_number(nvl, "nr");
3460
3461 NET_EPOCH_ENTER(et);
3462
3463 if ((ruleset = pf_find_keth_ruleset(
3464 nvlist_get_string(nvl, "path"))) == NULL) {
3465 NET_EPOCH_EXIT(et);
3466 ERROUT(ENOENT);
3467 }
3468
3469 nvlist_destroy(nvl);
3470 nvl = NULL;
3471 free(nvlpacked, M_NVLIST);
3472 nvlpacked = NULL;
3473
3474 nvl = nvlist_create(0);
3475 if (nvl == NULL) {
3476 NET_EPOCH_EXIT(et);
3477 ERROUT(ENOMEM);
3478 }
3479
3480 if (ruleset->anchor == NULL) {
3481 RB_FOREACH(anchor, pf_keth_anchor_global,
3482 &V_pf_keth_anchors) {
3483 if (anchor->parent == NULL && nr++ == req_nr) {
3484 found = true;
3485 break;
3486 }
3487 }
3488 } else {
3489 RB_FOREACH(anchor, pf_keth_anchor_node,
3490 &ruleset->anchor->children) {
3491 if (nr++ == req_nr) {
3492 found = true;
3493 break;
3494 }
3495 }
3496 }
3497
3498 NET_EPOCH_EXIT(et);
3499 if (found) {
3500 nvlist_add_number(nvl, "nr", nr);
3501 nvlist_add_string(nvl, "name", anchor->name);
3502 if (ruleset->anchor)
3503 nvlist_add_string(nvl, "path",
3504 ruleset->anchor->path);
3505 else
3506 nvlist_add_string(nvl, "path", "");
3507 } else {
3508 ERROUT(EBUSY);
3509 }
3510
3511 nvlpacked = nvlist_pack(nvl, &nv->len);
3512 if (nvlpacked == NULL)
3513 ERROUT(ENOMEM);
3514
3515 if (nv->size == 0)
3516 ERROUT(0);
3517 else if (nv->size < nv->len)
3518 ERROUT(ENOSPC);
3519
3520 error = copyout(nvlpacked, nv->data, nv->len);
3521
3522 #undef ERROUT
3523 DIOCGETETHRULESET_error:
3524 free(nvlpacked, M_NVLIST);
3525 nvlist_destroy(nvl);
3526 break;
3527 }
3528
3529 case DIOCADDRULENV: {
3530 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3531 nvlist_t *nvl = NULL;
3532 void *nvlpacked = NULL;
3533 struct pf_krule *rule = NULL;
3534 const char *anchor = "", *anchor_call = "";
3535 uint32_t ticket = 0, pool_ticket = 0;
3536
3537 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
3538
3539 if (nv->len > pf_ioctl_maxcount)
3540 ERROUT(ENOMEM);
3541
3542 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3543 error = copyin(nv->data, nvlpacked, nv->len);
3544 if (error)
3545 ERROUT(error);
3546
3547 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3548 if (nvl == NULL)
3549 ERROUT(EBADMSG);
3550
3551 if (! nvlist_exists_number(nvl, "ticket"))
3552 ERROUT(EINVAL);
3553 ticket = nvlist_get_number(nvl, "ticket");
3554
3555 if (! nvlist_exists_number(nvl, "pool_ticket"))
3556 ERROUT(EINVAL);
3557 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3558
3559 if (! nvlist_exists_nvlist(nvl, "rule"))
3560 ERROUT(EINVAL);
3561
3562 rule = pf_krule_alloc();
3563 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3564 rule);
3565 if (error)
3566 ERROUT(error);
3567
3568 if (nvlist_exists_string(nvl, "anchor"))
3569 anchor = nvlist_get_string(nvl, "anchor");
3570 if (nvlist_exists_string(nvl, "anchor_call"))
3571 anchor_call = nvlist_get_string(nvl, "anchor_call");
3572
3573 if ((error = nvlist_error(nvl)))
3574 ERROUT(error);
3575
3576 /* Frees rule on error */
3577 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3578 anchor_call, td->td_ucred->cr_ruid,
3579 td->td_proc ? td->td_proc->p_pid : 0);
3580
3581 nvlist_destroy(nvl);
3582 free(nvlpacked, M_NVLIST);
3583 break;
3584 #undef ERROUT
3585 DIOCADDRULENV_error:
3586 pf_krule_free(rule);
3587 nvlist_destroy(nvl);
3588 free(nvlpacked, M_NVLIST);
3589
3590 break;
3591 }
3592 case DIOCADDRULE: {
3593 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3594 struct pf_krule *rule;
3595
3596 rule = pf_krule_alloc();
3597 error = pf_rule_to_krule(&pr->rule, rule);
3598 if (error != 0) {
3599 pf_krule_free(rule);
3600 goto fail;
3601 }
3602
3603 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
3604
3605 /* Frees rule on error */
3606 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3607 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3608 td->td_proc ? td->td_proc->p_pid : 0);
3609 break;
3610 }
3611
3612 case DIOCGETRULES: {
3613 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3614
3615 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
3616
3617 error = pf_ioctl_getrules(pr);
3618
3619 break;
3620 }
3621
3622 case DIOCGETRULENV: {
3623 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3624 nvlist_t *nvrule = NULL;
3625 nvlist_t *nvl = NULL;
3626 struct pf_kruleset *ruleset;
3627 struct pf_krule *rule;
3628 void *nvlpacked = NULL;
3629 int rs_num, nr;
3630 bool clear_counter = false;
3631
3632 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
3633
3634 if (nv->len > pf_ioctl_maxcount)
3635 ERROUT(ENOMEM);
3636
3637 /* Copy the request in */
3638 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3639 error = copyin(nv->data, nvlpacked, nv->len);
3640 if (error)
3641 ERROUT(error);
3642
3643 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3644 if (nvl == NULL)
3645 ERROUT(EBADMSG);
3646
3647 if (! nvlist_exists_string(nvl, "anchor"))
3648 ERROUT(EBADMSG);
3649 if (! nvlist_exists_number(nvl, "ruleset"))
3650 ERROUT(EBADMSG);
3651 if (! nvlist_exists_number(nvl, "ticket"))
3652 ERROUT(EBADMSG);
3653 if (! nvlist_exists_number(nvl, "nr"))
3654 ERROUT(EBADMSG);
3655
3656 if (nvlist_exists_bool(nvl, "clear_counter"))
3657 clear_counter = nvlist_get_bool(nvl, "clear_counter");
3658
3659 if (clear_counter && !(flags & FWRITE))
3660 ERROUT(EACCES);
3661
3662 nr = nvlist_get_number(nvl, "nr");
3663
3664 PF_RULES_WLOCK();
3665 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3666 if (ruleset == NULL) {
3667 PF_RULES_WUNLOCK();
3668 ERROUT(ENOENT);
3669 }
3670
3671 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3672 if (rs_num >= PF_RULESET_MAX) {
3673 PF_RULES_WUNLOCK();
3674 ERROUT(EINVAL);
3675 }
3676
3677 if (nvlist_get_number(nvl, "ticket") !=
3678 ruleset->rules[rs_num].active.ticket) {
3679 PF_RULES_WUNLOCK();
3680 ERROUT(EBUSY);
3681 }
3682
3683 if ((error = nvlist_error(nvl))) {
3684 PF_RULES_WUNLOCK();
3685 ERROUT(error);
3686 }
3687
3688 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3689 while ((rule != NULL) && (rule->nr != nr))
3690 rule = TAILQ_NEXT(rule, entries);
3691 if (rule == NULL) {
3692 PF_RULES_WUNLOCK();
3693 ERROUT(EBUSY);
3694 }
3695
3696 nvrule = pf_krule_to_nvrule(rule);
3697
3698 nvlist_destroy(nvl);
3699 nvl = nvlist_create(0);
3700 if (nvl == NULL) {
3701 PF_RULES_WUNLOCK();
3702 ERROUT(ENOMEM);
3703 }
3704 nvlist_add_number(nvl, "nr", nr);
3705 nvlist_add_nvlist(nvl, "rule", nvrule);
3706 nvlist_destroy(nvrule);
3707 nvrule = NULL;
3708 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3709 PF_RULES_WUNLOCK();
3710 ERROUT(EBUSY);
3711 }
3712
3713 free(nvlpacked, M_NVLIST);
3714 nvlpacked = nvlist_pack(nvl, &nv->len);
3715 if (nvlpacked == NULL) {
3716 PF_RULES_WUNLOCK();
3717 ERROUT(ENOMEM);
3718 }
3719
3720 if (nv->size == 0) {
3721 PF_RULES_WUNLOCK();
3722 ERROUT(0);
3723 }
3724 else if (nv->size < nv->len) {
3725 PF_RULES_WUNLOCK();
3726 ERROUT(ENOSPC);
3727 }
3728
3729 if (clear_counter)
3730 pf_krule_clear_counters(rule);
3731
3732 PF_RULES_WUNLOCK();
3733
3734 error = copyout(nvlpacked, nv->data, nv->len);
3735
3736 #undef ERROUT
3737 DIOCGETRULENV_error:
3738 free(nvlpacked, M_NVLIST);
3739 nvlist_destroy(nvrule);
3740 nvlist_destroy(nvl);
3741
3742 break;
3743 }
3744
3745 case DIOCCHANGERULE: {
3746 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
3747 struct pf_kruleset *ruleset;
3748 struct pf_krule *oldrule = NULL, *newrule = NULL;
3749 struct pfi_kkif *kif = NULL;
3750 struct pf_kpooladdr *pa;
3751 u_int32_t nr = 0;
3752 int rs_num;
3753
3754 pcr->anchor[sizeof(pcr->anchor) - 1] = '\0';
3755
3756 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3757 pcr->action > PF_CHANGE_GET_TICKET) {
3758 error = EINVAL;
3759 goto fail;
3760 }
3761 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3762 error = EINVAL;
3763 goto fail;
3764 }
3765
3766 if (pcr->action != PF_CHANGE_REMOVE) {
3767 newrule = pf_krule_alloc();
3768 error = pf_rule_to_krule(&pcr->rule, newrule);
3769 if (error != 0) {
3770 pf_krule_free(newrule);
3771 goto fail;
3772 }
3773
3774 if ((error = pf_rule_checkaf(newrule))) {
3775 pf_krule_free(newrule);
3776 goto fail;
3777 }
3778 if (newrule->ifname[0])
3779 kif = pf_kkif_create(M_WAITOK);
3780 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3781 for (int i = 0; i < 2; i++) {
3782 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3783 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3784 }
3785 newrule->states_cur = counter_u64_alloc(M_WAITOK);
3786 newrule->states_tot = counter_u64_alloc(M_WAITOK);
3787 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
3788 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
3789 newrule->cuid = td->td_ucred->cr_ruid;
3790 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3791 TAILQ_INIT(&newrule->nat.list);
3792 TAILQ_INIT(&newrule->rdr.list);
3793 TAILQ_INIT(&newrule->route.list);
3794 }
3795 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3796
3797 PF_CONFIG_LOCK();
3798 PF_RULES_WLOCK();
3799 #ifdef PF_WANT_32_TO_64_COUNTER
3800 if (newrule != NULL) {
3801 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3802 newrule->allrulelinked = true;
3803 V_pf_allrulecount++;
3804 }
3805 #endif
3806
3807 if (!(pcr->action == PF_CHANGE_REMOVE ||
3808 pcr->action == PF_CHANGE_GET_TICKET) &&
3809 pcr->pool_ticket != V_ticket_pabuf)
3810 ERROUT(EBUSY);
3811
3812 ruleset = pf_find_kruleset(pcr->anchor);
3813 if (ruleset == NULL)
3814 ERROUT(EINVAL);
3815
3816 rs_num = pf_get_ruleset_number(pcr->rule.action);
3817 if (rs_num >= PF_RULESET_MAX)
3818 ERROUT(EINVAL);
3819
3820 /*
3821 * XXXMJG: there is no guarantee that the ruleset was
3822 * created by the usual route of calling DIOCXBEGIN.
3823 * As a result it is possible the rule tree will not
3824 * be allocated yet. Hack around it by doing it here.
3825 * Note it is fine to let the tree persist in case of
3826 * error as it will be freed down the road on future
3827 * updates (if need be).
3828 */
3829 if (ruleset->rules[rs_num].active.tree == NULL) {
3830 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3831 if (ruleset->rules[rs_num].active.tree == NULL) {
3832 ERROUT(ENOMEM);
3833 }
3834 }
3835
3836 if (pcr->action == PF_CHANGE_GET_TICKET) {
3837 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3838 ERROUT(0);
3839 } else if (pcr->ticket !=
3840 ruleset->rules[rs_num].active.ticket)
3841 ERROUT(EINVAL);
3842
3843 if (pcr->action != PF_CHANGE_REMOVE) {
3844 if (newrule->ifname[0]) {
3845 newrule->kif = pfi_kkif_attach(kif,
3846 newrule->ifname);
3847 kif = NULL;
3848 pfi_kkif_ref(newrule->kif);
3849 } else
3850 newrule->kif = NULL;
3851
3852 if (newrule->rtableid > 0 &&
3853 newrule->rtableid >= rt_numfibs)
3854 error = EBUSY;
3855
3856 #ifdef ALTQ
3857 /* set queue IDs */
3858 if (newrule->qname[0] != 0) {
3859 if ((newrule->qid =
3860 pf_qname2qid(newrule->qname)) == 0)
3861 error = EBUSY;
3862 else if (newrule->pqname[0] != 0) {
3863 if ((newrule->pqid =
3864 pf_qname2qid(newrule->pqname)) == 0)
3865 error = EBUSY;
3866 } else
3867 newrule->pqid = newrule->qid;
3868 }
3869 #endif /* ALTQ */
3870 if (newrule->tagname[0])
3871 if ((newrule->tag =
3872 pf_tagname2tag(newrule->tagname)) == 0)
3873 error = EBUSY;
3874 if (newrule->match_tagname[0])
3875 if ((newrule->match_tag = pf_tagname2tag(
3876 newrule->match_tagname)) == 0)
3877 error = EBUSY;
3878 if (newrule->rt && !newrule->direction)
3879 error = EINVAL;
3880 if (!newrule->log)
3881 newrule->logif = 0;
3882 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3883 error = ENOMEM;
3884 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3885 error = ENOMEM;
3886 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3887 error = EINVAL;
3888 for (int i = 0; i < 3; i++) {
3889 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
3890 if (pa->addr.type == PF_ADDR_TABLE) {
3891 pa->addr.p.tbl =
3892 pfr_attach_table(ruleset,
3893 pa->addr.v.tblname);
3894 if (pa->addr.p.tbl == NULL)
3895 error = ENOMEM;
3896 }
3897 }
3898
3899 newrule->overload_tbl = NULL;
3900 if (newrule->overload_tblname[0]) {
3901 if ((newrule->overload_tbl = pfr_attach_table(
3902 ruleset, newrule->overload_tblname)) ==
3903 NULL)
3904 error = EINVAL;
3905 else
3906 newrule->overload_tbl->pfrkt_flags |=
3907 PFR_TFLAG_ACTIVE;
3908 }
3909
3910 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
3911 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
3912 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list);
3913 if (((((newrule->action == PF_NAT) ||
3914 (newrule->action == PF_RDR) ||
3915 (newrule->action == PF_BINAT) ||
3916 (newrule->rt > PF_NOPFROUTE)) &&
3917 !newrule->anchor)) &&
3918 (TAILQ_FIRST(&newrule->rdr.list) == NULL))
3919 error = EINVAL;
3920
3921 if (error) {
3922 pf_free_rule(newrule);
3923 PF_RULES_WUNLOCK();
3924 PF_CONFIG_UNLOCK();
3925 goto fail;
3926 }
3927
3928 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
3929 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
3930 }
3931 pf_empty_kpool(&V_pf_pabuf[0]);
3932 pf_empty_kpool(&V_pf_pabuf[1]);
3933 pf_empty_kpool(&V_pf_pabuf[2]);
3934
3935 if (pcr->action == PF_CHANGE_ADD_HEAD)
3936 oldrule = TAILQ_FIRST(
3937 ruleset->rules[rs_num].active.ptr);
3938 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3939 oldrule = TAILQ_LAST(
3940 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3941 else {
3942 oldrule = TAILQ_FIRST(
3943 ruleset->rules[rs_num].active.ptr);
3944 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3945 oldrule = TAILQ_NEXT(oldrule, entries);
3946 if (oldrule == NULL) {
3947 if (newrule != NULL)
3948 pf_free_rule(newrule);
3949 PF_RULES_WUNLOCK();
3950 PF_CONFIG_UNLOCK();
3951 error = EINVAL;
3952 goto fail;
3953 }
3954 }
3955
3956 if (pcr->action == PF_CHANGE_REMOVE) {
3957 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3958 oldrule);
3959 RB_REMOVE(pf_krule_global,
3960 ruleset->rules[rs_num].active.tree, oldrule);
3961 ruleset->rules[rs_num].active.rcount--;
3962 } else {
3963 pf_hash_rule(newrule);
3964 if (RB_INSERT(pf_krule_global,
3965 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3966 pf_free_rule(newrule);
3967 PF_RULES_WUNLOCK();
3968 PF_CONFIG_UNLOCK();
3969 error = EEXIST;
3970 goto fail;
3971 }
3972
3973 if (oldrule == NULL)
3974 TAILQ_INSERT_TAIL(
3975 ruleset->rules[rs_num].active.ptr,
3976 newrule, entries);
3977 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3978 pcr->action == PF_CHANGE_ADD_BEFORE)
3979 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3980 else
3981 TAILQ_INSERT_AFTER(
3982 ruleset->rules[rs_num].active.ptr,
3983 oldrule, newrule, entries);
3984 ruleset->rules[rs_num].active.rcount++;
3985 }
3986
3987 nr = 0;
3988 TAILQ_FOREACH(oldrule,
3989 ruleset->rules[rs_num].active.ptr, entries)
3990 oldrule->nr = nr++;
3991
3992 ruleset->rules[rs_num].active.ticket++;
3993
3994 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3995 pf_remove_if_empty_kruleset(ruleset);
3996
3997 PF_RULES_WUNLOCK();
3998 PF_CONFIG_UNLOCK();
3999 break;
4000
4001 #undef ERROUT
4002 DIOCCHANGERULE_error:
4003 PF_RULES_WUNLOCK();
4004 PF_CONFIG_UNLOCK();
4005 pf_krule_free(newrule);
4006 pf_kkif_free(kif);
4007 break;
4008 }
4009
4010 case DIOCCLRSTATESNV: {
4011 error = pf_clearstates_nv((struct pfioc_nv *)addr);
4012 break;
4013 }
4014
4015 case DIOCKILLSTATESNV: {
4016 error = pf_killstates_nv((struct pfioc_nv *)addr);
4017 break;
4018 }
4019
4020 case DIOCADDSTATE: {
4021 struct pfioc_state *ps = (struct pfioc_state *)addr;
4022 struct pfsync_state_1301 *sp = &ps->state;
4023
4024 if (sp->timeout >= PFTM_MAX) {
4025 error = EINVAL;
4026 goto fail;
4027 }
4028 if (V_pfsync_state_import_ptr != NULL) {
4029 PF_RULES_RLOCK();
4030 error = V_pfsync_state_import_ptr(
4031 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
4032 PFSYNC_MSG_VERSION_1301);
4033 PF_RULES_RUNLOCK();
4034 } else
4035 error = EOPNOTSUPP;
4036 break;
4037 }
4038
4039 case DIOCGETSTATE: {
4040 struct pfioc_state *ps = (struct pfioc_state *)addr;
4041 struct pf_kstate *s;
4042
4043 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
4044 if (s == NULL) {
4045 error = ENOENT;
4046 goto fail;
4047 }
4048
4049 pfsync_state_export((union pfsync_state_union*)&ps->state,
4050 s, PFSYNC_MSG_VERSION_1301);
4051 PF_STATE_UNLOCK(s);
4052 break;
4053 }
4054
4055 case DIOCGETSTATENV: {
4056 error = pf_getstate((struct pfioc_nv *)addr);
4057 break;
4058 }
4059
4060 #ifdef COMPAT_FREEBSD14
4061 case DIOCGETSTATES: {
4062 struct pfioc_states *ps = (struct pfioc_states *)addr;
4063 struct pf_kstate *s;
4064 struct pfsync_state_1301 *pstore, *p;
4065 int i, nr;
4066 size_t slice_count = 16, count;
4067 void *out;
4068
4069 if (ps->ps_len <= 0) {
4070 nr = uma_zone_get_cur(V_pf_state_z);
4071 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4072 break;
4073 }
4074
4075 out = ps->ps_states;
4076 pstore = mallocarray(slice_count,
4077 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
4078 nr = 0;
4079
4080 for (i = 0; i <= V_pf_hashmask; i++) {
4081 struct pf_idhash *ih = &V_pf_idhash[i];
4082
4083 DIOCGETSTATES_retry:
4084 p = pstore;
4085
4086 if (LIST_EMPTY(&ih->states))
4087 continue;
4088
4089 PF_HASHROW_LOCK(ih);
4090 count = 0;
4091 LIST_FOREACH(s, &ih->states, entry) {
4092 if (s->timeout == PFTM_UNLINKED)
4093 continue;
4094 count++;
4095 }
4096
4097 if (count > slice_count) {
4098 PF_HASHROW_UNLOCK(ih);
4099 free(pstore, M_TEMP);
4100 slice_count = count * 2;
4101 pstore = mallocarray(slice_count,
4102 sizeof(struct pfsync_state_1301), M_TEMP,
4103 M_WAITOK | M_ZERO);
4104 goto DIOCGETSTATES_retry;
4105 }
4106
4107 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4108 PF_HASHROW_UNLOCK(ih);
4109 goto DIOCGETSTATES_full;
4110 }
4111
4112 LIST_FOREACH(s, &ih->states, entry) {
4113 if (s->timeout == PFTM_UNLINKED)
4114 continue;
4115
4116 pfsync_state_export((union pfsync_state_union*)p,
4117 s, PFSYNC_MSG_VERSION_1301);
4118 p++;
4119 nr++;
4120 }
4121 PF_HASHROW_UNLOCK(ih);
4122 error = copyout(pstore, out,
4123 sizeof(struct pfsync_state_1301) * count);
4124 if (error)
4125 goto fail;
4126 out = ps->ps_states + nr;
4127 }
4128 DIOCGETSTATES_full:
4129 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4130 free(pstore, M_TEMP);
4131
4132 break;
4133 }
4134
4135 case DIOCGETSTATESV2: {
4136 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
4137 struct pf_kstate *s;
4138 struct pf_state_export *pstore, *p;
4139 int i, nr;
4140 size_t slice_count = 16, count;
4141 void *out;
4142
4143 if (ps->ps_req_version > PF_STATE_VERSION) {
4144 error = ENOTSUP;
4145 goto fail;
4146 }
4147
4148 if (ps->ps_len <= 0) {
4149 nr = uma_zone_get_cur(V_pf_state_z);
4150 ps->ps_len = sizeof(struct pf_state_export) * nr;
4151 break;
4152 }
4153
4154 out = ps->ps_states;
4155 pstore = mallocarray(slice_count,
4156 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
4157 nr = 0;
4158
4159 for (i = 0; i <= V_pf_hashmask; i++) {
4160 struct pf_idhash *ih = &V_pf_idhash[i];
4161
4162 DIOCGETSTATESV2_retry:
4163 p = pstore;
4164
4165 if (LIST_EMPTY(&ih->states))
4166 continue;
4167
4168 PF_HASHROW_LOCK(ih);
4169 count = 0;
4170 LIST_FOREACH(s, &ih->states, entry) {
4171 if (s->timeout == PFTM_UNLINKED)
4172 continue;
4173 count++;
4174 }
4175
4176 if (count > slice_count) {
4177 PF_HASHROW_UNLOCK(ih);
4178 free(pstore, M_TEMP);
4179 slice_count = count * 2;
4180 pstore = mallocarray(slice_count,
4181 sizeof(struct pf_state_export), M_TEMP,
4182 M_WAITOK | M_ZERO);
4183 goto DIOCGETSTATESV2_retry;
4184 }
4185
4186 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4187 PF_HASHROW_UNLOCK(ih);
4188 goto DIOCGETSTATESV2_full;
4189 }
4190
4191 LIST_FOREACH(s, &ih->states, entry) {
4192 if (s->timeout == PFTM_UNLINKED)
4193 continue;
4194
4195 pf_state_export(p, s);
4196 p++;
4197 nr++;
4198 }
4199 PF_HASHROW_UNLOCK(ih);
4200 error = copyout(pstore, out,
4201 sizeof(struct pf_state_export) * count);
4202 if (error)
4203 goto fail;
4204 out = ps->ps_states + nr;
4205 }
4206 DIOCGETSTATESV2_full:
4207 ps->ps_len = nr * sizeof(struct pf_state_export);
4208 free(pstore, M_TEMP);
4209
4210 break;
4211 }
4212 #endif
4213 case DIOCGETSTATUSNV: {
4214 error = pf_getstatus((struct pfioc_nv *)addr);
4215 break;
4216 }
4217
4218 case DIOCSETSTATUSIF: {
4219 struct pfioc_if *pi = (struct pfioc_if *)addr;
4220
4221 if (pi->ifname[0] == 0) {
4222 bzero(V_pf_status.ifname, IFNAMSIZ);
4223 break;
4224 }
4225 PF_RULES_WLOCK();
4226 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
4227 PF_RULES_WUNLOCK();
4228 break;
4229 }
4230
4231 case DIOCCLRSTATUS: {
4232 pf_ioctl_clear_status();
4233 break;
4234 }
4235
4236 case DIOCNATLOOK: {
4237 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
4238
4239 error = pf_ioctl_natlook(pnl);
4240 break;
4241 }
4242
4243 case DIOCSETTIMEOUT: {
4244 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4245
4246 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4247 &pt->seconds);
4248 break;
4249 }
4250
4251 case DIOCGETTIMEOUT: {
4252 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4253
4254 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4255 break;
4256 }
4257
4258 case DIOCGETLIMIT: {
4259 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4260
4261 error = pf_ioctl_get_limit(pl->index, &pl->limit);
4262 break;
4263 }
4264
4265 case DIOCSETLIMIT: {
4266 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4267 unsigned int old_limit;
4268
4269 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4270 pl->limit = old_limit;
4271 break;
4272 }
4273
4274 case DIOCSETDEBUG: {
4275 u_int32_t *level = (u_int32_t *)addr;
4276
4277 PF_RULES_WLOCK();
4278 V_pf_status.debug = *level;
4279 PF_RULES_WUNLOCK();
4280 break;
4281 }
4282
4283 case DIOCCLRRULECTRS: {
4284 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4285 struct pf_kruleset *ruleset = &pf_main_ruleset;
4286 struct pf_krule *rule;
4287
4288 PF_RULES_WLOCK();
4289 TAILQ_FOREACH(rule,
4290 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4291 pf_counter_u64_zero(&rule->evaluations);
4292 for (int i = 0; i < 2; i++) {
4293 pf_counter_u64_zero(&rule->packets[i]);
4294 pf_counter_u64_zero(&rule->bytes[i]);
4295 }
4296 }
4297 PF_RULES_WUNLOCK();
4298 break;
4299 }
4300
4301 case DIOCGIFSPEEDV0:
4302 case DIOCGIFSPEEDV1: {
4303 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
4304 struct pf_ifspeed_v1 ps;
4305 struct ifnet *ifp;
4306
4307 if (psp->ifname[0] == '\0') {
4308 error = EINVAL;
4309 goto fail;
4310 }
4311
4312 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4313 if (error != 0)
4314 goto fail;
4315 ifp = ifunit(ps.ifname);
4316 if (ifp != NULL) {
4317 psp->baudrate32 =
4318 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4319 if (cmd == DIOCGIFSPEEDV1)
4320 psp->baudrate = ifp->if_baudrate;
4321 } else {
4322 error = EINVAL;
4323 }
4324 break;
4325 }
4326
4327 #ifdef ALTQ
4328 case DIOCSTARTALTQ: {
4329 struct pf_altq *altq;
4330
4331 PF_RULES_WLOCK();
4332 /* enable all altq interfaces on active list */
4333 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4334 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4335 error = pf_enable_altq(altq);
4336 if (error != 0)
4337 break;
4338 }
4339 }
4340 if (error == 0)
4341 V_pf_altq_running = 1;
4342 PF_RULES_WUNLOCK();
4343 DPFPRINTF(PF_DEBUG_MISC, "altq: started");
4344 break;
4345 }
4346
4347 case DIOCSTOPALTQ: {
4348 struct pf_altq *altq;
4349
4350 PF_RULES_WLOCK();
4351 /* disable all altq interfaces on active list */
4352 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4353 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4354 error = pf_disable_altq(altq);
4355 if (error != 0)
4356 break;
4357 }
4358 }
4359 if (error == 0)
4360 V_pf_altq_running = 0;
4361 PF_RULES_WUNLOCK();
4362 DPFPRINTF(PF_DEBUG_MISC, "altq: stopped");
4363 break;
4364 }
4365
4366 case DIOCADDALTQV0:
4367 case DIOCADDALTQV1: {
4368 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4369 struct pf_altq *altq, *a;
4370 struct ifnet *ifp;
4371
4372 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4373 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4374 if (error)
4375 goto fail;
4376 altq->local_flags = 0;
4377
4378 PF_RULES_WLOCK();
4379 if (pa->ticket != V_ticket_altqs_inactive) {
4380 PF_RULES_WUNLOCK();
4381 free(altq, M_PFALTQ);
4382 error = EBUSY;
4383 goto fail;
4384 }
4385
4386 /*
4387 * if this is for a queue, find the discipline and
4388 * copy the necessary fields
4389 */
4390 if (altq->qname[0] != 0) {
4391 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4392 PF_RULES_WUNLOCK();
4393 error = EBUSY;
4394 free(altq, M_PFALTQ);
4395 goto fail;
4396 }
4397 altq->altq_disc = NULL;
4398 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4399 if (strncmp(a->ifname, altq->ifname,
4400 IFNAMSIZ) == 0) {
4401 altq->altq_disc = a->altq_disc;
4402 break;
4403 }
4404 }
4405 }
4406
4407 if ((ifp = ifunit(altq->ifname)) == NULL)
4408 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4409 else
4410 error = altq_add(ifp, altq);
4411
4412 if (error) {
4413 PF_RULES_WUNLOCK();
4414 free(altq, M_PFALTQ);
4415 goto fail;
4416 }
4417
4418 if (altq->qname[0] != 0)
4419 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4420 else
4421 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4422 /* version error check done on import above */
4423 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4424 PF_RULES_WUNLOCK();
4425 break;
4426 }
4427
4428 case DIOCGETALTQSV0:
4429 case DIOCGETALTQSV1: {
4430 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4431 struct pf_altq *altq;
4432
4433 PF_RULES_RLOCK();
4434 pa->nr = 0;
4435 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4436 pa->nr++;
4437 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4438 pa->nr++;
4439 pa->ticket = V_ticket_altqs_active;
4440 PF_RULES_RUNLOCK();
4441 break;
4442 }
4443
4444 case DIOCGETALTQV0:
4445 case DIOCGETALTQV1: {
4446 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4447 struct pf_altq *altq;
4448
4449 PF_RULES_RLOCK();
4450 if (pa->ticket != V_ticket_altqs_active) {
4451 PF_RULES_RUNLOCK();
4452 error = EBUSY;
4453 goto fail;
4454 }
4455 altq = pf_altq_get_nth_active(pa->nr);
4456 if (altq == NULL) {
4457 PF_RULES_RUNLOCK();
4458 error = EBUSY;
4459 goto fail;
4460 }
4461 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4462 PF_RULES_RUNLOCK();
4463 break;
4464 }
4465
4466 case DIOCCHANGEALTQV0:
4467 case DIOCCHANGEALTQV1:
4468 /* CHANGEALTQ not supported yet! */
4469 error = ENODEV;
4470 break;
4471
4472 case DIOCGETQSTATSV0:
4473 case DIOCGETQSTATSV1: {
4474 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
4475 struct pf_altq *altq;
4476 int nbytes;
4477 u_int32_t version;
4478
4479 PF_RULES_RLOCK();
4480 if (pq->ticket != V_ticket_altqs_active) {
4481 PF_RULES_RUNLOCK();
4482 error = EBUSY;
4483 goto fail;
4484 }
4485 nbytes = pq->nbytes;
4486 altq = pf_altq_get_nth_active(pq->nr);
4487 if (altq == NULL) {
4488 PF_RULES_RUNLOCK();
4489 error = EBUSY;
4490 goto fail;
4491 }
4492
4493 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4494 PF_RULES_RUNLOCK();
4495 error = ENXIO;
4496 goto fail;
4497 }
4498 PF_RULES_RUNLOCK();
4499 if (cmd == DIOCGETQSTATSV0)
4500 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
4501 else
4502 version = pq->version;
4503 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4504 if (error == 0) {
4505 pq->scheduler = altq->scheduler;
4506 pq->nbytes = nbytes;
4507 }
4508 break;
4509 }
4510 #endif /* ALTQ */
4511
4512 case DIOCBEGINADDRS: {
4513 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4514
4515 error = pf_ioctl_begin_addrs(&pp->ticket);
4516 break;
4517 }
4518
4519 case DIOCADDADDR: {
4520 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4521 struct pf_nl_pooladdr npp = {};
4522
4523 npp.which = PF_RDR;
4524 memcpy(&npp, pp, sizeof(*pp));
4525 error = pf_ioctl_add_addr(&npp);
4526 break;
4527 }
4528
4529 case DIOCGETADDRS: {
4530 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4531 struct pf_nl_pooladdr npp = {};
4532
4533 npp.which = PF_RDR;
4534 memcpy(&npp, pp, sizeof(*pp));
4535 error = pf_ioctl_get_addrs(&npp);
4536 memcpy(pp, &npp, sizeof(*pp));
4537
4538 break;
4539 }
4540
4541 case DIOCGETADDR: {
4542 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4543 struct pf_nl_pooladdr npp = {};
4544
4545 npp.which = PF_RDR;
4546 memcpy(&npp, pp, sizeof(*pp));
4547 error = pf_ioctl_get_addr(&npp);
4548 memcpy(pp, &npp, sizeof(*pp));
4549
4550 break;
4551 }
4552
4553 case DIOCCHANGEADDR: {
4554 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
4555 struct pf_kpool *pool;
4556 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
4557 struct pf_kruleset *ruleset;
4558 struct pfi_kkif *kif = NULL;
4559
4560 pca->anchor[sizeof(pca->anchor) - 1] = '\0';
4561
4562 if (pca->action < PF_CHANGE_ADD_HEAD ||
4563 pca->action > PF_CHANGE_REMOVE) {
4564 error = EINVAL;
4565 goto fail;
4566 }
4567 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4568 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4569 pca->addr.addr.type != PF_ADDR_TABLE) {
4570 error = EINVAL;
4571 goto fail;
4572 }
4573 if (pca->addr.addr.p.dyn != NULL) {
4574 error = EINVAL;
4575 goto fail;
4576 }
4577
4578 if (pca->action != PF_CHANGE_REMOVE) {
4579 #ifndef INET
4580 if (pca->af == AF_INET) {
4581 error = EAFNOSUPPORT;
4582 goto fail;
4583 }
4584 #endif /* INET */
4585 #ifndef INET6
4586 if (pca->af == AF_INET6) {
4587 error = EAFNOSUPPORT;
4588 goto fail;
4589 }
4590 #endif /* INET6 */
4591 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4592 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4593 if (newpa->ifname[0])
4594 kif = pf_kkif_create(M_WAITOK);
4595 newpa->kif = NULL;
4596 }
4597 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4598 PF_RULES_WLOCK();
4599 ruleset = pf_find_kruleset(pca->anchor);
4600 if (ruleset == NULL)
4601 ERROUT(EBUSY);
4602
4603 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4604 pca->r_num, pca->r_last, 1, 1, PF_RDR);
4605 if (pool == NULL)
4606 ERROUT(EBUSY);
4607
4608 if (pca->action != PF_CHANGE_REMOVE) {
4609 if (newpa->ifname[0]) {
4610 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4611 pfi_kkif_ref(newpa->kif);
4612 kif = NULL;
4613 }
4614
4615 switch (newpa->addr.type) {
4616 case PF_ADDR_DYNIFTL:
4617 error = pfi_dynaddr_setup(&newpa->addr,
4618 pca->af);
4619 break;
4620 case PF_ADDR_TABLE:
4621 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4622 newpa->addr.v.tblname);
4623 if (newpa->addr.p.tbl == NULL)
4624 error = ENOMEM;
4625 break;
4626 }
4627 if (error)
4628 goto DIOCCHANGEADDR_error;
4629 }
4630
4631 switch (pca->action) {
4632 case PF_CHANGE_ADD_HEAD:
4633 oldpa = TAILQ_FIRST(&pool->list);
4634 break;
4635 case PF_CHANGE_ADD_TAIL:
4636 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4637 break;
4638 default:
4639 oldpa = TAILQ_FIRST(&pool->list);
4640 for (int i = 0; oldpa && i < pca->nr; i++)
4641 oldpa = TAILQ_NEXT(oldpa, entries);
4642
4643 if (oldpa == NULL)
4644 ERROUT(EINVAL);
4645 }
4646
4647 if (pca->action == PF_CHANGE_REMOVE) {
4648 TAILQ_REMOVE(&pool->list, oldpa, entries);
4649 switch (oldpa->addr.type) {
4650 case PF_ADDR_DYNIFTL:
4651 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4652 break;
4653 case PF_ADDR_TABLE:
4654 pfr_detach_table(oldpa->addr.p.tbl);
4655 break;
4656 }
4657 if (oldpa->kif)
4658 pfi_kkif_unref(oldpa->kif);
4659 free(oldpa, M_PFRULE);
4660 } else {
4661 if (oldpa == NULL)
4662 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4663 else if (pca->action == PF_CHANGE_ADD_HEAD ||
4664 pca->action == PF_CHANGE_ADD_BEFORE)
4665 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4666 else
4667 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4668 newpa, entries);
4669 }
4670
4671 pool->cur = TAILQ_FIRST(&pool->list);
4672 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4673 PF_RULES_WUNLOCK();
4674 break;
4675
4676 #undef ERROUT
4677 DIOCCHANGEADDR_error:
4678 if (newpa != NULL) {
4679 if (newpa->kif)
4680 pfi_kkif_unref(newpa->kif);
4681 free(newpa, M_PFRULE);
4682 }
4683 PF_RULES_WUNLOCK();
4684 pf_kkif_free(kif);
4685 break;
4686 }
4687
4688 case DIOCGETRULESETS: {
4689 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4690
4691 pr->path[sizeof(pr->path) - 1] = '\0';
4692
4693 error = pf_ioctl_get_rulesets(pr);
4694 break;
4695 }
4696
4697 case DIOCGETRULESET: {
4698 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4699
4700 pr->path[sizeof(pr->path) - 1] = '\0';
4701
4702 error = pf_ioctl_get_ruleset(pr);
4703 break;
4704 }
4705
4706 case DIOCRCLRTABLES: {
4707 struct pfioc_table *io = (struct pfioc_table *)addr;
4708
4709 if (io->pfrio_esize != 0) {
4710 error = ENODEV;
4711 goto fail;
4712 }
4713 PF_RULES_WLOCK();
4714 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4715 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4716 PF_RULES_WUNLOCK();
4717 break;
4718 }
4719
4720 case DIOCRADDTABLES: {
4721 struct pfioc_table *io = (struct pfioc_table *)addr;
4722 struct pfr_table *pfrts;
4723 size_t totlen;
4724
4725 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4726 error = ENODEV;
4727 goto fail;
4728 }
4729
4730 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4731 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4732 error = ENOMEM;
4733 goto fail;
4734 }
4735
4736 totlen = io->pfrio_size * sizeof(struct pfr_table);
4737 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4738 M_TEMP, M_WAITOK);
4739 error = copyin(io->pfrio_buffer, pfrts, totlen);
4740 if (error) {
4741 free(pfrts, M_TEMP);
4742 goto fail;
4743 }
4744 PF_RULES_WLOCK();
4745 error = pfr_add_tables(pfrts, io->pfrio_size,
4746 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4747 PF_RULES_WUNLOCK();
4748 free(pfrts, M_TEMP);
4749 break;
4750 }
4751
4752 case DIOCRDELTABLES: {
4753 struct pfioc_table *io = (struct pfioc_table *)addr;
4754 struct pfr_table *pfrts;
4755 size_t totlen;
4756
4757 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4758 error = ENODEV;
4759 goto fail;
4760 }
4761
4762 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4763 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4764 error = ENOMEM;
4765 goto fail;
4766 }
4767
4768 totlen = io->pfrio_size * sizeof(struct pfr_table);
4769 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4770 M_TEMP, M_WAITOK);
4771 error = copyin(io->pfrio_buffer, pfrts, totlen);
4772 if (error) {
4773 free(pfrts, M_TEMP);
4774 goto fail;
4775 }
4776 PF_RULES_WLOCK();
4777 error = pfr_del_tables(pfrts, io->pfrio_size,
4778 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4779 PF_RULES_WUNLOCK();
4780 free(pfrts, M_TEMP);
4781 break;
4782 }
4783
4784 case DIOCRGETTABLES: {
4785 struct pfioc_table *io = (struct pfioc_table *)addr;
4786 struct pfr_table *pfrts;
4787 size_t totlen;
4788 int n;
4789
4790 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4791 error = ENODEV;
4792 goto fail;
4793 }
4794 PF_RULES_RLOCK();
4795 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4796 if (n < 0) {
4797 PF_RULES_RUNLOCK();
4798 error = EINVAL;
4799 goto fail;
4800 }
4801 io->pfrio_size = min(io->pfrio_size, n);
4802
4803 totlen = io->pfrio_size * sizeof(struct pfr_table);
4804
4805 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4806 M_TEMP, M_NOWAIT | M_ZERO);
4807 if (pfrts == NULL) {
4808 error = ENOMEM;
4809 PF_RULES_RUNLOCK();
4810 goto fail;
4811 }
4812 error = pfr_get_tables(&io->pfrio_table, pfrts,
4813 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4814 PF_RULES_RUNLOCK();
4815 if (error == 0)
4816 error = copyout(pfrts, io->pfrio_buffer, totlen);
4817 free(pfrts, M_TEMP);
4818 break;
4819 }
4820
4821 case DIOCRGETTSTATS: {
4822 struct pfioc_table *io = (struct pfioc_table *)addr;
4823 struct pfr_tstats *pfrtstats;
4824 size_t totlen;
4825 int n;
4826
4827 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4828 error = ENODEV;
4829 goto fail;
4830 }
4831 PF_TABLE_STATS_LOCK();
4832 PF_RULES_RLOCK();
4833 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4834 if (n < 0) {
4835 PF_RULES_RUNLOCK();
4836 PF_TABLE_STATS_UNLOCK();
4837 error = EINVAL;
4838 goto fail;
4839 }
4840 io->pfrio_size = min(io->pfrio_size, n);
4841
4842 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4843 pfrtstats = mallocarray(io->pfrio_size,
4844 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4845 if (pfrtstats == NULL) {
4846 error = ENOMEM;
4847 PF_RULES_RUNLOCK();
4848 PF_TABLE_STATS_UNLOCK();
4849 goto fail;
4850 }
4851 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4852 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4853 PF_RULES_RUNLOCK();
4854 PF_TABLE_STATS_UNLOCK();
4855 if (error == 0)
4856 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4857 free(pfrtstats, M_TEMP);
4858 break;
4859 }
4860
4861 case DIOCRCLRTSTATS: {
4862 struct pfioc_table *io = (struct pfioc_table *)addr;
4863 struct pfr_table *pfrts;
4864 size_t totlen;
4865
4866 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4867 error = ENODEV;
4868 goto fail;
4869 }
4870
4871 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4872 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4873 /* We used to count tables and use the minimum required
4874 * size, so we didn't fail on overly large requests.
4875 * Keep doing so. */
4876 io->pfrio_size = pf_ioctl_maxcount;
4877 goto fail;
4878 }
4879
4880 totlen = io->pfrio_size * sizeof(struct pfr_table);
4881 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4882 M_TEMP, M_WAITOK);
4883 error = copyin(io->pfrio_buffer, pfrts, totlen);
4884 if (error) {
4885 free(pfrts, M_TEMP);
4886 goto fail;
4887 }
4888
4889 PF_TABLE_STATS_LOCK();
4890 PF_RULES_RLOCK();
4891 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4892 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4893 PF_RULES_RUNLOCK();
4894 PF_TABLE_STATS_UNLOCK();
4895 free(pfrts, M_TEMP);
4896 break;
4897 }
4898
4899 case DIOCRSETTFLAGS: {
4900 struct pfioc_table *io = (struct pfioc_table *)addr;
4901 struct pfr_table *pfrts;
4902 size_t totlen;
4903 int n;
4904
4905 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4906 error = ENODEV;
4907 goto fail;
4908 }
4909
4910 PF_RULES_RLOCK();
4911 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4912 if (n < 0) {
4913 PF_RULES_RUNLOCK();
4914 error = EINVAL;
4915 goto fail;
4916 }
4917
4918 io->pfrio_size = min(io->pfrio_size, n);
4919 PF_RULES_RUNLOCK();
4920
4921 totlen = io->pfrio_size * sizeof(struct pfr_table);
4922 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4923 M_TEMP, M_WAITOK);
4924 error = copyin(io->pfrio_buffer, pfrts, totlen);
4925 if (error) {
4926 free(pfrts, M_TEMP);
4927 goto fail;
4928 }
4929 PF_RULES_WLOCK();
4930 error = pfr_set_tflags(pfrts, io->pfrio_size,
4931 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4932 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4933 PF_RULES_WUNLOCK();
4934 free(pfrts, M_TEMP);
4935 break;
4936 }
4937
4938 case DIOCRCLRADDRS: {
4939 struct pfioc_table *io = (struct pfioc_table *)addr;
4940
4941 if (io->pfrio_esize != 0) {
4942 error = ENODEV;
4943 goto fail;
4944 }
4945 PF_RULES_WLOCK();
4946 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4947 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4948 PF_RULES_WUNLOCK();
4949 break;
4950 }
4951
4952 case DIOCRADDADDRS: {
4953 struct pfioc_table *io = (struct pfioc_table *)addr;
4954 struct pfr_addr *pfras;
4955 size_t totlen;
4956
4957 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4958 error = ENODEV;
4959 goto fail;
4960 }
4961 if (io->pfrio_size < 0 ||
4962 io->pfrio_size > pf_ioctl_maxcount ||
4963 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4964 error = EINVAL;
4965 goto fail;
4966 }
4967 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4968 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4969 M_TEMP, M_WAITOK);
4970 error = copyin(io->pfrio_buffer, pfras, totlen);
4971 if (error) {
4972 free(pfras, M_TEMP);
4973 goto fail;
4974 }
4975 PF_RULES_WLOCK();
4976 error = pfr_add_addrs(&io->pfrio_table, pfras,
4977 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4978 PFR_FLAG_USERIOCTL);
4979 PF_RULES_WUNLOCK();
4980 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4981 error = copyout(pfras, io->pfrio_buffer, totlen);
4982 free(pfras, M_TEMP);
4983 break;
4984 }
4985
4986 case DIOCRDELADDRS: {
4987 struct pfioc_table *io = (struct pfioc_table *)addr;
4988 struct pfr_addr *pfras;
4989 size_t totlen;
4990
4991 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4992 error = ENODEV;
4993 goto fail;
4994 }
4995 if (io->pfrio_size < 0 ||
4996 io->pfrio_size > pf_ioctl_maxcount ||
4997 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4998 error = EINVAL;
4999 goto fail;
5000 }
5001 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5002 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5003 M_TEMP, M_WAITOK);
5004 error = copyin(io->pfrio_buffer, pfras, totlen);
5005 if (error) {
5006 free(pfras, M_TEMP);
5007 goto fail;
5008 }
5009 PF_RULES_WLOCK();
5010 error = pfr_del_addrs(&io->pfrio_table, pfras,
5011 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
5012 PFR_FLAG_USERIOCTL);
5013 PF_RULES_WUNLOCK();
5014 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5015 error = copyout(pfras, io->pfrio_buffer, totlen);
5016 free(pfras, M_TEMP);
5017 break;
5018 }
5019
5020 case DIOCRSETADDRS: {
5021 struct pfioc_table *io = (struct pfioc_table *)addr;
5022 struct pfr_addr *pfras;
5023 size_t totlen, count;
5024
5025 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5026 error = ENODEV;
5027 goto fail;
5028 }
5029 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
5030 error = EINVAL;
5031 goto fail;
5032 }
5033 count = max(io->pfrio_size, io->pfrio_size2);
5034 if (count > pf_ioctl_maxcount ||
5035 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
5036 error = EINVAL;
5037 goto fail;
5038 }
5039 totlen = count * sizeof(struct pfr_addr);
5040 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
5041 M_WAITOK);
5042 error = copyin(io->pfrio_buffer, pfras, totlen);
5043 if (error) {
5044 free(pfras, M_TEMP);
5045 goto fail;
5046 }
5047 PF_RULES_WLOCK();
5048 error = pfr_set_addrs(&io->pfrio_table, pfras,
5049 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
5050 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
5051 PFR_FLAG_USERIOCTL, 0);
5052 PF_RULES_WUNLOCK();
5053 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5054 error = copyout(pfras, io->pfrio_buffer, totlen);
5055 free(pfras, M_TEMP);
5056 break;
5057 }
5058
5059 case DIOCRGETADDRS: {
5060 struct pfioc_table *io = (struct pfioc_table *)addr;
5061 struct pfr_addr *pfras;
5062 size_t totlen;
5063
5064 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5065 error = ENODEV;
5066 goto fail;
5067 }
5068 if (io->pfrio_size < 0 ||
5069 io->pfrio_size > pf_ioctl_maxcount ||
5070 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5071 error = EINVAL;
5072 goto fail;
5073 }
5074 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5075 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5076 M_TEMP, M_WAITOK | M_ZERO);
5077 PF_RULES_RLOCK();
5078 error = pfr_get_addrs(&io->pfrio_table, pfras,
5079 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5080 PF_RULES_RUNLOCK();
5081 if (error == 0)
5082 error = copyout(pfras, io->pfrio_buffer, totlen);
5083 free(pfras, M_TEMP);
5084 break;
5085 }
5086
5087 case DIOCRGETASTATS: {
5088 struct pfioc_table *io = (struct pfioc_table *)addr;
5089 struct pfr_astats *pfrastats;
5090 size_t totlen;
5091
5092 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
5093 error = ENODEV;
5094 goto fail;
5095 }
5096 if (io->pfrio_size < 0 ||
5097 io->pfrio_size > pf_ioctl_maxcount ||
5098 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5099 error = EINVAL;
5100 goto fail;
5101 }
5102 totlen = io->pfrio_size * sizeof(struct pfr_astats);
5103 pfrastats = mallocarray(io->pfrio_size,
5104 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
5105 PF_RULES_RLOCK();
5106 error = pfr_get_astats(&io->pfrio_table, pfrastats,
5107 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5108 PF_RULES_RUNLOCK();
5109 if (error == 0)
5110 error = copyout(pfrastats, io->pfrio_buffer, totlen);
5111 free(pfrastats, M_TEMP);
5112 break;
5113 }
5114
5115 case DIOCRCLRASTATS: {
5116 struct pfioc_table *io = (struct pfioc_table *)addr;
5117 struct pfr_addr *pfras;
5118 size_t totlen;
5119
5120 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5121 error = ENODEV;
5122 goto fail;
5123 }
5124 if (io->pfrio_size < 0 ||
5125 io->pfrio_size > pf_ioctl_maxcount ||
5126 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5127 error = EINVAL;
5128 goto fail;
5129 }
5130 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5131 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5132 M_TEMP, M_WAITOK);
5133 error = copyin(io->pfrio_buffer, pfras, totlen);
5134 if (error) {
5135 free(pfras, M_TEMP);
5136 goto fail;
5137 }
5138 PF_RULES_WLOCK();
5139 error = pfr_clr_astats(&io->pfrio_table, pfras,
5140 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5141 PFR_FLAG_USERIOCTL);
5142 PF_RULES_WUNLOCK();
5143 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5144 error = copyout(pfras, io->pfrio_buffer, totlen);
5145 free(pfras, M_TEMP);
5146 break;
5147 }
5148
5149 case DIOCRTSTADDRS: {
5150 struct pfioc_table *io = (struct pfioc_table *)addr;
5151 struct pfr_addr *pfras;
5152 size_t totlen;
5153
5154 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5155 error = ENODEV;
5156 goto fail;
5157 }
5158 if (io->pfrio_size < 0 ||
5159 io->pfrio_size > pf_ioctl_maxcount ||
5160 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5161 error = EINVAL;
5162 goto fail;
5163 }
5164 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5165 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5166 M_TEMP, M_WAITOK);
5167 error = copyin(io->pfrio_buffer, pfras, totlen);
5168 if (error) {
5169 free(pfras, M_TEMP);
5170 goto fail;
5171 }
5172 PF_RULES_RLOCK();
5173 error = pfr_tst_addrs(&io->pfrio_table, pfras,
5174 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5175 PFR_FLAG_USERIOCTL);
5176 PF_RULES_RUNLOCK();
5177 if (error == 0)
5178 error = copyout(pfras, io->pfrio_buffer, totlen);
5179 free(pfras, M_TEMP);
5180 break;
5181 }
5182
5183 case DIOCRINADEFINE: {
5184 struct pfioc_table *io = (struct pfioc_table *)addr;
5185 struct pfr_addr *pfras;
5186 size_t totlen;
5187
5188 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5189 error = ENODEV;
5190 goto fail;
5191 }
5192 if (io->pfrio_size < 0 ||
5193 io->pfrio_size > pf_ioctl_maxcount ||
5194 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5195 error = EINVAL;
5196 goto fail;
5197 }
5198 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5199 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5200 M_TEMP, M_WAITOK);
5201 error = copyin(io->pfrio_buffer, pfras, totlen);
5202 if (error) {
5203 free(pfras, M_TEMP);
5204 goto fail;
5205 }
5206 PF_RULES_WLOCK();
5207 error = pfr_ina_define(&io->pfrio_table, pfras,
5208 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5209 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5210 PF_RULES_WUNLOCK();
5211 free(pfras, M_TEMP);
5212 break;
5213 }
5214
5215 case DIOCOSFPADD: {
5216 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5217 PF_RULES_WLOCK();
5218 error = pf_osfp_add(io);
5219 PF_RULES_WUNLOCK();
5220 break;
5221 }
5222
5223 case DIOCOSFPGET: {
5224 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5225 PF_RULES_RLOCK();
5226 error = pf_osfp_get(io);
5227 PF_RULES_RUNLOCK();
5228 break;
5229 }
5230
5231 case DIOCXBEGIN: {
5232 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5233 struct pfioc_trans_e *ioes, *ioe;
5234 size_t totlen;
5235 int i;
5236
5237 if (io->esize != sizeof(*ioe)) {
5238 error = ENODEV;
5239 goto fail;
5240 }
5241 if (io->size < 0 ||
5242 io->size > pf_ioctl_maxcount ||
5243 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5244 error = EINVAL;
5245 goto fail;
5246 }
5247 totlen = sizeof(struct pfioc_trans_e) * io->size;
5248 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5249 M_TEMP, M_WAITOK);
5250 error = copyin(io->array, ioes, totlen);
5251 if (error) {
5252 free(ioes, M_TEMP);
5253 goto fail;
5254 }
5255 PF_RULES_WLOCK();
5256 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5257 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5258 switch (ioe->rs_num) {
5259 case PF_RULESET_ETH:
5260 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5261 PF_RULES_WUNLOCK();
5262 free(ioes, M_TEMP);
5263 goto fail;
5264 }
5265 break;
5266 #ifdef ALTQ
5267 case PF_RULESET_ALTQ:
5268 if (ioe->anchor[0]) {
5269 PF_RULES_WUNLOCK();
5270 free(ioes, M_TEMP);
5271 error = EINVAL;
5272 goto fail;
5273 }
5274 if ((error = pf_begin_altq(&ioe->ticket))) {
5275 PF_RULES_WUNLOCK();
5276 free(ioes, M_TEMP);
5277 goto fail;
5278 }
5279 break;
5280 #endif /* ALTQ */
5281 case PF_RULESET_TABLE:
5282 {
5283 struct pfr_table table;
5284
5285 bzero(&table, sizeof(table));
5286 strlcpy(table.pfrt_anchor, ioe->anchor,
5287 sizeof(table.pfrt_anchor));
5288 if ((error = pfr_ina_begin(&table,
5289 &ioe->ticket, NULL, 0))) {
5290 PF_RULES_WUNLOCK();
5291 free(ioes, M_TEMP);
5292 goto fail;
5293 }
5294 break;
5295 }
5296 default:
5297 if ((error = pf_begin_rules(&ioe->ticket,
5298 ioe->rs_num, ioe->anchor))) {
5299 PF_RULES_WUNLOCK();
5300 free(ioes, M_TEMP);
5301 goto fail;
5302 }
5303 break;
5304 }
5305 }
5306 PF_RULES_WUNLOCK();
5307 error = copyout(ioes, io->array, totlen);
5308 free(ioes, M_TEMP);
5309 break;
5310 }
5311
5312 case DIOCXROLLBACK: {
5313 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5314 struct pfioc_trans_e *ioe, *ioes;
5315 size_t totlen;
5316 int i;
5317
5318 if (io->esize != sizeof(*ioe)) {
5319 error = ENODEV;
5320 goto fail;
5321 }
5322 if (io->size < 0 ||
5323 io->size > pf_ioctl_maxcount ||
5324 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5325 error = EINVAL;
5326 goto fail;
5327 }
5328 totlen = sizeof(struct pfioc_trans_e) * io->size;
5329 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5330 M_TEMP, M_WAITOK);
5331 error = copyin(io->array, ioes, totlen);
5332 if (error) {
5333 free(ioes, M_TEMP);
5334 goto fail;
5335 }
5336 PF_RULES_WLOCK();
5337 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5338 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5339 switch (ioe->rs_num) {
5340 case PF_RULESET_ETH:
5341 if ((error = pf_rollback_eth(ioe->ticket,
5342 ioe->anchor))) {
5343 PF_RULES_WUNLOCK();
5344 free(ioes, M_TEMP);
5345 goto fail; /* really bad */
5346 }
5347 break;
5348 #ifdef ALTQ
5349 case PF_RULESET_ALTQ:
5350 if (ioe->anchor[0]) {
5351 PF_RULES_WUNLOCK();
5352 free(ioes, M_TEMP);
5353 error = EINVAL;
5354 goto fail;
5355 }
5356 if ((error = pf_rollback_altq(ioe->ticket))) {
5357 PF_RULES_WUNLOCK();
5358 free(ioes, M_TEMP);
5359 goto fail; /* really bad */
5360 }
5361 break;
5362 #endif /* ALTQ */
5363 case PF_RULESET_TABLE:
5364 {
5365 struct pfr_table table;
5366
5367 bzero(&table, sizeof(table));
5368 strlcpy(table.pfrt_anchor, ioe->anchor,
5369 sizeof(table.pfrt_anchor));
5370 if ((error = pfr_ina_rollback(&table,
5371 ioe->ticket, NULL, 0))) {
5372 PF_RULES_WUNLOCK();
5373 free(ioes, M_TEMP);
5374 goto fail; /* really bad */
5375 }
5376 break;
5377 }
5378 default:
5379 if ((error = pf_rollback_rules(ioe->ticket,
5380 ioe->rs_num, ioe->anchor))) {
5381 PF_RULES_WUNLOCK();
5382 free(ioes, M_TEMP);
5383 goto fail; /* really bad */
5384 }
5385 break;
5386 }
5387 }
5388 PF_RULES_WUNLOCK();
5389 free(ioes, M_TEMP);
5390 break;
5391 }
5392
5393 case DIOCXCOMMIT: {
5394 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5395 struct pfioc_trans_e *ioe, *ioes;
5396 struct pf_kruleset *rs;
5397 struct pf_keth_ruleset *ers;
5398 size_t totlen;
5399 int i;
5400
5401 if (io->esize != sizeof(*ioe)) {
5402 error = ENODEV;
5403 goto fail;
5404 }
5405
5406 if (io->size < 0 ||
5407 io->size > pf_ioctl_maxcount ||
5408 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5409 error = EINVAL;
5410 goto fail;
5411 }
5412
5413 totlen = sizeof(struct pfioc_trans_e) * io->size;
5414 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5415 M_TEMP, M_WAITOK);
5416 error = copyin(io->array, ioes, totlen);
5417 if (error) {
5418 free(ioes, M_TEMP);
5419 goto fail;
5420 }
5421 PF_RULES_WLOCK();
5422 /* First makes sure everything will succeed. */
5423 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5424 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5425 switch (ioe->rs_num) {
5426 case PF_RULESET_ETH:
5427 ers = pf_find_keth_ruleset(ioe->anchor);
5428 if (ers == NULL || ioe->ticket == 0 ||
5429 ioe->ticket != ers->inactive.ticket) {
5430 PF_RULES_WUNLOCK();
5431 free(ioes, M_TEMP);
5432 error = EINVAL;
5433 goto fail;
5434 }
5435 break;
5436 #ifdef ALTQ
5437 case PF_RULESET_ALTQ:
5438 if (ioe->anchor[0]) {
5439 PF_RULES_WUNLOCK();
5440 free(ioes, M_TEMP);
5441 error = EINVAL;
5442 goto fail;
5443 }
5444 if (!V_altqs_inactive_open || ioe->ticket !=
5445 V_ticket_altqs_inactive) {
5446 PF_RULES_WUNLOCK();
5447 free(ioes, M_TEMP);
5448 error = EBUSY;
5449 goto fail;
5450 }
5451 break;
5452 #endif /* ALTQ */
5453 case PF_RULESET_TABLE:
5454 rs = pf_find_kruleset(ioe->anchor);
5455 if (rs == NULL || !rs->topen || ioe->ticket !=
5456 rs->tticket) {
5457 PF_RULES_WUNLOCK();
5458 free(ioes, M_TEMP);
5459 error = EBUSY;
5460 goto fail;
5461 }
5462 break;
5463 default:
5464 if (ioe->rs_num < 0 || ioe->rs_num >=
5465 PF_RULESET_MAX) {
5466 PF_RULES_WUNLOCK();
5467 free(ioes, M_TEMP);
5468 error = EINVAL;
5469 goto fail;
5470 }
5471 rs = pf_find_kruleset(ioe->anchor);
5472 if (rs == NULL ||
5473 !rs->rules[ioe->rs_num].inactive.open ||
5474 rs->rules[ioe->rs_num].inactive.ticket !=
5475 ioe->ticket) {
5476 PF_RULES_WUNLOCK();
5477 free(ioes, M_TEMP);
5478 error = EBUSY;
5479 goto fail;
5480 }
5481 break;
5482 }
5483 }
5484 /* Now do the commit - no errors should happen here. */
5485 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5486 switch (ioe->rs_num) {
5487 case PF_RULESET_ETH:
5488 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5489 PF_RULES_WUNLOCK();
5490 free(ioes, M_TEMP);
5491 goto fail; /* really bad */
5492 }
5493 break;
5494 #ifdef ALTQ
5495 case PF_RULESET_ALTQ:
5496 if ((error = pf_commit_altq(ioe->ticket))) {
5497 PF_RULES_WUNLOCK();
5498 free(ioes, M_TEMP);
5499 goto fail; /* really bad */
5500 }
5501 break;
5502 #endif /* ALTQ */
5503 case PF_RULESET_TABLE:
5504 {
5505 struct pfr_table table;
5506
5507 bzero(&table, sizeof(table));
5508 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
5509 sizeof(table.pfrt_anchor));
5510 if ((error = pfr_ina_commit(&table,
5511 ioe->ticket, NULL, NULL, 0))) {
5512 PF_RULES_WUNLOCK();
5513 free(ioes, M_TEMP);
5514 goto fail; /* really bad */
5515 }
5516 break;
5517 }
5518 default:
5519 if ((error = pf_commit_rules(ioe->ticket,
5520 ioe->rs_num, ioe->anchor))) {
5521 PF_RULES_WUNLOCK();
5522 free(ioes, M_TEMP);
5523 goto fail; /* really bad */
5524 }
5525 break;
5526 }
5527 }
5528 PF_RULES_WUNLOCK();
5529
5530 /* Only hook into EtherNet taffic if we've got rules for it. */
5531 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5532 hook_pf_eth();
5533 else
5534 dehook_pf_eth();
5535
5536 free(ioes, M_TEMP);
5537 break;
5538 }
5539
5540 case DIOCGETSRCNODES: {
5541 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
5542 struct pf_srchash *sh;
5543 struct pf_ksrc_node *n;
5544 struct pf_src_node *p, *pstore;
5545 uint32_t i, nr = 0;
5546
5547 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5548 i++, sh++) {
5549 PF_HASHROW_LOCK(sh);
5550 LIST_FOREACH(n, &sh->nodes, entry)
5551 nr++;
5552 PF_HASHROW_UNLOCK(sh);
5553 }
5554
5555 psn->psn_len = min(psn->psn_len,
5556 sizeof(struct pf_src_node) * nr);
5557
5558 if (psn->psn_len == 0) {
5559 psn->psn_len = sizeof(struct pf_src_node) * nr;
5560 goto fail;
5561 }
5562
5563 nr = 0;
5564
5565 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5566 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5567 i++, sh++) {
5568 PF_HASHROW_LOCK(sh);
5569 LIST_FOREACH(n, &sh->nodes, entry) {
5570
5571 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5572 break;
5573
5574 pf_src_node_copy(n, p);
5575
5576 p++;
5577 nr++;
5578 }
5579 PF_HASHROW_UNLOCK(sh);
5580 }
5581 error = copyout(pstore, psn->psn_src_nodes,
5582 sizeof(struct pf_src_node) * nr);
5583 if (error) {
5584 free(pstore, M_TEMP);
5585 goto fail;
5586 }
5587 psn->psn_len = sizeof(struct pf_src_node) * nr;
5588 free(pstore, M_TEMP);
5589 break;
5590 }
5591
5592 case DIOCCLRSRCNODES: {
5593 pf_kill_srcnodes(NULL);
5594 break;
5595 }
5596
5597 case DIOCKILLSRCNODES:
5598 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5599 break;
5600
5601 #ifdef COMPAT_FREEBSD13
5602 case DIOCKEEPCOUNTERS_FREEBSD13:
5603 #endif
5604 case DIOCKEEPCOUNTERS:
5605 error = pf_keepcounters((struct pfioc_nv *)addr);
5606 break;
5607
5608 case DIOCGETSYNCOOKIES:
5609 error = pf_get_syncookies((struct pfioc_nv *)addr);
5610 break;
5611
5612 case DIOCSETSYNCOOKIES:
5613 error = pf_set_syncookies((struct pfioc_nv *)addr);
5614 break;
5615
5616 case DIOCSETHOSTID: {
5617 u_int32_t *hostid = (u_int32_t *)addr;
5618
5619 PF_RULES_WLOCK();
5620 if (*hostid == 0)
5621 V_pf_status.hostid = arc4random();
5622 else
5623 V_pf_status.hostid = *hostid;
5624 PF_RULES_WUNLOCK();
5625 break;
5626 }
5627
5628 case DIOCOSFPFLUSH:
5629 PF_RULES_WLOCK();
5630 pf_osfp_flush();
5631 PF_RULES_WUNLOCK();
5632 break;
5633
5634 case DIOCIGETIFACES: {
5635 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5636 struct pfi_kif *ifstore;
5637 size_t bufsiz;
5638
5639 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5640 error = ENODEV;
5641 goto fail;
5642 }
5643
5644 if (io->pfiio_size < 0 ||
5645 io->pfiio_size > pf_ioctl_maxcount ||
5646 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5647 error = EINVAL;
5648 goto fail;
5649 }
5650
5651 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5652
5653 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5654 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5655 M_TEMP, M_WAITOK | M_ZERO);
5656
5657 PF_RULES_RLOCK();
5658 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5659 PF_RULES_RUNLOCK();
5660 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5661 free(ifstore, M_TEMP);
5662 break;
5663 }
5664
5665 case DIOCSETIFFLAG: {
5666 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5667
5668 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5669
5670 PF_RULES_WLOCK();
5671 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5672 PF_RULES_WUNLOCK();
5673 break;
5674 }
5675
5676 case DIOCCLRIFFLAG: {
5677 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5678
5679 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5680
5681 PF_RULES_WLOCK();
5682 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5683 PF_RULES_WUNLOCK();
5684 break;
5685 }
5686
5687 case DIOCSETREASS: {
5688 u_int32_t *reass = (u_int32_t *)addr;
5689
5690 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5691 /* Removal of DF flag without reassembly enabled is not a
5692 * valid combination. Disable reassembly in such case. */
5693 if (!(V_pf_status.reass & PF_REASS_ENABLED))
5694 V_pf_status.reass = 0;
5695 break;
5696 }
5697
5698 default:
5699 error = ENODEV;
5700 break;
5701 }
5702 fail:
5703 CURVNET_RESTORE();
5704
5705 #undef ERROUT_IOCTL
5706
5707 return (error);
5708 }
5709
5710 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5711 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5712 {
5713 bzero(sp, sizeof(union pfsync_state_union));
5714
5715 /* copy from state key */
5716 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5717 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5718 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5719 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5720 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5721 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5722 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5723 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5724 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5725 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5726
5727 /* copy from state */
5728 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5729 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5730 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5731 sp->pfs_1301.expire = pf_state_expires(st);
5732 if (sp->pfs_1301.expire <= time_uptime)
5733 sp->pfs_1301.expire = htonl(0);
5734 else
5735 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5736
5737 sp->pfs_1301.direction = st->direction;
5738 sp->pfs_1301.log = st->act.log;
5739 sp->pfs_1301.timeout = st->timeout;
5740
5741 switch (msg_version) {
5742 case PFSYNC_MSG_VERSION_1301:
5743 sp->pfs_1301.state_flags = st->state_flags;
5744 break;
5745 case PFSYNC_MSG_VERSION_1400:
5746 sp->pfs_1400.state_flags = htons(st->state_flags);
5747 sp->pfs_1400.qid = htons(st->act.qid);
5748 sp->pfs_1400.pqid = htons(st->act.pqid);
5749 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5750 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5751 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5752 sp->pfs_1400.min_ttl = st->act.min_ttl;
5753 sp->pfs_1400.set_tos = st->act.set_tos;
5754 sp->pfs_1400.max_mss = htons(st->act.max_mss);
5755 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5756 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5757 sp->pfs_1400.rt = st->act.rt;
5758 if (st->act.rt_kif)
5759 strlcpy(sp->pfs_1400.rt_ifname,
5760 st->act.rt_kif->pfik_name,
5761 sizeof(sp->pfs_1400.rt_ifname));
5762 break;
5763 default:
5764 panic("%s: Unsupported pfsync_msg_version %d",
5765 __func__, msg_version);
5766 }
5767
5768 /*
5769 * XXX Why do we bother pfsyncing source node information if source
5770 * nodes are not synced? Showing users that there is source tracking
5771 * when there is none seems useless.
5772 */
5773 if (st->sns[PF_SN_LIMIT] != NULL)
5774 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5775 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
5776 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5777
5778 sp->pfs_1301.id = st->id;
5779 sp->pfs_1301.creatorid = st->creatorid;
5780 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5781 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5782
5783 if (st->rule == NULL)
5784 sp->pfs_1301.rule = htonl(-1);
5785 else
5786 sp->pfs_1301.rule = htonl(st->rule->nr);
5787 if (st->anchor == NULL)
5788 sp->pfs_1301.anchor = htonl(-1);
5789 else
5790 sp->pfs_1301.anchor = htonl(st->anchor->nr);
5791 if (st->nat_rule == NULL)
5792 sp->pfs_1301.nat_rule = htonl(-1);
5793 else
5794 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
5795
5796 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5797 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5798 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5799 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5800 }
5801
5802 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5803 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5804 {
5805 bzero(sp, sizeof(*sp));
5806
5807 sp->version = PF_STATE_VERSION;
5808
5809 /* copy from state key */
5810 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5811 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5812 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5813 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5814 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5815 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5816 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5817 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5818 sp->proto = st->key[PF_SK_WIRE]->proto;
5819 sp->af = st->key[PF_SK_WIRE]->af;
5820
5821 /* copy from state */
5822 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5823 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5824 sizeof(sp->orig_ifname));
5825 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr));
5826 sp->creation = htonl(time_uptime - (st->creation / 1000));
5827 sp->expire = pf_state_expires(st);
5828 if (sp->expire <= time_uptime)
5829 sp->expire = htonl(0);
5830 else
5831 sp->expire = htonl(sp->expire - time_uptime);
5832
5833 sp->direction = st->direction;
5834 sp->log = st->act.log;
5835 sp->timeout = st->timeout;
5836 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5837 sp->state_flags_compat = st->state_flags;
5838 sp->state_flags = htons(st->state_flags);
5839 if (st->sns[PF_SN_LIMIT] != NULL)
5840 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5841 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL)
5842 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5843 sp->id = st->id;
5844 sp->creatorid = st->creatorid;
5845 pf_state_peer_hton(&st->src, &sp->src);
5846 pf_state_peer_hton(&st->dst, &sp->dst);
5847
5848 if (st->rule == NULL)
5849 sp->rule = htonl(-1);
5850 else
5851 sp->rule = htonl(st->rule->nr);
5852 if (st->anchor == NULL)
5853 sp->anchor = htonl(-1);
5854 else
5855 sp->anchor = htonl(st->anchor->nr);
5856 if (st->nat_rule == NULL)
5857 sp->nat_rule = htonl(-1);
5858 else
5859 sp->nat_rule = htonl(st->nat_rule->nr);
5860
5861 sp->packets[0] = st->packets[0];
5862 sp->packets[1] = st->packets[1];
5863 sp->bytes[0] = st->bytes[0];
5864 sp->bytes[1] = st->bytes[1];
5865
5866 sp->qid = htons(st->act.qid);
5867 sp->pqid = htons(st->act.pqid);
5868 sp->dnpipe = htons(st->act.dnpipe);
5869 sp->dnrpipe = htons(st->act.dnrpipe);
5870 sp->rtableid = htonl(st->act.rtableid);
5871 sp->min_ttl = st->act.min_ttl;
5872 sp->set_tos = st->act.set_tos;
5873 sp->max_mss = htons(st->act.max_mss);
5874 sp->rt = st->act.rt;
5875 if (st->act.rt_kif)
5876 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
5877 sizeof(sp->rt_ifname));
5878 sp->set_prio[0] = st->act.set_prio[0];
5879 sp->set_prio[1] = st->act.set_prio[1];
5880
5881 }
5882
5883 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5884 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5885 {
5886 struct pfr_ktable *kt;
5887
5888 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5889
5890 kt = aw->p.tbl;
5891 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5892 kt = kt->pfrkt_root;
5893 aw->p.tbl = NULL;
5894 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5895 kt->pfrkt_cnt : -1;
5896 }
5897
5898 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5899 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5900 size_t number, char **names)
5901 {
5902 nvlist_t *nvc;
5903
5904 nvc = nvlist_create(0);
5905 if (nvc == NULL)
5906 return (ENOMEM);
5907
5908 for (int i = 0; i < number; i++) {
5909 nvlist_append_number_array(nvc, "counters",
5910 counter_u64_fetch(counters[i]));
5911 nvlist_append_string_array(nvc, "names",
5912 names[i]);
5913 nvlist_append_number_array(nvc, "ids",
5914 i);
5915 }
5916 nvlist_add_nvlist(nvl, name, nvc);
5917 nvlist_destroy(nvc);
5918
5919 return (0);
5920 }
5921
5922 static int
pf_getstatus(struct pfioc_nv * nv)5923 pf_getstatus(struct pfioc_nv *nv)
5924 {
5925 nvlist_t *nvl = NULL, *nvc = NULL;
5926 void *nvlpacked = NULL;
5927 int error;
5928 struct pf_status s;
5929 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5930 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5931 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5932 time_t since;
5933
5934 PF_RULES_RLOCK_TRACKER;
5935
5936 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
5937
5938 PF_RULES_RLOCK();
5939
5940 nvl = nvlist_create(0);
5941 if (nvl == NULL)
5942 ERROUT(ENOMEM);
5943
5944 since = time_second - (time_uptime - V_pf_status.since);
5945
5946 nvlist_add_bool(nvl, "running", V_pf_status.running);
5947 nvlist_add_number(nvl, "since", since);
5948 nvlist_add_number(nvl, "debug", V_pf_status.debug);
5949 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5950 nvlist_add_number(nvl, "states", V_pf_status.states);
5951 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5952 nvlist_add_number(nvl, "reass", V_pf_status.reass);
5953 nvlist_add_bool(nvl, "syncookies_active",
5954 V_pf_status.syncookies_active);
5955 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5956
5957 /* counters */
5958 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5959 PFRES_MAX, pf_reasons);
5960 if (error != 0)
5961 ERROUT(error);
5962
5963 /* lcounters */
5964 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5965 KLCNT_MAX, pf_lcounter);
5966 if (error != 0)
5967 ERROUT(error);
5968
5969 /* fcounters */
5970 nvc = nvlist_create(0);
5971 if (nvc == NULL)
5972 ERROUT(ENOMEM);
5973
5974 for (int i = 0; i < FCNT_MAX; i++) {
5975 nvlist_append_number_array(nvc, "counters",
5976 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5977 nvlist_append_string_array(nvc, "names",
5978 pf_fcounter[i]);
5979 nvlist_append_number_array(nvc, "ids",
5980 i);
5981 }
5982 nvlist_add_nvlist(nvl, "fcounters", nvc);
5983 nvlist_destroy(nvc);
5984 nvc = NULL;
5985
5986 /* scounters */
5987 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5988 SCNT_MAX, pf_fcounter);
5989 if (error != 0)
5990 ERROUT(error);
5991
5992 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5993 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5994 PF_MD5_DIGEST_LENGTH);
5995
5996 pfi_update_status(V_pf_status.ifname, &s);
5997
5998 /* pcounters / bcounters */
5999 for (int i = 0; i < 2; i++) {
6000 for (int j = 0; j < 2; j++) {
6001 for (int k = 0; k < 2; k++) {
6002 nvlist_append_number_array(nvl, "pcounters",
6003 s.pcounters[i][j][k]);
6004 }
6005 nvlist_append_number_array(nvl, "bcounters",
6006 s.bcounters[i][j]);
6007 }
6008 }
6009
6010 nvlpacked = nvlist_pack(nvl, &nv->len);
6011 if (nvlpacked == NULL)
6012 ERROUT(ENOMEM);
6013
6014 if (nv->size == 0)
6015 ERROUT(0);
6016 else if (nv->size < nv->len)
6017 ERROUT(ENOSPC);
6018
6019 PF_RULES_RUNLOCK();
6020 error = copyout(nvlpacked, nv->data, nv->len);
6021 goto done;
6022
6023 #undef ERROUT
6024 errout:
6025 PF_RULES_RUNLOCK();
6026 done:
6027 free(nvlpacked, M_NVLIST);
6028 nvlist_destroy(nvc);
6029 nvlist_destroy(nvl);
6030
6031 return (error);
6032 }
6033
6034 /*
6035 * XXX - Check for version mismatch!!!
6036 */
6037 static void
pf_clear_all_states(void)6038 pf_clear_all_states(void)
6039 {
6040 struct epoch_tracker et;
6041 struct pf_kstate *s;
6042 u_int i;
6043
6044 NET_EPOCH_ENTER(et);
6045 for (i = 0; i <= V_pf_hashmask; i++) {
6046 struct pf_idhash *ih = &V_pf_idhash[i];
6047 relock:
6048 PF_HASHROW_LOCK(ih);
6049 LIST_FOREACH(s, &ih->states, entry) {
6050 s->timeout = PFTM_PURGE;
6051 /* Don't send out individual delete messages. */
6052 s->state_flags |= PFSTATE_NOSYNC;
6053 pf_remove_state(s);
6054 goto relock;
6055 }
6056 PF_HASHROW_UNLOCK(ih);
6057 }
6058 NET_EPOCH_EXIT(et);
6059 }
6060
6061 static int
pf_clear_tables(void)6062 pf_clear_tables(void)
6063 {
6064 struct pfioc_table io;
6065 int error;
6066
6067 bzero(&io, sizeof(io));
6068 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
6069
6070 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
6071 io.pfrio_flags);
6072
6073 return (error);
6074 }
6075
6076 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)6077 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
6078 {
6079 struct pf_ksrc_node_list kill;
6080 u_int killed;
6081
6082 LIST_INIT(&kill);
6083 for (int i = 0; i <= V_pf_srchashmask; i++) {
6084 struct pf_srchash *sh = &V_pf_srchash[i];
6085 struct pf_ksrc_node *sn, *tmp;
6086
6087 PF_HASHROW_LOCK(sh);
6088 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
6089 if (psnk == NULL ||
6090 (pf_match_addr(psnk->psnk_src.neg,
6091 &psnk->psnk_src.addr.v.a.addr,
6092 &psnk->psnk_src.addr.v.a.mask,
6093 &sn->addr, sn->af) &&
6094 pf_match_addr(psnk->psnk_dst.neg,
6095 &psnk->psnk_dst.addr.v.a.addr,
6096 &psnk->psnk_dst.addr.v.a.mask,
6097 &sn->raddr, sn->af))) {
6098 pf_unlink_src_node(sn);
6099 LIST_INSERT_HEAD(&kill, sn, entry);
6100 sn->expire = 1;
6101 }
6102 PF_HASHROW_UNLOCK(sh);
6103 }
6104
6105 for (int i = 0; i <= V_pf_hashmask; i++) {
6106 struct pf_idhash *ih = &V_pf_idhash[i];
6107 struct pf_kstate *s;
6108
6109 PF_HASHROW_LOCK(ih);
6110 LIST_FOREACH(s, &ih->states, entry) {
6111 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX;
6112 sn_type++) {
6113 if (s->sns[sn_type] &&
6114 s->sns[sn_type]->expire == 1) {
6115 s->sns[sn_type] = NULL;
6116 }
6117 }
6118 }
6119 PF_HASHROW_UNLOCK(ih);
6120 }
6121
6122 killed = pf_free_src_nodes(&kill);
6123
6124 if (psnk != NULL)
6125 psnk->psnk_killed = killed;
6126 }
6127
6128 static int
pf_keepcounters(struct pfioc_nv * nv)6129 pf_keepcounters(struct pfioc_nv *nv)
6130 {
6131 nvlist_t *nvl = NULL;
6132 void *nvlpacked = NULL;
6133 int error = 0;
6134
6135 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6136
6137 if (nv->len > pf_ioctl_maxcount)
6138 ERROUT(ENOMEM);
6139
6140 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6141 error = copyin(nv->data, nvlpacked, nv->len);
6142 if (error)
6143 ERROUT(error);
6144
6145 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6146 if (nvl == NULL)
6147 ERROUT(EBADMSG);
6148
6149 if (! nvlist_exists_bool(nvl, "keep_counters"))
6150 ERROUT(EBADMSG);
6151
6152 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6153
6154 on_error:
6155 nvlist_destroy(nvl);
6156 free(nvlpacked, M_NVLIST);
6157 return (error);
6158 }
6159
6160 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6161 pf_clear_states(const struct pf_kstate_kill *kill)
6162 {
6163 struct pf_state_key_cmp match_key;
6164 struct pf_kstate *s;
6165 struct pfi_kkif *kif;
6166 int idx;
6167 unsigned int killed = 0, dir;
6168
6169 NET_EPOCH_ASSERT();
6170
6171 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6172 struct pf_idhash *ih = &V_pf_idhash[i];
6173
6174 relock_DIOCCLRSTATES:
6175 PF_HASHROW_LOCK(ih);
6176 LIST_FOREACH(s, &ih->states, entry) {
6177 /* For floating states look at the original kif. */
6178 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6179
6180 if (kill->psk_ifname[0] &&
6181 strcmp(kill->psk_ifname,
6182 kif->pfik_name))
6183 continue;
6184
6185 if (kill->psk_kill_match) {
6186 bzero(&match_key, sizeof(match_key));
6187
6188 if (s->direction == PF_OUT) {
6189 dir = PF_IN;
6190 idx = PF_SK_STACK;
6191 } else {
6192 dir = PF_OUT;
6193 idx = PF_SK_WIRE;
6194 }
6195
6196 match_key.af = s->key[idx]->af;
6197 match_key.proto = s->key[idx]->proto;
6198 pf_addrcpy(&match_key.addr[0],
6199 &s->key[idx]->addr[1], match_key.af);
6200 match_key.port[0] = s->key[idx]->port[1];
6201 pf_addrcpy(&match_key.addr[1],
6202 &s->key[idx]->addr[0], match_key.af);
6203 match_key.port[1] = s->key[idx]->port[0];
6204 }
6205
6206 /*
6207 * Don't send out individual
6208 * delete messages.
6209 */
6210 s->state_flags |= PFSTATE_NOSYNC;
6211 pf_remove_state(s);
6212 killed++;
6213
6214 if (kill->psk_kill_match)
6215 killed += pf_kill_matching_state(&match_key,
6216 dir);
6217
6218 goto relock_DIOCCLRSTATES;
6219 }
6220 PF_HASHROW_UNLOCK(ih);
6221 }
6222
6223 if (V_pfsync_clear_states_ptr != NULL)
6224 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6225
6226 return (killed);
6227 }
6228
6229 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6230 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6231 {
6232 struct pf_kstate *s;
6233
6234 NET_EPOCH_ASSERT();
6235 if (kill->psk_pfcmp.id) {
6236 if (kill->psk_pfcmp.creatorid == 0)
6237 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6238 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6239 kill->psk_pfcmp.creatorid))) {
6240 pf_remove_state(s);
6241 *killed = 1;
6242 }
6243 return;
6244 }
6245
6246 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6247 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6248 }
6249
6250 static int
pf_killstates_nv(struct pfioc_nv * nv)6251 pf_killstates_nv(struct pfioc_nv *nv)
6252 {
6253 struct pf_kstate_kill kill;
6254 struct epoch_tracker et;
6255 nvlist_t *nvl = NULL;
6256 void *nvlpacked = NULL;
6257 int error = 0;
6258 unsigned int killed = 0;
6259
6260 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6261
6262 if (nv->len > pf_ioctl_maxcount)
6263 ERROUT(ENOMEM);
6264
6265 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6266 error = copyin(nv->data, nvlpacked, nv->len);
6267 if (error)
6268 ERROUT(error);
6269
6270 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6271 if (nvl == NULL)
6272 ERROUT(EBADMSG);
6273
6274 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6275 if (error)
6276 ERROUT(error);
6277
6278 NET_EPOCH_ENTER(et);
6279 pf_killstates(&kill, &killed);
6280 NET_EPOCH_EXIT(et);
6281
6282 free(nvlpacked, M_NVLIST);
6283 nvlpacked = NULL;
6284 nvlist_destroy(nvl);
6285 nvl = nvlist_create(0);
6286 if (nvl == NULL)
6287 ERROUT(ENOMEM);
6288
6289 nvlist_add_number(nvl, "killed", killed);
6290
6291 nvlpacked = nvlist_pack(nvl, &nv->len);
6292 if (nvlpacked == NULL)
6293 ERROUT(ENOMEM);
6294
6295 if (nv->size == 0)
6296 ERROUT(0);
6297 else if (nv->size < nv->len)
6298 ERROUT(ENOSPC);
6299
6300 error = copyout(nvlpacked, nv->data, nv->len);
6301
6302 on_error:
6303 nvlist_destroy(nvl);
6304 free(nvlpacked, M_NVLIST);
6305 return (error);
6306 }
6307
6308 static int
pf_clearstates_nv(struct pfioc_nv * nv)6309 pf_clearstates_nv(struct pfioc_nv *nv)
6310 {
6311 struct pf_kstate_kill kill;
6312 struct epoch_tracker et;
6313 nvlist_t *nvl = NULL;
6314 void *nvlpacked = NULL;
6315 int error = 0;
6316 unsigned int killed;
6317
6318 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6319
6320 if (nv->len > pf_ioctl_maxcount)
6321 ERROUT(ENOMEM);
6322
6323 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6324 error = copyin(nv->data, nvlpacked, nv->len);
6325 if (error)
6326 ERROUT(error);
6327
6328 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6329 if (nvl == NULL)
6330 ERROUT(EBADMSG);
6331
6332 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6333 if (error)
6334 ERROUT(error);
6335
6336 NET_EPOCH_ENTER(et);
6337 killed = pf_clear_states(&kill);
6338 NET_EPOCH_EXIT(et);
6339
6340 free(nvlpacked, M_NVLIST);
6341 nvlpacked = NULL;
6342 nvlist_destroy(nvl);
6343 nvl = nvlist_create(0);
6344 if (nvl == NULL)
6345 ERROUT(ENOMEM);
6346
6347 nvlist_add_number(nvl, "killed", killed);
6348
6349 nvlpacked = nvlist_pack(nvl, &nv->len);
6350 if (nvlpacked == NULL)
6351 ERROUT(ENOMEM);
6352
6353 if (nv->size == 0)
6354 ERROUT(0);
6355 else if (nv->size < nv->len)
6356 ERROUT(ENOSPC);
6357
6358 error = copyout(nvlpacked, nv->data, nv->len);
6359
6360 #undef ERROUT
6361 on_error:
6362 nvlist_destroy(nvl);
6363 free(nvlpacked, M_NVLIST);
6364 return (error);
6365 }
6366
6367 static int
pf_getstate(struct pfioc_nv * nv)6368 pf_getstate(struct pfioc_nv *nv)
6369 {
6370 nvlist_t *nvl = NULL, *nvls;
6371 void *nvlpacked = NULL;
6372 struct pf_kstate *s = NULL;
6373 int error = 0;
6374 uint64_t id, creatorid;
6375
6376 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6377
6378 if (nv->len > pf_ioctl_maxcount)
6379 ERROUT(ENOMEM);
6380
6381 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6382 error = copyin(nv->data, nvlpacked, nv->len);
6383 if (error)
6384 ERROUT(error);
6385
6386 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6387 if (nvl == NULL)
6388 ERROUT(EBADMSG);
6389
6390 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6391 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6392
6393 s = pf_find_state_byid(id, creatorid);
6394 if (s == NULL)
6395 ERROUT(ENOENT);
6396
6397 free(nvlpacked, M_NVLIST);
6398 nvlpacked = NULL;
6399 nvlist_destroy(nvl);
6400 nvl = nvlist_create(0);
6401 if (nvl == NULL)
6402 ERROUT(ENOMEM);
6403
6404 nvls = pf_state_to_nvstate(s);
6405 if (nvls == NULL)
6406 ERROUT(ENOMEM);
6407
6408 nvlist_add_nvlist(nvl, "state", nvls);
6409 nvlist_destroy(nvls);
6410
6411 nvlpacked = nvlist_pack(nvl, &nv->len);
6412 if (nvlpacked == NULL)
6413 ERROUT(ENOMEM);
6414
6415 if (nv->size == 0)
6416 ERROUT(0);
6417 else if (nv->size < nv->len)
6418 ERROUT(ENOSPC);
6419
6420 error = copyout(nvlpacked, nv->data, nv->len);
6421
6422 #undef ERROUT
6423 errout:
6424 if (s != NULL)
6425 PF_STATE_UNLOCK(s);
6426 free(nvlpacked, M_NVLIST);
6427 nvlist_destroy(nvl);
6428 return (error);
6429 }
6430
6431 /*
6432 * XXX - Check for version mismatch!!!
6433 */
6434
6435 /*
6436 * Duplicate pfctl -Fa operation to get rid of as much as we can.
6437 */
6438 static int
shutdown_pf(void)6439 shutdown_pf(void)
6440 {
6441 int error = 0;
6442 u_int32_t t[5];
6443 char nn = '\0';
6444 struct pf_kanchor *anchor;
6445 struct pf_keth_anchor *eth_anchor;
6446 int rs_num;
6447
6448 do {
6449 /* Unlink rules of all user defined anchors */
6450 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6451 /* Wildcard based anchors may not have a respective
6452 * explicit anchor rule or they may be left empty
6453 * without rules. It leads to anchor.refcnt=0, and the
6454 * rest of the logic does not expect it. */
6455 if (anchor->refcnt == 0)
6456 anchor->refcnt = 1;
6457 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6458 if ((error = pf_begin_rules(&t[rs_num], rs_num,
6459 anchor->path)) != 0) {
6460 DPFPRINTF(PF_DEBUG_MISC, "%s: "
6461 "anchor.path=%s rs_num=%d",
6462 __func__, anchor->path, rs_num);
6463 goto error; /* XXX: rollback? */
6464 }
6465 }
6466 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6467 error = pf_commit_rules(t[rs_num], rs_num,
6468 anchor->path);
6469 MPASS(error == 0);
6470 }
6471 }
6472
6473 /* Unlink rules of all user defined ether anchors */
6474 RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6475 &V_pf_keth_anchors) {
6476 /* Wildcard based anchors may not have a respective
6477 * explicit anchor rule or they may be left empty
6478 * without rules. It leads to anchor.refcnt=0, and the
6479 * rest of the logic does not expect it. */
6480 if (eth_anchor->refcnt == 0)
6481 eth_anchor->refcnt = 1;
6482 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6483 != 0) {
6484 DPFPRINTF(PF_DEBUG_MISC, "%s: eth "
6485 "anchor.path=%s", __func__,
6486 eth_anchor->path);
6487 goto error;
6488 }
6489 error = pf_commit_eth(t[0], eth_anchor->path);
6490 MPASS(error == 0);
6491 }
6492
6493 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6494 != 0) {
6495 DPFPRINTF(PF_DEBUG_MISC, "%s: SCRUB", __func__);
6496 break;
6497 }
6498 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6499 != 0) {
6500 DPFPRINTF(PF_DEBUG_MISC, "%s: FILTER", __func__);
6501 break; /* XXX: rollback? */
6502 }
6503 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6504 != 0) {
6505 DPFPRINTF(PF_DEBUG_MISC, "%s: NAT", __func__);
6506 break; /* XXX: rollback? */
6507 }
6508 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6509 != 0) {
6510 DPFPRINTF(PF_DEBUG_MISC, "%s: BINAT", __func__);
6511 break; /* XXX: rollback? */
6512 }
6513 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6514 != 0) {
6515 DPFPRINTF(PF_DEBUG_MISC, "%s: RDR", __func__);
6516 break; /* XXX: rollback? */
6517 }
6518
6519 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6520 MPASS(error == 0);
6521 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6522 MPASS(error == 0);
6523 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6524 MPASS(error == 0);
6525 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6526 MPASS(error == 0);
6527 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6528 MPASS(error == 0);
6529
6530 if ((error = pf_clear_tables()) != 0)
6531 break;
6532
6533 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6534 DPFPRINTF(PF_DEBUG_MISC, "%s: eth", __func__);
6535 break;
6536 }
6537 error = pf_commit_eth(t[0], &nn);
6538 MPASS(error == 0);
6539
6540 #ifdef ALTQ
6541 if ((error = pf_begin_altq(&t[0])) != 0) {
6542 DPFPRINTF(PF_DEBUG_MISC, "%s: ALTQ", __func__);
6543 break;
6544 }
6545 pf_commit_altq(t[0]);
6546 #endif
6547
6548 pf_clear_all_states();
6549
6550 pf_kill_srcnodes(NULL);
6551
6552 /* status does not use malloced mem so no need to cleanup */
6553 /* fingerprints and interfaces have their own cleanup code */
6554 } while(0);
6555
6556 error:
6557 return (error);
6558 }
6559
6560 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6561 pf_check_return(int chk, struct mbuf **m)
6562 {
6563
6564 switch (chk) {
6565 case PF_PASS:
6566 if (*m == NULL)
6567 return (PFIL_CONSUMED);
6568 else
6569 return (PFIL_PASS);
6570 break;
6571 default:
6572 if (*m != NULL) {
6573 m_freem(*m);
6574 *m = NULL;
6575 }
6576 return (PFIL_DROPPED);
6577 }
6578 }
6579
6580 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6581 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6582 void *ruleset __unused, struct inpcb *inp)
6583 {
6584 int chk;
6585
6586 CURVNET_ASSERT_SET();
6587
6588 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6589
6590 return (pf_check_return(chk, m));
6591 }
6592
6593 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6594 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6595 void *ruleset __unused, struct inpcb *inp)
6596 {
6597 int chk;
6598
6599 CURVNET_ASSERT_SET();
6600
6601 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6602
6603 return (pf_check_return(chk, m));
6604 }
6605
6606 #ifdef INET
6607 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6608 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6609 void *ruleset __unused, struct inpcb *inp)
6610 {
6611 int chk;
6612
6613 CURVNET_ASSERT_SET();
6614
6615 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
6616
6617 return (pf_check_return(chk, m));
6618 }
6619
6620 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6621 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6622 void *ruleset __unused, struct inpcb *inp)
6623 {
6624 int chk;
6625
6626 CURVNET_ASSERT_SET();
6627
6628 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
6629
6630 return (pf_check_return(chk, m));
6631 }
6632 #endif
6633
6634 #ifdef INET6
6635 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6636 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6637 void *ruleset __unused, struct inpcb *inp)
6638 {
6639 int chk;
6640
6641 CURVNET_ASSERT_SET();
6642
6643 /*
6644 * In case of loopback traffic IPv6 uses the real interface in
6645 * order to support scoped addresses. In order to support stateful
6646 * filtering we have change this to lo0 as it is the case in IPv4.
6647 */
6648 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6649 m, inp, NULL);
6650
6651 return (pf_check_return(chk, m));
6652 }
6653
6654 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6655 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6656 void *ruleset __unused, struct inpcb *inp)
6657 {
6658 int chk;
6659
6660 CURVNET_ASSERT_SET();
6661
6662 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL);
6663
6664 return (pf_check_return(chk, m));
6665 }
6666 #endif /* INET6 */
6667
6668 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6669 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6670 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
6671 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
6672
6673 #ifdef INET
6674 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6675 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6676 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
6677 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
6678 #endif
6679 #ifdef INET6
6680 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6681 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6682 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
6683 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
6684 #endif
6685
6686 static void
hook_pf_eth(void)6687 hook_pf_eth(void)
6688 {
6689 struct pfil_hook_args pha = {
6690 .pa_version = PFIL_VERSION,
6691 .pa_modname = "pf",
6692 .pa_type = PFIL_TYPE_ETHERNET,
6693 };
6694 struct pfil_link_args pla = {
6695 .pa_version = PFIL_VERSION,
6696 };
6697 int ret __diagused;
6698
6699 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6700 return;
6701
6702 pha.pa_mbuf_chk = pf_eth_check_in;
6703 pha.pa_flags = PFIL_IN;
6704 pha.pa_rulname = "eth-in";
6705 V_pf_eth_in_hook = pfil_add_hook(&pha);
6706 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6707 pla.pa_head = V_link_pfil_head;
6708 pla.pa_hook = V_pf_eth_in_hook;
6709 ret = pfil_link(&pla);
6710 MPASS(ret == 0);
6711 pha.pa_mbuf_chk = pf_eth_check_out;
6712 pha.pa_flags = PFIL_OUT;
6713 pha.pa_rulname = "eth-out";
6714 V_pf_eth_out_hook = pfil_add_hook(&pha);
6715 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6716 pla.pa_head = V_link_pfil_head;
6717 pla.pa_hook = V_pf_eth_out_hook;
6718 ret = pfil_link(&pla);
6719 MPASS(ret == 0);
6720
6721 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6722 }
6723
6724 static void
hook_pf(void)6725 hook_pf(void)
6726 {
6727 struct pfil_hook_args pha = {
6728 .pa_version = PFIL_VERSION,
6729 .pa_modname = "pf",
6730 };
6731 struct pfil_link_args pla = {
6732 .pa_version = PFIL_VERSION,
6733 };
6734 int ret __diagused;
6735
6736 if (atomic_load_bool(&V_pf_pfil_hooked))
6737 return;
6738
6739 #ifdef INET
6740 pha.pa_type = PFIL_TYPE_IP4;
6741 pha.pa_mbuf_chk = pf_check_in;
6742 pha.pa_flags = PFIL_IN;
6743 pha.pa_rulname = "default-in";
6744 V_pf_ip4_in_hook = pfil_add_hook(&pha);
6745 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6746 pla.pa_head = V_inet_pfil_head;
6747 pla.pa_hook = V_pf_ip4_in_hook;
6748 ret = pfil_link(&pla);
6749 MPASS(ret == 0);
6750 pha.pa_mbuf_chk = pf_check_out;
6751 pha.pa_flags = PFIL_OUT;
6752 pha.pa_rulname = "default-out";
6753 V_pf_ip4_out_hook = pfil_add_hook(&pha);
6754 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6755 pla.pa_head = V_inet_pfil_head;
6756 pla.pa_hook = V_pf_ip4_out_hook;
6757 ret = pfil_link(&pla);
6758 MPASS(ret == 0);
6759 if (V_pf_filter_local) {
6760 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6761 pla.pa_head = V_inet_local_pfil_head;
6762 pla.pa_hook = V_pf_ip4_out_hook;
6763 ret = pfil_link(&pla);
6764 MPASS(ret == 0);
6765 }
6766 #endif
6767 #ifdef INET6
6768 pha.pa_type = PFIL_TYPE_IP6;
6769 pha.pa_mbuf_chk = pf_check6_in;
6770 pha.pa_flags = PFIL_IN;
6771 pha.pa_rulname = "default-in6";
6772 V_pf_ip6_in_hook = pfil_add_hook(&pha);
6773 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6774 pla.pa_head = V_inet6_pfil_head;
6775 pla.pa_hook = V_pf_ip6_in_hook;
6776 ret = pfil_link(&pla);
6777 MPASS(ret == 0);
6778 pha.pa_mbuf_chk = pf_check6_out;
6779 pha.pa_rulname = "default-out6";
6780 pha.pa_flags = PFIL_OUT;
6781 V_pf_ip6_out_hook = pfil_add_hook(&pha);
6782 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6783 pla.pa_head = V_inet6_pfil_head;
6784 pla.pa_hook = V_pf_ip6_out_hook;
6785 ret = pfil_link(&pla);
6786 MPASS(ret == 0);
6787 if (V_pf_filter_local) {
6788 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6789 pla.pa_head = V_inet6_local_pfil_head;
6790 pla.pa_hook = V_pf_ip6_out_hook;
6791 ret = pfil_link(&pla);
6792 MPASS(ret == 0);
6793 }
6794 #endif
6795
6796 atomic_store_bool(&V_pf_pfil_hooked, true);
6797 }
6798
6799 static void
dehook_pf_eth(void)6800 dehook_pf_eth(void)
6801 {
6802
6803 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6804 return;
6805
6806 pfil_remove_hook(V_pf_eth_in_hook);
6807 pfil_remove_hook(V_pf_eth_out_hook);
6808
6809 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6810 }
6811
6812 static void
dehook_pf(void)6813 dehook_pf(void)
6814 {
6815
6816 if (!atomic_load_bool(&V_pf_pfil_hooked))
6817 return;
6818
6819 #ifdef INET
6820 pfil_remove_hook(V_pf_ip4_in_hook);
6821 pfil_remove_hook(V_pf_ip4_out_hook);
6822 #endif
6823 #ifdef INET6
6824 pfil_remove_hook(V_pf_ip6_in_hook);
6825 pfil_remove_hook(V_pf_ip6_out_hook);
6826 #endif
6827
6828 atomic_store_bool(&V_pf_pfil_hooked, false);
6829 }
6830
6831 static void
pf_load_vnet(void)6832 pf_load_vnet(void)
6833 {
6834 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6835 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6836
6837 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6838 sx_init(&V_pf_ioctl_lock, "pf ioctl");
6839
6840 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6841 PF_RULE_TAG_HASH_SIZE_DEFAULT);
6842 #ifdef ALTQ
6843 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6844 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6845 #endif
6846
6847 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6848
6849 pfattach_vnet();
6850 V_pf_vnet_active = 1;
6851 }
6852
6853 static int
pf_load(void)6854 pf_load(void)
6855 {
6856 int error;
6857
6858 sx_init(&pf_end_lock, "pf end thread");
6859
6860 pf_mtag_initialize();
6861
6862 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6863 if (pf_dev == NULL)
6864 return (ENOMEM);
6865
6866 pf_end_threads = 0;
6867 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6868 if (error != 0)
6869 return (error);
6870
6871 pfi_initialize();
6872
6873 return (0);
6874 }
6875
6876 static void
pf_unload_vnet(void)6877 pf_unload_vnet(void)
6878 {
6879 int ret __diagused;
6880
6881 V_pf_vnet_active = 0;
6882 V_pf_status.running = 0;
6883 dehook_pf();
6884 dehook_pf_eth();
6885
6886 PF_RULES_WLOCK();
6887 pf_syncookies_cleanup();
6888 shutdown_pf();
6889 PF_RULES_WUNLOCK();
6890
6891 ret = swi_remove(V_pf_swi_cookie);
6892 MPASS(ret == 0);
6893 ret = intr_event_destroy(V_pf_swi_ie);
6894 MPASS(ret == 0);
6895
6896 pf_unload_vnet_purge();
6897
6898 pf_normalize_cleanup();
6899 PF_RULES_WLOCK();
6900 pfi_cleanup_vnet();
6901 PF_RULES_WUNLOCK();
6902 pfr_cleanup();
6903 pf_osfp_flush();
6904 pf_cleanup();
6905 if (IS_DEFAULT_VNET(curvnet))
6906 pf_mtag_cleanup();
6907
6908 pf_cleanup_tagset(&V_pf_tags);
6909 #ifdef ALTQ
6910 pf_cleanup_tagset(&V_pf_qids);
6911 #endif
6912 uma_zdestroy(V_pf_tag_z);
6913
6914 #ifdef PF_WANT_32_TO_64_COUNTER
6915 PF_RULES_WLOCK();
6916 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6917
6918 MPASS(LIST_EMPTY(&V_pf_allkiflist));
6919 MPASS(V_pf_allkifcount == 0);
6920
6921 LIST_REMOVE(&V_pf_default_rule, allrulelist);
6922 V_pf_allrulecount--;
6923 LIST_REMOVE(V_pf_rulemarker, allrulelist);
6924
6925 MPASS(LIST_EMPTY(&V_pf_allrulelist));
6926 MPASS(V_pf_allrulecount == 0);
6927
6928 PF_RULES_WUNLOCK();
6929
6930 free(V_pf_kifmarker, PFI_MTYPE);
6931 free(V_pf_rulemarker, M_PFRULE);
6932 #endif
6933
6934 /* Free counters last as we updated them during shutdown. */
6935 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6936 for (int i = 0; i < 2; i++) {
6937 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6938 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6939 }
6940 counter_u64_free(V_pf_default_rule.states_cur);
6941 counter_u64_free(V_pf_default_rule.states_tot);
6942 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
6943 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]);
6944 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6945
6946 for (int i = 0; i < PFRES_MAX; i++)
6947 counter_u64_free(V_pf_status.counters[i]);
6948 for (int i = 0; i < KLCNT_MAX; i++)
6949 counter_u64_free(V_pf_status.lcounters[i]);
6950 for (int i = 0; i < FCNT_MAX; i++)
6951 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6952 for (int i = 0; i < SCNT_MAX; i++)
6953 counter_u64_free(V_pf_status.scounters[i]);
6954
6955 rm_destroy(&V_pf_rules_lock);
6956 sx_destroy(&V_pf_ioctl_lock);
6957 }
6958
6959 static void
pf_unload(void)6960 pf_unload(void)
6961 {
6962
6963 sx_xlock(&pf_end_lock);
6964 pf_end_threads = 1;
6965 while (pf_end_threads < 2) {
6966 wakeup_one(pf_purge_thread);
6967 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6968 }
6969 sx_xunlock(&pf_end_lock);
6970
6971 pf_nl_unregister();
6972
6973 if (pf_dev != NULL)
6974 destroy_dev(pf_dev);
6975
6976 pfi_cleanup();
6977
6978 sx_destroy(&pf_end_lock);
6979 }
6980
6981 static void
vnet_pf_init(void * unused __unused)6982 vnet_pf_init(void *unused __unused)
6983 {
6984
6985 pf_load_vnet();
6986 }
6987 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6988 vnet_pf_init, NULL);
6989
6990 static void
vnet_pf_uninit(const void * unused __unused)6991 vnet_pf_uninit(const void *unused __unused)
6992 {
6993
6994 pf_unload_vnet();
6995 }
6996 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6997 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6998 vnet_pf_uninit, NULL);
6999
7000 static int
pf_modevent(module_t mod,int type,void * data)7001 pf_modevent(module_t mod, int type, void *data)
7002 {
7003 int error = 0;
7004
7005 switch(type) {
7006 case MOD_LOAD:
7007 error = pf_load();
7008 pf_nl_register();
7009 break;
7010 case MOD_UNLOAD:
7011 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
7012 * the vnet_pf_uninit()s */
7013 break;
7014 default:
7015 error = EINVAL;
7016 break;
7017 }
7018
7019 return (error);
7020 }
7021
7022 static moduledata_t pf_mod = {
7023 "pf",
7024 pf_modevent,
7025 0
7026 };
7027
7028 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
7029 MODULE_DEPEND(pf, netlink, 1, 1, 1);
7030 MODULE_DEPEND(pf, crypto, 1, 1, 1);
7031 MODULE_VERSION(pf, PF_MODVER);
7032