xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision d909f06b907d53ac4cbf9444ae75ca278900b0f3)
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);
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 void		 pf_rollback_eth_cb(struct epoch_context *);
111 static int		 pf_rollback_eth(uint32_t, const char *);
112 static int		 pf_commit_eth(uint32_t, const char *);
113 static void		 pf_free_eth_rule(struct pf_keth_rule *);
114 #ifdef ALTQ
115 static int		 pf_begin_altq(u_int32_t *);
116 static int		 pf_rollback_altq(u_int32_t);
117 static int		 pf_commit_altq(u_int32_t);
118 static int		 pf_enable_altq(struct pf_altq *);
119 static int		 pf_disable_altq(struct pf_altq *);
120 static uint16_t		 pf_qname2qid(const char *);
121 static void		 pf_qid_unref(uint16_t);
122 #endif /* ALTQ */
123 static int		 pf_begin_rules(u_int32_t *, int, const char *);
124 static int		 pf_rollback_rules(u_int32_t, int, char *);
125 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
126 static void		 pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
127 static void		 pf_hash_rule(struct pf_krule *);
128 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
129 static int		 pf_commit_rules(u_int32_t, int, char *);
130 static int		 pf_addr_setup(struct pf_kruleset *,
131 			    struct pf_addr_wrap *, sa_family_t);
132 static void		 pf_addr_copyout(struct pf_addr_wrap *);
133 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
134 			    struct pf_src_node *);
135 #ifdef ALTQ
136 static int		 pf_export_kaltq(struct pf_altq *,
137 			    struct pfioc_altq_v1 *, size_t);
138 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
139 			    struct pf_altq *, size_t);
140 #endif /* ALTQ */
141 
142 VNET_DEFINE(struct pf_krule,	pf_default_rule);
143 
144 static __inline int             pf_krule_compare(struct pf_krule *,
145 				    struct pf_krule *);
146 
147 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
148 
149 #ifdef ALTQ
150 VNET_DEFINE_STATIC(int,		pf_altq_running);
151 #define	V_pf_altq_running	VNET(pf_altq_running)
152 #endif
153 
154 #define	TAGID_MAX	 50000
155 struct pf_tagname {
156 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
157 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
158 	char			name[PF_TAG_NAME_SIZE];
159 	uint16_t		tag;
160 	int			ref;
161 };
162 
163 struct pf_tagset {
164 	TAILQ_HEAD(, pf_tagname)	*namehash;
165 	TAILQ_HEAD(, pf_tagname)	*taghash;
166 	unsigned int			 mask;
167 	uint32_t			 seed;
168 	BITSET_DEFINE(, TAGID_MAX)	 avail;
169 };
170 
171 VNET_DEFINE(struct pf_tagset, pf_tags);
172 #define	V_pf_tags	VNET(pf_tags)
173 static unsigned int	pf_rule_tag_hashsize;
174 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
175 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
176     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
177     "Size of pf(4) rule tag hashtable");
178 
179 #ifdef ALTQ
180 VNET_DEFINE(struct pf_tagset, pf_qids);
181 #define	V_pf_qids	VNET(pf_qids)
182 static unsigned int	pf_queue_tag_hashsize;
183 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
184 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
185     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
186     "Size of pf(4) queue tag hashtable");
187 #endif
188 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
189 #define	V_pf_tag_z		 VNET(pf_tag_z)
190 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
191 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
192 
193 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
194 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
195 #endif
196 
197 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
198 #define V_pf_filter_local	VNET(pf_filter_local)
199 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
200     &VNET_NAME(pf_filter_local), false,
201     "Enable filtering for packets delivered to local network stack");
202 
203 #ifdef PF_DEFAULT_TO_DROP
204 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
205 #else
206 VNET_DEFINE_STATIC(bool, default_to_drop);
207 #endif
208 #define	V_default_to_drop VNET(default_to_drop)
209 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
210     &VNET_NAME(default_to_drop), false,
211     "Make the default rule drop all packets.");
212 
213 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
214 			    unsigned int);
215 static void		 pf_cleanup_tagset(struct pf_tagset *);
216 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
217 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
218 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
219 static u_int16_t	 pf_tagname2tag(const char *);
220 static void		 tag_unref(struct pf_tagset *, u_int16_t);
221 
222 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
223 
224 struct cdev *pf_dev;
225 
226 /*
227  * XXX - These are new and need to be checked when moveing to a new version
228  */
229 static void		 pf_clear_all_states(void);
230 static int		 pf_killstates_row(struct pf_kstate_kill *,
231 			    struct pf_idhash *);
232 static int		 pf_killstates_nv(struct pfioc_nv *);
233 static int		 pf_clearstates_nv(struct pfioc_nv *);
234 static int		 pf_getstate(struct pfioc_nv *);
235 static int		 pf_getstatus(struct pfioc_nv *);
236 static int		 pf_clear_tables(void);
237 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
238 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
239 static int		 pf_keepcounters(struct pfioc_nv *);
240 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
241 
242 /*
243  * Wrapper functions for pfil(9) hooks
244  */
245 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
246     int flags, void *ruleset __unused, struct inpcb *inp);
247 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
248     int flags, void *ruleset __unused, struct inpcb *inp);
249 #ifdef INET
250 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
251     int flags, void *ruleset __unused, struct inpcb *inp);
252 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
253     int flags, void *ruleset __unused, struct inpcb *inp);
254 #endif
255 #ifdef INET6
256 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
257     int flags, void *ruleset __unused, struct inpcb *inp);
258 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
259     int flags, void *ruleset __unused, struct inpcb *inp);
260 #endif
261 
262 static void		hook_pf_eth(void);
263 static void		hook_pf(void);
264 static void		dehook_pf_eth(void);
265 static void		dehook_pf(void);
266 static int		shutdown_pf(void);
267 static int		pf_load(void);
268 static void		pf_unload(void);
269 
270 static struct cdevsw pf_cdevsw = {
271 	.d_ioctl =	pfioctl,
272 	.d_name =	PF_NAME,
273 	.d_version =	D_VERSION,
274 };
275 
276 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
277 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
278 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
279 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
280 
281 /*
282  * We need a flag that is neither hooked nor running to know when
283  * the VNET is "valid".  We primarily need this to control (global)
284  * external event, e.g., eventhandlers.
285  */
286 VNET_DEFINE(int, pf_vnet_active);
287 #define V_pf_vnet_active	VNET(pf_vnet_active)
288 
289 int pf_end_threads;
290 struct proc *pf_purge_proc;
291 
292 VNET_DEFINE(struct rmlock, pf_rules_lock);
293 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
294 #define	V_pf_ioctl_lock		VNET(pf_ioctl_lock)
295 struct sx			pf_end_lock;
296 
297 /* pfsync */
298 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
299 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
300 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
301 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
302 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
303 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
304 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
305 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
306 
307 /* pflog */
308 pflog_packet_t			*pflog_packet_ptr = NULL;
309 
310 /*
311  * Copy a user-provided string, returning an error if truncation would occur.
312  * Avoid scanning past "sz" bytes in the source string since there's no
313  * guarantee that it's nul-terminated.
314  */
315 static int
316 pf_user_strcpy(char *dst, const char *src, size_t sz)
317 {
318 	if (strnlen(src, sz) == sz)
319 		return (EINVAL);
320 	(void)strlcpy(dst, src, sz);
321 	return (0);
322 }
323 
324 static void
325 pfattach_vnet(void)
326 {
327 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
328 
329 	bzero(&V_pf_status, sizeof(V_pf_status));
330 
331 	pf_initialize();
332 	pfr_initialize();
333 	pfi_initialize_vnet();
334 	pf_normalize_init();
335 	pf_syncookies_init();
336 
337 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
338 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
339 
340 	RB_INIT(&V_pf_anchors);
341 	pf_init_kruleset(&pf_main_ruleset);
342 
343 	pf_init_keth(V_pf_keth);
344 
345 	/* default rule should never be garbage collected */
346 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
347 	V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
348 	V_pf_default_rule.nr = -1;
349 	V_pf_default_rule.rtableid = -1;
350 
351 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
352 	for (int i = 0; i < 2; i++) {
353 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
354 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
355 	}
356 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
357 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
358 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
359 
360 	V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
361 	    M_WAITOK | M_ZERO);
362 
363 #ifdef PF_WANT_32_TO_64_COUNTER
364 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
365 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
366 	PF_RULES_WLOCK();
367 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
368 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
369 	V_pf_allrulecount++;
370 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
371 	PF_RULES_WUNLOCK();
372 #endif
373 
374 	/* initialize default timeouts */
375 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
376 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
377 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
378 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
379 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
380 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
381 	my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
382 	my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
383 	my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
384 	my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
385 	my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
386 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
387 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
388 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
389 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
390 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
391 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
392 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
393 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
394 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
395 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
396 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
397 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
398 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
399 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
400 
401 	V_pf_status.debug = PF_DEBUG_URGENT;
402 	/*
403 	 * XXX This is different than in OpenBSD where reassembly is enabled by
404 	 * defult. In FreeBSD we expect people to still use scrub rules and
405 	 * switch to the new syntax later. Only when they switch they must
406 	 * explicitly enable reassemle. We could change the default once the
407 	 * scrub rule functionality is hopefully removed some day in future.
408 	 */
409 	V_pf_status.reass = 0;
410 
411 	V_pf_pfil_hooked = false;
412 	V_pf_pfil_eth_hooked = false;
413 
414 	/* XXX do our best to avoid a conflict */
415 	V_pf_status.hostid = arc4random();
416 
417 	for (int i = 0; i < PFRES_MAX; i++)
418 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
419 	for (int i = 0; i < KLCNT_MAX; i++)
420 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
421 	for (int i = 0; i < FCNT_MAX; i++)
422 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
423 	for (int i = 0; i < SCNT_MAX; i++)
424 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
425 
426 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
427 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
428 		/* XXXGL: leaked all above. */
429 		return;
430 }
431 
432 static struct pf_kpool *
433 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
434     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
435     u_int8_t check_ticket)
436 {
437 	struct pf_kruleset	*ruleset;
438 	struct pf_krule		*rule;
439 	int			 rs_num;
440 
441 	ruleset = pf_find_kruleset(anchor);
442 	if (ruleset == NULL)
443 		return (NULL);
444 	rs_num = pf_get_ruleset_number(rule_action);
445 	if (rs_num >= PF_RULESET_MAX)
446 		return (NULL);
447 	if (active) {
448 		if (check_ticket && ticket !=
449 		    ruleset->rules[rs_num].active.ticket)
450 			return (NULL);
451 		if (r_last)
452 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
453 			    pf_krulequeue);
454 		else
455 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
456 	} else {
457 		if (check_ticket && ticket !=
458 		    ruleset->rules[rs_num].inactive.ticket)
459 			return (NULL);
460 		if (r_last)
461 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
462 			    pf_krulequeue);
463 		else
464 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
465 	}
466 	if (!r_last) {
467 		while ((rule != NULL) && (rule->nr != rule_number))
468 			rule = TAILQ_NEXT(rule, entries);
469 	}
470 	if (rule == NULL)
471 		return (NULL);
472 
473 	return (&rule->rpool);
474 }
475 
476 static void
477 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
478 {
479 	struct pf_kpooladdr	*mv_pool_pa;
480 
481 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
482 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
483 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
484 	}
485 }
486 
487 static void
488 pf_empty_kpool(struct pf_kpalist *poola)
489 {
490 	struct pf_kpooladdr *pa;
491 
492 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
493 		switch (pa->addr.type) {
494 		case PF_ADDR_DYNIFTL:
495 			pfi_dynaddr_remove(pa->addr.p.dyn);
496 			break;
497 		case PF_ADDR_TABLE:
498 			/* XXX: this could be unfinished pooladdr on pabuf */
499 			if (pa->addr.p.tbl != NULL)
500 				pfr_detach_table(pa->addr.p.tbl);
501 			break;
502 		}
503 		if (pa->kif)
504 			pfi_kkif_unref(pa->kif);
505 		TAILQ_REMOVE(poola, pa, entries);
506 		free(pa, M_PFRULE);
507 	}
508 }
509 
510 static void
511 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
512 {
513 
514 	PF_RULES_WASSERT();
515 	PF_UNLNKDRULES_ASSERT();
516 
517 	TAILQ_REMOVE(rulequeue, rule, entries);
518 
519 	rule->rule_ref |= PFRULE_REFS;
520 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
521 }
522 
523 static void
524 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
525 {
526 
527 	PF_RULES_WASSERT();
528 
529 	PF_UNLNKDRULES_LOCK();
530 	pf_unlink_rule_locked(rulequeue, rule);
531 	PF_UNLNKDRULES_UNLOCK();
532 }
533 
534 static void
535 pf_free_eth_rule(struct pf_keth_rule *rule)
536 {
537 	PF_RULES_WASSERT();
538 
539 	if (rule == NULL)
540 		return;
541 
542 	if (rule->tag)
543 		tag_unref(&V_pf_tags, rule->tag);
544 	if (rule->match_tag)
545 		tag_unref(&V_pf_tags, rule->match_tag);
546 #ifdef ALTQ
547 	pf_qid_unref(rule->qid);
548 #endif
549 
550 	if (rule->bridge_to)
551 		pfi_kkif_unref(rule->bridge_to);
552 	if (rule->kif)
553 		pfi_kkif_unref(rule->kif);
554 
555 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
556 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
557 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
558 		pfr_detach_table(rule->ipdst.addr.p.tbl);
559 
560 	counter_u64_free(rule->evaluations);
561 	for (int i = 0; i < 2; i++) {
562 		counter_u64_free(rule->packets[i]);
563 		counter_u64_free(rule->bytes[i]);
564 	}
565 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
566 	pf_keth_anchor_remove(rule);
567 
568 	free(rule, M_PFRULE);
569 }
570 
571 void
572 pf_free_rule(struct pf_krule *rule)
573 {
574 
575 	PF_RULES_WASSERT();
576 	PF_CONFIG_ASSERT();
577 
578 	if (rule->tag)
579 		tag_unref(&V_pf_tags, rule->tag);
580 	if (rule->match_tag)
581 		tag_unref(&V_pf_tags, rule->match_tag);
582 #ifdef ALTQ
583 	if (rule->pqid != rule->qid)
584 		pf_qid_unref(rule->pqid);
585 	pf_qid_unref(rule->qid);
586 #endif
587 	switch (rule->src.addr.type) {
588 	case PF_ADDR_DYNIFTL:
589 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
590 		break;
591 	case PF_ADDR_TABLE:
592 		pfr_detach_table(rule->src.addr.p.tbl);
593 		break;
594 	}
595 	switch (rule->dst.addr.type) {
596 	case PF_ADDR_DYNIFTL:
597 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
598 		break;
599 	case PF_ADDR_TABLE:
600 		pfr_detach_table(rule->dst.addr.p.tbl);
601 		break;
602 	}
603 	if (rule->overload_tbl)
604 		pfr_detach_table(rule->overload_tbl);
605 	if (rule->kif)
606 		pfi_kkif_unref(rule->kif);
607 	pf_kanchor_remove(rule);
608 	pf_empty_kpool(&rule->rpool.list);
609 
610 	pf_krule_free(rule);
611 }
612 
613 static void
614 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
615     unsigned int default_size)
616 {
617 	unsigned int i;
618 	unsigned int hashsize;
619 
620 	if (*tunable_size == 0 || !powerof2(*tunable_size))
621 		*tunable_size = default_size;
622 
623 	hashsize = *tunable_size;
624 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
625 	    M_WAITOK);
626 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
627 	    M_WAITOK);
628 	ts->mask = hashsize - 1;
629 	ts->seed = arc4random();
630 	for (i = 0; i < hashsize; i++) {
631 		TAILQ_INIT(&ts->namehash[i]);
632 		TAILQ_INIT(&ts->taghash[i]);
633 	}
634 	BIT_FILL(TAGID_MAX, &ts->avail);
635 }
636 
637 static void
638 pf_cleanup_tagset(struct pf_tagset *ts)
639 {
640 	unsigned int i;
641 	unsigned int hashsize;
642 	struct pf_tagname *t, *tmp;
643 
644 	/*
645 	 * Only need to clean up one of the hashes as each tag is hashed
646 	 * into each table.
647 	 */
648 	hashsize = ts->mask + 1;
649 	for (i = 0; i < hashsize; i++)
650 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
651 			uma_zfree(V_pf_tag_z, t);
652 
653 	free(ts->namehash, M_PFHASH);
654 	free(ts->taghash, M_PFHASH);
655 }
656 
657 static uint16_t
658 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
659 {
660 	size_t len;
661 
662 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
663 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
664 }
665 
666 static uint16_t
667 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
668 {
669 
670 	return (tag & ts->mask);
671 }
672 
673 static u_int16_t
674 tagname2tag(struct pf_tagset *ts, const char *tagname)
675 {
676 	struct pf_tagname	*tag;
677 	u_int32_t		 index;
678 	u_int16_t		 new_tagid;
679 
680 	PF_RULES_WASSERT();
681 
682 	index = tagname2hashindex(ts, tagname);
683 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
684 		if (strcmp(tagname, tag->name) == 0) {
685 			tag->ref++;
686 			return (tag->tag);
687 		}
688 
689 	/*
690 	 * new entry
691 	 *
692 	 * to avoid fragmentation, we do a linear search from the beginning
693 	 * and take the first free slot we find.
694 	 */
695 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
696 	/*
697 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
698 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
699 	 * set.  It may also return a bit number greater than TAGID_MAX due
700 	 * to rounding of the number of bits in the vector up to a multiple
701 	 * of the vector word size at declaration/allocation time.
702 	 */
703 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
704 		return (0);
705 
706 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
707 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
708 
709 	/* allocate and fill new struct pf_tagname */
710 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
711 	if (tag == NULL)
712 		return (0);
713 	strlcpy(tag->name, tagname, sizeof(tag->name));
714 	tag->tag = new_tagid;
715 	tag->ref = 1;
716 
717 	/* Insert into namehash */
718 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
719 
720 	/* Insert into taghash */
721 	index = tag2hashindex(ts, new_tagid);
722 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
723 
724 	return (tag->tag);
725 }
726 
727 static void
728 tag_unref(struct pf_tagset *ts, u_int16_t tag)
729 {
730 	struct pf_tagname	*t;
731 	uint16_t		 index;
732 
733 	PF_RULES_WASSERT();
734 
735 	index = tag2hashindex(ts, tag);
736 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
737 		if (tag == t->tag) {
738 			if (--t->ref == 0) {
739 				TAILQ_REMOVE(&ts->taghash[index], t,
740 				    taghash_entries);
741 				index = tagname2hashindex(ts, t->name);
742 				TAILQ_REMOVE(&ts->namehash[index], t,
743 				    namehash_entries);
744 				/* Bits are 0-based for BIT_SET() */
745 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
746 				uma_zfree(V_pf_tag_z, t);
747 			}
748 			break;
749 		}
750 }
751 
752 static uint16_t
753 pf_tagname2tag(const char *tagname)
754 {
755 	return (tagname2tag(&V_pf_tags, tagname));
756 }
757 
758 static int
759 pf_begin_eth(uint32_t *ticket, const char *anchor)
760 {
761 	struct pf_keth_rule *rule, *tmp;
762 	struct pf_keth_ruleset *rs;
763 
764 	PF_RULES_WASSERT();
765 
766 	rs = pf_find_or_create_keth_ruleset(anchor);
767 	if (rs == NULL)
768 		return (EINVAL);
769 
770 	/* Purge old inactive rules. */
771 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
772 	    tmp) {
773 		TAILQ_REMOVE(rs->inactive.rules, rule,
774 		    entries);
775 		pf_free_eth_rule(rule);
776 	}
777 
778 	*ticket = ++rs->inactive.ticket;
779 	rs->inactive.open = 1;
780 
781 	return (0);
782 }
783 
784 static void
785 pf_rollback_eth_cb(struct epoch_context *ctx)
786 {
787 	struct pf_keth_ruleset *rs;
788 
789 	rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
790 
791 	CURVNET_SET(rs->vnet);
792 
793 	PF_RULES_WLOCK();
794 	pf_rollback_eth(rs->inactive.ticket,
795 	    rs->anchor ? rs->anchor->path : "");
796 	PF_RULES_WUNLOCK();
797 
798 	CURVNET_RESTORE();
799 }
800 
801 static int
802 pf_rollback_eth(uint32_t ticket, const char *anchor)
803 {
804 	struct pf_keth_rule *rule, *tmp;
805 	struct pf_keth_ruleset *rs;
806 
807 	PF_RULES_WASSERT();
808 
809 	rs = pf_find_keth_ruleset(anchor);
810 	if (rs == NULL)
811 		return (EINVAL);
812 
813 	if (!rs->inactive.open ||
814 	    ticket != rs->inactive.ticket)
815 		return (0);
816 
817 	/* Purge old inactive rules. */
818 	TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
819 	    tmp) {
820 		TAILQ_REMOVE(rs->inactive.rules, rule, entries);
821 		pf_free_eth_rule(rule);
822 	}
823 
824 	rs->inactive.open = 0;
825 
826 	pf_remove_if_empty_keth_ruleset(rs);
827 
828 	return (0);
829 }
830 
831 #define	PF_SET_SKIP_STEPS(i)					\
832 	do {							\
833 		while (head[i] != cur) {			\
834 			head[i]->skip[i].ptr = cur;		\
835 			head[i] = TAILQ_NEXT(head[i], entries);	\
836 		}						\
837 	} while (0)
838 
839 static void
840 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
841 {
842 	struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
843 	int i;
844 
845 	cur = TAILQ_FIRST(rules);
846 	prev = cur;
847 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
848 		head[i] = cur;
849 	while (cur != NULL) {
850 		if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
851 			PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
852 		if (cur->direction != prev->direction)
853 			PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
854 		if (cur->proto != prev->proto)
855 			PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
856 		if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
857 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
858 		if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
859 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
860 		if (cur->ipsrc.neg != prev->ipsrc.neg ||
861 		    pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
862 			PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
863 		if (cur->ipdst.neg != prev->ipdst.neg ||
864 		    pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
865 			PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
866 
867 		prev = cur;
868 		cur = TAILQ_NEXT(cur, entries);
869 	}
870 	for (i = 0; i < PFE_SKIP_COUNT; ++i)
871 		PF_SET_SKIP_STEPS(i);
872 }
873 
874 static int
875 pf_commit_eth(uint32_t ticket, const char *anchor)
876 {
877 	struct pf_keth_ruleq *rules;
878 	struct pf_keth_ruleset *rs;
879 
880 	rs = pf_find_keth_ruleset(anchor);
881 	if (rs == NULL) {
882 		return (EINVAL);
883 	}
884 
885 	if (!rs->inactive.open ||
886 	    ticket != rs->inactive.ticket)
887 		return (EBUSY);
888 
889 	PF_RULES_WASSERT();
890 
891 	pf_eth_calc_skip_steps(rs->inactive.rules);
892 
893 	rules = rs->active.rules;
894 	ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
895 	rs->inactive.rules = rules;
896 	rs->inactive.ticket = rs->active.ticket;
897 
898 	/* Clean up inactive rules (i.e. previously active rules), only when
899 	 * we're sure they're no longer used. */
900 	NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
901 
902 	return (0);
903 }
904 
905 #ifdef ALTQ
906 static uint16_t
907 pf_qname2qid(const char *qname)
908 {
909 	return (tagname2tag(&V_pf_qids, qname));
910 }
911 
912 static void
913 pf_qid_unref(uint16_t qid)
914 {
915 	tag_unref(&V_pf_qids, qid);
916 }
917 
918 static int
919 pf_begin_altq(u_int32_t *ticket)
920 {
921 	struct pf_altq	*altq, *tmp;
922 	int		 error = 0;
923 
924 	PF_RULES_WASSERT();
925 
926 	/* Purge the old altq lists */
927 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
928 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
929 			/* detach and destroy the discipline */
930 			error = altq_remove(altq);
931 		}
932 		free(altq, M_PFALTQ);
933 	}
934 	TAILQ_INIT(V_pf_altq_ifs_inactive);
935 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
936 		pf_qid_unref(altq->qid);
937 		free(altq, M_PFALTQ);
938 	}
939 	TAILQ_INIT(V_pf_altqs_inactive);
940 	if (error)
941 		return (error);
942 	*ticket = ++V_ticket_altqs_inactive;
943 	V_altqs_inactive_open = 1;
944 	return (0);
945 }
946 
947 static int
948 pf_rollback_altq(u_int32_t ticket)
949 {
950 	struct pf_altq	*altq, *tmp;
951 	int		 error = 0;
952 
953 	PF_RULES_WASSERT();
954 
955 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
956 		return (0);
957 	/* Purge the old altq lists */
958 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
959 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
960 			/* detach and destroy the discipline */
961 			error = altq_remove(altq);
962 		}
963 		free(altq, M_PFALTQ);
964 	}
965 	TAILQ_INIT(V_pf_altq_ifs_inactive);
966 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
967 		pf_qid_unref(altq->qid);
968 		free(altq, M_PFALTQ);
969 	}
970 	TAILQ_INIT(V_pf_altqs_inactive);
971 	V_altqs_inactive_open = 0;
972 	return (error);
973 }
974 
975 static int
976 pf_commit_altq(u_int32_t ticket)
977 {
978 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
979 	struct pf_altq		*altq, *tmp;
980 	int			 err, error = 0;
981 
982 	PF_RULES_WASSERT();
983 
984 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
985 		return (EBUSY);
986 
987 	/* swap altqs, keep the old. */
988 	old_altqs = V_pf_altqs_active;
989 	old_altq_ifs = V_pf_altq_ifs_active;
990 	V_pf_altqs_active = V_pf_altqs_inactive;
991 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
992 	V_pf_altqs_inactive = old_altqs;
993 	V_pf_altq_ifs_inactive = old_altq_ifs;
994 	V_ticket_altqs_active = V_ticket_altqs_inactive;
995 
996 	/* Attach new disciplines */
997 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
998 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
999 			/* attach the discipline */
1000 			error = altq_pfattach(altq);
1001 			if (error == 0 && V_pf_altq_running)
1002 				error = pf_enable_altq(altq);
1003 			if (error != 0)
1004 				return (error);
1005 		}
1006 	}
1007 
1008 	/* Purge the old altq lists */
1009 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1010 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1011 			/* detach and destroy the discipline */
1012 			if (V_pf_altq_running)
1013 				error = pf_disable_altq(altq);
1014 			err = altq_pfdetach(altq);
1015 			if (err != 0 && error == 0)
1016 				error = err;
1017 			err = altq_remove(altq);
1018 			if (err != 0 && error == 0)
1019 				error = err;
1020 		}
1021 		free(altq, M_PFALTQ);
1022 	}
1023 	TAILQ_INIT(V_pf_altq_ifs_inactive);
1024 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1025 		pf_qid_unref(altq->qid);
1026 		free(altq, M_PFALTQ);
1027 	}
1028 	TAILQ_INIT(V_pf_altqs_inactive);
1029 
1030 	V_altqs_inactive_open = 0;
1031 	return (error);
1032 }
1033 
1034 static int
1035 pf_enable_altq(struct pf_altq *altq)
1036 {
1037 	struct ifnet		*ifp;
1038 	struct tb_profile	 tb;
1039 	int			 error = 0;
1040 
1041 	if ((ifp = ifunit(altq->ifname)) == NULL)
1042 		return (EINVAL);
1043 
1044 	if (ifp->if_snd.altq_type != ALTQT_NONE)
1045 		error = altq_enable(&ifp->if_snd);
1046 
1047 	/* set tokenbucket regulator */
1048 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1049 		tb.rate = altq->ifbandwidth;
1050 		tb.depth = altq->tbrsize;
1051 		error = tbr_set(&ifp->if_snd, &tb);
1052 	}
1053 
1054 	return (error);
1055 }
1056 
1057 static int
1058 pf_disable_altq(struct pf_altq *altq)
1059 {
1060 	struct ifnet		*ifp;
1061 	struct tb_profile	 tb;
1062 	int			 error;
1063 
1064 	if ((ifp = ifunit(altq->ifname)) == NULL)
1065 		return (EINVAL);
1066 
1067 	/*
1068 	 * when the discipline is no longer referenced, it was overridden
1069 	 * by a new one.  if so, just return.
1070 	 */
1071 	if (altq->altq_disc != ifp->if_snd.altq_disc)
1072 		return (0);
1073 
1074 	error = altq_disable(&ifp->if_snd);
1075 
1076 	if (error == 0) {
1077 		/* clear tokenbucket regulator */
1078 		tb.rate = 0;
1079 		error = tbr_set(&ifp->if_snd, &tb);
1080 	}
1081 
1082 	return (error);
1083 }
1084 
1085 static int
1086 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1087     struct pf_altq *altq)
1088 {
1089 	struct ifnet	*ifp1;
1090 	int		 error = 0;
1091 
1092 	/* Deactivate the interface in question */
1093 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1094 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1095 	    (remove && ifp1 == ifp)) {
1096 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1097 	} else {
1098 		error = altq_add(ifp1, altq);
1099 
1100 		if (ticket != V_ticket_altqs_inactive)
1101 			error = EBUSY;
1102 
1103 		if (error)
1104 			free(altq, M_PFALTQ);
1105 	}
1106 
1107 	return (error);
1108 }
1109 
1110 void
1111 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1112 {
1113 	struct pf_altq	*a1, *a2, *a3;
1114 	u_int32_t	 ticket;
1115 	int		 error = 0;
1116 
1117 	/*
1118 	 * No need to re-evaluate the configuration for events on interfaces
1119 	 * that do not support ALTQ, as it's not possible for such
1120 	 * interfaces to be part of the configuration.
1121 	 */
1122 	if (!ALTQ_IS_READY(&ifp->if_snd))
1123 		return;
1124 
1125 	/* Interrupt userland queue modifications */
1126 	if (V_altqs_inactive_open)
1127 		pf_rollback_altq(V_ticket_altqs_inactive);
1128 
1129 	/* Start new altq ruleset */
1130 	if (pf_begin_altq(&ticket))
1131 		return;
1132 
1133 	/* Copy the current active set */
1134 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1135 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1136 		if (a2 == NULL) {
1137 			error = ENOMEM;
1138 			break;
1139 		}
1140 		bcopy(a1, a2, sizeof(struct pf_altq));
1141 
1142 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1143 		if (error)
1144 			break;
1145 
1146 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1147 	}
1148 	if (error)
1149 		goto out;
1150 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1151 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1152 		if (a2 == NULL) {
1153 			error = ENOMEM;
1154 			break;
1155 		}
1156 		bcopy(a1, a2, sizeof(struct pf_altq));
1157 
1158 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1159 			error = EBUSY;
1160 			free(a2, M_PFALTQ);
1161 			break;
1162 		}
1163 		a2->altq_disc = NULL;
1164 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1165 			if (strncmp(a3->ifname, a2->ifname,
1166 				IFNAMSIZ) == 0) {
1167 				a2->altq_disc = a3->altq_disc;
1168 				break;
1169 			}
1170 		}
1171 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1172 		if (error)
1173 			break;
1174 
1175 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1176 	}
1177 
1178 out:
1179 	if (error != 0)
1180 		pf_rollback_altq(ticket);
1181 	else
1182 		pf_commit_altq(ticket);
1183 }
1184 #endif /* ALTQ */
1185 
1186 static struct pf_krule_global *
1187 pf_rule_tree_alloc(int flags)
1188 {
1189 	struct pf_krule_global *tree;
1190 
1191 	tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1192 	if (tree == NULL)
1193 		return (NULL);
1194 	RB_INIT(tree);
1195 	return (tree);
1196 }
1197 
1198 static void
1199 pf_rule_tree_free(struct pf_krule_global *tree)
1200 {
1201 
1202 	free(tree, M_TEMP);
1203 }
1204 
1205 static int
1206 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1207 {
1208 	struct pf_krule_global *tree;
1209 	struct pf_kruleset	*rs;
1210 	struct pf_krule		*rule;
1211 
1212 	PF_RULES_WASSERT();
1213 
1214 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1215 		return (EINVAL);
1216 	tree = pf_rule_tree_alloc(M_NOWAIT);
1217 	if (tree == NULL)
1218 		return (ENOMEM);
1219 	rs = pf_find_or_create_kruleset(anchor);
1220 	if (rs == NULL) {
1221 		free(tree, M_TEMP);
1222 		return (EINVAL);
1223 	}
1224 	pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1225 	rs->rules[rs_num].inactive.tree = tree;
1226 
1227 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1228 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1229 		rs->rules[rs_num].inactive.rcount--;
1230 	}
1231 	*ticket = ++rs->rules[rs_num].inactive.ticket;
1232 	rs->rules[rs_num].inactive.open = 1;
1233 	return (0);
1234 }
1235 
1236 static int
1237 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1238 {
1239 	struct pf_kruleset	*rs;
1240 	struct pf_krule		*rule;
1241 
1242 	PF_RULES_WASSERT();
1243 
1244 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1245 		return (EINVAL);
1246 	rs = pf_find_kruleset(anchor);
1247 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1248 	    rs->rules[rs_num].inactive.ticket != ticket)
1249 		return (0);
1250 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1251 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1252 		rs->rules[rs_num].inactive.rcount--;
1253 	}
1254 	rs->rules[rs_num].inactive.open = 0;
1255 	return (0);
1256 }
1257 
1258 #define PF_MD5_UPD(st, elm)						\
1259 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1260 
1261 #define PF_MD5_UPD_STR(st, elm)						\
1262 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1263 
1264 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
1265 		(stor) = htonl((st)->elm);				\
1266 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1267 } while (0)
1268 
1269 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
1270 		(stor) = htons((st)->elm);				\
1271 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1272 } while (0)
1273 
1274 static void
1275 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1276 {
1277 	PF_MD5_UPD(pfr, addr.type);
1278 	switch (pfr->addr.type) {
1279 		case PF_ADDR_DYNIFTL:
1280 			PF_MD5_UPD(pfr, addr.v.ifname);
1281 			PF_MD5_UPD(pfr, addr.iflags);
1282 			break;
1283 		case PF_ADDR_TABLE:
1284 			PF_MD5_UPD(pfr, addr.v.tblname);
1285 			break;
1286 		case PF_ADDR_ADDRMASK:
1287 			/* XXX ignore af? */
1288 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1289 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1290 			break;
1291 	}
1292 
1293 	PF_MD5_UPD(pfr, port[0]);
1294 	PF_MD5_UPD(pfr, port[1]);
1295 	PF_MD5_UPD(pfr, neg);
1296 	PF_MD5_UPD(pfr, port_op);
1297 }
1298 
1299 static void
1300 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1301 {
1302 	u_int16_t x;
1303 	u_int32_t y;
1304 
1305 	pf_hash_rule_addr(ctx, &rule->src);
1306 	pf_hash_rule_addr(ctx, &rule->dst);
1307 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1308 		PF_MD5_UPD_STR(rule, label[i]);
1309 	PF_MD5_UPD_STR(rule, ifname);
1310 	PF_MD5_UPD_STR(rule, match_tagname);
1311 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1312 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1313 	PF_MD5_UPD_HTONL(rule, prob, y);
1314 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1315 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1316 	PF_MD5_UPD(rule, uid.op);
1317 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1318 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1319 	PF_MD5_UPD(rule, gid.op);
1320 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1321 	PF_MD5_UPD(rule, action);
1322 	PF_MD5_UPD(rule, direction);
1323 	PF_MD5_UPD(rule, af);
1324 	PF_MD5_UPD(rule, quick);
1325 	PF_MD5_UPD(rule, ifnot);
1326 	PF_MD5_UPD(rule, match_tag_not);
1327 	PF_MD5_UPD(rule, natpass);
1328 	PF_MD5_UPD(rule, keep_state);
1329 	PF_MD5_UPD(rule, proto);
1330 	PF_MD5_UPD(rule, type);
1331 	PF_MD5_UPD(rule, code);
1332 	PF_MD5_UPD(rule, flags);
1333 	PF_MD5_UPD(rule, flagset);
1334 	PF_MD5_UPD(rule, allow_opts);
1335 	PF_MD5_UPD(rule, rt);
1336 	PF_MD5_UPD(rule, tos);
1337 	PF_MD5_UPD(rule, scrub_flags);
1338 	PF_MD5_UPD(rule, min_ttl);
1339 	PF_MD5_UPD(rule, set_tos);
1340 	if (rule->anchor != NULL)
1341 		PF_MD5_UPD_STR(rule, anchor->path);
1342 }
1343 
1344 static void
1345 pf_hash_rule(struct pf_krule *rule)
1346 {
1347 	MD5_CTX		ctx;
1348 
1349 	MD5Init(&ctx);
1350 	pf_hash_rule_rolling(&ctx, rule);
1351 	MD5Final(rule->md5sum, &ctx);
1352 }
1353 
1354 static int
1355 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1356 {
1357 
1358 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1359 }
1360 
1361 static int
1362 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1363 {
1364 	struct pf_kruleset	*rs;
1365 	struct pf_krule		*rule, **old_array, *old_rule;
1366 	struct pf_krulequeue	*old_rules;
1367 	struct pf_krule_global  *old_tree;
1368 	int			 error;
1369 	u_int32_t		 old_rcount;
1370 
1371 	PF_RULES_WASSERT();
1372 
1373 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1374 		return (EINVAL);
1375 	rs = pf_find_kruleset(anchor);
1376 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1377 	    ticket != rs->rules[rs_num].inactive.ticket)
1378 		return (EBUSY);
1379 
1380 	/* Calculate checksum for the main ruleset */
1381 	if (rs == &pf_main_ruleset) {
1382 		error = pf_setup_pfsync_matching(rs);
1383 		if (error != 0)
1384 			return (error);
1385 	}
1386 
1387 	/* Swap rules, keep the old. */
1388 	old_rules = rs->rules[rs_num].active.ptr;
1389 	old_rcount = rs->rules[rs_num].active.rcount;
1390 	old_array = rs->rules[rs_num].active.ptr_array;
1391 	old_tree = rs->rules[rs_num].active.tree;
1392 
1393 	rs->rules[rs_num].active.ptr =
1394 	    rs->rules[rs_num].inactive.ptr;
1395 	rs->rules[rs_num].active.ptr_array =
1396 	    rs->rules[rs_num].inactive.ptr_array;
1397 	rs->rules[rs_num].active.tree =
1398 	    rs->rules[rs_num].inactive.tree;
1399 	rs->rules[rs_num].active.rcount =
1400 	    rs->rules[rs_num].inactive.rcount;
1401 
1402 	/* Attempt to preserve counter information. */
1403 	if (V_pf_status.keep_counters && old_tree != NULL) {
1404 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1405 		    entries) {
1406 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1407 			if (old_rule == NULL) {
1408 				continue;
1409 			}
1410 			pf_counter_u64_critical_enter();
1411 			pf_counter_u64_rollup_protected(&rule->evaluations,
1412 			    pf_counter_u64_fetch(&old_rule->evaluations));
1413 			pf_counter_u64_rollup_protected(&rule->packets[0],
1414 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1415 			pf_counter_u64_rollup_protected(&rule->packets[1],
1416 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1417 			pf_counter_u64_rollup_protected(&rule->bytes[0],
1418 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1419 			pf_counter_u64_rollup_protected(&rule->bytes[1],
1420 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1421 			pf_counter_u64_critical_exit();
1422 		}
1423 	}
1424 
1425 	rs->rules[rs_num].inactive.ptr = old_rules;
1426 	rs->rules[rs_num].inactive.ptr_array = old_array;
1427 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1428 	rs->rules[rs_num].inactive.rcount = old_rcount;
1429 
1430 	rs->rules[rs_num].active.ticket =
1431 	    rs->rules[rs_num].inactive.ticket;
1432 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1433 
1434 	/* Purge the old rule list. */
1435 	PF_UNLNKDRULES_LOCK();
1436 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1437 		pf_unlink_rule_locked(old_rules, rule);
1438 	PF_UNLNKDRULES_UNLOCK();
1439 	if (rs->rules[rs_num].inactive.ptr_array)
1440 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1441 	rs->rules[rs_num].inactive.ptr_array = NULL;
1442 	rs->rules[rs_num].inactive.rcount = 0;
1443 	rs->rules[rs_num].inactive.open = 0;
1444 	pf_remove_if_empty_kruleset(rs);
1445 	free(old_tree, M_TEMP);
1446 
1447 	return (0);
1448 }
1449 
1450 static int
1451 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1452 {
1453 	MD5_CTX			 ctx;
1454 	struct pf_krule		*rule;
1455 	int			 rs_cnt;
1456 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1457 
1458 	MD5Init(&ctx);
1459 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1460 		/* XXX PF_RULESET_SCRUB as well? */
1461 		if (rs_cnt == PF_RULESET_SCRUB)
1462 			continue;
1463 
1464 		if (rs->rules[rs_cnt].inactive.ptr_array)
1465 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1466 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1467 
1468 		if (rs->rules[rs_cnt].inactive.rcount) {
1469 			rs->rules[rs_cnt].inactive.ptr_array =
1470 			    mallocarray(rs->rules[rs_cnt].inactive.rcount,
1471 			    sizeof(struct pf_rule **),
1472 			    M_TEMP, M_NOWAIT);
1473 
1474 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1475 				return (ENOMEM);
1476 		}
1477 
1478 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1479 		    entries) {
1480 			pf_hash_rule_rolling(&ctx, rule);
1481 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1482 		}
1483 	}
1484 
1485 	MD5Final(digest, &ctx);
1486 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1487 	return (0);
1488 }
1489 
1490 static int
1491 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1492 {
1493 	int error = 0;
1494 
1495 	switch (addr->type) {
1496 	case PF_ADDR_TABLE:
1497 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1498 		if (addr->p.tbl == NULL)
1499 			error = ENOMEM;
1500 		break;
1501 	default:
1502 		error = EINVAL;
1503 	}
1504 
1505 	return (error);
1506 }
1507 
1508 static int
1509 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1510     sa_family_t af)
1511 {
1512 	int error = 0;
1513 
1514 	switch (addr->type) {
1515 	case PF_ADDR_TABLE:
1516 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1517 		if (addr->p.tbl == NULL)
1518 			error = ENOMEM;
1519 		break;
1520 	case PF_ADDR_DYNIFTL:
1521 		error = pfi_dynaddr_setup(addr, af);
1522 		break;
1523 	}
1524 
1525 	return (error);
1526 }
1527 
1528 static void
1529 pf_addr_copyout(struct pf_addr_wrap *addr)
1530 {
1531 
1532 	switch (addr->type) {
1533 	case PF_ADDR_DYNIFTL:
1534 		pfi_dynaddr_copyout(addr);
1535 		break;
1536 	case PF_ADDR_TABLE:
1537 		pf_tbladdr_copyout(addr);
1538 		break;
1539 	}
1540 }
1541 
1542 static void
1543 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1544 {
1545 	int	secs = time_uptime, diff;
1546 
1547 	bzero(out, sizeof(struct pf_src_node));
1548 
1549 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1550 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1551 
1552 	if (in->rule.ptr != NULL)
1553 		out->rule.nr = in->rule.ptr->nr;
1554 
1555 	for (int i = 0; i < 2; i++) {
1556 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1557 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1558 	}
1559 
1560 	out->states = in->states;
1561 	out->conn = in->conn;
1562 	out->af = in->af;
1563 	out->ruletype = in->ruletype;
1564 
1565 	out->creation = secs - in->creation;
1566 	if (out->expire > secs)
1567 		out->expire -= secs;
1568 	else
1569 		out->expire = 0;
1570 
1571 	/* Adjust the connection rate estimate. */
1572 	diff = secs - in->conn_rate.last;
1573 	if (diff >= in->conn_rate.seconds)
1574 		out->conn_rate.count = 0;
1575 	else
1576 		out->conn_rate.count -=
1577 		    in->conn_rate.count * diff /
1578 		    in->conn_rate.seconds;
1579 }
1580 
1581 #ifdef ALTQ
1582 /*
1583  * Handle export of struct pf_kaltq to user binaries that may be using any
1584  * version of struct pf_altq.
1585  */
1586 static int
1587 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1588 {
1589 	u_int32_t version;
1590 
1591 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1592 		version = 0;
1593 	else
1594 		version = pa->version;
1595 
1596 	if (version > PFIOC_ALTQ_VERSION)
1597 		return (EINVAL);
1598 
1599 #define ASSIGN(x) exported_q->x = q->x
1600 #define COPY(x) \
1601 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1602 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1603 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1604 
1605 	switch (version) {
1606 	case 0: {
1607 		struct pf_altq_v0 *exported_q =
1608 		    &((struct pfioc_altq_v0 *)pa)->altq;
1609 
1610 		COPY(ifname);
1611 
1612 		ASSIGN(scheduler);
1613 		ASSIGN(tbrsize);
1614 		exported_q->tbrsize = SATU16(q->tbrsize);
1615 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1616 
1617 		COPY(qname);
1618 		COPY(parent);
1619 		ASSIGN(parent_qid);
1620 		exported_q->bandwidth = SATU32(q->bandwidth);
1621 		ASSIGN(priority);
1622 		ASSIGN(local_flags);
1623 
1624 		ASSIGN(qlimit);
1625 		ASSIGN(flags);
1626 
1627 		if (q->scheduler == ALTQT_HFSC) {
1628 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1629 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1630 			    SATU32(q->pq_u.hfsc_opts.x)
1631 
1632 			ASSIGN_OPT_SATU32(rtsc_m1);
1633 			ASSIGN_OPT(rtsc_d);
1634 			ASSIGN_OPT_SATU32(rtsc_m2);
1635 
1636 			ASSIGN_OPT_SATU32(lssc_m1);
1637 			ASSIGN_OPT(lssc_d);
1638 			ASSIGN_OPT_SATU32(lssc_m2);
1639 
1640 			ASSIGN_OPT_SATU32(ulsc_m1);
1641 			ASSIGN_OPT(ulsc_d);
1642 			ASSIGN_OPT_SATU32(ulsc_m2);
1643 
1644 			ASSIGN_OPT(flags);
1645 
1646 #undef ASSIGN_OPT
1647 #undef ASSIGN_OPT_SATU32
1648 		} else
1649 			COPY(pq_u);
1650 
1651 		ASSIGN(qid);
1652 		break;
1653 	}
1654 	case 1:	{
1655 		struct pf_altq_v1 *exported_q =
1656 		    &((struct pfioc_altq_v1 *)pa)->altq;
1657 
1658 		COPY(ifname);
1659 
1660 		ASSIGN(scheduler);
1661 		ASSIGN(tbrsize);
1662 		ASSIGN(ifbandwidth);
1663 
1664 		COPY(qname);
1665 		COPY(parent);
1666 		ASSIGN(parent_qid);
1667 		ASSIGN(bandwidth);
1668 		ASSIGN(priority);
1669 		ASSIGN(local_flags);
1670 
1671 		ASSIGN(qlimit);
1672 		ASSIGN(flags);
1673 		COPY(pq_u);
1674 
1675 		ASSIGN(qid);
1676 		break;
1677 	}
1678 	default:
1679 		panic("%s: unhandled struct pfioc_altq version", __func__);
1680 		break;
1681 	}
1682 
1683 #undef ASSIGN
1684 #undef COPY
1685 #undef SATU16
1686 #undef SATU32
1687 
1688 	return (0);
1689 }
1690 
1691 /*
1692  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1693  * that may be using any version of it.
1694  */
1695 static int
1696 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1697 {
1698 	u_int32_t version;
1699 
1700 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1701 		version = 0;
1702 	else
1703 		version = pa->version;
1704 
1705 	if (version > PFIOC_ALTQ_VERSION)
1706 		return (EINVAL);
1707 
1708 #define ASSIGN(x) q->x = imported_q->x
1709 #define COPY(x) \
1710 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1711 
1712 	switch (version) {
1713 	case 0: {
1714 		struct pf_altq_v0 *imported_q =
1715 		    &((struct pfioc_altq_v0 *)pa)->altq;
1716 
1717 		COPY(ifname);
1718 
1719 		ASSIGN(scheduler);
1720 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1721 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1722 
1723 		COPY(qname);
1724 		COPY(parent);
1725 		ASSIGN(parent_qid);
1726 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1727 		ASSIGN(priority);
1728 		ASSIGN(local_flags);
1729 
1730 		ASSIGN(qlimit);
1731 		ASSIGN(flags);
1732 
1733 		if (imported_q->scheduler == ALTQT_HFSC) {
1734 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1735 
1736 			/*
1737 			 * The m1 and m2 parameters are being copied from
1738 			 * 32-bit to 64-bit.
1739 			 */
1740 			ASSIGN_OPT(rtsc_m1);
1741 			ASSIGN_OPT(rtsc_d);
1742 			ASSIGN_OPT(rtsc_m2);
1743 
1744 			ASSIGN_OPT(lssc_m1);
1745 			ASSIGN_OPT(lssc_d);
1746 			ASSIGN_OPT(lssc_m2);
1747 
1748 			ASSIGN_OPT(ulsc_m1);
1749 			ASSIGN_OPT(ulsc_d);
1750 			ASSIGN_OPT(ulsc_m2);
1751 
1752 			ASSIGN_OPT(flags);
1753 
1754 #undef ASSIGN_OPT
1755 		} else
1756 			COPY(pq_u);
1757 
1758 		ASSIGN(qid);
1759 		break;
1760 	}
1761 	case 1: {
1762 		struct pf_altq_v1 *imported_q =
1763 		    &((struct pfioc_altq_v1 *)pa)->altq;
1764 
1765 		COPY(ifname);
1766 
1767 		ASSIGN(scheduler);
1768 		ASSIGN(tbrsize);
1769 		ASSIGN(ifbandwidth);
1770 
1771 		COPY(qname);
1772 		COPY(parent);
1773 		ASSIGN(parent_qid);
1774 		ASSIGN(bandwidth);
1775 		ASSIGN(priority);
1776 		ASSIGN(local_flags);
1777 
1778 		ASSIGN(qlimit);
1779 		ASSIGN(flags);
1780 		COPY(pq_u);
1781 
1782 		ASSIGN(qid);
1783 		break;
1784 	}
1785 	default:
1786 		panic("%s: unhandled struct pfioc_altq version", __func__);
1787 		break;
1788 	}
1789 
1790 #undef ASSIGN
1791 #undef COPY
1792 
1793 	return (0);
1794 }
1795 
1796 static struct pf_altq *
1797 pf_altq_get_nth_active(u_int32_t n)
1798 {
1799 	struct pf_altq		*altq;
1800 	u_int32_t		 nr;
1801 
1802 	nr = 0;
1803 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1804 		if (nr == n)
1805 			return (altq);
1806 		nr++;
1807 	}
1808 
1809 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1810 		if (nr == n)
1811 			return (altq);
1812 		nr++;
1813 	}
1814 
1815 	return (NULL);
1816 }
1817 #endif /* ALTQ */
1818 
1819 struct pf_krule *
1820 pf_krule_alloc(void)
1821 {
1822 	struct pf_krule *rule;
1823 
1824 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1825 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1826 	rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1827 	    M_WAITOK | M_ZERO);
1828 	return (rule);
1829 }
1830 
1831 void
1832 pf_krule_free(struct pf_krule *rule)
1833 {
1834 #ifdef PF_WANT_32_TO_64_COUNTER
1835 	bool wowned;
1836 #endif
1837 
1838 	if (rule == NULL)
1839 		return;
1840 
1841 #ifdef PF_WANT_32_TO_64_COUNTER
1842 	if (rule->allrulelinked) {
1843 		wowned = PF_RULES_WOWNED();
1844 		if (!wowned)
1845 			PF_RULES_WLOCK();
1846 		LIST_REMOVE(rule, allrulelist);
1847 		V_pf_allrulecount--;
1848 		if (!wowned)
1849 			PF_RULES_WUNLOCK();
1850 	}
1851 #endif
1852 
1853 	pf_counter_u64_deinit(&rule->evaluations);
1854 	for (int i = 0; i < 2; i++) {
1855 		pf_counter_u64_deinit(&rule->packets[i]);
1856 		pf_counter_u64_deinit(&rule->bytes[i]);
1857 	}
1858 	counter_u64_free(rule->states_cur);
1859 	counter_u64_free(rule->states_tot);
1860 	counter_u64_free(rule->src_nodes);
1861 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1862 
1863 	mtx_destroy(&rule->rpool.mtx);
1864 	free(rule, M_PFRULE);
1865 }
1866 
1867 void
1868 pf_krule_clear_counters(struct pf_krule *rule)
1869 {
1870 	pf_counter_u64_zero(&rule->evaluations);
1871 	for (int i = 0; i < 2; i++) {
1872 		pf_counter_u64_zero(&rule->packets[i]);
1873 		pf_counter_u64_zero(&rule->bytes[i]);
1874 	}
1875 	counter_u64_zero(rule->states_tot);
1876 }
1877 
1878 static void
1879 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1880     struct pf_pooladdr *pool)
1881 {
1882 
1883 	bzero(pool, sizeof(*pool));
1884 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1885 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1886 }
1887 
1888 static int
1889 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1890     struct pf_kpooladdr *kpool)
1891 {
1892 	int ret;
1893 
1894 	bzero(kpool, sizeof(*kpool));
1895 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1896 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1897 	    sizeof(kpool->ifname));
1898 	return (ret);
1899 }
1900 
1901 static void
1902 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1903 {
1904 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1905 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1906 
1907 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1908 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1909 
1910 	kpool->tblidx = pool->tblidx;
1911 	kpool->proxy_port[0] = pool->proxy_port[0];
1912 	kpool->proxy_port[1] = pool->proxy_port[1];
1913 	kpool->opts = pool->opts;
1914 }
1915 
1916 static int
1917 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1918 {
1919 	int ret;
1920 
1921 #ifndef INET
1922 	if (rule->af == AF_INET) {
1923 		return (EAFNOSUPPORT);
1924 	}
1925 #endif /* INET */
1926 #ifndef INET6
1927 	if (rule->af == AF_INET6) {
1928 		return (EAFNOSUPPORT);
1929 	}
1930 #endif /* INET6 */
1931 
1932 	ret = pf_check_rule_addr(&rule->src);
1933 	if (ret != 0)
1934 		return (ret);
1935 	ret = pf_check_rule_addr(&rule->dst);
1936 	if (ret != 0)
1937 		return (ret);
1938 
1939 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1940 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1941 
1942 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1943 	if (ret != 0)
1944 		return (ret);
1945 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1946 	if (ret != 0)
1947 		return (ret);
1948 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1949 	if (ret != 0)
1950 		return (ret);
1951 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1952 	if (ret != 0)
1953 		return (ret);
1954 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
1955 	    sizeof(rule->tagname));
1956 	if (ret != 0)
1957 		return (ret);
1958 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1959 	    sizeof(rule->match_tagname));
1960 	if (ret != 0)
1961 		return (ret);
1962 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1963 	    sizeof(rule->overload_tblname));
1964 	if (ret != 0)
1965 		return (ret);
1966 
1967 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1968 
1969 	/* Don't allow userspace to set evaluations, packets or bytes. */
1970 	/* kif, anchor, overload_tbl are not copied over. */
1971 
1972 	krule->os_fingerprint = rule->os_fingerprint;
1973 
1974 	krule->rtableid = rule->rtableid;
1975 	/* pf_rule->timeout is smaller than pf_krule->timeout */
1976 	bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1977 	krule->max_states = rule->max_states;
1978 	krule->max_src_nodes = rule->max_src_nodes;
1979 	krule->max_src_states = rule->max_src_states;
1980 	krule->max_src_conn = rule->max_src_conn;
1981 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1982 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1983 	krule->qid = rule->qid;
1984 	krule->pqid = rule->pqid;
1985 	krule->nr = rule->nr;
1986 	krule->prob = rule->prob;
1987 	krule->cuid = rule->cuid;
1988 	krule->cpid = rule->cpid;
1989 
1990 	krule->return_icmp = rule->return_icmp;
1991 	krule->return_icmp6 = rule->return_icmp6;
1992 	krule->max_mss = rule->max_mss;
1993 	krule->tag = rule->tag;
1994 	krule->match_tag = rule->match_tag;
1995 	krule->scrub_flags = rule->scrub_flags;
1996 
1997 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1998 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1999 
2000 	krule->rule_flag = rule->rule_flag;
2001 	krule->action = rule->action;
2002 	krule->direction = rule->direction;
2003 	krule->log = rule->log;
2004 	krule->logif = rule->logif;
2005 	krule->quick = rule->quick;
2006 	krule->ifnot = rule->ifnot;
2007 	krule->match_tag_not = rule->match_tag_not;
2008 	krule->natpass = rule->natpass;
2009 
2010 	krule->keep_state = rule->keep_state;
2011 	krule->af = rule->af;
2012 	krule->proto = rule->proto;
2013 	krule->type = rule->type;
2014 	krule->code = rule->code;
2015 	krule->flags = rule->flags;
2016 	krule->flagset = rule->flagset;
2017 	krule->min_ttl = rule->min_ttl;
2018 	krule->allow_opts = rule->allow_opts;
2019 	krule->rt = rule->rt;
2020 	krule->return_ttl = rule->return_ttl;
2021 	krule->tos = rule->tos;
2022 	krule->set_tos = rule->set_tos;
2023 
2024 	krule->flush = rule->flush;
2025 	krule->prio = rule->prio;
2026 	krule->set_prio[0] = rule->set_prio[0];
2027 	krule->set_prio[1] = rule->set_prio[1];
2028 
2029 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2030 
2031 	return (0);
2032 }
2033 
2034 int
2035 pf_ioctl_getrules(struct pfioc_rule *pr)
2036 {
2037 	struct pf_kruleset	*ruleset;
2038 	struct pf_krule		*tail;
2039 	int			 rs_num;
2040 
2041 	PF_RULES_WLOCK();
2042 	ruleset = pf_find_kruleset(pr->anchor);
2043 	if (ruleset == NULL) {
2044 		PF_RULES_WUNLOCK();
2045 		return (EINVAL);
2046 	}
2047 	rs_num = pf_get_ruleset_number(pr->rule.action);
2048 	if (rs_num >= PF_RULESET_MAX) {
2049 		PF_RULES_WUNLOCK();
2050 		return (EINVAL);
2051 	}
2052 	tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2053 	    pf_krulequeue);
2054 	if (tail)
2055 		pr->nr = tail->nr + 1;
2056 	else
2057 		pr->nr = 0;
2058 	pr->ticket = ruleset->rules[rs_num].active.ticket;
2059 	PF_RULES_WUNLOCK();
2060 
2061 	return (0);
2062 }
2063 
2064 int
2065 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2066     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2067     uid_t uid, pid_t pid)
2068 {
2069 	struct pf_kruleset	*ruleset;
2070 	struct pf_krule		*tail;
2071 	struct pf_kpooladdr	*pa;
2072 	struct pfi_kkif		*kif = NULL;
2073 	int			 rs_num;
2074 	int			 error = 0;
2075 
2076 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2077 		error = EINVAL;
2078 		goto errout_unlocked;
2079 	}
2080 
2081 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2082 
2083 	if (rule->ifname[0])
2084 		kif = pf_kkif_create(M_WAITOK);
2085 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2086 	for (int i = 0; i < 2; i++) {
2087 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2088 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2089 	}
2090 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2091 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2092 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2093 	rule->cuid = uid;
2094 	rule->cpid = pid;
2095 	TAILQ_INIT(&rule->rpool.list);
2096 
2097 	PF_CONFIG_LOCK();
2098 	PF_RULES_WLOCK();
2099 #ifdef PF_WANT_32_TO_64_COUNTER
2100 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2101 	MPASS(!rule->allrulelinked);
2102 	rule->allrulelinked = true;
2103 	V_pf_allrulecount++;
2104 #endif
2105 	ruleset = pf_find_kruleset(anchor);
2106 	if (ruleset == NULL)
2107 		ERROUT(EINVAL);
2108 	rs_num = pf_get_ruleset_number(rule->action);
2109 	if (rs_num >= PF_RULESET_MAX)
2110 		ERROUT(EINVAL);
2111 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2112 		DPFPRINTF(PF_DEBUG_MISC,
2113 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2114 		    ruleset->rules[rs_num].inactive.ticket));
2115 		ERROUT(EBUSY);
2116 	}
2117 	if (pool_ticket != V_ticket_pabuf) {
2118 		DPFPRINTF(PF_DEBUG_MISC,
2119 		    ("pool_ticket: %d != %d\n", pool_ticket,
2120 		    V_ticket_pabuf));
2121 		ERROUT(EBUSY);
2122 	}
2123 	/*
2124 	 * XXXMJG hack: there is no mechanism to ensure they started the
2125 	 * transaction. Ticket checked above may happen to match by accident,
2126 	 * even if nobody called DIOCXBEGIN, let alone this process.
2127 	 * Partially work around it by checking if the RB tree got allocated,
2128 	 * see pf_begin_rules.
2129 	 */
2130 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2131 		ERROUT(EINVAL);
2132 	}
2133 
2134 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2135 	    pf_krulequeue);
2136 	if (tail)
2137 		rule->nr = tail->nr + 1;
2138 	else
2139 		rule->nr = 0;
2140 	if (rule->ifname[0]) {
2141 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2142 		kif = NULL;
2143 		pfi_kkif_ref(rule->kif);
2144 	} else
2145 		rule->kif = NULL;
2146 
2147 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2148 		error = EBUSY;
2149 
2150 #ifdef ALTQ
2151 	/* set queue IDs */
2152 	if (rule->qname[0] != 0) {
2153 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2154 			error = EBUSY;
2155 		else if (rule->pqname[0] != 0) {
2156 			if ((rule->pqid =
2157 			    pf_qname2qid(rule->pqname)) == 0)
2158 				error = EBUSY;
2159 		} else
2160 			rule->pqid = rule->qid;
2161 	}
2162 #endif
2163 	if (rule->tagname[0])
2164 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2165 			error = EBUSY;
2166 	if (rule->match_tagname[0])
2167 		if ((rule->match_tag =
2168 		    pf_tagname2tag(rule->match_tagname)) == 0)
2169 			error = EBUSY;
2170 	if (rule->rt && !rule->direction)
2171 		error = EINVAL;
2172 	if (!rule->log)
2173 		rule->logif = 0;
2174 	if (rule->logif >= PFLOGIFS_MAX)
2175 		error = EINVAL;
2176 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2177 		error = ENOMEM;
2178 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2179 		error = ENOMEM;
2180 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2181 		error = EINVAL;
2182 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2183 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2184 	    rule->set_prio[1] > PF_PRIO_MAX))
2185 		error = EINVAL;
2186 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2187 		if (pa->addr.type == PF_ADDR_TABLE) {
2188 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2189 			    pa->addr.v.tblname);
2190 			if (pa->addr.p.tbl == NULL)
2191 				error = ENOMEM;
2192 		}
2193 
2194 	rule->overload_tbl = NULL;
2195 	if (rule->overload_tblname[0]) {
2196 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2197 		    rule->overload_tblname)) == NULL)
2198 			error = EINVAL;
2199 		else
2200 			rule->overload_tbl->pfrkt_flags |=
2201 			    PFR_TFLAG_ACTIVE;
2202 	}
2203 
2204 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2205 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2206 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2207 	    (rule->rt > PF_NOPFROUTE)) &&
2208 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2209 		error = EINVAL;
2210 
2211 	if (error) {
2212 		pf_free_rule(rule);
2213 		rule = NULL;
2214 		ERROUT(error);
2215 	}
2216 
2217 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2218 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2219 	    rule, entries);
2220 	ruleset->rules[rs_num].inactive.rcount++;
2221 
2222 	PF_RULES_WUNLOCK();
2223 	pf_hash_rule(rule);
2224 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2225 		PF_RULES_WLOCK();
2226 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2227 		ruleset->rules[rs_num].inactive.rcount--;
2228 		pf_free_rule(rule);
2229 		rule = NULL;
2230 		ERROUT(EEXIST);
2231 	}
2232 	PF_CONFIG_UNLOCK();
2233 
2234 	return (0);
2235 
2236 #undef ERROUT
2237 errout:
2238 	PF_RULES_WUNLOCK();
2239 	PF_CONFIG_UNLOCK();
2240 errout_unlocked:
2241 	pf_kkif_free(kif);
2242 	pf_krule_free(rule);
2243 	return (error);
2244 }
2245 
2246 static bool
2247 pf_label_match(const struct pf_krule *rule, const char *label)
2248 {
2249 	int i = 0;
2250 
2251 	while (*rule->label[i]) {
2252 		if (strcmp(rule->label[i], label) == 0)
2253 			return (true);
2254 		i++;
2255 	}
2256 
2257 	return (false);
2258 }
2259 
2260 static unsigned int
2261 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2262 {
2263 	struct pf_kstate *s;
2264 	int more = 0;
2265 
2266 	s = pf_find_state_all(key, dir, &more);
2267 	if (s == NULL)
2268 		return (0);
2269 
2270 	if (more) {
2271 		PF_STATE_UNLOCK(s);
2272 		return (0);
2273 	}
2274 
2275 	pf_unlink_state(s);
2276 	return (1);
2277 }
2278 
2279 static int
2280 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2281 {
2282 	struct pf_kstate	*s;
2283 	struct pf_state_key	*sk;
2284 	struct pf_addr		*srcaddr, *dstaddr;
2285 	struct pf_state_key_cmp	 match_key;
2286 	int			 idx, killed = 0;
2287 	unsigned int		 dir;
2288 	u_int16_t		 srcport, dstport;
2289 	struct pfi_kkif		*kif;
2290 
2291 relock_DIOCKILLSTATES:
2292 	PF_HASHROW_LOCK(ih);
2293 	LIST_FOREACH(s, &ih->states, entry) {
2294 		/* For floating states look at the original kif. */
2295 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2296 
2297 		sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2298 		if (s->direction == PF_OUT) {
2299 			srcaddr = &sk->addr[1];
2300 			dstaddr = &sk->addr[0];
2301 			srcport = sk->port[1];
2302 			dstport = sk->port[0];
2303 		} else {
2304 			srcaddr = &sk->addr[0];
2305 			dstaddr = &sk->addr[1];
2306 			srcport = sk->port[0];
2307 			dstport = sk->port[1];
2308 		}
2309 
2310 		if (psk->psk_af && sk->af != psk->psk_af)
2311 			continue;
2312 
2313 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2314 			continue;
2315 
2316 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2317 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2318 			continue;
2319 
2320 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2321 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2322 			continue;
2323 
2324 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2325 		    &psk->psk_rt_addr.addr.v.a.addr,
2326 		    &psk->psk_rt_addr.addr.v.a.mask,
2327 		    &s->rt_addr, sk->af))
2328 			continue;
2329 
2330 		if (psk->psk_src.port_op != 0 &&
2331 		    ! pf_match_port(psk->psk_src.port_op,
2332 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2333 			continue;
2334 
2335 		if (psk->psk_dst.port_op != 0 &&
2336 		    ! pf_match_port(psk->psk_dst.port_op,
2337 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2338 			continue;
2339 
2340 		if (psk->psk_label[0] &&
2341 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2342 			continue;
2343 
2344 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2345 		    kif->pfik_name))
2346 			continue;
2347 
2348 		if (psk->psk_kill_match) {
2349 			/* Create the key to find matching states, with lock
2350 			 * held. */
2351 
2352 			bzero(&match_key, sizeof(match_key));
2353 
2354 			if (s->direction == PF_OUT) {
2355 				dir = PF_IN;
2356 				idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2357 			} else {
2358 				dir = PF_OUT;
2359 				idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2360 			}
2361 
2362 			match_key.af = s->key[idx]->af;
2363 			match_key.proto = s->key[idx]->proto;
2364 			PF_ACPY(&match_key.addr[0],
2365 			    &s->key[idx]->addr[1], match_key.af);
2366 			match_key.port[0] = s->key[idx]->port[1];
2367 			PF_ACPY(&match_key.addr[1],
2368 			    &s->key[idx]->addr[0], match_key.af);
2369 			match_key.port[1] = s->key[idx]->port[0];
2370 		}
2371 
2372 		pf_unlink_state(s);
2373 		killed++;
2374 
2375 		if (psk->psk_kill_match)
2376 			killed += pf_kill_matching_state(&match_key, dir);
2377 
2378 		goto relock_DIOCKILLSTATES;
2379 	}
2380 	PF_HASHROW_UNLOCK(ih);
2381 
2382 	return (killed);
2383 }
2384 
2385 int
2386 pf_start(void)
2387 {
2388 	int error = 0;
2389 
2390 	sx_xlock(&V_pf_ioctl_lock);
2391 	if (V_pf_status.running)
2392 		error = EEXIST;
2393 	else {
2394 		hook_pf();
2395 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2396 			hook_pf_eth();
2397 		V_pf_status.running = 1;
2398 		V_pf_status.since = time_second;
2399 		new_unrhdr64(&V_pf_stateid, time_second);
2400 
2401 		DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2402 	}
2403 	sx_xunlock(&V_pf_ioctl_lock);
2404 
2405 	return (error);
2406 }
2407 
2408 int
2409 pf_stop(void)
2410 {
2411 	int error = 0;
2412 
2413 	sx_xlock(&V_pf_ioctl_lock);
2414 	if (!V_pf_status.running)
2415 		error = ENOENT;
2416 	else {
2417 		V_pf_status.running = 0;
2418 		dehook_pf();
2419 		dehook_pf_eth();
2420 		V_pf_status.since = time_second;
2421 		DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2422 	}
2423 	sx_xunlock(&V_pf_ioctl_lock);
2424 
2425 	return (error);
2426 }
2427 
2428 void
2429 pf_ioctl_clear_status(void)
2430 {
2431 	PF_RULES_WLOCK();
2432 	for (int i = 0; i < PFRES_MAX; i++)
2433 		counter_u64_zero(V_pf_status.counters[i]);
2434 	for (int i = 0; i < FCNT_MAX; i++)
2435 		pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2436 	for (int i = 0; i < SCNT_MAX; i++)
2437 		counter_u64_zero(V_pf_status.scounters[i]);
2438 	for (int i = 0; i < KLCNT_MAX; i++)
2439 		counter_u64_zero(V_pf_status.lcounters[i]);
2440 	V_pf_status.since = time_second;
2441 	if (*V_pf_status.ifname)
2442 		pfi_update_status(V_pf_status.ifname, NULL);
2443 	PF_RULES_WUNLOCK();
2444 }
2445 
2446 int
2447 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2448 {
2449 	uint32_t old;
2450 
2451 	if (timeout < 0 || timeout >= PFTM_MAX ||
2452 	    seconds < 0)
2453 		return (EINVAL);
2454 
2455 	PF_RULES_WLOCK();
2456 	old = V_pf_default_rule.timeout[timeout];
2457 	if (timeout == PFTM_INTERVAL && seconds == 0)
2458 		seconds = 1;
2459 	V_pf_default_rule.timeout[timeout] = seconds;
2460 	if (timeout == PFTM_INTERVAL && seconds < old)
2461 		wakeup(pf_purge_thread);
2462 
2463 	if (prev_seconds != NULL)
2464 		*prev_seconds = old;
2465 
2466 	PF_RULES_WUNLOCK();
2467 
2468 	return (0);
2469 }
2470 
2471 int
2472 pf_ioctl_get_timeout(int timeout, int *seconds)
2473 {
2474 	PF_RULES_RLOCK_TRACKER;
2475 
2476 	if (timeout < 0 || timeout >= PFTM_MAX)
2477 		return (EINVAL);
2478 
2479 	PF_RULES_RLOCK();
2480 	*seconds = V_pf_default_rule.timeout[timeout];
2481 	PF_RULES_RUNLOCK();
2482 
2483 	return (0);
2484 }
2485 
2486 int
2487 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2488 {
2489 
2490 	PF_RULES_WLOCK();
2491 	if (index < 0 || index >= PF_LIMIT_MAX ||
2492 	    V_pf_limits[index].zone == NULL) {
2493 		PF_RULES_WUNLOCK();
2494 		return (EINVAL);
2495 	}
2496 	uma_zone_set_max(V_pf_limits[index].zone, limit);
2497 	if (old_limit != NULL)
2498 		*old_limit = V_pf_limits[index].limit;
2499 	V_pf_limits[index].limit = limit;
2500 	PF_RULES_WUNLOCK();
2501 
2502 	return (0);
2503 }
2504 
2505 int
2506 pf_ioctl_get_limit(int index, unsigned int *limit)
2507 {
2508 	PF_RULES_RLOCK_TRACKER;
2509 
2510 	if (index < 0 || index >= PF_LIMIT_MAX)
2511 		return (EINVAL);
2512 
2513 	PF_RULES_RLOCK();
2514 	*limit = V_pf_limits[index].limit;
2515 	PF_RULES_RUNLOCK();
2516 
2517 	return (0);
2518 }
2519 
2520 int
2521 pf_ioctl_begin_addrs(uint32_t *ticket)
2522 {
2523 	PF_RULES_WLOCK();
2524 	pf_empty_kpool(&V_pf_pabuf);
2525 	*ticket = ++V_ticket_pabuf;
2526 	PF_RULES_WUNLOCK();
2527 
2528 	return (0);
2529 }
2530 
2531 int
2532 pf_ioctl_add_addr(struct pfioc_pooladdr *pp)
2533 {
2534 	struct pf_kpooladdr	*pa = NULL;
2535 	struct pfi_kkif		*kif = NULL;
2536 	int error;
2537 
2538 #ifndef INET
2539 	if (pp->af == AF_INET)
2540 		return (EAFNOSUPPORT);
2541 #endif /* INET */
2542 #ifndef INET6
2543 	if (pp->af == AF_INET6)
2544 		return (EAFNOSUPPORT);
2545 #endif /* INET6 */
2546 
2547 	if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2548 	    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2549 	    pp->addr.addr.type != PF_ADDR_TABLE)
2550 		return (EINVAL);
2551 
2552 	if (pp->addr.addr.p.dyn != NULL)
2553 		return (EINVAL);
2554 
2555 	pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2556 	error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2557 	if (error != 0)
2558 		goto out;
2559 	if (pa->ifname[0])
2560 		kif = pf_kkif_create(M_WAITOK);
2561 	PF_RULES_WLOCK();
2562 	if (pp->ticket != V_ticket_pabuf) {
2563 		PF_RULES_WUNLOCK();
2564 		if (pa->ifname[0])
2565 			pf_kkif_free(kif);
2566 		error = EBUSY;
2567 		goto out;
2568 	}
2569 	if (pa->ifname[0]) {
2570 		pa->kif = pfi_kkif_attach(kif, pa->ifname);
2571 		kif = NULL;
2572 		pfi_kkif_ref(pa->kif);
2573 	} else
2574 		pa->kif = NULL;
2575 	if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2576 	    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2577 		if (pa->ifname[0])
2578 			pfi_kkif_unref(pa->kif);
2579 		PF_RULES_WUNLOCK();
2580 		goto out;
2581 	}
2582 	TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2583 	PF_RULES_WUNLOCK();
2584 
2585 	return (0);
2586 
2587 out:
2588 	free(pa, M_PFRULE);
2589 	return (error);
2590 }
2591 
2592 static int
2593 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2594 {
2595 	int			 error = 0;
2596 	PF_RULES_RLOCK_TRACKER;
2597 
2598 #define	ERROUT_IOCTL(target, x)					\
2599     do {								\
2600 	    error = (x);						\
2601 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2602 	    goto target;						\
2603     } while (0)
2604 
2605 
2606 	/* XXX keep in sync with switch() below */
2607 	if (securelevel_gt(td->td_ucred, 2))
2608 		switch (cmd) {
2609 		case DIOCGETRULES:
2610 		case DIOCGETRULENV:
2611 		case DIOCGETADDRS:
2612 		case DIOCGETADDR:
2613 		case DIOCGETSTATE:
2614 		case DIOCGETSTATENV:
2615 		case DIOCSETSTATUSIF:
2616 		case DIOCGETSTATUSNV:
2617 		case DIOCCLRSTATUS:
2618 		case DIOCNATLOOK:
2619 		case DIOCSETDEBUG:
2620 #ifdef COMPAT_FREEBSD14
2621 		case DIOCGETSTATES:
2622 		case DIOCGETSTATESV2:
2623 #endif
2624 		case DIOCGETTIMEOUT:
2625 		case DIOCCLRRULECTRS:
2626 		case DIOCGETLIMIT:
2627 		case DIOCGETALTQSV0:
2628 		case DIOCGETALTQSV1:
2629 		case DIOCGETALTQV0:
2630 		case DIOCGETALTQV1:
2631 		case DIOCGETQSTATSV0:
2632 		case DIOCGETQSTATSV1:
2633 		case DIOCGETRULESETS:
2634 		case DIOCGETRULESET:
2635 		case DIOCRGETTABLES:
2636 		case DIOCRGETTSTATS:
2637 		case DIOCRCLRTSTATS:
2638 		case DIOCRCLRADDRS:
2639 		case DIOCRADDADDRS:
2640 		case DIOCRDELADDRS:
2641 		case DIOCRSETADDRS:
2642 		case DIOCRGETADDRS:
2643 		case DIOCRGETASTATS:
2644 		case DIOCRCLRASTATS:
2645 		case DIOCRTSTADDRS:
2646 		case DIOCOSFPGET:
2647 		case DIOCGETSRCNODES:
2648 		case DIOCCLRSRCNODES:
2649 		case DIOCGETSYNCOOKIES:
2650 		case DIOCIGETIFACES:
2651 		case DIOCGIFSPEEDV0:
2652 		case DIOCGIFSPEEDV1:
2653 		case DIOCSETIFFLAG:
2654 		case DIOCCLRIFFLAG:
2655 		case DIOCGETETHRULES:
2656 		case DIOCGETETHRULE:
2657 		case DIOCGETETHRULESETS:
2658 		case DIOCGETETHRULESET:
2659 			break;
2660 		case DIOCRCLRTABLES:
2661 		case DIOCRADDTABLES:
2662 		case DIOCRDELTABLES:
2663 		case DIOCRSETTFLAGS:
2664 			if (((struct pfioc_table *)addr)->pfrio_flags &
2665 			    PFR_FLAG_DUMMY)
2666 				break; /* dummy operation ok */
2667 			return (EPERM);
2668 		default:
2669 			return (EPERM);
2670 		}
2671 
2672 	if (!(flags & FWRITE))
2673 		switch (cmd) {
2674 		case DIOCGETRULES:
2675 		case DIOCGETADDRS:
2676 		case DIOCGETADDR:
2677 		case DIOCGETSTATE:
2678 		case DIOCGETSTATENV:
2679 		case DIOCGETSTATUSNV:
2680 #ifdef COMPAT_FREEBSD14
2681 		case DIOCGETSTATES:
2682 		case DIOCGETSTATESV2:
2683 #endif
2684 		case DIOCGETTIMEOUT:
2685 		case DIOCGETLIMIT:
2686 		case DIOCGETALTQSV0:
2687 		case DIOCGETALTQSV1:
2688 		case DIOCGETALTQV0:
2689 		case DIOCGETALTQV1:
2690 		case DIOCGETQSTATSV0:
2691 		case DIOCGETQSTATSV1:
2692 		case DIOCGETRULESETS:
2693 		case DIOCGETRULESET:
2694 		case DIOCNATLOOK:
2695 		case DIOCRGETTABLES:
2696 		case DIOCRGETTSTATS:
2697 		case DIOCRGETADDRS:
2698 		case DIOCRGETASTATS:
2699 		case DIOCRTSTADDRS:
2700 		case DIOCOSFPGET:
2701 		case DIOCGETSRCNODES:
2702 		case DIOCGETSYNCOOKIES:
2703 		case DIOCIGETIFACES:
2704 		case DIOCGIFSPEEDV1:
2705 		case DIOCGIFSPEEDV0:
2706 		case DIOCGETRULENV:
2707 		case DIOCGETETHRULES:
2708 		case DIOCGETETHRULE:
2709 		case DIOCGETETHRULESETS:
2710 		case DIOCGETETHRULESET:
2711 			break;
2712 		case DIOCRCLRTABLES:
2713 		case DIOCRADDTABLES:
2714 		case DIOCRDELTABLES:
2715 		case DIOCRCLRTSTATS:
2716 		case DIOCRCLRADDRS:
2717 		case DIOCRADDADDRS:
2718 		case DIOCRDELADDRS:
2719 		case DIOCRSETADDRS:
2720 		case DIOCRSETTFLAGS:
2721 			if (((struct pfioc_table *)addr)->pfrio_flags &
2722 			    PFR_FLAG_DUMMY) {
2723 				flags |= FWRITE; /* need write lock for dummy */
2724 				break; /* dummy operation ok */
2725 			}
2726 			return (EACCES);
2727 		default:
2728 			return (EACCES);
2729 		}
2730 
2731 	CURVNET_SET(TD_TO_VNET(td));
2732 
2733 	switch (cmd) {
2734 #ifdef COMPAT_FREEBSD14
2735 	case DIOCSTART:
2736 		error = pf_start();
2737 		break;
2738 
2739 	case DIOCSTOP:
2740 		error = pf_stop();
2741 		break;
2742 #endif
2743 
2744 	case DIOCGETETHRULES: {
2745 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2746 		nvlist_t		*nvl;
2747 		void			*packed;
2748 		struct pf_keth_rule	*tail;
2749 		struct pf_keth_ruleset	*rs;
2750 		u_int32_t		 ticket, nr;
2751 		const char		*anchor = "";
2752 
2753 		nvl = NULL;
2754 		packed = NULL;
2755 
2756 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2757 
2758 		if (nv->len > pf_ioctl_maxcount)
2759 			ERROUT(ENOMEM);
2760 
2761 		/* Copy the request in */
2762 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2763 		if (packed == NULL)
2764 			ERROUT(ENOMEM);
2765 
2766 		error = copyin(nv->data, packed, nv->len);
2767 		if (error)
2768 			ERROUT(error);
2769 
2770 		nvl = nvlist_unpack(packed, nv->len, 0);
2771 		if (nvl == NULL)
2772 			ERROUT(EBADMSG);
2773 
2774 		if (! nvlist_exists_string(nvl, "anchor"))
2775 			ERROUT(EBADMSG);
2776 
2777 		anchor = nvlist_get_string(nvl, "anchor");
2778 
2779 		rs = pf_find_keth_ruleset(anchor);
2780 
2781 		nvlist_destroy(nvl);
2782 		nvl = NULL;
2783 		free(packed, M_NVLIST);
2784 		packed = NULL;
2785 
2786 		if (rs == NULL)
2787 			ERROUT(ENOENT);
2788 
2789 		/* Reply */
2790 		nvl = nvlist_create(0);
2791 		if (nvl == NULL)
2792 			ERROUT(ENOMEM);
2793 
2794 		PF_RULES_RLOCK();
2795 
2796 		ticket = rs->active.ticket;
2797 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2798 		if (tail)
2799 			nr = tail->nr + 1;
2800 		else
2801 			nr = 0;
2802 
2803 		PF_RULES_RUNLOCK();
2804 
2805 		nvlist_add_number(nvl, "ticket", ticket);
2806 		nvlist_add_number(nvl, "nr", nr);
2807 
2808 		packed = nvlist_pack(nvl, &nv->len);
2809 		if (packed == NULL)
2810 			ERROUT(ENOMEM);
2811 
2812 		if (nv->size == 0)
2813 			ERROUT(0);
2814 		else if (nv->size < nv->len)
2815 			ERROUT(ENOSPC);
2816 
2817 		error = copyout(packed, nv->data, nv->len);
2818 
2819 #undef ERROUT
2820 DIOCGETETHRULES_error:
2821 		free(packed, M_NVLIST);
2822 		nvlist_destroy(nvl);
2823 		break;
2824 	}
2825 
2826 	case DIOCGETETHRULE: {
2827 		struct epoch_tracker	 et;
2828 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2829 		nvlist_t		*nvl = NULL;
2830 		void			*nvlpacked = NULL;
2831 		struct pf_keth_rule	*rule = NULL;
2832 		struct pf_keth_ruleset	*rs;
2833 		u_int32_t		 ticket, nr;
2834 		bool			 clear = false;
2835 		const char		*anchor;
2836 
2837 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2838 
2839 		if (nv->len > pf_ioctl_maxcount)
2840 			ERROUT(ENOMEM);
2841 
2842 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2843 		if (nvlpacked == NULL)
2844 			ERROUT(ENOMEM);
2845 
2846 		error = copyin(nv->data, nvlpacked, nv->len);
2847 		if (error)
2848 			ERROUT(error);
2849 
2850 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2851 		if (nvl == NULL)
2852 			ERROUT(EBADMSG);
2853 		if (! nvlist_exists_number(nvl, "ticket"))
2854 			ERROUT(EBADMSG);
2855 		ticket = nvlist_get_number(nvl, "ticket");
2856 		if (! nvlist_exists_string(nvl, "anchor"))
2857 			ERROUT(EBADMSG);
2858 		anchor = nvlist_get_string(nvl, "anchor");
2859 
2860 		if (nvlist_exists_bool(nvl, "clear"))
2861 			clear = nvlist_get_bool(nvl, "clear");
2862 
2863 		if (clear && !(flags & FWRITE))
2864 			ERROUT(EACCES);
2865 
2866 		if (! nvlist_exists_number(nvl, "nr"))
2867 			ERROUT(EBADMSG);
2868 		nr = nvlist_get_number(nvl, "nr");
2869 
2870 		PF_RULES_RLOCK();
2871 		rs = pf_find_keth_ruleset(anchor);
2872 		if (rs == NULL) {
2873 			PF_RULES_RUNLOCK();
2874 			ERROUT(ENOENT);
2875 		}
2876 		if (ticket != rs->active.ticket) {
2877 			PF_RULES_RUNLOCK();
2878 			ERROUT(EBUSY);
2879 		}
2880 
2881 		nvlist_destroy(nvl);
2882 		nvl = NULL;
2883 		free(nvlpacked, M_NVLIST);
2884 		nvlpacked = NULL;
2885 
2886 		rule = TAILQ_FIRST(rs->active.rules);
2887 		while ((rule != NULL) && (rule->nr != nr))
2888 			rule = TAILQ_NEXT(rule, entries);
2889 		if (rule == NULL) {
2890 			PF_RULES_RUNLOCK();
2891 			ERROUT(ENOENT);
2892 		}
2893 		/* Make sure rule can't go away. */
2894 		NET_EPOCH_ENTER(et);
2895 		PF_RULES_RUNLOCK();
2896 		nvl = pf_keth_rule_to_nveth_rule(rule);
2897 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2898 			ERROUT(EBUSY);
2899 		NET_EPOCH_EXIT(et);
2900 		if (nvl == NULL)
2901 			ERROUT(ENOMEM);
2902 
2903 		nvlpacked = nvlist_pack(nvl, &nv->len);
2904 		if (nvlpacked == NULL)
2905 			ERROUT(ENOMEM);
2906 
2907 		if (nv->size == 0)
2908 			ERROUT(0);
2909 		else if (nv->size < nv->len)
2910 			ERROUT(ENOSPC);
2911 
2912 		error = copyout(nvlpacked, nv->data, nv->len);
2913 		if (error == 0 && clear) {
2914 			counter_u64_zero(rule->evaluations);
2915 			for (int i = 0; i < 2; i++) {
2916 				counter_u64_zero(rule->packets[i]);
2917 				counter_u64_zero(rule->bytes[i]);
2918 			}
2919 		}
2920 
2921 #undef ERROUT
2922 DIOCGETETHRULE_error:
2923 		free(nvlpacked, M_NVLIST);
2924 		nvlist_destroy(nvl);
2925 		break;
2926 	}
2927 
2928 	case DIOCADDETHRULE: {
2929 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2930 		nvlist_t		*nvl = NULL;
2931 		void			*nvlpacked = NULL;
2932 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2933 		struct pf_keth_ruleset	*ruleset = NULL;
2934 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
2935 		const char		*anchor = "", *anchor_call = "";
2936 
2937 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2938 
2939 		if (nv->len > pf_ioctl_maxcount)
2940 			ERROUT(ENOMEM);
2941 
2942 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2943 		if (nvlpacked == NULL)
2944 			ERROUT(ENOMEM);
2945 
2946 		error = copyin(nv->data, nvlpacked, nv->len);
2947 		if (error)
2948 			ERROUT(error);
2949 
2950 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2951 		if (nvl == NULL)
2952 			ERROUT(EBADMSG);
2953 
2954 		if (! nvlist_exists_number(nvl, "ticket"))
2955 			ERROUT(EBADMSG);
2956 
2957 		if (nvlist_exists_string(nvl, "anchor"))
2958 			anchor = nvlist_get_string(nvl, "anchor");
2959 		if (nvlist_exists_string(nvl, "anchor_call"))
2960 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2961 
2962 		ruleset = pf_find_keth_ruleset(anchor);
2963 		if (ruleset == NULL)
2964 			ERROUT(EINVAL);
2965 
2966 		if (nvlist_get_number(nvl, "ticket") !=
2967 		    ruleset->inactive.ticket) {
2968 			DPFPRINTF(PF_DEBUG_MISC,
2969 			    ("ticket: %d != %d\n",
2970 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2971 			    ruleset->inactive.ticket));
2972 			ERROUT(EBUSY);
2973 		}
2974 
2975 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
2976 		if (rule == NULL)
2977 			ERROUT(ENOMEM);
2978 		rule->timestamp = NULL;
2979 
2980 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
2981 		if (error != 0)
2982 			ERROUT(error);
2983 
2984 		if (rule->ifname[0])
2985 			kif = pf_kkif_create(M_WAITOK);
2986 		if (rule->bridge_to_name[0])
2987 			bridge_to_kif = pf_kkif_create(M_WAITOK);
2988 		rule->evaluations = counter_u64_alloc(M_WAITOK);
2989 		for (int i = 0; i < 2; i++) {
2990 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
2991 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
2992 		}
2993 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
2994 		    M_WAITOK | M_ZERO);
2995 
2996 		PF_RULES_WLOCK();
2997 
2998 		if (rule->ifname[0]) {
2999 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
3000 			pfi_kkif_ref(rule->kif);
3001 		} else
3002 			rule->kif = NULL;
3003 		if (rule->bridge_to_name[0]) {
3004 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3005 			    rule->bridge_to_name);
3006 			pfi_kkif_ref(rule->bridge_to);
3007 		} else
3008 			rule->bridge_to = NULL;
3009 
3010 #ifdef ALTQ
3011 		/* set queue IDs */
3012 		if (rule->qname[0] != 0) {
3013 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3014 				error = EBUSY;
3015 			else
3016 				rule->qid = rule->qid;
3017 		}
3018 #endif
3019 		if (rule->tagname[0])
3020 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3021 				error = EBUSY;
3022 		if (rule->match_tagname[0])
3023 			if ((rule->match_tag = pf_tagname2tag(
3024 			    rule->match_tagname)) == 0)
3025 				error = EBUSY;
3026 
3027 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3028 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3029 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3030 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3031 
3032 		if (error) {
3033 			pf_free_eth_rule(rule);
3034 			PF_RULES_WUNLOCK();
3035 			ERROUT(error);
3036 		}
3037 
3038 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3039 			pf_free_eth_rule(rule);
3040 			PF_RULES_WUNLOCK();
3041 			ERROUT(EINVAL);
3042 		}
3043 
3044 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3045 		if (tail)
3046 			rule->nr = tail->nr + 1;
3047 		else
3048 			rule->nr = 0;
3049 
3050 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3051 
3052 		PF_RULES_WUNLOCK();
3053 
3054 #undef ERROUT
3055 DIOCADDETHRULE_error:
3056 		nvlist_destroy(nvl);
3057 		free(nvlpacked, M_NVLIST);
3058 		break;
3059 	}
3060 
3061 	case DIOCGETETHRULESETS: {
3062 		struct epoch_tracker	 et;
3063 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3064 		nvlist_t		*nvl = NULL;
3065 		void			*nvlpacked = NULL;
3066 		struct pf_keth_ruleset	*ruleset;
3067 		struct pf_keth_anchor	*anchor;
3068 		int			 nr = 0;
3069 
3070 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3071 
3072 		if (nv->len > pf_ioctl_maxcount)
3073 			ERROUT(ENOMEM);
3074 
3075 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3076 		if (nvlpacked == NULL)
3077 			ERROUT(ENOMEM);
3078 
3079 		error = copyin(nv->data, nvlpacked, nv->len);
3080 		if (error)
3081 			ERROUT(error);
3082 
3083 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3084 		if (nvl == NULL)
3085 			ERROUT(EBADMSG);
3086 		if (! nvlist_exists_string(nvl, "path"))
3087 			ERROUT(EBADMSG);
3088 
3089 		NET_EPOCH_ENTER(et);
3090 
3091 		if ((ruleset = pf_find_keth_ruleset(
3092 		    nvlist_get_string(nvl, "path"))) == NULL) {
3093 			NET_EPOCH_EXIT(et);
3094 			ERROUT(ENOENT);
3095 		}
3096 
3097 		if (ruleset->anchor == NULL) {
3098 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3099 				if (anchor->parent == NULL)
3100 					nr++;
3101 		} else {
3102 			RB_FOREACH(anchor, pf_keth_anchor_node,
3103 			    &ruleset->anchor->children)
3104 				nr++;
3105 		}
3106 
3107 		NET_EPOCH_EXIT(et);
3108 
3109 		nvlist_destroy(nvl);
3110 		nvl = NULL;
3111 		free(nvlpacked, M_NVLIST);
3112 		nvlpacked = NULL;
3113 
3114 		nvl = nvlist_create(0);
3115 		if (nvl == NULL)
3116 			ERROUT(ENOMEM);
3117 
3118 		nvlist_add_number(nvl, "nr", nr);
3119 
3120 		nvlpacked = nvlist_pack(nvl, &nv->len);
3121 		if (nvlpacked == NULL)
3122 			ERROUT(ENOMEM);
3123 
3124 		if (nv->size == 0)
3125 			ERROUT(0);
3126 		else if (nv->size < nv->len)
3127 			ERROUT(ENOSPC);
3128 
3129 		error = copyout(nvlpacked, nv->data, nv->len);
3130 
3131 #undef ERROUT
3132 DIOCGETETHRULESETS_error:
3133 		free(nvlpacked, M_NVLIST);
3134 		nvlist_destroy(nvl);
3135 		break;
3136 	}
3137 
3138 	case DIOCGETETHRULESET: {
3139 		struct epoch_tracker	 et;
3140 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3141 		nvlist_t		*nvl = NULL;
3142 		void			*nvlpacked = NULL;
3143 		struct pf_keth_ruleset	*ruleset;
3144 		struct pf_keth_anchor	*anchor;
3145 		int			 nr = 0, req_nr = 0;
3146 		bool			 found = false;
3147 
3148 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3149 
3150 		if (nv->len > pf_ioctl_maxcount)
3151 			ERROUT(ENOMEM);
3152 
3153 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3154 		if (nvlpacked == NULL)
3155 			ERROUT(ENOMEM);
3156 
3157 		error = copyin(nv->data, nvlpacked, nv->len);
3158 		if (error)
3159 			ERROUT(error);
3160 
3161 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3162 		if (nvl == NULL)
3163 			ERROUT(EBADMSG);
3164 		if (! nvlist_exists_string(nvl, "path"))
3165 			ERROUT(EBADMSG);
3166 		if (! nvlist_exists_number(nvl, "nr"))
3167 			ERROUT(EBADMSG);
3168 
3169 		req_nr = nvlist_get_number(nvl, "nr");
3170 
3171 		NET_EPOCH_ENTER(et);
3172 
3173 		if ((ruleset = pf_find_keth_ruleset(
3174 		    nvlist_get_string(nvl, "path"))) == NULL) {
3175 			NET_EPOCH_EXIT(et);
3176 			ERROUT(ENOENT);
3177 		}
3178 
3179 		nvlist_destroy(nvl);
3180 		nvl = NULL;
3181 		free(nvlpacked, M_NVLIST);
3182 		nvlpacked = NULL;
3183 
3184 		nvl = nvlist_create(0);
3185 		if (nvl == NULL) {
3186 			NET_EPOCH_EXIT(et);
3187 			ERROUT(ENOMEM);
3188 		}
3189 
3190 		if (ruleset->anchor == NULL) {
3191 			RB_FOREACH(anchor, pf_keth_anchor_global,
3192 			    &V_pf_keth_anchors) {
3193 				if (anchor->parent == NULL && nr++ == req_nr) {
3194 					found = true;
3195 					break;
3196 				}
3197 			}
3198 		} else {
3199 			RB_FOREACH(anchor, pf_keth_anchor_node,
3200 			     &ruleset->anchor->children) {
3201 				if (nr++ == req_nr) {
3202 					found = true;
3203 					break;
3204 				}
3205 			}
3206 		}
3207 
3208 		NET_EPOCH_EXIT(et);
3209 		if (found) {
3210 			nvlist_add_number(nvl, "nr", nr);
3211 			nvlist_add_string(nvl, "name", anchor->name);
3212 			if (ruleset->anchor)
3213 				nvlist_add_string(nvl, "path",
3214 				    ruleset->anchor->path);
3215 			else
3216 				nvlist_add_string(nvl, "path", "");
3217 		} else {
3218 			ERROUT(EBUSY);
3219 		}
3220 
3221 		nvlpacked = nvlist_pack(nvl, &nv->len);
3222 		if (nvlpacked == NULL)
3223 			ERROUT(ENOMEM);
3224 
3225 		if (nv->size == 0)
3226 			ERROUT(0);
3227 		else if (nv->size < nv->len)
3228 			ERROUT(ENOSPC);
3229 
3230 		error = copyout(nvlpacked, nv->data, nv->len);
3231 
3232 #undef ERROUT
3233 DIOCGETETHRULESET_error:
3234 		free(nvlpacked, M_NVLIST);
3235 		nvlist_destroy(nvl);
3236 		break;
3237 	}
3238 
3239 	case DIOCADDRULENV: {
3240 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3241 		nvlist_t	*nvl = NULL;
3242 		void		*nvlpacked = NULL;
3243 		struct pf_krule	*rule = NULL;
3244 		const char	*anchor = "", *anchor_call = "";
3245 		uint32_t	 ticket = 0, pool_ticket = 0;
3246 
3247 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3248 
3249 		if (nv->len > pf_ioctl_maxcount)
3250 			ERROUT(ENOMEM);
3251 
3252 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3253 		error = copyin(nv->data, nvlpacked, nv->len);
3254 		if (error)
3255 			ERROUT(error);
3256 
3257 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3258 		if (nvl == NULL)
3259 			ERROUT(EBADMSG);
3260 
3261 		if (! nvlist_exists_number(nvl, "ticket"))
3262 			ERROUT(EINVAL);
3263 		ticket = nvlist_get_number(nvl, "ticket");
3264 
3265 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3266 			ERROUT(EINVAL);
3267 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3268 
3269 		if (! nvlist_exists_nvlist(nvl, "rule"))
3270 			ERROUT(EINVAL);
3271 
3272 		rule = pf_krule_alloc();
3273 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3274 		    rule);
3275 		if (error)
3276 			ERROUT(error);
3277 
3278 		if (nvlist_exists_string(nvl, "anchor"))
3279 			anchor = nvlist_get_string(nvl, "anchor");
3280 		if (nvlist_exists_string(nvl, "anchor_call"))
3281 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3282 
3283 		if ((error = nvlist_error(nvl)))
3284 			ERROUT(error);
3285 
3286 		/* Frees rule on error */
3287 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3288 		    anchor_call, td->td_ucred->cr_ruid,
3289 		    td->td_proc ? td->td_proc->p_pid : 0);
3290 
3291 		nvlist_destroy(nvl);
3292 		free(nvlpacked, M_NVLIST);
3293 		break;
3294 #undef ERROUT
3295 DIOCADDRULENV_error:
3296 		pf_krule_free(rule);
3297 		nvlist_destroy(nvl);
3298 		free(nvlpacked, M_NVLIST);
3299 
3300 		break;
3301 	}
3302 	case DIOCADDRULE: {
3303 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3304 		struct pf_krule		*rule;
3305 
3306 		rule = pf_krule_alloc();
3307 		error = pf_rule_to_krule(&pr->rule, rule);
3308 		if (error != 0) {
3309 			pf_krule_free(rule);
3310 			break;
3311 		}
3312 
3313 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3314 
3315 		/* Frees rule on error */
3316 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3317 		    pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3318 		    td->td_proc ? td->td_proc->p_pid : 0);
3319 		break;
3320 	}
3321 
3322 	case DIOCGETRULES: {
3323 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3324 
3325 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3326 
3327 		error = pf_ioctl_getrules(pr);
3328 
3329 		break;
3330 	}
3331 
3332 	case DIOCGETRULENV: {
3333 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3334 		nvlist_t		*nvrule = NULL;
3335 		nvlist_t		*nvl = NULL;
3336 		struct pf_kruleset	*ruleset;
3337 		struct pf_krule		*rule;
3338 		void			*nvlpacked = NULL;
3339 		int			 rs_num, nr;
3340 		bool			 clear_counter = false;
3341 
3342 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3343 
3344 		if (nv->len > pf_ioctl_maxcount)
3345 			ERROUT(ENOMEM);
3346 
3347 		/* Copy the request in */
3348 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3349 		if (nvlpacked == NULL)
3350 			ERROUT(ENOMEM);
3351 
3352 		error = copyin(nv->data, nvlpacked, nv->len);
3353 		if (error)
3354 			ERROUT(error);
3355 
3356 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3357 		if (nvl == NULL)
3358 			ERROUT(EBADMSG);
3359 
3360 		if (! nvlist_exists_string(nvl, "anchor"))
3361 			ERROUT(EBADMSG);
3362 		if (! nvlist_exists_number(nvl, "ruleset"))
3363 			ERROUT(EBADMSG);
3364 		if (! nvlist_exists_number(nvl, "ticket"))
3365 			ERROUT(EBADMSG);
3366 		if (! nvlist_exists_number(nvl, "nr"))
3367 			ERROUT(EBADMSG);
3368 
3369 		if (nvlist_exists_bool(nvl, "clear_counter"))
3370 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3371 
3372 		if (clear_counter && !(flags & FWRITE))
3373 			ERROUT(EACCES);
3374 
3375 		nr = nvlist_get_number(nvl, "nr");
3376 
3377 		PF_RULES_WLOCK();
3378 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3379 		if (ruleset == NULL) {
3380 			PF_RULES_WUNLOCK();
3381 			ERROUT(ENOENT);
3382 		}
3383 
3384 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3385 		if (rs_num >= PF_RULESET_MAX) {
3386 			PF_RULES_WUNLOCK();
3387 			ERROUT(EINVAL);
3388 		}
3389 
3390 		if (nvlist_get_number(nvl, "ticket") !=
3391 		    ruleset->rules[rs_num].active.ticket) {
3392 			PF_RULES_WUNLOCK();
3393 			ERROUT(EBUSY);
3394 		}
3395 
3396 		if ((error = nvlist_error(nvl))) {
3397 			PF_RULES_WUNLOCK();
3398 			ERROUT(error);
3399 		}
3400 
3401 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3402 		while ((rule != NULL) && (rule->nr != nr))
3403 			rule = TAILQ_NEXT(rule, entries);
3404 		if (rule == NULL) {
3405 			PF_RULES_WUNLOCK();
3406 			ERROUT(EBUSY);
3407 		}
3408 
3409 		nvrule = pf_krule_to_nvrule(rule);
3410 
3411 		nvlist_destroy(nvl);
3412 		nvl = nvlist_create(0);
3413 		if (nvl == NULL) {
3414 			PF_RULES_WUNLOCK();
3415 			ERROUT(ENOMEM);
3416 		}
3417 		nvlist_add_number(nvl, "nr", nr);
3418 		nvlist_add_nvlist(nvl, "rule", nvrule);
3419 		nvlist_destroy(nvrule);
3420 		nvrule = NULL;
3421 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3422 			PF_RULES_WUNLOCK();
3423 			ERROUT(EBUSY);
3424 		}
3425 
3426 		free(nvlpacked, M_NVLIST);
3427 		nvlpacked = nvlist_pack(nvl, &nv->len);
3428 		if (nvlpacked == NULL) {
3429 			PF_RULES_WUNLOCK();
3430 			ERROUT(ENOMEM);
3431 		}
3432 
3433 		if (nv->size == 0) {
3434 			PF_RULES_WUNLOCK();
3435 			ERROUT(0);
3436 		}
3437 		else if (nv->size < nv->len) {
3438 			PF_RULES_WUNLOCK();
3439 			ERROUT(ENOSPC);
3440 		}
3441 
3442 		if (clear_counter)
3443 			pf_krule_clear_counters(rule);
3444 
3445 		PF_RULES_WUNLOCK();
3446 
3447 		error = copyout(nvlpacked, nv->data, nv->len);
3448 
3449 #undef ERROUT
3450 DIOCGETRULENV_error:
3451 		free(nvlpacked, M_NVLIST);
3452 		nvlist_destroy(nvrule);
3453 		nvlist_destroy(nvl);
3454 
3455 		break;
3456 	}
3457 
3458 	case DIOCCHANGERULE: {
3459 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3460 		struct pf_kruleset	*ruleset;
3461 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3462 		struct pfi_kkif		*kif = NULL;
3463 		struct pf_kpooladdr	*pa;
3464 		u_int32_t		 nr = 0;
3465 		int			 rs_num;
3466 
3467 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3468 
3469 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3470 		    pcr->action > PF_CHANGE_GET_TICKET) {
3471 			error = EINVAL;
3472 			break;
3473 		}
3474 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3475 			error = EINVAL;
3476 			break;
3477 		}
3478 
3479 		if (pcr->action != PF_CHANGE_REMOVE) {
3480 			newrule = pf_krule_alloc();
3481 			error = pf_rule_to_krule(&pcr->rule, newrule);
3482 			if (error != 0) {
3483 				pf_krule_free(newrule);
3484 				break;
3485 			}
3486 
3487 			if (newrule->ifname[0])
3488 				kif = pf_kkif_create(M_WAITOK);
3489 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3490 			for (int i = 0; i < 2; i++) {
3491 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3492 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3493 			}
3494 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3495 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3496 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3497 			newrule->cuid = td->td_ucred->cr_ruid;
3498 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3499 			TAILQ_INIT(&newrule->rpool.list);
3500 		}
3501 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3502 
3503 		PF_CONFIG_LOCK();
3504 		PF_RULES_WLOCK();
3505 #ifdef PF_WANT_32_TO_64_COUNTER
3506 		if (newrule != NULL) {
3507 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3508 			newrule->allrulelinked = true;
3509 			V_pf_allrulecount++;
3510 		}
3511 #endif
3512 
3513 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3514 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3515 		    pcr->pool_ticket != V_ticket_pabuf)
3516 			ERROUT(EBUSY);
3517 
3518 		ruleset = pf_find_kruleset(pcr->anchor);
3519 		if (ruleset == NULL)
3520 			ERROUT(EINVAL);
3521 
3522 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3523 		if (rs_num >= PF_RULESET_MAX)
3524 			ERROUT(EINVAL);
3525 
3526 		/*
3527 		 * XXXMJG: there is no guarantee that the ruleset was
3528 		 * created by the usual route of calling DIOCXBEGIN.
3529 		 * As a result it is possible the rule tree will not
3530 		 * be allocated yet. Hack around it by doing it here.
3531 		 * Note it is fine to let the tree persist in case of
3532 		 * error as it will be freed down the road on future
3533 		 * updates (if need be).
3534 		 */
3535 		if (ruleset->rules[rs_num].active.tree == NULL) {
3536 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3537 			if (ruleset->rules[rs_num].active.tree == NULL) {
3538 				ERROUT(ENOMEM);
3539 			}
3540 		}
3541 
3542 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3543 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3544 			ERROUT(0);
3545 		} else if (pcr->ticket !=
3546 			    ruleset->rules[rs_num].active.ticket)
3547 				ERROUT(EINVAL);
3548 
3549 		if (pcr->action != PF_CHANGE_REMOVE) {
3550 			if (newrule->ifname[0]) {
3551 				newrule->kif = pfi_kkif_attach(kif,
3552 				    newrule->ifname);
3553 				kif = NULL;
3554 				pfi_kkif_ref(newrule->kif);
3555 			} else
3556 				newrule->kif = NULL;
3557 
3558 			if (newrule->rtableid > 0 &&
3559 			    newrule->rtableid >= rt_numfibs)
3560 				error = EBUSY;
3561 
3562 #ifdef ALTQ
3563 			/* set queue IDs */
3564 			if (newrule->qname[0] != 0) {
3565 				if ((newrule->qid =
3566 				    pf_qname2qid(newrule->qname)) == 0)
3567 					error = EBUSY;
3568 				else if (newrule->pqname[0] != 0) {
3569 					if ((newrule->pqid =
3570 					    pf_qname2qid(newrule->pqname)) == 0)
3571 						error = EBUSY;
3572 				} else
3573 					newrule->pqid = newrule->qid;
3574 			}
3575 #endif /* ALTQ */
3576 			if (newrule->tagname[0])
3577 				if ((newrule->tag =
3578 				    pf_tagname2tag(newrule->tagname)) == 0)
3579 					error = EBUSY;
3580 			if (newrule->match_tagname[0])
3581 				if ((newrule->match_tag = pf_tagname2tag(
3582 				    newrule->match_tagname)) == 0)
3583 					error = EBUSY;
3584 			if (newrule->rt && !newrule->direction)
3585 				error = EINVAL;
3586 			if (!newrule->log)
3587 				newrule->logif = 0;
3588 			if (newrule->logif >= PFLOGIFS_MAX)
3589 				error = EINVAL;
3590 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3591 				error = ENOMEM;
3592 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3593 				error = ENOMEM;
3594 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3595 				error = EINVAL;
3596 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3597 				if (pa->addr.type == PF_ADDR_TABLE) {
3598 					pa->addr.p.tbl =
3599 					    pfr_attach_table(ruleset,
3600 					    pa->addr.v.tblname);
3601 					if (pa->addr.p.tbl == NULL)
3602 						error = ENOMEM;
3603 				}
3604 
3605 			newrule->overload_tbl = NULL;
3606 			if (newrule->overload_tblname[0]) {
3607 				if ((newrule->overload_tbl = pfr_attach_table(
3608 				    ruleset, newrule->overload_tblname)) ==
3609 				    NULL)
3610 					error = EINVAL;
3611 				else
3612 					newrule->overload_tbl->pfrkt_flags |=
3613 					    PFR_TFLAG_ACTIVE;
3614 			}
3615 
3616 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3617 			if (((((newrule->action == PF_NAT) ||
3618 			    (newrule->action == PF_RDR) ||
3619 			    (newrule->action == PF_BINAT) ||
3620 			    (newrule->rt > PF_NOPFROUTE)) &&
3621 			    !newrule->anchor)) &&
3622 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3623 				error = EINVAL;
3624 
3625 			if (error) {
3626 				pf_free_rule(newrule);
3627 				PF_RULES_WUNLOCK();
3628 				PF_CONFIG_UNLOCK();
3629 				break;
3630 			}
3631 
3632 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3633 		}
3634 		pf_empty_kpool(&V_pf_pabuf);
3635 
3636 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3637 			oldrule = TAILQ_FIRST(
3638 			    ruleset->rules[rs_num].active.ptr);
3639 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3640 			oldrule = TAILQ_LAST(
3641 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3642 		else {
3643 			oldrule = TAILQ_FIRST(
3644 			    ruleset->rules[rs_num].active.ptr);
3645 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3646 				oldrule = TAILQ_NEXT(oldrule, entries);
3647 			if (oldrule == NULL) {
3648 				if (newrule != NULL)
3649 					pf_free_rule(newrule);
3650 				PF_RULES_WUNLOCK();
3651 				PF_CONFIG_UNLOCK();
3652 				error = EINVAL;
3653 				break;
3654 			}
3655 		}
3656 
3657 		if (pcr->action == PF_CHANGE_REMOVE) {
3658 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3659 			    oldrule);
3660 			RB_REMOVE(pf_krule_global,
3661 			    ruleset->rules[rs_num].active.tree, oldrule);
3662 			ruleset->rules[rs_num].active.rcount--;
3663 		} else {
3664 			pf_hash_rule(newrule);
3665 			if (RB_INSERT(pf_krule_global,
3666 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3667 				pf_free_rule(newrule);
3668 				PF_RULES_WUNLOCK();
3669 				PF_CONFIG_UNLOCK();
3670 				error = EEXIST;
3671 				break;
3672 			}
3673 
3674 			if (oldrule == NULL)
3675 				TAILQ_INSERT_TAIL(
3676 				    ruleset->rules[rs_num].active.ptr,
3677 				    newrule, entries);
3678 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3679 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3680 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3681 			else
3682 				TAILQ_INSERT_AFTER(
3683 				    ruleset->rules[rs_num].active.ptr,
3684 				    oldrule, newrule, entries);
3685 			ruleset->rules[rs_num].active.rcount++;
3686 		}
3687 
3688 		nr = 0;
3689 		TAILQ_FOREACH(oldrule,
3690 		    ruleset->rules[rs_num].active.ptr, entries)
3691 			oldrule->nr = nr++;
3692 
3693 		ruleset->rules[rs_num].active.ticket++;
3694 
3695 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3696 		pf_remove_if_empty_kruleset(ruleset);
3697 
3698 		PF_RULES_WUNLOCK();
3699 		PF_CONFIG_UNLOCK();
3700 		break;
3701 
3702 #undef ERROUT
3703 DIOCCHANGERULE_error:
3704 		PF_RULES_WUNLOCK();
3705 		PF_CONFIG_UNLOCK();
3706 		pf_krule_free(newrule);
3707 		pf_kkif_free(kif);
3708 		break;
3709 	}
3710 
3711 	case DIOCCLRSTATESNV: {
3712 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3713 		break;
3714 	}
3715 
3716 	case DIOCKILLSTATESNV: {
3717 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3718 		break;
3719 	}
3720 
3721 	case DIOCADDSTATE: {
3722 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3723 		struct pfsync_state_1301	*sp = &ps->state;
3724 
3725 		if (sp->timeout >= PFTM_MAX) {
3726 			error = EINVAL;
3727 			break;
3728 		}
3729 		if (V_pfsync_state_import_ptr != NULL) {
3730 			PF_RULES_RLOCK();
3731 			error = V_pfsync_state_import_ptr(
3732 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3733 			    PFSYNC_MSG_VERSION_1301);
3734 			PF_RULES_RUNLOCK();
3735 		} else
3736 			error = EOPNOTSUPP;
3737 		break;
3738 	}
3739 
3740 	case DIOCGETSTATE: {
3741 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3742 		struct pf_kstate	*s;
3743 
3744 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3745 		if (s == NULL) {
3746 			error = ENOENT;
3747 			break;
3748 		}
3749 
3750 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3751 		    s, PFSYNC_MSG_VERSION_1301);
3752 		PF_STATE_UNLOCK(s);
3753 		break;
3754 	}
3755 
3756 	case DIOCGETSTATENV: {
3757 		error = pf_getstate((struct pfioc_nv *)addr);
3758 		break;
3759 	}
3760 
3761 #ifdef COMPAT_FREEBSD14
3762 	case DIOCGETSTATES: {
3763 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3764 		struct pf_kstate	*s;
3765 		struct pfsync_state_1301	*pstore, *p;
3766 		int			 i, nr;
3767 		size_t			 slice_count = 16, count;
3768 		void			*out;
3769 
3770 		if (ps->ps_len <= 0) {
3771 			nr = uma_zone_get_cur(V_pf_state_z);
3772 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3773 			break;
3774 		}
3775 
3776 		out = ps->ps_states;
3777 		pstore = mallocarray(slice_count,
3778 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3779 		nr = 0;
3780 
3781 		for (i = 0; i <= pf_hashmask; i++) {
3782 			struct pf_idhash *ih = &V_pf_idhash[i];
3783 
3784 DIOCGETSTATES_retry:
3785 			p = pstore;
3786 
3787 			if (LIST_EMPTY(&ih->states))
3788 				continue;
3789 
3790 			PF_HASHROW_LOCK(ih);
3791 			count = 0;
3792 			LIST_FOREACH(s, &ih->states, entry) {
3793 				if (s->timeout == PFTM_UNLINKED)
3794 					continue;
3795 				count++;
3796 			}
3797 
3798 			if (count > slice_count) {
3799 				PF_HASHROW_UNLOCK(ih);
3800 				free(pstore, M_TEMP);
3801 				slice_count = count * 2;
3802 				pstore = mallocarray(slice_count,
3803 				    sizeof(struct pfsync_state_1301), M_TEMP,
3804 				    M_WAITOK | M_ZERO);
3805 				goto DIOCGETSTATES_retry;
3806 			}
3807 
3808 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3809 				PF_HASHROW_UNLOCK(ih);
3810 				goto DIOCGETSTATES_full;
3811 			}
3812 
3813 			LIST_FOREACH(s, &ih->states, entry) {
3814 				if (s->timeout == PFTM_UNLINKED)
3815 					continue;
3816 
3817 				pfsync_state_export((union pfsync_state_union*)p,
3818 				    s, PFSYNC_MSG_VERSION_1301);
3819 				p++;
3820 				nr++;
3821 			}
3822 			PF_HASHROW_UNLOCK(ih);
3823 			error = copyout(pstore, out,
3824 			    sizeof(struct pfsync_state_1301) * count);
3825 			if (error)
3826 				break;
3827 			out = ps->ps_states + nr;
3828 		}
3829 DIOCGETSTATES_full:
3830 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3831 		free(pstore, M_TEMP);
3832 
3833 		break;
3834 	}
3835 
3836 	case DIOCGETSTATESV2: {
3837 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3838 		struct pf_kstate	*s;
3839 		struct pf_state_export	*pstore, *p;
3840 		int i, nr;
3841 		size_t slice_count = 16, count;
3842 		void *out;
3843 
3844 		if (ps->ps_req_version > PF_STATE_VERSION) {
3845 			error = ENOTSUP;
3846 			break;
3847 		}
3848 
3849 		if (ps->ps_len <= 0) {
3850 			nr = uma_zone_get_cur(V_pf_state_z);
3851 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3852 			break;
3853 		}
3854 
3855 		out = ps->ps_states;
3856 		pstore = mallocarray(slice_count,
3857 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3858 		nr = 0;
3859 
3860 		for (i = 0; i <= pf_hashmask; i++) {
3861 			struct pf_idhash *ih = &V_pf_idhash[i];
3862 
3863 DIOCGETSTATESV2_retry:
3864 			p = pstore;
3865 
3866 			if (LIST_EMPTY(&ih->states))
3867 				continue;
3868 
3869 			PF_HASHROW_LOCK(ih);
3870 			count = 0;
3871 			LIST_FOREACH(s, &ih->states, entry) {
3872 				if (s->timeout == PFTM_UNLINKED)
3873 					continue;
3874 				count++;
3875 			}
3876 
3877 			if (count > slice_count) {
3878 				PF_HASHROW_UNLOCK(ih);
3879 				free(pstore, M_TEMP);
3880 				slice_count = count * 2;
3881 				pstore = mallocarray(slice_count,
3882 				    sizeof(struct pf_state_export), M_TEMP,
3883 				    M_WAITOK | M_ZERO);
3884 				goto DIOCGETSTATESV2_retry;
3885 			}
3886 
3887 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3888 				PF_HASHROW_UNLOCK(ih);
3889 				goto DIOCGETSTATESV2_full;
3890 			}
3891 
3892 			LIST_FOREACH(s, &ih->states, entry) {
3893 				if (s->timeout == PFTM_UNLINKED)
3894 					continue;
3895 
3896 				pf_state_export(p, s);
3897 				p++;
3898 				nr++;
3899 			}
3900 			PF_HASHROW_UNLOCK(ih);
3901 			error = copyout(pstore, out,
3902 			    sizeof(struct pf_state_export) * count);
3903 			if (error)
3904 				break;
3905 			out = ps->ps_states + nr;
3906 		}
3907 DIOCGETSTATESV2_full:
3908 		ps->ps_len = nr * sizeof(struct pf_state_export);
3909 		free(pstore, M_TEMP);
3910 
3911 		break;
3912 	}
3913 #endif
3914 	case DIOCGETSTATUSNV: {
3915 		error = pf_getstatus((struct pfioc_nv *)addr);
3916 		break;
3917 	}
3918 
3919 	case DIOCSETSTATUSIF: {
3920 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3921 
3922 		if (pi->ifname[0] == 0) {
3923 			bzero(V_pf_status.ifname, IFNAMSIZ);
3924 			break;
3925 		}
3926 		PF_RULES_WLOCK();
3927 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3928 		PF_RULES_WUNLOCK();
3929 		break;
3930 	}
3931 
3932 	case DIOCCLRSTATUS: {
3933 		pf_ioctl_clear_status();
3934 		break;
3935 	}
3936 
3937 	case DIOCNATLOOK: {
3938 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3939 		struct pf_state_key	*sk;
3940 		struct pf_kstate	*state;
3941 		struct pf_state_key_cmp	 key;
3942 		int			 m = 0, direction = pnl->direction;
3943 		int			 sidx, didx;
3944 
3945 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3946 		sidx = (direction == PF_IN) ? 1 : 0;
3947 		didx = (direction == PF_IN) ? 0 : 1;
3948 
3949 		if (!pnl->proto ||
3950 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3951 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3952 		    ((pnl->proto == IPPROTO_TCP ||
3953 		    pnl->proto == IPPROTO_UDP) &&
3954 		    (!pnl->dport || !pnl->sport)))
3955 			error = EINVAL;
3956 		else {
3957 			bzero(&key, sizeof(key));
3958 			key.af = pnl->af;
3959 			key.proto = pnl->proto;
3960 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3961 			key.port[sidx] = pnl->sport;
3962 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3963 			key.port[didx] = pnl->dport;
3964 
3965 			state = pf_find_state_all(&key, direction, &m);
3966 			if (state == NULL) {
3967 				error = ENOENT;
3968 			} else {
3969 				if (m > 1) {
3970 					PF_STATE_UNLOCK(state);
3971 					error = E2BIG;	/* more than one state */
3972 				} else {
3973 					sk = state->key[sidx];
3974 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3975 					pnl->rsport = sk->port[sidx];
3976 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3977 					pnl->rdport = sk->port[didx];
3978 					PF_STATE_UNLOCK(state);
3979 				}
3980 			}
3981 		}
3982 		break;
3983 	}
3984 
3985 	case DIOCSETTIMEOUT: {
3986 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3987 
3988 		error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
3989 		    &pt->seconds);
3990 		break;
3991 	}
3992 
3993 	case DIOCGETTIMEOUT: {
3994 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3995 
3996 		error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
3997 		break;
3998 	}
3999 
4000 	case DIOCGETLIMIT: {
4001 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4002 
4003 		error = pf_ioctl_get_limit(pl->index, &pl->limit);
4004 		break;
4005 	}
4006 
4007 	case DIOCSETLIMIT: {
4008 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4009 		unsigned int old_limit;
4010 
4011 		error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4012 		pl->limit = old_limit;
4013 		break;
4014 	}
4015 
4016 	case DIOCSETDEBUG: {
4017 		u_int32_t	*level = (u_int32_t *)addr;
4018 
4019 		PF_RULES_WLOCK();
4020 		V_pf_status.debug = *level;
4021 		PF_RULES_WUNLOCK();
4022 		break;
4023 	}
4024 
4025 	case DIOCCLRRULECTRS: {
4026 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4027 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
4028 		struct pf_krule		*rule;
4029 
4030 		PF_RULES_WLOCK();
4031 		TAILQ_FOREACH(rule,
4032 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4033 			pf_counter_u64_zero(&rule->evaluations);
4034 			for (int i = 0; i < 2; i++) {
4035 				pf_counter_u64_zero(&rule->packets[i]);
4036 				pf_counter_u64_zero(&rule->bytes[i]);
4037 			}
4038 		}
4039 		PF_RULES_WUNLOCK();
4040 		break;
4041 	}
4042 
4043 	case DIOCGIFSPEEDV0:
4044 	case DIOCGIFSPEEDV1: {
4045 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
4046 		struct pf_ifspeed_v1	ps;
4047 		struct ifnet		*ifp;
4048 
4049 		if (psp->ifname[0] == '\0') {
4050 			error = EINVAL;
4051 			break;
4052 		}
4053 
4054 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4055 		if (error != 0)
4056 			break;
4057 		ifp = ifunit(ps.ifname);
4058 		if (ifp != NULL) {
4059 			psp->baudrate32 =
4060 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4061 			if (cmd == DIOCGIFSPEEDV1)
4062 				psp->baudrate = ifp->if_baudrate;
4063 		} else {
4064 			error = EINVAL;
4065 		}
4066 		break;
4067 	}
4068 
4069 #ifdef ALTQ
4070 	case DIOCSTARTALTQ: {
4071 		struct pf_altq		*altq;
4072 
4073 		PF_RULES_WLOCK();
4074 		/* enable all altq interfaces on active list */
4075 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4076 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4077 				error = pf_enable_altq(altq);
4078 				if (error != 0)
4079 					break;
4080 			}
4081 		}
4082 		if (error == 0)
4083 			V_pf_altq_running = 1;
4084 		PF_RULES_WUNLOCK();
4085 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4086 		break;
4087 	}
4088 
4089 	case DIOCSTOPALTQ: {
4090 		struct pf_altq		*altq;
4091 
4092 		PF_RULES_WLOCK();
4093 		/* disable all altq interfaces on active list */
4094 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4095 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4096 				error = pf_disable_altq(altq);
4097 				if (error != 0)
4098 					break;
4099 			}
4100 		}
4101 		if (error == 0)
4102 			V_pf_altq_running = 0;
4103 		PF_RULES_WUNLOCK();
4104 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4105 		break;
4106 	}
4107 
4108 	case DIOCADDALTQV0:
4109 	case DIOCADDALTQV1: {
4110 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4111 		struct pf_altq		*altq, *a;
4112 		struct ifnet		*ifp;
4113 
4114 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4115 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4116 		if (error)
4117 			break;
4118 		altq->local_flags = 0;
4119 
4120 		PF_RULES_WLOCK();
4121 		if (pa->ticket != V_ticket_altqs_inactive) {
4122 			PF_RULES_WUNLOCK();
4123 			free(altq, M_PFALTQ);
4124 			error = EBUSY;
4125 			break;
4126 		}
4127 
4128 		/*
4129 		 * if this is for a queue, find the discipline and
4130 		 * copy the necessary fields
4131 		 */
4132 		if (altq->qname[0] != 0) {
4133 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4134 				PF_RULES_WUNLOCK();
4135 				error = EBUSY;
4136 				free(altq, M_PFALTQ);
4137 				break;
4138 			}
4139 			altq->altq_disc = NULL;
4140 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4141 				if (strncmp(a->ifname, altq->ifname,
4142 				    IFNAMSIZ) == 0) {
4143 					altq->altq_disc = a->altq_disc;
4144 					break;
4145 				}
4146 			}
4147 		}
4148 
4149 		if ((ifp = ifunit(altq->ifname)) == NULL)
4150 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4151 		else
4152 			error = altq_add(ifp, altq);
4153 
4154 		if (error) {
4155 			PF_RULES_WUNLOCK();
4156 			free(altq, M_PFALTQ);
4157 			break;
4158 		}
4159 
4160 		if (altq->qname[0] != 0)
4161 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4162 		else
4163 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4164 		/* version error check done on import above */
4165 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4166 		PF_RULES_WUNLOCK();
4167 		break;
4168 	}
4169 
4170 	case DIOCGETALTQSV0:
4171 	case DIOCGETALTQSV1: {
4172 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4173 		struct pf_altq		*altq;
4174 
4175 		PF_RULES_RLOCK();
4176 		pa->nr = 0;
4177 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4178 			pa->nr++;
4179 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4180 			pa->nr++;
4181 		pa->ticket = V_ticket_altqs_active;
4182 		PF_RULES_RUNLOCK();
4183 		break;
4184 	}
4185 
4186 	case DIOCGETALTQV0:
4187 	case DIOCGETALTQV1: {
4188 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4189 		struct pf_altq		*altq;
4190 
4191 		PF_RULES_RLOCK();
4192 		if (pa->ticket != V_ticket_altqs_active) {
4193 			PF_RULES_RUNLOCK();
4194 			error = EBUSY;
4195 			break;
4196 		}
4197 		altq = pf_altq_get_nth_active(pa->nr);
4198 		if (altq == NULL) {
4199 			PF_RULES_RUNLOCK();
4200 			error = EBUSY;
4201 			break;
4202 		}
4203 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4204 		PF_RULES_RUNLOCK();
4205 		break;
4206 	}
4207 
4208 	case DIOCCHANGEALTQV0:
4209 	case DIOCCHANGEALTQV1:
4210 		/* CHANGEALTQ not supported yet! */
4211 		error = ENODEV;
4212 		break;
4213 
4214 	case DIOCGETQSTATSV0:
4215 	case DIOCGETQSTATSV1: {
4216 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4217 		struct pf_altq		*altq;
4218 		int			 nbytes;
4219 		u_int32_t		 version;
4220 
4221 		PF_RULES_RLOCK();
4222 		if (pq->ticket != V_ticket_altqs_active) {
4223 			PF_RULES_RUNLOCK();
4224 			error = EBUSY;
4225 			break;
4226 		}
4227 		nbytes = pq->nbytes;
4228 		altq = pf_altq_get_nth_active(pq->nr);
4229 		if (altq == NULL) {
4230 			PF_RULES_RUNLOCK();
4231 			error = EBUSY;
4232 			break;
4233 		}
4234 
4235 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4236 			PF_RULES_RUNLOCK();
4237 			error = ENXIO;
4238 			break;
4239 		}
4240 		PF_RULES_RUNLOCK();
4241 		if (cmd == DIOCGETQSTATSV0)
4242 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4243 		else
4244 			version = pq->version;
4245 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4246 		if (error == 0) {
4247 			pq->scheduler = altq->scheduler;
4248 			pq->nbytes = nbytes;
4249 		}
4250 		break;
4251 	}
4252 #endif /* ALTQ */
4253 
4254 	case DIOCBEGINADDRS: {
4255 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4256 
4257 		error = pf_ioctl_begin_addrs(&pp->ticket);
4258 		break;
4259 	}
4260 
4261 	case DIOCADDADDR: {
4262 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4263 
4264 		error = pf_ioctl_add_addr(pp);
4265 		break;
4266 	}
4267 
4268 	case DIOCGETADDRS: {
4269 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4270 		struct pf_kpool		*pool;
4271 		struct pf_kpooladdr	*pa;
4272 
4273 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4274 		pp->nr = 0;
4275 
4276 		PF_RULES_RLOCK();
4277 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4278 		    pp->r_num, 0, 1, 0);
4279 		if (pool == NULL) {
4280 			PF_RULES_RUNLOCK();
4281 			error = EBUSY;
4282 			break;
4283 		}
4284 		TAILQ_FOREACH(pa, &pool->list, entries)
4285 			pp->nr++;
4286 		PF_RULES_RUNLOCK();
4287 		break;
4288 	}
4289 
4290 	case DIOCGETADDR: {
4291 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4292 		struct pf_kpool		*pool;
4293 		struct pf_kpooladdr	*pa;
4294 		u_int32_t		 nr = 0;
4295 
4296 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4297 
4298 		PF_RULES_RLOCK();
4299 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4300 		    pp->r_num, 0, 1, 1);
4301 		if (pool == NULL) {
4302 			PF_RULES_RUNLOCK();
4303 			error = EBUSY;
4304 			break;
4305 		}
4306 		pa = TAILQ_FIRST(&pool->list);
4307 		while ((pa != NULL) && (nr < pp->nr)) {
4308 			pa = TAILQ_NEXT(pa, entries);
4309 			nr++;
4310 		}
4311 		if (pa == NULL) {
4312 			PF_RULES_RUNLOCK();
4313 			error = EBUSY;
4314 			break;
4315 		}
4316 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4317 		pf_addr_copyout(&pp->addr.addr);
4318 		PF_RULES_RUNLOCK();
4319 		break;
4320 	}
4321 
4322 	case DIOCCHANGEADDR: {
4323 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4324 		struct pf_kpool		*pool;
4325 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4326 		struct pf_kruleset	*ruleset;
4327 		struct pfi_kkif		*kif = NULL;
4328 
4329 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4330 
4331 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4332 		    pca->action > PF_CHANGE_REMOVE) {
4333 			error = EINVAL;
4334 			break;
4335 		}
4336 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4337 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4338 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4339 			error = EINVAL;
4340 			break;
4341 		}
4342 		if (pca->addr.addr.p.dyn != NULL) {
4343 			error = EINVAL;
4344 			break;
4345 		}
4346 
4347 		if (pca->action != PF_CHANGE_REMOVE) {
4348 #ifndef INET
4349 			if (pca->af == AF_INET) {
4350 				error = EAFNOSUPPORT;
4351 				break;
4352 			}
4353 #endif /* INET */
4354 #ifndef INET6
4355 			if (pca->af == AF_INET6) {
4356 				error = EAFNOSUPPORT;
4357 				break;
4358 			}
4359 #endif /* INET6 */
4360 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4361 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4362 			if (newpa->ifname[0])
4363 				kif = pf_kkif_create(M_WAITOK);
4364 			newpa->kif = NULL;
4365 		}
4366 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4367 		PF_RULES_WLOCK();
4368 		ruleset = pf_find_kruleset(pca->anchor);
4369 		if (ruleset == NULL)
4370 			ERROUT(EBUSY);
4371 
4372 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4373 		    pca->r_num, pca->r_last, 1, 1);
4374 		if (pool == NULL)
4375 			ERROUT(EBUSY);
4376 
4377 		if (pca->action != PF_CHANGE_REMOVE) {
4378 			if (newpa->ifname[0]) {
4379 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4380 				pfi_kkif_ref(newpa->kif);
4381 				kif = NULL;
4382 			}
4383 
4384 			switch (newpa->addr.type) {
4385 			case PF_ADDR_DYNIFTL:
4386 				error = pfi_dynaddr_setup(&newpa->addr,
4387 				    pca->af);
4388 				break;
4389 			case PF_ADDR_TABLE:
4390 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4391 				    newpa->addr.v.tblname);
4392 				if (newpa->addr.p.tbl == NULL)
4393 					error = ENOMEM;
4394 				break;
4395 			}
4396 			if (error)
4397 				goto DIOCCHANGEADDR_error;
4398 		}
4399 
4400 		switch (pca->action) {
4401 		case PF_CHANGE_ADD_HEAD:
4402 			oldpa = TAILQ_FIRST(&pool->list);
4403 			break;
4404 		case PF_CHANGE_ADD_TAIL:
4405 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4406 			break;
4407 		default:
4408 			oldpa = TAILQ_FIRST(&pool->list);
4409 			for (int i = 0; oldpa && i < pca->nr; i++)
4410 				oldpa = TAILQ_NEXT(oldpa, entries);
4411 
4412 			if (oldpa == NULL)
4413 				ERROUT(EINVAL);
4414 		}
4415 
4416 		if (pca->action == PF_CHANGE_REMOVE) {
4417 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4418 			switch (oldpa->addr.type) {
4419 			case PF_ADDR_DYNIFTL:
4420 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4421 				break;
4422 			case PF_ADDR_TABLE:
4423 				pfr_detach_table(oldpa->addr.p.tbl);
4424 				break;
4425 			}
4426 			if (oldpa->kif)
4427 				pfi_kkif_unref(oldpa->kif);
4428 			free(oldpa, M_PFRULE);
4429 		} else {
4430 			if (oldpa == NULL)
4431 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4432 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4433 			    pca->action == PF_CHANGE_ADD_BEFORE)
4434 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4435 			else
4436 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4437 				    newpa, entries);
4438 		}
4439 
4440 		pool->cur = TAILQ_FIRST(&pool->list);
4441 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4442 		PF_RULES_WUNLOCK();
4443 		break;
4444 
4445 #undef ERROUT
4446 DIOCCHANGEADDR_error:
4447 		if (newpa != NULL) {
4448 			if (newpa->kif)
4449 				pfi_kkif_unref(newpa->kif);
4450 			free(newpa, M_PFRULE);
4451 		}
4452 		PF_RULES_WUNLOCK();
4453 		pf_kkif_free(kif);
4454 		break;
4455 	}
4456 
4457 	case DIOCGETRULESETS: {
4458 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4459 		struct pf_kruleset	*ruleset;
4460 		struct pf_kanchor	*anchor;
4461 
4462 		pr->path[sizeof(pr->path) - 1] = 0;
4463 
4464 		PF_RULES_RLOCK();
4465 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4466 			PF_RULES_RUNLOCK();
4467 			error = ENOENT;
4468 			break;
4469 		}
4470 		pr->nr = 0;
4471 		if (ruleset->anchor == NULL) {
4472 			/* XXX kludge for pf_main_ruleset */
4473 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4474 				if (anchor->parent == NULL)
4475 					pr->nr++;
4476 		} else {
4477 			RB_FOREACH(anchor, pf_kanchor_node,
4478 			    &ruleset->anchor->children)
4479 				pr->nr++;
4480 		}
4481 		PF_RULES_RUNLOCK();
4482 		break;
4483 	}
4484 
4485 	case DIOCGETRULESET: {
4486 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4487 		struct pf_kruleset	*ruleset;
4488 		struct pf_kanchor	*anchor;
4489 		u_int32_t		 nr = 0;
4490 
4491 		pr->path[sizeof(pr->path) - 1] = 0;
4492 
4493 		PF_RULES_RLOCK();
4494 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4495 			PF_RULES_RUNLOCK();
4496 			error = ENOENT;
4497 			break;
4498 		}
4499 		pr->name[0] = 0;
4500 		if (ruleset->anchor == NULL) {
4501 			/* XXX kludge for pf_main_ruleset */
4502 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4503 				if (anchor->parent == NULL && nr++ == pr->nr) {
4504 					strlcpy(pr->name, anchor->name,
4505 					    sizeof(pr->name));
4506 					break;
4507 				}
4508 		} else {
4509 			RB_FOREACH(anchor, pf_kanchor_node,
4510 			    &ruleset->anchor->children)
4511 				if (nr++ == pr->nr) {
4512 					strlcpy(pr->name, anchor->name,
4513 					    sizeof(pr->name));
4514 					break;
4515 				}
4516 		}
4517 		if (!pr->name[0])
4518 			error = EBUSY;
4519 		PF_RULES_RUNLOCK();
4520 		break;
4521 	}
4522 
4523 	case DIOCRCLRTABLES: {
4524 		struct pfioc_table *io = (struct pfioc_table *)addr;
4525 
4526 		if (io->pfrio_esize != 0) {
4527 			error = ENODEV;
4528 			break;
4529 		}
4530 		PF_RULES_WLOCK();
4531 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4532 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4533 		PF_RULES_WUNLOCK();
4534 		break;
4535 	}
4536 
4537 	case DIOCRADDTABLES: {
4538 		struct pfioc_table *io = (struct pfioc_table *)addr;
4539 		struct pfr_table *pfrts;
4540 		size_t totlen;
4541 
4542 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4543 			error = ENODEV;
4544 			break;
4545 		}
4546 
4547 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4548 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4549 			error = ENOMEM;
4550 			break;
4551 		}
4552 
4553 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4554 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4555 		    M_TEMP, M_WAITOK);
4556 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4557 		if (error) {
4558 			free(pfrts, M_TEMP);
4559 			break;
4560 		}
4561 		PF_RULES_WLOCK();
4562 		error = pfr_add_tables(pfrts, io->pfrio_size,
4563 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4564 		PF_RULES_WUNLOCK();
4565 		free(pfrts, M_TEMP);
4566 		break;
4567 	}
4568 
4569 	case DIOCRDELTABLES: {
4570 		struct pfioc_table *io = (struct pfioc_table *)addr;
4571 		struct pfr_table *pfrts;
4572 		size_t totlen;
4573 
4574 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4575 			error = ENODEV;
4576 			break;
4577 		}
4578 
4579 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4580 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4581 			error = ENOMEM;
4582 			break;
4583 		}
4584 
4585 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4586 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4587 		    M_TEMP, M_WAITOK);
4588 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4589 		if (error) {
4590 			free(pfrts, M_TEMP);
4591 			break;
4592 		}
4593 		PF_RULES_WLOCK();
4594 		error = pfr_del_tables(pfrts, io->pfrio_size,
4595 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4596 		PF_RULES_WUNLOCK();
4597 		free(pfrts, M_TEMP);
4598 		break;
4599 	}
4600 
4601 	case DIOCRGETTABLES: {
4602 		struct pfioc_table *io = (struct pfioc_table *)addr;
4603 		struct pfr_table *pfrts;
4604 		size_t totlen;
4605 		int n;
4606 
4607 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4608 			error = ENODEV;
4609 			break;
4610 		}
4611 		PF_RULES_RLOCK();
4612 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4613 		if (n < 0) {
4614 			PF_RULES_RUNLOCK();
4615 			error = EINVAL;
4616 			break;
4617 		}
4618 		io->pfrio_size = min(io->pfrio_size, n);
4619 
4620 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4621 
4622 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4623 		    M_TEMP, M_NOWAIT | M_ZERO);
4624 		if (pfrts == NULL) {
4625 			error = ENOMEM;
4626 			PF_RULES_RUNLOCK();
4627 			break;
4628 		}
4629 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4630 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4631 		PF_RULES_RUNLOCK();
4632 		if (error == 0)
4633 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4634 		free(pfrts, M_TEMP);
4635 		break;
4636 	}
4637 
4638 	case DIOCRGETTSTATS: {
4639 		struct pfioc_table *io = (struct pfioc_table *)addr;
4640 		struct pfr_tstats *pfrtstats;
4641 		size_t totlen;
4642 		int n;
4643 
4644 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4645 			error = ENODEV;
4646 			break;
4647 		}
4648 		PF_TABLE_STATS_LOCK();
4649 		PF_RULES_RLOCK();
4650 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4651 		if (n < 0) {
4652 			PF_RULES_RUNLOCK();
4653 			PF_TABLE_STATS_UNLOCK();
4654 			error = EINVAL;
4655 			break;
4656 		}
4657 		io->pfrio_size = min(io->pfrio_size, n);
4658 
4659 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4660 		pfrtstats = mallocarray(io->pfrio_size,
4661 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4662 		if (pfrtstats == NULL) {
4663 			error = ENOMEM;
4664 			PF_RULES_RUNLOCK();
4665 			PF_TABLE_STATS_UNLOCK();
4666 			break;
4667 		}
4668 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4669 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4670 		PF_RULES_RUNLOCK();
4671 		PF_TABLE_STATS_UNLOCK();
4672 		if (error == 0)
4673 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4674 		free(pfrtstats, M_TEMP);
4675 		break;
4676 	}
4677 
4678 	case DIOCRCLRTSTATS: {
4679 		struct pfioc_table *io = (struct pfioc_table *)addr;
4680 		struct pfr_table *pfrts;
4681 		size_t totlen;
4682 
4683 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4684 			error = ENODEV;
4685 			break;
4686 		}
4687 
4688 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4689 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4690 			/* We used to count tables and use the minimum required
4691 			 * size, so we didn't fail on overly large requests.
4692 			 * Keep doing so. */
4693 			io->pfrio_size = pf_ioctl_maxcount;
4694 			break;
4695 		}
4696 
4697 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4698 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4699 		    M_TEMP, M_WAITOK);
4700 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4701 		if (error) {
4702 			free(pfrts, M_TEMP);
4703 			break;
4704 		}
4705 
4706 		PF_TABLE_STATS_LOCK();
4707 		PF_RULES_RLOCK();
4708 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4709 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4710 		PF_RULES_RUNLOCK();
4711 		PF_TABLE_STATS_UNLOCK();
4712 		free(pfrts, M_TEMP);
4713 		break;
4714 	}
4715 
4716 	case DIOCRSETTFLAGS: {
4717 		struct pfioc_table *io = (struct pfioc_table *)addr;
4718 		struct pfr_table *pfrts;
4719 		size_t totlen;
4720 		int n;
4721 
4722 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4723 			error = ENODEV;
4724 			break;
4725 		}
4726 
4727 		PF_RULES_RLOCK();
4728 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4729 		if (n < 0) {
4730 			PF_RULES_RUNLOCK();
4731 			error = EINVAL;
4732 			break;
4733 		}
4734 
4735 		io->pfrio_size = min(io->pfrio_size, n);
4736 		PF_RULES_RUNLOCK();
4737 
4738 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4739 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4740 		    M_TEMP, M_WAITOK);
4741 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4742 		if (error) {
4743 			free(pfrts, M_TEMP);
4744 			break;
4745 		}
4746 		PF_RULES_WLOCK();
4747 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4748 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4749 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4750 		PF_RULES_WUNLOCK();
4751 		free(pfrts, M_TEMP);
4752 		break;
4753 	}
4754 
4755 	case DIOCRCLRADDRS: {
4756 		struct pfioc_table *io = (struct pfioc_table *)addr;
4757 
4758 		if (io->pfrio_esize != 0) {
4759 			error = ENODEV;
4760 			break;
4761 		}
4762 		PF_RULES_WLOCK();
4763 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4764 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4765 		PF_RULES_WUNLOCK();
4766 		break;
4767 	}
4768 
4769 	case DIOCRADDADDRS: {
4770 		struct pfioc_table *io = (struct pfioc_table *)addr;
4771 		struct pfr_addr *pfras;
4772 		size_t totlen;
4773 
4774 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4775 			error = ENODEV;
4776 			break;
4777 		}
4778 		if (io->pfrio_size < 0 ||
4779 		    io->pfrio_size > pf_ioctl_maxcount ||
4780 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4781 			error = EINVAL;
4782 			break;
4783 		}
4784 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4785 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4786 		    M_TEMP, M_WAITOK);
4787 		error = copyin(io->pfrio_buffer, pfras, totlen);
4788 		if (error) {
4789 			free(pfras, M_TEMP);
4790 			break;
4791 		}
4792 		PF_RULES_WLOCK();
4793 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4794 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4795 		    PFR_FLAG_USERIOCTL);
4796 		PF_RULES_WUNLOCK();
4797 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4798 			error = copyout(pfras, io->pfrio_buffer, totlen);
4799 		free(pfras, M_TEMP);
4800 		break;
4801 	}
4802 
4803 	case DIOCRDELADDRS: {
4804 		struct pfioc_table *io = (struct pfioc_table *)addr;
4805 		struct pfr_addr *pfras;
4806 		size_t totlen;
4807 
4808 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4809 			error = ENODEV;
4810 			break;
4811 		}
4812 		if (io->pfrio_size < 0 ||
4813 		    io->pfrio_size > pf_ioctl_maxcount ||
4814 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4815 			error = EINVAL;
4816 			break;
4817 		}
4818 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4819 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4820 		    M_TEMP, M_WAITOK);
4821 		error = copyin(io->pfrio_buffer, pfras, totlen);
4822 		if (error) {
4823 			free(pfras, M_TEMP);
4824 			break;
4825 		}
4826 		PF_RULES_WLOCK();
4827 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4828 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4829 		    PFR_FLAG_USERIOCTL);
4830 		PF_RULES_WUNLOCK();
4831 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4832 			error = copyout(pfras, io->pfrio_buffer, totlen);
4833 		free(pfras, M_TEMP);
4834 		break;
4835 	}
4836 
4837 	case DIOCRSETADDRS: {
4838 		struct pfioc_table *io = (struct pfioc_table *)addr;
4839 		struct pfr_addr *pfras;
4840 		size_t totlen, count;
4841 
4842 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4843 			error = ENODEV;
4844 			break;
4845 		}
4846 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4847 			error = EINVAL;
4848 			break;
4849 		}
4850 		count = max(io->pfrio_size, io->pfrio_size2);
4851 		if (count > pf_ioctl_maxcount ||
4852 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4853 			error = EINVAL;
4854 			break;
4855 		}
4856 		totlen = count * sizeof(struct pfr_addr);
4857 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4858 		    M_WAITOK);
4859 		error = copyin(io->pfrio_buffer, pfras, totlen);
4860 		if (error) {
4861 			free(pfras, M_TEMP);
4862 			break;
4863 		}
4864 		PF_RULES_WLOCK();
4865 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4866 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4867 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4868 		    PFR_FLAG_USERIOCTL, 0);
4869 		PF_RULES_WUNLOCK();
4870 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4871 			error = copyout(pfras, io->pfrio_buffer, totlen);
4872 		free(pfras, M_TEMP);
4873 		break;
4874 	}
4875 
4876 	case DIOCRGETADDRS: {
4877 		struct pfioc_table *io = (struct pfioc_table *)addr;
4878 		struct pfr_addr *pfras;
4879 		size_t totlen;
4880 
4881 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4882 			error = ENODEV;
4883 			break;
4884 		}
4885 		if (io->pfrio_size < 0 ||
4886 		    io->pfrio_size > pf_ioctl_maxcount ||
4887 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4888 			error = EINVAL;
4889 			break;
4890 		}
4891 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4892 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4893 		    M_TEMP, M_WAITOK | M_ZERO);
4894 		PF_RULES_RLOCK();
4895 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4896 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4897 		PF_RULES_RUNLOCK();
4898 		if (error == 0)
4899 			error = copyout(pfras, io->pfrio_buffer, totlen);
4900 		free(pfras, M_TEMP);
4901 		break;
4902 	}
4903 
4904 	case DIOCRGETASTATS: {
4905 		struct pfioc_table *io = (struct pfioc_table *)addr;
4906 		struct pfr_astats *pfrastats;
4907 		size_t totlen;
4908 
4909 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4910 			error = ENODEV;
4911 			break;
4912 		}
4913 		if (io->pfrio_size < 0 ||
4914 		    io->pfrio_size > pf_ioctl_maxcount ||
4915 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4916 			error = EINVAL;
4917 			break;
4918 		}
4919 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4920 		pfrastats = mallocarray(io->pfrio_size,
4921 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4922 		PF_RULES_RLOCK();
4923 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4924 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4925 		PF_RULES_RUNLOCK();
4926 		if (error == 0)
4927 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4928 		free(pfrastats, M_TEMP);
4929 		break;
4930 	}
4931 
4932 	case DIOCRCLRASTATS: {
4933 		struct pfioc_table *io = (struct pfioc_table *)addr;
4934 		struct pfr_addr *pfras;
4935 		size_t totlen;
4936 
4937 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4938 			error = ENODEV;
4939 			break;
4940 		}
4941 		if (io->pfrio_size < 0 ||
4942 		    io->pfrio_size > pf_ioctl_maxcount ||
4943 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4944 			error = EINVAL;
4945 			break;
4946 		}
4947 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4948 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4949 		    M_TEMP, M_WAITOK);
4950 		error = copyin(io->pfrio_buffer, pfras, totlen);
4951 		if (error) {
4952 			free(pfras, M_TEMP);
4953 			break;
4954 		}
4955 		PF_RULES_WLOCK();
4956 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4957 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4958 		    PFR_FLAG_USERIOCTL);
4959 		PF_RULES_WUNLOCK();
4960 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4961 			error = copyout(pfras, io->pfrio_buffer, totlen);
4962 		free(pfras, M_TEMP);
4963 		break;
4964 	}
4965 
4966 	case DIOCRTSTADDRS: {
4967 		struct pfioc_table *io = (struct pfioc_table *)addr;
4968 		struct pfr_addr *pfras;
4969 		size_t totlen;
4970 
4971 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4972 			error = ENODEV;
4973 			break;
4974 		}
4975 		if (io->pfrio_size < 0 ||
4976 		    io->pfrio_size > pf_ioctl_maxcount ||
4977 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4978 			error = EINVAL;
4979 			break;
4980 		}
4981 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4982 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4983 		    M_TEMP, M_WAITOK);
4984 		error = copyin(io->pfrio_buffer, pfras, totlen);
4985 		if (error) {
4986 			free(pfras, M_TEMP);
4987 			break;
4988 		}
4989 		PF_RULES_RLOCK();
4990 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
4991 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4992 		    PFR_FLAG_USERIOCTL);
4993 		PF_RULES_RUNLOCK();
4994 		if (error == 0)
4995 			error = copyout(pfras, io->pfrio_buffer, totlen);
4996 		free(pfras, M_TEMP);
4997 		break;
4998 	}
4999 
5000 	case DIOCRINADEFINE: {
5001 		struct pfioc_table *io = (struct pfioc_table *)addr;
5002 		struct pfr_addr *pfras;
5003 		size_t totlen;
5004 
5005 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5006 			error = ENODEV;
5007 			break;
5008 		}
5009 		if (io->pfrio_size < 0 ||
5010 		    io->pfrio_size > pf_ioctl_maxcount ||
5011 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5012 			error = EINVAL;
5013 			break;
5014 		}
5015 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5016 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5017 		    M_TEMP, M_WAITOK);
5018 		error = copyin(io->pfrio_buffer, pfras, totlen);
5019 		if (error) {
5020 			free(pfras, M_TEMP);
5021 			break;
5022 		}
5023 		PF_RULES_WLOCK();
5024 		error = pfr_ina_define(&io->pfrio_table, pfras,
5025 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5026 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5027 		PF_RULES_WUNLOCK();
5028 		free(pfras, M_TEMP);
5029 		break;
5030 	}
5031 
5032 	case DIOCOSFPADD: {
5033 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5034 		PF_RULES_WLOCK();
5035 		error = pf_osfp_add(io);
5036 		PF_RULES_WUNLOCK();
5037 		break;
5038 	}
5039 
5040 	case DIOCOSFPGET: {
5041 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5042 		PF_RULES_RLOCK();
5043 		error = pf_osfp_get(io);
5044 		PF_RULES_RUNLOCK();
5045 		break;
5046 	}
5047 
5048 	case DIOCXBEGIN: {
5049 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5050 		struct pfioc_trans_e	*ioes, *ioe;
5051 		size_t			 totlen;
5052 		int			 i;
5053 
5054 		if (io->esize != sizeof(*ioe)) {
5055 			error = ENODEV;
5056 			break;
5057 		}
5058 		if (io->size < 0 ||
5059 		    io->size > pf_ioctl_maxcount ||
5060 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5061 			error = EINVAL;
5062 			break;
5063 		}
5064 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5065 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5066 		    M_TEMP, M_WAITOK);
5067 		error = copyin(io->array, ioes, totlen);
5068 		if (error) {
5069 			free(ioes, M_TEMP);
5070 			break;
5071 		}
5072 		/* Ensure there's no more ethernet rules to clean up. */
5073 		NET_EPOCH_DRAIN_CALLBACKS();
5074 		PF_RULES_WLOCK();
5075 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5076 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5077 			switch (ioe->rs_num) {
5078 			case PF_RULESET_ETH:
5079 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5080 					PF_RULES_WUNLOCK();
5081 					free(ioes, M_TEMP);
5082 					goto fail;
5083 				}
5084 				break;
5085 #ifdef ALTQ
5086 			case PF_RULESET_ALTQ:
5087 				if (ioe->anchor[0]) {
5088 					PF_RULES_WUNLOCK();
5089 					free(ioes, M_TEMP);
5090 					error = EINVAL;
5091 					goto fail;
5092 				}
5093 				if ((error = pf_begin_altq(&ioe->ticket))) {
5094 					PF_RULES_WUNLOCK();
5095 					free(ioes, M_TEMP);
5096 					goto fail;
5097 				}
5098 				break;
5099 #endif /* ALTQ */
5100 			case PF_RULESET_TABLE:
5101 			    {
5102 				struct pfr_table table;
5103 
5104 				bzero(&table, sizeof(table));
5105 				strlcpy(table.pfrt_anchor, ioe->anchor,
5106 				    sizeof(table.pfrt_anchor));
5107 				if ((error = pfr_ina_begin(&table,
5108 				    &ioe->ticket, NULL, 0))) {
5109 					PF_RULES_WUNLOCK();
5110 					free(ioes, M_TEMP);
5111 					goto fail;
5112 				}
5113 				break;
5114 			    }
5115 			default:
5116 				if ((error = pf_begin_rules(&ioe->ticket,
5117 				    ioe->rs_num, ioe->anchor))) {
5118 					PF_RULES_WUNLOCK();
5119 					free(ioes, M_TEMP);
5120 					goto fail;
5121 				}
5122 				break;
5123 			}
5124 		}
5125 		PF_RULES_WUNLOCK();
5126 		error = copyout(ioes, io->array, totlen);
5127 		free(ioes, M_TEMP);
5128 		break;
5129 	}
5130 
5131 	case DIOCXROLLBACK: {
5132 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5133 		struct pfioc_trans_e	*ioe, *ioes;
5134 		size_t			 totlen;
5135 		int			 i;
5136 
5137 		if (io->esize != sizeof(*ioe)) {
5138 			error = ENODEV;
5139 			break;
5140 		}
5141 		if (io->size < 0 ||
5142 		    io->size > pf_ioctl_maxcount ||
5143 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5144 			error = EINVAL;
5145 			break;
5146 		}
5147 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5148 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5149 		    M_TEMP, M_WAITOK);
5150 		error = copyin(io->array, ioes, totlen);
5151 		if (error) {
5152 			free(ioes, M_TEMP);
5153 			break;
5154 		}
5155 		PF_RULES_WLOCK();
5156 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5157 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5158 			switch (ioe->rs_num) {
5159 			case PF_RULESET_ETH:
5160 				if ((error = pf_rollback_eth(ioe->ticket,
5161 				    ioe->anchor))) {
5162 					PF_RULES_WUNLOCK();
5163 					free(ioes, M_TEMP);
5164 					goto fail; /* really bad */
5165 				}
5166 				break;
5167 #ifdef ALTQ
5168 			case PF_RULESET_ALTQ:
5169 				if (ioe->anchor[0]) {
5170 					PF_RULES_WUNLOCK();
5171 					free(ioes, M_TEMP);
5172 					error = EINVAL;
5173 					goto fail;
5174 				}
5175 				if ((error = pf_rollback_altq(ioe->ticket))) {
5176 					PF_RULES_WUNLOCK();
5177 					free(ioes, M_TEMP);
5178 					goto fail; /* really bad */
5179 				}
5180 				break;
5181 #endif /* ALTQ */
5182 			case PF_RULESET_TABLE:
5183 			    {
5184 				struct pfr_table table;
5185 
5186 				bzero(&table, sizeof(table));
5187 				strlcpy(table.pfrt_anchor, ioe->anchor,
5188 				    sizeof(table.pfrt_anchor));
5189 				if ((error = pfr_ina_rollback(&table,
5190 				    ioe->ticket, NULL, 0))) {
5191 					PF_RULES_WUNLOCK();
5192 					free(ioes, M_TEMP);
5193 					goto fail; /* really bad */
5194 				}
5195 				break;
5196 			    }
5197 			default:
5198 				if ((error = pf_rollback_rules(ioe->ticket,
5199 				    ioe->rs_num, ioe->anchor))) {
5200 					PF_RULES_WUNLOCK();
5201 					free(ioes, M_TEMP);
5202 					goto fail; /* really bad */
5203 				}
5204 				break;
5205 			}
5206 		}
5207 		PF_RULES_WUNLOCK();
5208 		free(ioes, M_TEMP);
5209 		break;
5210 	}
5211 
5212 	case DIOCXCOMMIT: {
5213 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5214 		struct pfioc_trans_e	*ioe, *ioes;
5215 		struct pf_kruleset	*rs;
5216 		struct pf_keth_ruleset	*ers;
5217 		size_t			 totlen;
5218 		int			 i;
5219 
5220 		if (io->esize != sizeof(*ioe)) {
5221 			error = ENODEV;
5222 			break;
5223 		}
5224 
5225 		if (io->size < 0 ||
5226 		    io->size > pf_ioctl_maxcount ||
5227 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5228 			error = EINVAL;
5229 			break;
5230 		}
5231 
5232 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5233 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5234 		    M_TEMP, M_WAITOK);
5235 		error = copyin(io->array, ioes, totlen);
5236 		if (error) {
5237 			free(ioes, M_TEMP);
5238 			break;
5239 		}
5240 		PF_RULES_WLOCK();
5241 		/* First makes sure everything will succeed. */
5242 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5243 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5244 			switch (ioe->rs_num) {
5245 			case PF_RULESET_ETH:
5246 				ers = pf_find_keth_ruleset(ioe->anchor);
5247 				if (ers == NULL || ioe->ticket == 0 ||
5248 				    ioe->ticket != ers->inactive.ticket) {
5249 					PF_RULES_WUNLOCK();
5250 					free(ioes, M_TEMP);
5251 					error = EINVAL;
5252 					goto fail;
5253 				}
5254 				break;
5255 #ifdef ALTQ
5256 			case PF_RULESET_ALTQ:
5257 				if (ioe->anchor[0]) {
5258 					PF_RULES_WUNLOCK();
5259 					free(ioes, M_TEMP);
5260 					error = EINVAL;
5261 					goto fail;
5262 				}
5263 				if (!V_altqs_inactive_open || ioe->ticket !=
5264 				    V_ticket_altqs_inactive) {
5265 					PF_RULES_WUNLOCK();
5266 					free(ioes, M_TEMP);
5267 					error = EBUSY;
5268 					goto fail;
5269 				}
5270 				break;
5271 #endif /* ALTQ */
5272 			case PF_RULESET_TABLE:
5273 				rs = pf_find_kruleset(ioe->anchor);
5274 				if (rs == NULL || !rs->topen || ioe->ticket !=
5275 				    rs->tticket) {
5276 					PF_RULES_WUNLOCK();
5277 					free(ioes, M_TEMP);
5278 					error = EBUSY;
5279 					goto fail;
5280 				}
5281 				break;
5282 			default:
5283 				if (ioe->rs_num < 0 || ioe->rs_num >=
5284 				    PF_RULESET_MAX) {
5285 					PF_RULES_WUNLOCK();
5286 					free(ioes, M_TEMP);
5287 					error = EINVAL;
5288 					goto fail;
5289 				}
5290 				rs = pf_find_kruleset(ioe->anchor);
5291 				if (rs == NULL ||
5292 				    !rs->rules[ioe->rs_num].inactive.open ||
5293 				    rs->rules[ioe->rs_num].inactive.ticket !=
5294 				    ioe->ticket) {
5295 					PF_RULES_WUNLOCK();
5296 					free(ioes, M_TEMP);
5297 					error = EBUSY;
5298 					goto fail;
5299 				}
5300 				break;
5301 			}
5302 		}
5303 		/* Now do the commit - no errors should happen here. */
5304 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5305 			switch (ioe->rs_num) {
5306 			case PF_RULESET_ETH:
5307 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5308 					PF_RULES_WUNLOCK();
5309 					free(ioes, M_TEMP);
5310 					goto fail; /* really bad */
5311 				}
5312 				break;
5313 #ifdef ALTQ
5314 			case PF_RULESET_ALTQ:
5315 				if ((error = pf_commit_altq(ioe->ticket))) {
5316 					PF_RULES_WUNLOCK();
5317 					free(ioes, M_TEMP);
5318 					goto fail; /* really bad */
5319 				}
5320 				break;
5321 #endif /* ALTQ */
5322 			case PF_RULESET_TABLE:
5323 			    {
5324 				struct pfr_table table;
5325 
5326 				bzero(&table, sizeof(table));
5327 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5328 				    sizeof(table.pfrt_anchor));
5329 				if ((error = pfr_ina_commit(&table,
5330 				    ioe->ticket, NULL, NULL, 0))) {
5331 					PF_RULES_WUNLOCK();
5332 					free(ioes, M_TEMP);
5333 					goto fail; /* really bad */
5334 				}
5335 				break;
5336 			    }
5337 			default:
5338 				if ((error = pf_commit_rules(ioe->ticket,
5339 				    ioe->rs_num, ioe->anchor))) {
5340 					PF_RULES_WUNLOCK();
5341 					free(ioes, M_TEMP);
5342 					goto fail; /* really bad */
5343 				}
5344 				break;
5345 			}
5346 		}
5347 		PF_RULES_WUNLOCK();
5348 
5349 		/* Only hook into EtherNet taffic if we've got rules for it. */
5350 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5351 			hook_pf_eth();
5352 		else
5353 			dehook_pf_eth();
5354 
5355 		free(ioes, M_TEMP);
5356 		break;
5357 	}
5358 
5359 	case DIOCGETSRCNODES: {
5360 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5361 		struct pf_srchash	*sh;
5362 		struct pf_ksrc_node	*n;
5363 		struct pf_src_node	*p, *pstore;
5364 		uint32_t		 i, nr = 0;
5365 
5366 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5367 				i++, sh++) {
5368 			PF_HASHROW_LOCK(sh);
5369 			LIST_FOREACH(n, &sh->nodes, entry)
5370 				nr++;
5371 			PF_HASHROW_UNLOCK(sh);
5372 		}
5373 
5374 		psn->psn_len = min(psn->psn_len,
5375 		    sizeof(struct pf_src_node) * nr);
5376 
5377 		if (psn->psn_len == 0) {
5378 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5379 			break;
5380 		}
5381 
5382 		nr = 0;
5383 
5384 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5385 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5386 		    i++, sh++) {
5387 		    PF_HASHROW_LOCK(sh);
5388 		    LIST_FOREACH(n, &sh->nodes, entry) {
5389 
5390 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5391 				break;
5392 
5393 			pf_src_node_copy(n, p);
5394 
5395 			p++;
5396 			nr++;
5397 		    }
5398 		    PF_HASHROW_UNLOCK(sh);
5399 		}
5400 		error = copyout(pstore, psn->psn_src_nodes,
5401 		    sizeof(struct pf_src_node) * nr);
5402 		if (error) {
5403 			free(pstore, M_TEMP);
5404 			break;
5405 		}
5406 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5407 		free(pstore, M_TEMP);
5408 		break;
5409 	}
5410 
5411 	case DIOCCLRSRCNODES: {
5412 		pf_clear_srcnodes(NULL);
5413 		pf_purge_expired_src_nodes();
5414 		break;
5415 	}
5416 
5417 	case DIOCKILLSRCNODES:
5418 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5419 		break;
5420 
5421 #ifdef COMPAT_FREEBSD13
5422 	case DIOCKEEPCOUNTERS_FREEBSD13:
5423 #endif
5424 	case DIOCKEEPCOUNTERS:
5425 		error = pf_keepcounters((struct pfioc_nv *)addr);
5426 		break;
5427 
5428 	case DIOCGETSYNCOOKIES:
5429 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5430 		break;
5431 
5432 	case DIOCSETSYNCOOKIES:
5433 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5434 		break;
5435 
5436 	case DIOCSETHOSTID: {
5437 		u_int32_t	*hostid = (u_int32_t *)addr;
5438 
5439 		PF_RULES_WLOCK();
5440 		if (*hostid == 0)
5441 			V_pf_status.hostid = arc4random();
5442 		else
5443 			V_pf_status.hostid = *hostid;
5444 		PF_RULES_WUNLOCK();
5445 		break;
5446 	}
5447 
5448 	case DIOCOSFPFLUSH:
5449 		PF_RULES_WLOCK();
5450 		pf_osfp_flush();
5451 		PF_RULES_WUNLOCK();
5452 		break;
5453 
5454 	case DIOCIGETIFACES: {
5455 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5456 		struct pfi_kif *ifstore;
5457 		size_t bufsiz;
5458 
5459 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5460 			error = ENODEV;
5461 			break;
5462 		}
5463 
5464 		if (io->pfiio_size < 0 ||
5465 		    io->pfiio_size > pf_ioctl_maxcount ||
5466 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5467 			error = EINVAL;
5468 			break;
5469 		}
5470 
5471 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5472 
5473 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5474 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5475 		    M_TEMP, M_WAITOK | M_ZERO);
5476 
5477 		PF_RULES_RLOCK();
5478 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5479 		PF_RULES_RUNLOCK();
5480 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5481 		free(ifstore, M_TEMP);
5482 		break;
5483 	}
5484 
5485 	case DIOCSETIFFLAG: {
5486 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5487 
5488 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5489 
5490 		PF_RULES_WLOCK();
5491 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5492 		PF_RULES_WUNLOCK();
5493 		break;
5494 	}
5495 
5496 	case DIOCCLRIFFLAG: {
5497 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5498 
5499 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5500 
5501 		PF_RULES_WLOCK();
5502 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5503 		PF_RULES_WUNLOCK();
5504 		break;
5505 	}
5506 
5507 	case DIOCSETREASS: {
5508 		u_int32_t	*reass = (u_int32_t *)addr;
5509 
5510 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5511 		/* Removal of DF flag without reassembly enabled is not a
5512 		 * valid combination. Disable reassembly in such case. */
5513 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5514 			V_pf_status.reass = 0;
5515 		break;
5516 	}
5517 
5518 	default:
5519 		error = ENODEV;
5520 		break;
5521 	}
5522 fail:
5523 	CURVNET_RESTORE();
5524 
5525 #undef ERROUT_IOCTL
5526 
5527 	return (error);
5528 }
5529 
5530 void
5531 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5532 {
5533 	bzero(sp, sizeof(union pfsync_state_union));
5534 
5535 	/* copy from state key */
5536 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5537 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5538 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5539 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5540 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5541 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5542 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5543 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5544 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5545 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5546 
5547 	/* copy from state */
5548 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5549 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5550 	sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5551 	sp->pfs_1301.expire = pf_state_expires(st);
5552 	if (sp->pfs_1301.expire <= time_uptime)
5553 		sp->pfs_1301.expire = htonl(0);
5554 	else
5555 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5556 
5557 	sp->pfs_1301.direction = st->direction;
5558 	sp->pfs_1301.log = st->act.log;
5559 	sp->pfs_1301.timeout = st->timeout;
5560 
5561 	switch (msg_version) {
5562 		case PFSYNC_MSG_VERSION_1301:
5563 			sp->pfs_1301.state_flags = st->state_flags;
5564 			break;
5565 		case PFSYNC_MSG_VERSION_1400:
5566 			sp->pfs_1400.state_flags = htons(st->state_flags);
5567 			sp->pfs_1400.qid = htons(st->act.qid);
5568 			sp->pfs_1400.pqid = htons(st->act.pqid);
5569 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5570 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5571 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5572 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5573 			sp->pfs_1400.set_tos = st->act.set_tos;
5574 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5575 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5576 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5577 			sp->pfs_1400.rt = st->rt;
5578 			if (st->rt_kif)
5579 				strlcpy(sp->pfs_1400.rt_ifname,
5580 				    st->rt_kif->pfik_name,
5581 				    sizeof(sp->pfs_1400.rt_ifname));
5582 			break;
5583 		default:
5584 			panic("%s: Unsupported pfsync_msg_version %d",
5585 			    __func__, msg_version);
5586 	}
5587 
5588 	if (st->src_node)
5589 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5590 	if (st->nat_src_node)
5591 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5592 
5593 	sp->pfs_1301.id = st->id;
5594 	sp->pfs_1301.creatorid = st->creatorid;
5595 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5596 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5597 
5598 	if (st->rule.ptr == NULL)
5599 		sp->pfs_1301.rule = htonl(-1);
5600 	else
5601 		sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5602 	if (st->anchor.ptr == NULL)
5603 		sp->pfs_1301.anchor = htonl(-1);
5604 	else
5605 		sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5606 	if (st->nat_rule.ptr == NULL)
5607 		sp->pfs_1301.nat_rule = htonl(-1);
5608 	else
5609 		sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5610 
5611 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5612 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5613 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5614 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5615 }
5616 
5617 void
5618 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5619 {
5620 	bzero(sp, sizeof(*sp));
5621 
5622 	sp->version = PF_STATE_VERSION;
5623 
5624 	/* copy from state key */
5625 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5626 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5627 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5628 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5629 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5630 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5631 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5632 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5633 	sp->proto = st->key[PF_SK_WIRE]->proto;
5634 	sp->af = st->key[PF_SK_WIRE]->af;
5635 
5636 	/* copy from state */
5637 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5638 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5639 	    sizeof(sp->orig_ifname));
5640 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5641 	sp->creation = htonl(time_uptime - (st->creation / 1000));
5642 	sp->expire = pf_state_expires(st);
5643 	if (sp->expire <= time_uptime)
5644 		sp->expire = htonl(0);
5645 	else
5646 		sp->expire = htonl(sp->expire - time_uptime);
5647 
5648 	sp->direction = st->direction;
5649 	sp->log = st->act.log;
5650 	sp->timeout = st->timeout;
5651 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5652 	sp->state_flags_compat = st->state_flags;
5653 	sp->state_flags = htons(st->state_flags);
5654 	if (st->src_node)
5655 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5656 	if (st->nat_src_node)
5657 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5658 
5659 	sp->id = st->id;
5660 	sp->creatorid = st->creatorid;
5661 	pf_state_peer_hton(&st->src, &sp->src);
5662 	pf_state_peer_hton(&st->dst, &sp->dst);
5663 
5664 	if (st->rule.ptr == NULL)
5665 		sp->rule = htonl(-1);
5666 	else
5667 		sp->rule = htonl(st->rule.ptr->nr);
5668 	if (st->anchor.ptr == NULL)
5669 		sp->anchor = htonl(-1);
5670 	else
5671 		sp->anchor = htonl(st->anchor.ptr->nr);
5672 	if (st->nat_rule.ptr == NULL)
5673 		sp->nat_rule = htonl(-1);
5674 	else
5675 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5676 
5677 	sp->packets[0] = st->packets[0];
5678 	sp->packets[1] = st->packets[1];
5679 	sp->bytes[0] = st->bytes[0];
5680 	sp->bytes[1] = st->bytes[1];
5681 
5682 	sp->qid = htons(st->act.qid);
5683 	sp->pqid = htons(st->act.pqid);
5684 	sp->dnpipe = htons(st->act.dnpipe);
5685 	sp->dnrpipe = htons(st->act.dnrpipe);
5686 	sp->rtableid = htonl(st->act.rtableid);
5687 	sp->min_ttl = st->act.min_ttl;
5688 	sp->set_tos = st->act.set_tos;
5689 	sp->max_mss = htons(st->act.max_mss);
5690 	sp->rt = st->rt;
5691 	if (st->rt_kif)
5692 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5693 		    sizeof(sp->rt_ifname));
5694 	sp->set_prio[0] = st->act.set_prio[0];
5695 	sp->set_prio[1] = st->act.set_prio[1];
5696 
5697 }
5698 
5699 static void
5700 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5701 {
5702 	struct pfr_ktable *kt;
5703 
5704 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5705 
5706 	kt = aw->p.tbl;
5707 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5708 		kt = kt->pfrkt_root;
5709 	aw->p.tbl = NULL;
5710 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5711 		kt->pfrkt_cnt : -1;
5712 }
5713 
5714 static int
5715 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5716     size_t number, char **names)
5717 {
5718 	nvlist_t        *nvc;
5719 
5720 	nvc = nvlist_create(0);
5721 	if (nvc == NULL)
5722 		return (ENOMEM);
5723 
5724 	for (int i = 0; i < number; i++) {
5725 		nvlist_append_number_array(nvc, "counters",
5726 		    counter_u64_fetch(counters[i]));
5727 		nvlist_append_string_array(nvc, "names",
5728 		    names[i]);
5729 		nvlist_append_number_array(nvc, "ids",
5730 		    i);
5731 	}
5732 	nvlist_add_nvlist(nvl, name, nvc);
5733 	nvlist_destroy(nvc);
5734 
5735 	return (0);
5736 }
5737 
5738 static int
5739 pf_getstatus(struct pfioc_nv *nv)
5740 {
5741 	nvlist_t        *nvl = NULL, *nvc = NULL;
5742 	void            *nvlpacked = NULL;
5743 	int              error;
5744 	struct pf_status s;
5745 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5746 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5747 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5748 	PF_RULES_RLOCK_TRACKER;
5749 
5750 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5751 
5752 	PF_RULES_RLOCK();
5753 
5754 	nvl = nvlist_create(0);
5755 	if (nvl == NULL)
5756 		ERROUT(ENOMEM);
5757 
5758 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5759 	nvlist_add_number(nvl, "since", V_pf_status.since);
5760 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5761 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5762 	nvlist_add_number(nvl, "states", V_pf_status.states);
5763 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5764 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5765 	nvlist_add_bool(nvl, "syncookies_active",
5766 	    V_pf_status.syncookies_active);
5767 	nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5768 
5769 	/* counters */
5770 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5771 	    PFRES_MAX, pf_reasons);
5772 	if (error != 0)
5773 		ERROUT(error);
5774 
5775 	/* lcounters */
5776 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5777 	    KLCNT_MAX, pf_lcounter);
5778 	if (error != 0)
5779 		ERROUT(error);
5780 
5781 	/* fcounters */
5782 	nvc = nvlist_create(0);
5783 	if (nvc == NULL)
5784 		ERROUT(ENOMEM);
5785 
5786 	for (int i = 0; i < FCNT_MAX; i++) {
5787 		nvlist_append_number_array(nvc, "counters",
5788 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5789 		nvlist_append_string_array(nvc, "names",
5790 		    pf_fcounter[i]);
5791 		nvlist_append_number_array(nvc, "ids",
5792 		    i);
5793 	}
5794 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5795 	nvlist_destroy(nvc);
5796 	nvc = NULL;
5797 
5798 	/* scounters */
5799 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5800 	    SCNT_MAX, pf_fcounter);
5801 	if (error != 0)
5802 		ERROUT(error);
5803 
5804 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5805 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5806 	    PF_MD5_DIGEST_LENGTH);
5807 
5808 	pfi_update_status(V_pf_status.ifname, &s);
5809 
5810 	/* pcounters / bcounters */
5811 	for (int i = 0; i < 2; i++) {
5812 		for (int j = 0; j < 2; j++) {
5813 			for (int k = 0; k < 2; k++) {
5814 				nvlist_append_number_array(nvl, "pcounters",
5815 				    s.pcounters[i][j][k]);
5816 			}
5817 			nvlist_append_number_array(nvl, "bcounters",
5818 			    s.bcounters[i][j]);
5819 		}
5820 	}
5821 
5822 	nvlpacked = nvlist_pack(nvl, &nv->len);
5823 	if (nvlpacked == NULL)
5824 		ERROUT(ENOMEM);
5825 
5826 	if (nv->size == 0)
5827 		ERROUT(0);
5828 	else if (nv->size < nv->len)
5829 		ERROUT(ENOSPC);
5830 
5831 	PF_RULES_RUNLOCK();
5832 	error = copyout(nvlpacked, nv->data, nv->len);
5833 	goto done;
5834 
5835 #undef ERROUT
5836 errout:
5837 	PF_RULES_RUNLOCK();
5838 done:
5839 	free(nvlpacked, M_NVLIST);
5840 	nvlist_destroy(nvc);
5841 	nvlist_destroy(nvl);
5842 
5843 	return (error);
5844 }
5845 
5846 /*
5847  * XXX - Check for version mismatch!!!
5848  */
5849 static void
5850 pf_clear_all_states(void)
5851 {
5852 	struct epoch_tracker	 et;
5853 	struct pf_kstate	*s;
5854 	u_int i;
5855 
5856 	NET_EPOCH_ENTER(et);
5857 	for (i = 0; i <= pf_hashmask; i++) {
5858 		struct pf_idhash *ih = &V_pf_idhash[i];
5859 relock:
5860 		PF_HASHROW_LOCK(ih);
5861 		LIST_FOREACH(s, &ih->states, entry) {
5862 			s->timeout = PFTM_PURGE;
5863 			/* Don't send out individual delete messages. */
5864 			s->state_flags |= PFSTATE_NOSYNC;
5865 			pf_unlink_state(s);
5866 			goto relock;
5867 		}
5868 		PF_HASHROW_UNLOCK(ih);
5869 	}
5870 	NET_EPOCH_EXIT(et);
5871 }
5872 
5873 static int
5874 pf_clear_tables(void)
5875 {
5876 	struct pfioc_table io;
5877 	int error;
5878 
5879 	bzero(&io, sizeof(io));
5880 	io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5881 
5882 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5883 	    io.pfrio_flags);
5884 
5885 	return (error);
5886 }
5887 
5888 static void
5889 pf_clear_srcnodes(struct pf_ksrc_node *n)
5890 {
5891 	struct pf_kstate *s;
5892 	int i;
5893 
5894 	for (i = 0; i <= pf_hashmask; i++) {
5895 		struct pf_idhash *ih = &V_pf_idhash[i];
5896 
5897 		PF_HASHROW_LOCK(ih);
5898 		LIST_FOREACH(s, &ih->states, entry) {
5899 			if (n == NULL || n == s->src_node)
5900 				s->src_node = NULL;
5901 			if (n == NULL || n == s->nat_src_node)
5902 				s->nat_src_node = NULL;
5903 		}
5904 		PF_HASHROW_UNLOCK(ih);
5905 	}
5906 
5907 	if (n == NULL) {
5908 		struct pf_srchash *sh;
5909 
5910 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5911 		    i++, sh++) {
5912 			PF_HASHROW_LOCK(sh);
5913 			LIST_FOREACH(n, &sh->nodes, entry) {
5914 				n->expire = 1;
5915 				n->states = 0;
5916 			}
5917 			PF_HASHROW_UNLOCK(sh);
5918 		}
5919 	} else {
5920 		/* XXX: hash slot should already be locked here. */
5921 		n->expire = 1;
5922 		n->states = 0;
5923 	}
5924 }
5925 
5926 static void
5927 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5928 {
5929 	struct pf_ksrc_node_list	 kill;
5930 
5931 	LIST_INIT(&kill);
5932 	for (int i = 0; i <= pf_srchashmask; i++) {
5933 		struct pf_srchash *sh = &V_pf_srchash[i];
5934 		struct pf_ksrc_node *sn, *tmp;
5935 
5936 		PF_HASHROW_LOCK(sh);
5937 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5938 			if (PF_MATCHA(psnk->psnk_src.neg,
5939 			      &psnk->psnk_src.addr.v.a.addr,
5940 			      &psnk->psnk_src.addr.v.a.mask,
5941 			      &sn->addr, sn->af) &&
5942 			    PF_MATCHA(psnk->psnk_dst.neg,
5943 			      &psnk->psnk_dst.addr.v.a.addr,
5944 			      &psnk->psnk_dst.addr.v.a.mask,
5945 			      &sn->raddr, sn->af)) {
5946 				pf_unlink_src_node(sn);
5947 				LIST_INSERT_HEAD(&kill, sn, entry);
5948 				sn->expire = 1;
5949 			}
5950 		PF_HASHROW_UNLOCK(sh);
5951 	}
5952 
5953 	for (int i = 0; i <= pf_hashmask; i++) {
5954 		struct pf_idhash *ih = &V_pf_idhash[i];
5955 		struct pf_kstate *s;
5956 
5957 		PF_HASHROW_LOCK(ih);
5958 		LIST_FOREACH(s, &ih->states, entry) {
5959 			if (s->src_node && s->src_node->expire == 1)
5960 				s->src_node = NULL;
5961 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5962 				s->nat_src_node = NULL;
5963 		}
5964 		PF_HASHROW_UNLOCK(ih);
5965 	}
5966 
5967 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5968 }
5969 
5970 static int
5971 pf_keepcounters(struct pfioc_nv *nv)
5972 {
5973 	nvlist_t	*nvl = NULL;
5974 	void		*nvlpacked = NULL;
5975 	int		 error = 0;
5976 
5977 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5978 
5979 	if (nv->len > pf_ioctl_maxcount)
5980 		ERROUT(ENOMEM);
5981 
5982 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5983 	if (nvlpacked == NULL)
5984 		ERROUT(ENOMEM);
5985 
5986 	error = copyin(nv->data, nvlpacked, nv->len);
5987 	if (error)
5988 		ERROUT(error);
5989 
5990 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5991 	if (nvl == NULL)
5992 		ERROUT(EBADMSG);
5993 
5994 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5995 		ERROUT(EBADMSG);
5996 
5997 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5998 
5999 on_error:
6000 	nvlist_destroy(nvl);
6001 	free(nvlpacked, M_NVLIST);
6002 	return (error);
6003 }
6004 
6005 unsigned int
6006 pf_clear_states(const struct pf_kstate_kill *kill)
6007 {
6008 	struct pf_state_key_cmp	 match_key;
6009 	struct pf_kstate	*s;
6010 	struct pfi_kkif	*kif;
6011 	int		 idx;
6012 	unsigned int	 killed = 0, dir;
6013 
6014 	NET_EPOCH_ASSERT();
6015 
6016 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
6017 		struct pf_idhash *ih = &V_pf_idhash[i];
6018 
6019 relock_DIOCCLRSTATES:
6020 		PF_HASHROW_LOCK(ih);
6021 		LIST_FOREACH(s, &ih->states, entry) {
6022 			/* For floating states look at the original kif. */
6023 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6024 
6025 			if (kill->psk_ifname[0] &&
6026 			    strcmp(kill->psk_ifname,
6027 			    kif->pfik_name))
6028 				continue;
6029 
6030 			if (kill->psk_kill_match) {
6031 				bzero(&match_key, sizeof(match_key));
6032 
6033 				if (s->direction == PF_OUT) {
6034 					dir = PF_IN;
6035 					idx = PF_SK_STACK;
6036 				} else {
6037 					dir = PF_OUT;
6038 					idx = PF_SK_WIRE;
6039 				}
6040 
6041 				match_key.af = s->key[idx]->af;
6042 				match_key.proto = s->key[idx]->proto;
6043 				PF_ACPY(&match_key.addr[0],
6044 				    &s->key[idx]->addr[1], match_key.af);
6045 				match_key.port[0] = s->key[idx]->port[1];
6046 				PF_ACPY(&match_key.addr[1],
6047 				    &s->key[idx]->addr[0], match_key.af);
6048 				match_key.port[1] = s->key[idx]->port[0];
6049 			}
6050 
6051 			/*
6052 			 * Don't send out individual
6053 			 * delete messages.
6054 			 */
6055 			s->state_flags |= PFSTATE_NOSYNC;
6056 			pf_unlink_state(s);
6057 			killed++;
6058 
6059 			if (kill->psk_kill_match)
6060 				killed += pf_kill_matching_state(&match_key,
6061 				    dir);
6062 
6063 			goto relock_DIOCCLRSTATES;
6064 		}
6065 		PF_HASHROW_UNLOCK(ih);
6066 	}
6067 
6068 	if (V_pfsync_clear_states_ptr != NULL)
6069 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6070 
6071 	return (killed);
6072 }
6073 
6074 void
6075 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6076 {
6077 	struct pf_kstate	*s;
6078 
6079 	NET_EPOCH_ASSERT();
6080 	if (kill->psk_pfcmp.id) {
6081 		if (kill->psk_pfcmp.creatorid == 0)
6082 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6083 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6084 		    kill->psk_pfcmp.creatorid))) {
6085 			pf_unlink_state(s);
6086 			*killed = 1;
6087 		}
6088 		return;
6089 	}
6090 
6091 	for (unsigned int i = 0; i <= pf_hashmask; i++)
6092 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6093 }
6094 
6095 static int
6096 pf_killstates_nv(struct pfioc_nv *nv)
6097 {
6098 	struct pf_kstate_kill	 kill;
6099 	struct epoch_tracker	 et;
6100 	nvlist_t		*nvl = NULL;
6101 	void			*nvlpacked = NULL;
6102 	int			 error = 0;
6103 	unsigned int		 killed = 0;
6104 
6105 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6106 
6107 	if (nv->len > pf_ioctl_maxcount)
6108 		ERROUT(ENOMEM);
6109 
6110 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6111 	if (nvlpacked == NULL)
6112 		ERROUT(ENOMEM);
6113 
6114 	error = copyin(nv->data, nvlpacked, nv->len);
6115 	if (error)
6116 		ERROUT(error);
6117 
6118 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6119 	if (nvl == NULL)
6120 		ERROUT(EBADMSG);
6121 
6122 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6123 	if (error)
6124 		ERROUT(error);
6125 
6126 	NET_EPOCH_ENTER(et);
6127 	pf_killstates(&kill, &killed);
6128 	NET_EPOCH_EXIT(et);
6129 
6130 	free(nvlpacked, M_NVLIST);
6131 	nvlpacked = NULL;
6132 	nvlist_destroy(nvl);
6133 	nvl = nvlist_create(0);
6134 	if (nvl == NULL)
6135 		ERROUT(ENOMEM);
6136 
6137 	nvlist_add_number(nvl, "killed", killed);
6138 
6139 	nvlpacked = nvlist_pack(nvl, &nv->len);
6140 	if (nvlpacked == NULL)
6141 		ERROUT(ENOMEM);
6142 
6143 	if (nv->size == 0)
6144 		ERROUT(0);
6145 	else if (nv->size < nv->len)
6146 		ERROUT(ENOSPC);
6147 
6148 	error = copyout(nvlpacked, nv->data, nv->len);
6149 
6150 on_error:
6151 	nvlist_destroy(nvl);
6152 	free(nvlpacked, M_NVLIST);
6153 	return (error);
6154 }
6155 
6156 static int
6157 pf_clearstates_nv(struct pfioc_nv *nv)
6158 {
6159 	struct pf_kstate_kill	 kill;
6160 	struct epoch_tracker	 et;
6161 	nvlist_t		*nvl = NULL;
6162 	void			*nvlpacked = NULL;
6163 	int			 error = 0;
6164 	unsigned int		 killed;
6165 
6166 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6167 
6168 	if (nv->len > pf_ioctl_maxcount)
6169 		ERROUT(ENOMEM);
6170 
6171 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6172 	if (nvlpacked == NULL)
6173 		ERROUT(ENOMEM);
6174 
6175 	error = copyin(nv->data, nvlpacked, nv->len);
6176 	if (error)
6177 		ERROUT(error);
6178 
6179 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6180 	if (nvl == NULL)
6181 		ERROUT(EBADMSG);
6182 
6183 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6184 	if (error)
6185 		ERROUT(error);
6186 
6187 	NET_EPOCH_ENTER(et);
6188 	killed = pf_clear_states(&kill);
6189 	NET_EPOCH_EXIT(et);
6190 
6191 	free(nvlpacked, M_NVLIST);
6192 	nvlpacked = NULL;
6193 	nvlist_destroy(nvl);
6194 	nvl = nvlist_create(0);
6195 	if (nvl == NULL)
6196 		ERROUT(ENOMEM);
6197 
6198 	nvlist_add_number(nvl, "killed", killed);
6199 
6200 	nvlpacked = nvlist_pack(nvl, &nv->len);
6201 	if (nvlpacked == NULL)
6202 		ERROUT(ENOMEM);
6203 
6204 	if (nv->size == 0)
6205 		ERROUT(0);
6206 	else if (nv->size < nv->len)
6207 		ERROUT(ENOSPC);
6208 
6209 	error = copyout(nvlpacked, nv->data, nv->len);
6210 
6211 #undef ERROUT
6212 on_error:
6213 	nvlist_destroy(nvl);
6214 	free(nvlpacked, M_NVLIST);
6215 	return (error);
6216 }
6217 
6218 static int
6219 pf_getstate(struct pfioc_nv *nv)
6220 {
6221 	nvlist_t		*nvl = NULL, *nvls;
6222 	void			*nvlpacked = NULL;
6223 	struct pf_kstate	*s = NULL;
6224 	int			 error = 0;
6225 	uint64_t		 id, creatorid;
6226 
6227 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6228 
6229 	if (nv->len > pf_ioctl_maxcount)
6230 		ERROUT(ENOMEM);
6231 
6232 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6233 	if (nvlpacked == NULL)
6234 		ERROUT(ENOMEM);
6235 
6236 	error = copyin(nv->data, nvlpacked, nv->len);
6237 	if (error)
6238 		ERROUT(error);
6239 
6240 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6241 	if (nvl == NULL)
6242 		ERROUT(EBADMSG);
6243 
6244 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6245 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6246 
6247 	s = pf_find_state_byid(id, creatorid);
6248 	if (s == NULL)
6249 		ERROUT(ENOENT);
6250 
6251 	free(nvlpacked, M_NVLIST);
6252 	nvlpacked = NULL;
6253 	nvlist_destroy(nvl);
6254 	nvl = nvlist_create(0);
6255 	if (nvl == NULL)
6256 		ERROUT(ENOMEM);
6257 
6258 	nvls = pf_state_to_nvstate(s);
6259 	if (nvls == NULL)
6260 		ERROUT(ENOMEM);
6261 
6262 	nvlist_add_nvlist(nvl, "state", nvls);
6263 	nvlist_destroy(nvls);
6264 
6265 	nvlpacked = nvlist_pack(nvl, &nv->len);
6266 	if (nvlpacked == NULL)
6267 		ERROUT(ENOMEM);
6268 
6269 	if (nv->size == 0)
6270 		ERROUT(0);
6271 	else if (nv->size < nv->len)
6272 		ERROUT(ENOSPC);
6273 
6274 	error = copyout(nvlpacked, nv->data, nv->len);
6275 
6276 #undef ERROUT
6277 errout:
6278 	if (s != NULL)
6279 		PF_STATE_UNLOCK(s);
6280 	free(nvlpacked, M_NVLIST);
6281 	nvlist_destroy(nvl);
6282 	return (error);
6283 }
6284 
6285 /*
6286  * XXX - Check for version mismatch!!!
6287  */
6288 
6289 /*
6290  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6291  */
6292 static int
6293 shutdown_pf(void)
6294 {
6295 	int error = 0;
6296 	u_int32_t t[5];
6297 	char nn = '\0';
6298 	struct pf_kanchor *anchor;
6299 	struct pf_keth_anchor *eth_anchor;
6300 	int rs_num;
6301 
6302 	do {
6303 		/* Unlink rules of all user defined anchors */
6304 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6305 			/* Wildcard based anchors may not have a respective
6306 			 * explicit anchor rule or they may be left empty
6307 			 * without rules. It leads to anchor.refcnt=0, and the
6308 			 * rest of the logic does not expect it. */
6309 			if (anchor->refcnt == 0)
6310 				anchor->refcnt = 1;
6311 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6312 				if ((error = pf_begin_rules(&t[rs_num], rs_num,
6313 				    anchor->path)) != 0) {
6314 					DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6315 					    "anchor.path=%s rs_num=%d\n",
6316 					    anchor->path, rs_num));
6317 					goto error;	/* XXX: rollback? */
6318 				}
6319 			}
6320 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6321 				error = pf_commit_rules(t[rs_num], rs_num,
6322 				    anchor->path);
6323 				MPASS(error == 0);
6324 			}
6325 		}
6326 
6327 		/* Unlink rules of all user defined ether anchors */
6328 		RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6329 		    &V_pf_keth_anchors) {
6330 			/* Wildcard based anchors may not have a respective
6331 			 * explicit anchor rule or they may be left empty
6332 			 * without rules. It leads to anchor.refcnt=0, and the
6333 			 * rest of the logic does not expect it. */
6334 			if (eth_anchor->refcnt == 0)
6335 				eth_anchor->refcnt = 1;
6336 			if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6337 			    != 0) {
6338 				DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6339 				    "anchor.path=%s\n", eth_anchor->path));
6340 				goto error;
6341 			}
6342 			error = pf_commit_eth(t[0], eth_anchor->path);
6343 			MPASS(error == 0);
6344 		}
6345 
6346 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6347 		    != 0) {
6348 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6349 			break;
6350 		}
6351 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6352 		    != 0) {
6353 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6354 			break;		/* XXX: rollback? */
6355 		}
6356 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6357 		    != 0) {
6358 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6359 			break;		/* XXX: rollback? */
6360 		}
6361 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6362 		    != 0) {
6363 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6364 			break;		/* XXX: rollback? */
6365 		}
6366 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6367 		    != 0) {
6368 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6369 			break;		/* XXX: rollback? */
6370 		}
6371 
6372 		error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6373 		MPASS(error == 0);
6374 		error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6375 		MPASS(error == 0);
6376 		error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6377 		MPASS(error == 0);
6378 		error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6379 		MPASS(error == 0);
6380 		error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6381 		MPASS(error == 0);
6382 
6383 		if ((error = pf_clear_tables()) != 0)
6384 			break;
6385 
6386 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6387 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6388 			break;
6389 		}
6390 		error = pf_commit_eth(t[0], &nn);
6391 		MPASS(error == 0);
6392 
6393 #ifdef ALTQ
6394 		if ((error = pf_begin_altq(&t[0])) != 0) {
6395 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6396 			break;
6397 		}
6398 		pf_commit_altq(t[0]);
6399 #endif
6400 
6401 		pf_clear_all_states();
6402 
6403 		pf_clear_srcnodes(NULL);
6404 
6405 		/* status does not use malloced mem so no need to cleanup */
6406 		/* fingerprints and interfaces have their own cleanup code */
6407 	} while(0);
6408 
6409 error:
6410 	return (error);
6411 }
6412 
6413 static pfil_return_t
6414 pf_check_return(int chk, struct mbuf **m)
6415 {
6416 
6417 	switch (chk) {
6418 	case PF_PASS:
6419 		if (*m == NULL)
6420 			return (PFIL_CONSUMED);
6421 		else
6422 			return (PFIL_PASS);
6423 		break;
6424 	default:
6425 		if (*m != NULL) {
6426 			m_freem(*m);
6427 			*m = NULL;
6428 		}
6429 		return (PFIL_DROPPED);
6430 	}
6431 }
6432 
6433 static pfil_return_t
6434 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6435     void *ruleset __unused, struct inpcb *inp)
6436 {
6437 	int chk;
6438 
6439 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6440 
6441 	return (pf_check_return(chk, m));
6442 }
6443 
6444 static pfil_return_t
6445 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6446     void *ruleset __unused, struct inpcb *inp)
6447 {
6448 	int chk;
6449 
6450 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6451 
6452 	return (pf_check_return(chk, m));
6453 }
6454 
6455 #ifdef INET
6456 static pfil_return_t
6457 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6458     void *ruleset __unused, struct inpcb *inp)
6459 {
6460 	int chk;
6461 
6462 	chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6463 
6464 	return (pf_check_return(chk, m));
6465 }
6466 
6467 static pfil_return_t
6468 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6469     void *ruleset __unused,  struct inpcb *inp)
6470 {
6471 	int chk;
6472 
6473 	chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6474 
6475 	return (pf_check_return(chk, m));
6476 }
6477 #endif
6478 
6479 #ifdef INET6
6480 static pfil_return_t
6481 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6482     void *ruleset __unused,  struct inpcb *inp)
6483 {
6484 	int chk;
6485 
6486 	/*
6487 	 * In case of loopback traffic IPv6 uses the real interface in
6488 	 * order to support scoped addresses. In order to support stateful
6489 	 * filtering we have change this to lo0 as it is the case in IPv4.
6490 	 */
6491 	CURVNET_SET(ifp->if_vnet);
6492 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6493 	    m, inp, NULL);
6494 	CURVNET_RESTORE();
6495 
6496 	return (pf_check_return(chk, m));
6497 }
6498 
6499 static pfil_return_t
6500 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6501     void *ruleset __unused,  struct inpcb *inp)
6502 {
6503 	int chk;
6504 
6505 	CURVNET_SET(ifp->if_vnet);
6506 	chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6507 	CURVNET_RESTORE();
6508 
6509 	return (pf_check_return(chk, m));
6510 }
6511 #endif /* INET6 */
6512 
6513 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6514 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6515 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6516 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6517 
6518 #ifdef INET
6519 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6520 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6521 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6522 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6523 #endif
6524 #ifdef INET6
6525 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6526 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6527 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6528 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6529 #endif
6530 
6531 static void
6532 hook_pf_eth(void)
6533 {
6534 	struct pfil_hook_args pha = {
6535 		.pa_version = PFIL_VERSION,
6536 		.pa_modname = "pf",
6537 		.pa_type = PFIL_TYPE_ETHERNET,
6538 	};
6539 	struct pfil_link_args pla = {
6540 		.pa_version = PFIL_VERSION,
6541 	};
6542 	int ret __diagused;
6543 
6544 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6545 		return;
6546 
6547 	pha.pa_mbuf_chk = pf_eth_check_in;
6548 	pha.pa_flags = PFIL_IN;
6549 	pha.pa_rulname = "eth-in";
6550 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6551 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6552 	pla.pa_head = V_link_pfil_head;
6553 	pla.pa_hook = V_pf_eth_in_hook;
6554 	ret = pfil_link(&pla);
6555 	MPASS(ret == 0);
6556 	pha.pa_mbuf_chk = pf_eth_check_out;
6557 	pha.pa_flags = PFIL_OUT;
6558 	pha.pa_rulname = "eth-out";
6559 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6560 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6561 	pla.pa_head = V_link_pfil_head;
6562 	pla.pa_hook = V_pf_eth_out_hook;
6563 	ret = pfil_link(&pla);
6564 	MPASS(ret == 0);
6565 
6566 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6567 }
6568 
6569 static void
6570 hook_pf(void)
6571 {
6572 	struct pfil_hook_args pha = {
6573 		.pa_version = PFIL_VERSION,
6574 		.pa_modname = "pf",
6575 	};
6576 	struct pfil_link_args pla = {
6577 		.pa_version = PFIL_VERSION,
6578 	};
6579 	int ret __diagused;
6580 
6581 	if (atomic_load_bool(&V_pf_pfil_hooked))
6582 		return;
6583 
6584 #ifdef INET
6585 	pha.pa_type = PFIL_TYPE_IP4;
6586 	pha.pa_mbuf_chk = pf_check_in;
6587 	pha.pa_flags = PFIL_IN;
6588 	pha.pa_rulname = "default-in";
6589 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6590 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6591 	pla.pa_head = V_inet_pfil_head;
6592 	pla.pa_hook = V_pf_ip4_in_hook;
6593 	ret = pfil_link(&pla);
6594 	MPASS(ret == 0);
6595 	pha.pa_mbuf_chk = pf_check_out;
6596 	pha.pa_flags = PFIL_OUT;
6597 	pha.pa_rulname = "default-out";
6598 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6599 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6600 	pla.pa_head = V_inet_pfil_head;
6601 	pla.pa_hook = V_pf_ip4_out_hook;
6602 	ret = pfil_link(&pla);
6603 	MPASS(ret == 0);
6604 	if (V_pf_filter_local) {
6605 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6606 		pla.pa_head = V_inet_local_pfil_head;
6607 		pla.pa_hook = V_pf_ip4_out_hook;
6608 		ret = pfil_link(&pla);
6609 		MPASS(ret == 0);
6610 	}
6611 #endif
6612 #ifdef INET6
6613 	pha.pa_type = PFIL_TYPE_IP6;
6614 	pha.pa_mbuf_chk = pf_check6_in;
6615 	pha.pa_flags = PFIL_IN;
6616 	pha.pa_rulname = "default-in6";
6617 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6618 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6619 	pla.pa_head = V_inet6_pfil_head;
6620 	pla.pa_hook = V_pf_ip6_in_hook;
6621 	ret = pfil_link(&pla);
6622 	MPASS(ret == 0);
6623 	pha.pa_mbuf_chk = pf_check6_out;
6624 	pha.pa_rulname = "default-out6";
6625 	pha.pa_flags = PFIL_OUT;
6626 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6627 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6628 	pla.pa_head = V_inet6_pfil_head;
6629 	pla.pa_hook = V_pf_ip6_out_hook;
6630 	ret = pfil_link(&pla);
6631 	MPASS(ret == 0);
6632 	if (V_pf_filter_local) {
6633 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6634 		pla.pa_head = V_inet6_local_pfil_head;
6635 		pla.pa_hook = V_pf_ip6_out_hook;
6636 		ret = pfil_link(&pla);
6637 		MPASS(ret == 0);
6638 	}
6639 #endif
6640 
6641 	atomic_store_bool(&V_pf_pfil_hooked, true);
6642 }
6643 
6644 static void
6645 dehook_pf_eth(void)
6646 {
6647 
6648 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6649 		return;
6650 
6651 	pfil_remove_hook(V_pf_eth_in_hook);
6652 	pfil_remove_hook(V_pf_eth_out_hook);
6653 
6654 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6655 }
6656 
6657 static void
6658 dehook_pf(void)
6659 {
6660 
6661 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6662 		return;
6663 
6664 #ifdef INET
6665 	pfil_remove_hook(V_pf_ip4_in_hook);
6666 	pfil_remove_hook(V_pf_ip4_out_hook);
6667 #endif
6668 #ifdef INET6
6669 	pfil_remove_hook(V_pf_ip6_in_hook);
6670 	pfil_remove_hook(V_pf_ip6_out_hook);
6671 #endif
6672 
6673 	atomic_store_bool(&V_pf_pfil_hooked, false);
6674 }
6675 
6676 static void
6677 pf_load_vnet(void)
6678 {
6679 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6680 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6681 
6682 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6683 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6684 
6685 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6686 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6687 #ifdef ALTQ
6688 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6689 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6690 #endif
6691 
6692 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6693 
6694 	pfattach_vnet();
6695 	V_pf_vnet_active = 1;
6696 }
6697 
6698 static int
6699 pf_load(void)
6700 {
6701 	int error;
6702 
6703 	sx_init(&pf_end_lock, "pf end thread");
6704 
6705 	pf_mtag_initialize();
6706 
6707 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6708 	if (pf_dev == NULL)
6709 		return (ENOMEM);
6710 
6711 	pf_end_threads = 0;
6712 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6713 	if (error != 0)
6714 		return (error);
6715 
6716 	pfi_initialize();
6717 
6718 	return (0);
6719 }
6720 
6721 static void
6722 pf_unload_vnet(void)
6723 {
6724 	int ret __diagused;
6725 
6726 	V_pf_vnet_active = 0;
6727 	V_pf_status.running = 0;
6728 	dehook_pf();
6729 	dehook_pf_eth();
6730 
6731 	PF_RULES_WLOCK();
6732 	pf_syncookies_cleanup();
6733 	shutdown_pf();
6734 	PF_RULES_WUNLOCK();
6735 
6736 	/* Make sure we've cleaned up ethernet rules before we continue. */
6737 	NET_EPOCH_DRAIN_CALLBACKS();
6738 
6739 	ret = swi_remove(V_pf_swi_cookie);
6740 	MPASS(ret == 0);
6741 	ret = intr_event_destroy(V_pf_swi_ie);
6742 	MPASS(ret == 0);
6743 
6744 	pf_unload_vnet_purge();
6745 
6746 	pf_normalize_cleanup();
6747 	PF_RULES_WLOCK();
6748 	pfi_cleanup_vnet();
6749 	PF_RULES_WUNLOCK();
6750 	pfr_cleanup();
6751 	pf_osfp_flush();
6752 	pf_cleanup();
6753 	if (IS_DEFAULT_VNET(curvnet))
6754 		pf_mtag_cleanup();
6755 
6756 	pf_cleanup_tagset(&V_pf_tags);
6757 #ifdef ALTQ
6758 	pf_cleanup_tagset(&V_pf_qids);
6759 #endif
6760 	uma_zdestroy(V_pf_tag_z);
6761 
6762 #ifdef PF_WANT_32_TO_64_COUNTER
6763 	PF_RULES_WLOCK();
6764 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6765 
6766 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6767 	MPASS(V_pf_allkifcount == 0);
6768 
6769 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6770 	V_pf_allrulecount--;
6771 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6772 
6773 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6774 	MPASS(V_pf_allrulecount == 0);
6775 
6776 	PF_RULES_WUNLOCK();
6777 
6778 	free(V_pf_kifmarker, PFI_MTYPE);
6779 	free(V_pf_rulemarker, M_PFRULE);
6780 #endif
6781 
6782 	/* Free counters last as we updated them during shutdown. */
6783 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6784 	for (int i = 0; i < 2; i++) {
6785 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6786 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6787 	}
6788 	counter_u64_free(V_pf_default_rule.states_cur);
6789 	counter_u64_free(V_pf_default_rule.states_tot);
6790 	counter_u64_free(V_pf_default_rule.src_nodes);
6791 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6792 
6793 	for (int i = 0; i < PFRES_MAX; i++)
6794 		counter_u64_free(V_pf_status.counters[i]);
6795 	for (int i = 0; i < KLCNT_MAX; i++)
6796 		counter_u64_free(V_pf_status.lcounters[i]);
6797 	for (int i = 0; i < FCNT_MAX; i++)
6798 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6799 	for (int i = 0; i < SCNT_MAX; i++)
6800 		counter_u64_free(V_pf_status.scounters[i]);
6801 
6802 	rm_destroy(&V_pf_rules_lock);
6803 	sx_destroy(&V_pf_ioctl_lock);
6804 }
6805 
6806 static void
6807 pf_unload(void)
6808 {
6809 
6810 	sx_xlock(&pf_end_lock);
6811 	pf_end_threads = 1;
6812 	while (pf_end_threads < 2) {
6813 		wakeup_one(pf_purge_thread);
6814 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6815 	}
6816 	sx_xunlock(&pf_end_lock);
6817 
6818 	pf_nl_unregister();
6819 
6820 	if (pf_dev != NULL)
6821 		destroy_dev(pf_dev);
6822 
6823 	pfi_cleanup();
6824 
6825 	sx_destroy(&pf_end_lock);
6826 }
6827 
6828 static void
6829 vnet_pf_init(void *unused __unused)
6830 {
6831 
6832 	pf_load_vnet();
6833 }
6834 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6835     vnet_pf_init, NULL);
6836 
6837 static void
6838 vnet_pf_uninit(const void *unused __unused)
6839 {
6840 
6841 	pf_unload_vnet();
6842 }
6843 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6844 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6845     vnet_pf_uninit, NULL);
6846 
6847 static int
6848 pf_modevent(module_t mod, int type, void *data)
6849 {
6850 	int error = 0;
6851 
6852 	switch(type) {
6853 	case MOD_LOAD:
6854 		error = pf_load();
6855 		pf_nl_register();
6856 		break;
6857 	case MOD_UNLOAD:
6858 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6859 		 * the vnet_pf_uninit()s */
6860 		break;
6861 	default:
6862 		error = EINVAL;
6863 		break;
6864 	}
6865 
6866 	return (error);
6867 }
6868 
6869 static moduledata_t pf_mod = {
6870 	"pf",
6871 	pf_modevent,
6872 	0
6873 };
6874 
6875 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6876 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6877 MODULE_VERSION(pf, PF_MODVER);
6878