xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 452e6f549cdb33393e05fabbaee0aed7eb744e68)
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_src_node_copy(const struct pf_ksrc_node *,
133 			    struct pf_src_node *);
134 #ifdef ALTQ
135 static int		 pf_export_kaltq(struct pf_altq *,
136 			    struct pfioc_altq_v1 *, size_t);
137 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
138 			    struct pf_altq *, size_t);
139 #endif /* ALTQ */
140 
141 VNET_DEFINE(struct pf_krule,	pf_default_rule);
142 
143 static __inline int             pf_krule_compare(struct pf_krule *,
144 				    struct pf_krule *);
145 
146 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
147 
148 #ifdef ALTQ
149 VNET_DEFINE_STATIC(int,		pf_altq_running);
150 #define	V_pf_altq_running	VNET(pf_altq_running)
151 #endif
152 
153 #define	TAGID_MAX	 50000
154 struct pf_tagname {
155 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
156 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
157 	char			name[PF_TAG_NAME_SIZE];
158 	uint16_t		tag;
159 	int			ref;
160 };
161 
162 struct pf_tagset {
163 	TAILQ_HEAD(, pf_tagname)	*namehash;
164 	TAILQ_HEAD(, pf_tagname)	*taghash;
165 	unsigned int			 mask;
166 	uint32_t			 seed;
167 	BITSET_DEFINE(, TAGID_MAX)	 avail;
168 };
169 
170 VNET_DEFINE(struct pf_tagset, pf_tags);
171 #define	V_pf_tags	VNET(pf_tags)
172 static unsigned int	pf_rule_tag_hashsize;
173 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
174 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
175     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
176     "Size of pf(4) rule tag hashtable");
177 
178 #ifdef ALTQ
179 VNET_DEFINE(struct pf_tagset, pf_qids);
180 #define	V_pf_qids	VNET(pf_qids)
181 static unsigned int	pf_queue_tag_hashsize;
182 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
183 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
184     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
185     "Size of pf(4) queue tag hashtable");
186 #endif
187 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
188 #define	V_pf_tag_z		 VNET(pf_tag_z)
189 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
190 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
191 
192 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
193 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
194 #endif
195 
196 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
197 #define V_pf_filter_local	VNET(pf_filter_local)
198 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
199     &VNET_NAME(pf_filter_local), false,
200     "Enable filtering for packets delivered to local network stack");
201 
202 #ifdef PF_DEFAULT_TO_DROP
203 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
204 #else
205 VNET_DEFINE_STATIC(bool, default_to_drop);
206 #endif
207 #define	V_default_to_drop VNET(default_to_drop)
208 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
209     &VNET_NAME(default_to_drop), false,
210     "Make the default rule drop all packets.");
211 
212 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
213 			    unsigned int);
214 static void		 pf_cleanup_tagset(struct pf_tagset *);
215 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
216 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
217 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
218 static u_int16_t	 pf_tagname2tag(const char *);
219 static void		 tag_unref(struct pf_tagset *, u_int16_t);
220 
221 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
222 
223 struct cdev *pf_dev;
224 
225 /*
226  * XXX - These are new and need to be checked when moveing to a new version
227  */
228 static void		 pf_clear_all_states(void);
229 static int		 pf_killstates_row(struct pf_kstate_kill *,
230 			    struct pf_idhash *);
231 static int		 pf_killstates_nv(struct pfioc_nv *);
232 static int		 pf_clearstates_nv(struct pfioc_nv *);
233 static int		 pf_getstate(struct pfioc_nv *);
234 static int		 pf_getstatus(struct pfioc_nv *);
235 static int		 pf_clear_tables(void);
236 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
237 static int		 pf_keepcounters(struct pfioc_nv *);
238 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
239 
240 /*
241  * Wrapper functions for pfil(9) hooks
242  */
243 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
244     int flags, void *ruleset __unused, struct inpcb *inp);
245 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
246     int flags, void *ruleset __unused, struct inpcb *inp);
247 #ifdef INET
248 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
249     int flags, void *ruleset __unused, struct inpcb *inp);
250 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
251     int flags, void *ruleset __unused, struct inpcb *inp);
252 #endif
253 #ifdef INET6
254 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
255     int flags, void *ruleset __unused, struct inpcb *inp);
256 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
257     int flags, void *ruleset __unused, struct inpcb *inp);
258 #endif
259 
260 static void		hook_pf_eth(void);
261 static void		hook_pf(void);
262 static void		dehook_pf_eth(void);
263 static void		dehook_pf(void);
264 static int		shutdown_pf(void);
265 static int		pf_load(void);
266 static void		pf_unload(void);
267 
268 static struct cdevsw pf_cdevsw = {
269 	.d_ioctl =	pfioctl,
270 	.d_name =	PF_NAME,
271 	.d_version =	D_VERSION,
272 };
273 
274 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
275 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
276 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
277 #define V_pf_pfil_eth_hooked	VNET(pf_pfil_eth_hooked)
278 
279 /*
280  * We need a flag that is neither hooked nor running to know when
281  * the VNET is "valid".  We primarily need this to control (global)
282  * external event, e.g., eventhandlers.
283  */
284 VNET_DEFINE(int, pf_vnet_active);
285 #define V_pf_vnet_active	VNET(pf_vnet_active)
286 
287 int pf_end_threads;
288 struct proc *pf_purge_proc;
289 
290 VNET_DEFINE(struct rmlock, pf_rules_lock);
291 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
292 #define	V_pf_ioctl_lock		VNET(pf_ioctl_lock)
293 struct sx			pf_end_lock;
294 
295 /* pfsync */
296 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
297 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
298 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
299 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
300 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
301 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
302 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
303 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
304 
305 /* pflog */
306 pflog_packet_t			*pflog_packet_ptr = NULL;
307 
308 /*
309  * Copy a user-provided string, returning an error if truncation would occur.
310  * Avoid scanning past "sz" bytes in the source string since there's no
311  * guarantee that it's nul-terminated.
312  */
313 static int
pf_user_strcpy(char * dst,const char * src,size_t sz)314 pf_user_strcpy(char *dst, const char *src, size_t sz)
315 {
316 	if (strnlen(src, sz) == sz)
317 		return (EINVAL);
318 	(void)strlcpy(dst, src, sz);
319 	return (0);
320 }
321 
322 static void
pfattach_vnet(void)323 pfattach_vnet(void)
324 {
325 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
326 
327 	bzero(&V_pf_status, sizeof(V_pf_status));
328 
329 	pf_initialize();
330 	pfr_initialize();
331 	pfi_initialize_vnet();
332 	pf_normalize_init();
333 	pf_syncookies_init();
334 
335 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
336 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
337 
338 	RB_INIT(&V_pf_anchors);
339 	pf_init_kruleset(&pf_main_ruleset);
340 
341 	pf_init_keth(V_pf_keth);
342 
343 	/* default rule should never be garbage collected */
344 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
345 	V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
346 	V_pf_default_rule.nr = -1;
347 	V_pf_default_rule.rtableid = -1;
348 
349 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
350 	for (int i = 0; i < 2; i++) {
351 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
352 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
353 	}
354 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
355 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
356 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
357 
358 	V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
359 	    M_WAITOK | M_ZERO);
360 
361 #ifdef PF_WANT_32_TO_64_COUNTER
362 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
363 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
364 	PF_RULES_WLOCK();
365 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
366 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
367 	V_pf_allrulecount++;
368 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
369 	PF_RULES_WUNLOCK();
370 #endif
371 
372 	/* initialize default timeouts */
373 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
374 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
375 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
376 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
377 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
378 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
379 	my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
380 	my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
381 	my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
382 	my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
383 	my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
384 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
385 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
386 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
387 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
388 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
389 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
390 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
391 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
392 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
393 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
394 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
395 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
396 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
397 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
398 
399 	V_pf_status.debug = PF_DEBUG_URGENT;
400 	/*
401 	 * XXX This is different than in OpenBSD where reassembly is enabled by
402 	 * defult. In FreeBSD we expect people to still use scrub rules and
403 	 * switch to the new syntax later. Only when they switch they must
404 	 * explicitly enable reassemle. We could change the default once the
405 	 * scrub rule functionality is hopefully removed some day in future.
406 	 */
407 	V_pf_status.reass = 0;
408 
409 	V_pf_pfil_hooked = false;
410 	V_pf_pfil_eth_hooked = false;
411 
412 	/* XXX do our best to avoid a conflict */
413 	V_pf_status.hostid = arc4random();
414 
415 	for (int i = 0; i < PFRES_MAX; i++)
416 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
417 	for (int i = 0; i < KLCNT_MAX; i++)
418 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
419 	for (int i = 0; i < FCNT_MAX; i++)
420 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
421 	for (int i = 0; i < SCNT_MAX; i++)
422 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
423 
424 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
425 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
426 		/* XXXGL: leaked all above. */
427 		return;
428 }
429 
430 static struct pf_kpool *
pf_get_kpool(const char * anchor,u_int32_t ticket,u_int8_t rule_action,u_int32_t rule_number,u_int8_t r_last,u_int8_t active,u_int8_t check_ticket)431 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
432     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
433     u_int8_t check_ticket)
434 {
435 	struct pf_kruleset	*ruleset;
436 	struct pf_krule		*rule;
437 	int			 rs_num;
438 
439 	ruleset = pf_find_kruleset(anchor);
440 	if (ruleset == NULL)
441 		return (NULL);
442 	rs_num = pf_get_ruleset_number(rule_action);
443 	if (rs_num >= PF_RULESET_MAX)
444 		return (NULL);
445 	if (active) {
446 		if (check_ticket && ticket !=
447 		    ruleset->rules[rs_num].active.ticket)
448 			return (NULL);
449 		if (r_last)
450 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
451 			    pf_krulequeue);
452 		else
453 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
454 	} else {
455 		if (check_ticket && ticket !=
456 		    ruleset->rules[rs_num].inactive.ticket)
457 			return (NULL);
458 		if (r_last)
459 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
460 			    pf_krulequeue);
461 		else
462 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
463 	}
464 	if (!r_last) {
465 		while ((rule != NULL) && (rule->nr != rule_number))
466 			rule = TAILQ_NEXT(rule, entries);
467 	}
468 	if (rule == NULL)
469 		return (NULL);
470 
471 	return (&rule->rpool);
472 }
473 
474 static void
pf_mv_kpool(struct pf_kpalist * poola,struct pf_kpalist * poolb)475 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
476 {
477 	struct pf_kpooladdr	*mv_pool_pa;
478 
479 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
480 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
481 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
482 	}
483 }
484 
485 static void
pf_empty_kpool(struct pf_kpalist * poola)486 pf_empty_kpool(struct pf_kpalist *poola)
487 {
488 	struct pf_kpooladdr *pa;
489 
490 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
491 		switch (pa->addr.type) {
492 		case PF_ADDR_DYNIFTL:
493 			pfi_dynaddr_remove(pa->addr.p.dyn);
494 			break;
495 		case PF_ADDR_TABLE:
496 			/* XXX: this could be unfinished pooladdr on pabuf */
497 			if (pa->addr.p.tbl != NULL)
498 				pfr_detach_table(pa->addr.p.tbl);
499 			break;
500 		}
501 		if (pa->kif)
502 			pfi_kkif_unref(pa->kif);
503 		TAILQ_REMOVE(poola, pa, entries);
504 		free(pa, M_PFRULE);
505 	}
506 }
507 
508 static void
pf_unlink_rule_locked(struct pf_krulequeue * rulequeue,struct pf_krule * rule)509 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
510 {
511 
512 	PF_RULES_WASSERT();
513 	PF_UNLNKDRULES_ASSERT();
514 
515 	TAILQ_REMOVE(rulequeue, rule, entries);
516 
517 	rule->rule_ref |= PFRULE_REFS;
518 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
519 }
520 
521 static void
pf_unlink_rule(struct pf_krulequeue * rulequeue,struct pf_krule * rule)522 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
523 {
524 
525 	PF_RULES_WASSERT();
526 
527 	PF_UNLNKDRULES_LOCK();
528 	pf_unlink_rule_locked(rulequeue, rule);
529 	PF_UNLNKDRULES_UNLOCK();
530 }
531 
532 static void
pf_free_eth_rule(struct pf_keth_rule * rule)533 pf_free_eth_rule(struct pf_keth_rule *rule)
534 {
535 	PF_RULES_WASSERT();
536 
537 	if (rule == NULL)
538 		return;
539 
540 	if (rule->tag)
541 		tag_unref(&V_pf_tags, rule->tag);
542 	if (rule->match_tag)
543 		tag_unref(&V_pf_tags, rule->match_tag);
544 #ifdef ALTQ
545 	pf_qid_unref(rule->qid);
546 #endif
547 
548 	if (rule->bridge_to)
549 		pfi_kkif_unref(rule->bridge_to);
550 	if (rule->kif)
551 		pfi_kkif_unref(rule->kif);
552 
553 	if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
554 		pfr_detach_table(rule->ipsrc.addr.p.tbl);
555 	if (rule->ipdst.addr.type == PF_ADDR_TABLE)
556 		pfr_detach_table(rule->ipdst.addr.p.tbl);
557 
558 	counter_u64_free(rule->evaluations);
559 	for (int i = 0; i < 2; i++) {
560 		counter_u64_free(rule->packets[i]);
561 		counter_u64_free(rule->bytes[i]);
562 	}
563 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
564 	pf_keth_anchor_remove(rule);
565 
566 	free(rule, M_PFRULE);
567 }
568 
569 void
pf_free_rule(struct pf_krule * rule)570 pf_free_rule(struct pf_krule *rule)
571 {
572 
573 	PF_RULES_WASSERT();
574 	PF_CONFIG_ASSERT();
575 
576 	if (rule->tag)
577 		tag_unref(&V_pf_tags, rule->tag);
578 	if (rule->match_tag)
579 		tag_unref(&V_pf_tags, rule->match_tag);
580 #ifdef ALTQ
581 	if (rule->pqid != rule->qid)
582 		pf_qid_unref(rule->pqid);
583 	pf_qid_unref(rule->qid);
584 #endif
585 	switch (rule->src.addr.type) {
586 	case PF_ADDR_DYNIFTL:
587 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
588 		break;
589 	case PF_ADDR_TABLE:
590 		pfr_detach_table(rule->src.addr.p.tbl);
591 		break;
592 	}
593 	switch (rule->dst.addr.type) {
594 	case PF_ADDR_DYNIFTL:
595 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
596 		break;
597 	case PF_ADDR_TABLE:
598 		pfr_detach_table(rule->dst.addr.p.tbl);
599 		break;
600 	}
601 	if (rule->overload_tbl)
602 		pfr_detach_table(rule->overload_tbl);
603 	if (rule->kif)
604 		pfi_kkif_unref(rule->kif);
605 	if (rule->rcv_kif)
606 		pfi_kkif_unref(rule->rcv_kif);
607 	pf_kanchor_remove(rule);
608 	pf_empty_kpool(&rule->rpool.list);
609 
610 	pf_krule_free(rule);
611 }
612 
613 static void
pf_init_tagset(struct pf_tagset * ts,unsigned int * tunable_size,unsigned int default_size)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
pf_cleanup_tagset(struct pf_tagset * ts)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
tagname2hashindex(const struct pf_tagset * ts,const char * tagname)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
tag2hashindex(const struct pf_tagset * ts,uint16_t tag)667 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
668 {
669 
670 	return (tag & ts->mask);
671 }
672 
673 static u_int16_t
tagname2tag(struct pf_tagset * ts,const char * tagname)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
tag_unref(struct pf_tagset * ts,u_int16_t tag)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
pf_tagname2tag(const char * tagname)753 pf_tagname2tag(const char *tagname)
754 {
755 	return (tagname2tag(&V_pf_tags, tagname));
756 }
757 
758 static int
pf_begin_eth(uint32_t * ticket,const char * anchor)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
pf_rollback_eth_cb(struct epoch_context * ctx)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
pf_rollback_eth(uint32_t ticket,const char * anchor)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
pf_eth_calc_skip_steps(struct pf_keth_ruleq * rules)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
pf_commit_eth(uint32_t ticket,const char * anchor)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
pf_qname2qid(const char * qname)907 pf_qname2qid(const char *qname)
908 {
909 	return (tagname2tag(&V_pf_qids, qname));
910 }
911 
912 static void
pf_qid_unref(uint16_t qid)913 pf_qid_unref(uint16_t qid)
914 {
915 	tag_unref(&V_pf_qids, qid);
916 }
917 
918 static int
pf_begin_altq(u_int32_t * ticket)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
pf_rollback_altq(u_int32_t ticket)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
pf_commit_altq(u_int32_t ticket)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
pf_enable_altq(struct pf_altq * altq)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
pf_disable_altq(struct pf_altq * altq)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
pf_altq_ifnet_event_add(struct ifnet * ifp,int remove,u_int32_t ticket,struct pf_altq * altq)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
pf_altq_ifnet_event(struct ifnet * ifp,int remove)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 *
pf_rule_tree_alloc(int flags)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
pf_rule_tree_free(struct pf_krule_global * tree)1199 pf_rule_tree_free(struct pf_krule_global *tree)
1200 {
1201 
1202 	free(tree, M_TEMP);
1203 }
1204 
1205 static int
pf_begin_rules(u_int32_t * ticket,int rs_num,const char * anchor)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
pf_rollback_rules(u_int32_t ticket,int rs_num,char * anchor)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
pf_hash_rule_addr(MD5_CTX * ctx,struct pf_rule_addr * pfr)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
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)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, rcv_ifname);
1311 	PF_MD5_UPD_STR(rule, match_tagname);
1312 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1313 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1314 	PF_MD5_UPD_HTONL(rule, prob, y);
1315 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1316 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1317 	PF_MD5_UPD(rule, uid.op);
1318 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1319 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1320 	PF_MD5_UPD(rule, gid.op);
1321 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1322 	PF_MD5_UPD(rule, action);
1323 	PF_MD5_UPD(rule, direction);
1324 	PF_MD5_UPD(rule, af);
1325 	PF_MD5_UPD(rule, quick);
1326 	PF_MD5_UPD(rule, ifnot);
1327 	PF_MD5_UPD(rule, match_tag_not);
1328 	PF_MD5_UPD(rule, natpass);
1329 	PF_MD5_UPD(rule, keep_state);
1330 	PF_MD5_UPD(rule, proto);
1331 	PF_MD5_UPD(rule, type);
1332 	PF_MD5_UPD(rule, code);
1333 	PF_MD5_UPD(rule, flags);
1334 	PF_MD5_UPD(rule, flagset);
1335 	PF_MD5_UPD(rule, allow_opts);
1336 	PF_MD5_UPD(rule, rt);
1337 	PF_MD5_UPD(rule, tos);
1338 	PF_MD5_UPD(rule, scrub_flags);
1339 	PF_MD5_UPD(rule, min_ttl);
1340 	PF_MD5_UPD(rule, set_tos);
1341 	if (rule->anchor != NULL)
1342 		PF_MD5_UPD_STR(rule, anchor->path);
1343 }
1344 
1345 static void
pf_hash_rule(struct pf_krule * rule)1346 pf_hash_rule(struct pf_krule *rule)
1347 {
1348 	MD5_CTX		ctx;
1349 
1350 	MD5Init(&ctx);
1351 	pf_hash_rule_rolling(&ctx, rule);
1352 	MD5Final(rule->md5sum, &ctx);
1353 }
1354 
1355 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1356 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1357 {
1358 
1359 	return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1360 }
1361 
1362 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1363 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1364 {
1365 	struct pf_kruleset	*rs;
1366 	struct pf_krule		*rule, **old_array, *old_rule;
1367 	struct pf_krulequeue	*old_rules;
1368 	struct pf_krule_global  *old_tree;
1369 	int			 error;
1370 	u_int32_t		 old_rcount;
1371 
1372 	PF_RULES_WASSERT();
1373 
1374 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1375 		return (EINVAL);
1376 	rs = pf_find_kruleset(anchor);
1377 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1378 	    ticket != rs->rules[rs_num].inactive.ticket)
1379 		return (EBUSY);
1380 
1381 	/* Calculate checksum for the main ruleset */
1382 	if (rs == &pf_main_ruleset) {
1383 		error = pf_setup_pfsync_matching(rs);
1384 		if (error != 0)
1385 			return (error);
1386 	}
1387 
1388 	/* Swap rules, keep the old. */
1389 	old_rules = rs->rules[rs_num].active.ptr;
1390 	old_rcount = rs->rules[rs_num].active.rcount;
1391 	old_array = rs->rules[rs_num].active.ptr_array;
1392 	old_tree = rs->rules[rs_num].active.tree;
1393 
1394 	rs->rules[rs_num].active.ptr =
1395 	    rs->rules[rs_num].inactive.ptr;
1396 	rs->rules[rs_num].active.ptr_array =
1397 	    rs->rules[rs_num].inactive.ptr_array;
1398 	rs->rules[rs_num].active.tree =
1399 	    rs->rules[rs_num].inactive.tree;
1400 	rs->rules[rs_num].active.rcount =
1401 	    rs->rules[rs_num].inactive.rcount;
1402 
1403 	/* Attempt to preserve counter information. */
1404 	if (V_pf_status.keep_counters && old_tree != NULL) {
1405 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1406 		    entries) {
1407 			old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1408 			if (old_rule == NULL) {
1409 				continue;
1410 			}
1411 			pf_counter_u64_critical_enter();
1412 			pf_counter_u64_rollup_protected(&rule->evaluations,
1413 			    pf_counter_u64_fetch(&old_rule->evaluations));
1414 			pf_counter_u64_rollup_protected(&rule->packets[0],
1415 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1416 			pf_counter_u64_rollup_protected(&rule->packets[1],
1417 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1418 			pf_counter_u64_rollup_protected(&rule->bytes[0],
1419 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1420 			pf_counter_u64_rollup_protected(&rule->bytes[1],
1421 			    pf_counter_u64_fetch(&old_rule->bytes[1]));
1422 			pf_counter_u64_critical_exit();
1423 		}
1424 	}
1425 
1426 	rs->rules[rs_num].inactive.ptr = old_rules;
1427 	rs->rules[rs_num].inactive.ptr_array = old_array;
1428 	rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1429 	rs->rules[rs_num].inactive.rcount = old_rcount;
1430 
1431 	rs->rules[rs_num].active.ticket =
1432 	    rs->rules[rs_num].inactive.ticket;
1433 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1434 
1435 	/* Purge the old rule list. */
1436 	PF_UNLNKDRULES_LOCK();
1437 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1438 		pf_unlink_rule_locked(old_rules, rule);
1439 	PF_UNLNKDRULES_UNLOCK();
1440 	if (rs->rules[rs_num].inactive.ptr_array)
1441 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1442 	rs->rules[rs_num].inactive.ptr_array = NULL;
1443 	rs->rules[rs_num].inactive.rcount = 0;
1444 	rs->rules[rs_num].inactive.open = 0;
1445 	pf_remove_if_empty_kruleset(rs);
1446 	free(old_tree, M_TEMP);
1447 
1448 	return (0);
1449 }
1450 
1451 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1452 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1453 {
1454 	MD5_CTX			 ctx;
1455 	struct pf_krule		*rule;
1456 	int			 rs_cnt;
1457 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1458 
1459 	MD5Init(&ctx);
1460 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1461 		/* XXX PF_RULESET_SCRUB as well? */
1462 		if (rs_cnt == PF_RULESET_SCRUB)
1463 			continue;
1464 
1465 		if (rs->rules[rs_cnt].inactive.ptr_array)
1466 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1467 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1468 
1469 		if (rs->rules[rs_cnt].inactive.rcount) {
1470 			rs->rules[rs_cnt].inactive.ptr_array =
1471 			    mallocarray(rs->rules[rs_cnt].inactive.rcount,
1472 			    sizeof(struct pf_rule **),
1473 			    M_TEMP, M_NOWAIT);
1474 
1475 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1476 				return (ENOMEM);
1477 		}
1478 
1479 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1480 		    entries) {
1481 			pf_hash_rule_rolling(&ctx, rule);
1482 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1483 		}
1484 	}
1485 
1486 	MD5Final(digest, &ctx);
1487 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1488 	return (0);
1489 }
1490 
1491 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1492 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1493 {
1494 	int error = 0;
1495 
1496 	switch (addr->type) {
1497 	case PF_ADDR_TABLE:
1498 		addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1499 		if (addr->p.tbl == NULL)
1500 			error = ENOMEM;
1501 		break;
1502 	default:
1503 		error = EINVAL;
1504 	}
1505 
1506 	return (error);
1507 }
1508 
1509 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1510 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1511     sa_family_t af)
1512 {
1513 	int error = 0;
1514 
1515 	switch (addr->type) {
1516 	case PF_ADDR_TABLE:
1517 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1518 		if (addr->p.tbl == NULL)
1519 			error = ENOMEM;
1520 		break;
1521 	case PF_ADDR_DYNIFTL:
1522 		error = pfi_dynaddr_setup(addr, af);
1523 		break;
1524 	}
1525 
1526 	return (error);
1527 }
1528 
1529 void
pf_addr_copyout(struct pf_addr_wrap * addr)1530 pf_addr_copyout(struct pf_addr_wrap *addr)
1531 {
1532 
1533 	switch (addr->type) {
1534 	case PF_ADDR_DYNIFTL:
1535 		pfi_dynaddr_copyout(addr);
1536 		break;
1537 	case PF_ADDR_TABLE:
1538 		pf_tbladdr_copyout(addr);
1539 		break;
1540 	}
1541 }
1542 
1543 static void
pf_src_node_copy(const struct pf_ksrc_node * in,struct pf_src_node * out)1544 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1545 {
1546 	int	secs = time_uptime, diff;
1547 
1548 	bzero(out, sizeof(struct pf_src_node));
1549 
1550 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1551 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1552 
1553 	if (in->rule != NULL)
1554 		out->rule.nr = in->rule->nr;
1555 
1556 	for (int i = 0; i < 2; i++) {
1557 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1558 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1559 	}
1560 
1561 	out->states = in->states;
1562 	out->conn = in->conn;
1563 	out->af = in->af;
1564 	out->ruletype = in->ruletype;
1565 
1566 	out->creation = secs - in->creation;
1567 	if (out->expire > secs)
1568 		out->expire -= secs;
1569 	else
1570 		out->expire = 0;
1571 
1572 	/* Adjust the connection rate estimate. */
1573 	diff = secs - in->conn_rate.last;
1574 	if (diff >= in->conn_rate.seconds)
1575 		out->conn_rate.count = 0;
1576 	else
1577 		out->conn_rate.count -=
1578 		    in->conn_rate.count * diff /
1579 		    in->conn_rate.seconds;
1580 }
1581 
1582 #ifdef ALTQ
1583 /*
1584  * Handle export of struct pf_kaltq to user binaries that may be using any
1585  * version of struct pf_altq.
1586  */
1587 static int
pf_export_kaltq(struct pf_altq * q,struct pfioc_altq_v1 * pa,size_t ioc_size)1588 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1589 {
1590 	u_int32_t version;
1591 
1592 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1593 		version = 0;
1594 	else
1595 		version = pa->version;
1596 
1597 	if (version > PFIOC_ALTQ_VERSION)
1598 		return (EINVAL);
1599 
1600 #define ASSIGN(x) exported_q->x = q->x
1601 #define COPY(x) \
1602 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1603 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1604 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1605 
1606 	switch (version) {
1607 	case 0: {
1608 		struct pf_altq_v0 *exported_q =
1609 		    &((struct pfioc_altq_v0 *)pa)->altq;
1610 
1611 		COPY(ifname);
1612 
1613 		ASSIGN(scheduler);
1614 		ASSIGN(tbrsize);
1615 		exported_q->tbrsize = SATU16(q->tbrsize);
1616 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1617 
1618 		COPY(qname);
1619 		COPY(parent);
1620 		ASSIGN(parent_qid);
1621 		exported_q->bandwidth = SATU32(q->bandwidth);
1622 		ASSIGN(priority);
1623 		ASSIGN(local_flags);
1624 
1625 		ASSIGN(qlimit);
1626 		ASSIGN(flags);
1627 
1628 		if (q->scheduler == ALTQT_HFSC) {
1629 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1630 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1631 			    SATU32(q->pq_u.hfsc_opts.x)
1632 
1633 			ASSIGN_OPT_SATU32(rtsc_m1);
1634 			ASSIGN_OPT(rtsc_d);
1635 			ASSIGN_OPT_SATU32(rtsc_m2);
1636 
1637 			ASSIGN_OPT_SATU32(lssc_m1);
1638 			ASSIGN_OPT(lssc_d);
1639 			ASSIGN_OPT_SATU32(lssc_m2);
1640 
1641 			ASSIGN_OPT_SATU32(ulsc_m1);
1642 			ASSIGN_OPT(ulsc_d);
1643 			ASSIGN_OPT_SATU32(ulsc_m2);
1644 
1645 			ASSIGN_OPT(flags);
1646 
1647 #undef ASSIGN_OPT
1648 #undef ASSIGN_OPT_SATU32
1649 		} else
1650 			COPY(pq_u);
1651 
1652 		ASSIGN(qid);
1653 		break;
1654 	}
1655 	case 1:	{
1656 		struct pf_altq_v1 *exported_q =
1657 		    &((struct pfioc_altq_v1 *)pa)->altq;
1658 
1659 		COPY(ifname);
1660 
1661 		ASSIGN(scheduler);
1662 		ASSIGN(tbrsize);
1663 		ASSIGN(ifbandwidth);
1664 
1665 		COPY(qname);
1666 		COPY(parent);
1667 		ASSIGN(parent_qid);
1668 		ASSIGN(bandwidth);
1669 		ASSIGN(priority);
1670 		ASSIGN(local_flags);
1671 
1672 		ASSIGN(qlimit);
1673 		ASSIGN(flags);
1674 		COPY(pq_u);
1675 
1676 		ASSIGN(qid);
1677 		break;
1678 	}
1679 	default:
1680 		panic("%s: unhandled struct pfioc_altq version", __func__);
1681 		break;
1682 	}
1683 
1684 #undef ASSIGN
1685 #undef COPY
1686 #undef SATU16
1687 #undef SATU32
1688 
1689 	return (0);
1690 }
1691 
1692 /*
1693  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1694  * that may be using any version of it.
1695  */
1696 static int
pf_import_kaltq(struct pfioc_altq_v1 * pa,struct pf_altq * q,size_t ioc_size)1697 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1698 {
1699 	u_int32_t version;
1700 
1701 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1702 		version = 0;
1703 	else
1704 		version = pa->version;
1705 
1706 	if (version > PFIOC_ALTQ_VERSION)
1707 		return (EINVAL);
1708 
1709 #define ASSIGN(x) q->x = imported_q->x
1710 #define COPY(x) \
1711 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1712 
1713 	switch (version) {
1714 	case 0: {
1715 		struct pf_altq_v0 *imported_q =
1716 		    &((struct pfioc_altq_v0 *)pa)->altq;
1717 
1718 		COPY(ifname);
1719 
1720 		ASSIGN(scheduler);
1721 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1722 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1723 
1724 		COPY(qname);
1725 		COPY(parent);
1726 		ASSIGN(parent_qid);
1727 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1728 		ASSIGN(priority);
1729 		ASSIGN(local_flags);
1730 
1731 		ASSIGN(qlimit);
1732 		ASSIGN(flags);
1733 
1734 		if (imported_q->scheduler == ALTQT_HFSC) {
1735 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1736 
1737 			/*
1738 			 * The m1 and m2 parameters are being copied from
1739 			 * 32-bit to 64-bit.
1740 			 */
1741 			ASSIGN_OPT(rtsc_m1);
1742 			ASSIGN_OPT(rtsc_d);
1743 			ASSIGN_OPT(rtsc_m2);
1744 
1745 			ASSIGN_OPT(lssc_m1);
1746 			ASSIGN_OPT(lssc_d);
1747 			ASSIGN_OPT(lssc_m2);
1748 
1749 			ASSIGN_OPT(ulsc_m1);
1750 			ASSIGN_OPT(ulsc_d);
1751 			ASSIGN_OPT(ulsc_m2);
1752 
1753 			ASSIGN_OPT(flags);
1754 
1755 #undef ASSIGN_OPT
1756 		} else
1757 			COPY(pq_u);
1758 
1759 		ASSIGN(qid);
1760 		break;
1761 	}
1762 	case 1: {
1763 		struct pf_altq_v1 *imported_q =
1764 		    &((struct pfioc_altq_v1 *)pa)->altq;
1765 
1766 		COPY(ifname);
1767 
1768 		ASSIGN(scheduler);
1769 		ASSIGN(tbrsize);
1770 		ASSIGN(ifbandwidth);
1771 
1772 		COPY(qname);
1773 		COPY(parent);
1774 		ASSIGN(parent_qid);
1775 		ASSIGN(bandwidth);
1776 		ASSIGN(priority);
1777 		ASSIGN(local_flags);
1778 
1779 		ASSIGN(qlimit);
1780 		ASSIGN(flags);
1781 		COPY(pq_u);
1782 
1783 		ASSIGN(qid);
1784 		break;
1785 	}
1786 	default:
1787 		panic("%s: unhandled struct pfioc_altq version", __func__);
1788 		break;
1789 	}
1790 
1791 #undef ASSIGN
1792 #undef COPY
1793 
1794 	return (0);
1795 }
1796 
1797 static struct pf_altq *
pf_altq_get_nth_active(u_int32_t n)1798 pf_altq_get_nth_active(u_int32_t n)
1799 {
1800 	struct pf_altq		*altq;
1801 	u_int32_t		 nr;
1802 
1803 	nr = 0;
1804 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1805 		if (nr == n)
1806 			return (altq);
1807 		nr++;
1808 	}
1809 
1810 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1811 		if (nr == n)
1812 			return (altq);
1813 		nr++;
1814 	}
1815 
1816 	return (NULL);
1817 }
1818 #endif /* ALTQ */
1819 
1820 struct pf_krule *
pf_krule_alloc(void)1821 pf_krule_alloc(void)
1822 {
1823 	struct pf_krule *rule;
1824 
1825 	rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1826 	mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1827 	rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1828 	    M_WAITOK | M_ZERO);
1829 	return (rule);
1830 }
1831 
1832 void
pf_krule_free(struct pf_krule * rule)1833 pf_krule_free(struct pf_krule *rule)
1834 {
1835 #ifdef PF_WANT_32_TO_64_COUNTER
1836 	bool wowned;
1837 #endif
1838 
1839 	if (rule == NULL)
1840 		return;
1841 
1842 #ifdef PF_WANT_32_TO_64_COUNTER
1843 	if (rule->allrulelinked) {
1844 		wowned = PF_RULES_WOWNED();
1845 		if (!wowned)
1846 			PF_RULES_WLOCK();
1847 		LIST_REMOVE(rule, allrulelist);
1848 		V_pf_allrulecount--;
1849 		if (!wowned)
1850 			PF_RULES_WUNLOCK();
1851 	}
1852 #endif
1853 
1854 	pf_counter_u64_deinit(&rule->evaluations);
1855 	for (int i = 0; i < 2; i++) {
1856 		pf_counter_u64_deinit(&rule->packets[i]);
1857 		pf_counter_u64_deinit(&rule->bytes[i]);
1858 	}
1859 	counter_u64_free(rule->states_cur);
1860 	counter_u64_free(rule->states_tot);
1861 	counter_u64_free(rule->src_nodes);
1862 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1863 
1864 	mtx_destroy(&rule->rpool.mtx);
1865 	free(rule, M_PFRULE);
1866 }
1867 
1868 void
pf_krule_clear_counters(struct pf_krule * rule)1869 pf_krule_clear_counters(struct pf_krule *rule)
1870 {
1871 	pf_counter_u64_zero(&rule->evaluations);
1872 	for (int i = 0; i < 2; i++) {
1873 		pf_counter_u64_zero(&rule->packets[i]);
1874 		pf_counter_u64_zero(&rule->bytes[i]);
1875 	}
1876 	counter_u64_zero(rule->states_tot);
1877 }
1878 
1879 static void
pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr * kpool,struct pf_pooladdr * pool)1880 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1881     struct pf_pooladdr *pool)
1882 {
1883 
1884 	bzero(pool, sizeof(*pool));
1885 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1886 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1887 }
1888 
1889 static int
pf_pooladdr_to_kpooladdr(const struct pf_pooladdr * pool,struct pf_kpooladdr * kpool)1890 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1891     struct pf_kpooladdr *kpool)
1892 {
1893 	int ret;
1894 
1895 	bzero(kpool, sizeof(*kpool));
1896 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1897 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1898 	    sizeof(kpool->ifname));
1899 	return (ret);
1900 }
1901 
1902 static void
pf_pool_to_kpool(const struct pf_pool * pool,struct pf_kpool * kpool)1903 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1904 {
1905 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1906 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1907 
1908 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1909 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1910 
1911 	kpool->tblidx = pool->tblidx;
1912 	kpool->proxy_port[0] = pool->proxy_port[0];
1913 	kpool->proxy_port[1] = pool->proxy_port[1];
1914 	kpool->opts = pool->opts;
1915 }
1916 
1917 static int
pf_rule_to_krule(const struct pf_rule * rule,struct pf_krule * krule)1918 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1919 {
1920 	int ret;
1921 
1922 #ifndef INET
1923 	if (rule->af == AF_INET) {
1924 		return (EAFNOSUPPORT);
1925 	}
1926 #endif /* INET */
1927 #ifndef INET6
1928 	if (rule->af == AF_INET6) {
1929 		return (EAFNOSUPPORT);
1930 	}
1931 #endif /* INET6 */
1932 
1933 	ret = pf_check_rule_addr(&rule->src);
1934 	if (ret != 0)
1935 		return (ret);
1936 	ret = pf_check_rule_addr(&rule->dst);
1937 	if (ret != 0)
1938 		return (ret);
1939 
1940 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1941 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1942 
1943 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1944 	if (ret != 0)
1945 		return (ret);
1946 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1947 	if (ret != 0)
1948 		return (ret);
1949 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1950 	if (ret != 0)
1951 		return (ret);
1952 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1953 	if (ret != 0)
1954 		return (ret);
1955 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
1956 	    sizeof(rule->tagname));
1957 	if (ret != 0)
1958 		return (ret);
1959 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1960 	    sizeof(rule->match_tagname));
1961 	if (ret != 0)
1962 		return (ret);
1963 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1964 	    sizeof(rule->overload_tblname));
1965 	if (ret != 0)
1966 		return (ret);
1967 
1968 	pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1969 
1970 	/* Don't allow userspace to set evaluations, packets or bytes. */
1971 	/* kif, anchor, overload_tbl are not copied over. */
1972 
1973 	krule->os_fingerprint = rule->os_fingerprint;
1974 
1975 	krule->rtableid = rule->rtableid;
1976 	/* pf_rule->timeout is smaller than pf_krule->timeout */
1977 	bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1978 	krule->max_states = rule->max_states;
1979 	krule->max_src_nodes = rule->max_src_nodes;
1980 	krule->max_src_states = rule->max_src_states;
1981 	krule->max_src_conn = rule->max_src_conn;
1982 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1983 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1984 	krule->qid = rule->qid;
1985 	krule->pqid = rule->pqid;
1986 	krule->nr = rule->nr;
1987 	krule->prob = rule->prob;
1988 	krule->cuid = rule->cuid;
1989 	krule->cpid = rule->cpid;
1990 
1991 	krule->return_icmp = rule->return_icmp;
1992 	krule->return_icmp6 = rule->return_icmp6;
1993 	krule->max_mss = rule->max_mss;
1994 	krule->tag = rule->tag;
1995 	krule->match_tag = rule->match_tag;
1996 	krule->scrub_flags = rule->scrub_flags;
1997 
1998 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1999 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2000 
2001 	krule->rule_flag = rule->rule_flag;
2002 	krule->action = rule->action;
2003 	krule->direction = rule->direction;
2004 	krule->log = rule->log;
2005 	krule->logif = rule->logif;
2006 	krule->quick = rule->quick;
2007 	krule->ifnot = rule->ifnot;
2008 	krule->match_tag_not = rule->match_tag_not;
2009 	krule->natpass = rule->natpass;
2010 
2011 	krule->keep_state = rule->keep_state;
2012 	krule->af = rule->af;
2013 	krule->proto = rule->proto;
2014 	krule->type = rule->type;
2015 	krule->code = rule->code;
2016 	krule->flags = rule->flags;
2017 	krule->flagset = rule->flagset;
2018 	krule->min_ttl = rule->min_ttl;
2019 	krule->allow_opts = rule->allow_opts;
2020 	krule->rt = rule->rt;
2021 	krule->return_ttl = rule->return_ttl;
2022 	krule->tos = rule->tos;
2023 	krule->set_tos = rule->set_tos;
2024 
2025 	krule->flush = rule->flush;
2026 	krule->prio = rule->prio;
2027 	krule->set_prio[0] = rule->set_prio[0];
2028 	krule->set_prio[1] = rule->set_prio[1];
2029 
2030 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2031 
2032 	return (0);
2033 }
2034 
2035 int
pf_ioctl_getrules(struct pfioc_rule * pr)2036 pf_ioctl_getrules(struct pfioc_rule *pr)
2037 {
2038 	struct pf_kruleset	*ruleset;
2039 	struct pf_krule		*tail;
2040 	int			 rs_num;
2041 
2042 	PF_RULES_WLOCK();
2043 	ruleset = pf_find_kruleset(pr->anchor);
2044 	if (ruleset == NULL) {
2045 		PF_RULES_WUNLOCK();
2046 		return (EINVAL);
2047 	}
2048 	rs_num = pf_get_ruleset_number(pr->rule.action);
2049 	if (rs_num >= PF_RULESET_MAX) {
2050 		PF_RULES_WUNLOCK();
2051 		return (EINVAL);
2052 	}
2053 	tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2054 	    pf_krulequeue);
2055 	if (tail)
2056 		pr->nr = tail->nr + 1;
2057 	else
2058 		pr->nr = 0;
2059 	pr->ticket = ruleset->rules[rs_num].active.ticket;
2060 	PF_RULES_WUNLOCK();
2061 
2062 	return (0);
2063 }
2064 
2065 int
pf_ioctl_addrule(struct pf_krule * rule,uint32_t ticket,uint32_t pool_ticket,const char * anchor,const char * anchor_call,uid_t uid,pid_t pid)2066 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2067     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2068     uid_t uid, pid_t pid)
2069 {
2070 	struct pf_kruleset	*ruleset;
2071 	struct pf_krule		*tail;
2072 	struct pf_kpooladdr	*pa;
2073 	struct pfi_kkif		*kif = NULL, *rcv_kif = NULL;
2074 	int			 rs_num;
2075 	int			 error = 0;
2076 
2077 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2078 		error = EINVAL;
2079 		goto errout_unlocked;
2080 	}
2081 
2082 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
2083 
2084 	if (rule->ifname[0])
2085 		kif = pf_kkif_create(M_WAITOK);
2086 	if (rule->rcv_ifname[0])
2087 		rcv_kif = pf_kkif_create(M_WAITOK);
2088 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2089 	for (int i = 0; i < 2; i++) {
2090 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2091 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2092 	}
2093 	rule->states_cur = counter_u64_alloc(M_WAITOK);
2094 	rule->states_tot = counter_u64_alloc(M_WAITOK);
2095 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
2096 	rule->cuid = uid;
2097 	rule->cpid = pid;
2098 	TAILQ_INIT(&rule->rpool.list);
2099 
2100 	PF_CONFIG_LOCK();
2101 	PF_RULES_WLOCK();
2102 #ifdef PF_WANT_32_TO_64_COUNTER
2103 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2104 	MPASS(!rule->allrulelinked);
2105 	rule->allrulelinked = true;
2106 	V_pf_allrulecount++;
2107 #endif
2108 	ruleset = pf_find_kruleset(anchor);
2109 	if (ruleset == NULL)
2110 		ERROUT(EINVAL);
2111 	rs_num = pf_get_ruleset_number(rule->action);
2112 	if (rs_num >= PF_RULESET_MAX)
2113 		ERROUT(EINVAL);
2114 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2115 		DPFPRINTF(PF_DEBUG_MISC,
2116 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
2117 		    ruleset->rules[rs_num].inactive.ticket));
2118 		ERROUT(EBUSY);
2119 	}
2120 	if (pool_ticket != V_ticket_pabuf) {
2121 		DPFPRINTF(PF_DEBUG_MISC,
2122 		    ("pool_ticket: %d != %d\n", pool_ticket,
2123 		    V_ticket_pabuf));
2124 		ERROUT(EBUSY);
2125 	}
2126 	/*
2127 	 * XXXMJG hack: there is no mechanism to ensure they started the
2128 	 * transaction. Ticket checked above may happen to match by accident,
2129 	 * even if nobody called DIOCXBEGIN, let alone this process.
2130 	 * Partially work around it by checking if the RB tree got allocated,
2131 	 * see pf_begin_rules.
2132 	 */
2133 	if (ruleset->rules[rs_num].inactive.tree == NULL) {
2134 		ERROUT(EINVAL);
2135 	}
2136 
2137 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2138 	    pf_krulequeue);
2139 	if (tail)
2140 		rule->nr = tail->nr + 1;
2141 	else
2142 		rule->nr = 0;
2143 	if (rule->ifname[0]) {
2144 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
2145 		kif = NULL;
2146 		pfi_kkif_ref(rule->kif);
2147 	} else
2148 		rule->kif = NULL;
2149 
2150 	if (rule->rcv_ifname[0]) {
2151 		rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname);
2152 		rcv_kif = NULL;
2153 		pfi_kkif_ref(rule->rcv_kif);
2154 	} else
2155 		rule->rcv_kif = NULL;
2156 
2157 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2158 		error = EBUSY;
2159 
2160 #ifdef ALTQ
2161 	/* set queue IDs */
2162 	if (rule->qname[0] != 0) {
2163 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2164 			error = EBUSY;
2165 		else if (rule->pqname[0] != 0) {
2166 			if ((rule->pqid =
2167 			    pf_qname2qid(rule->pqname)) == 0)
2168 				error = EBUSY;
2169 		} else
2170 			rule->pqid = rule->qid;
2171 	}
2172 #endif
2173 	if (rule->tagname[0])
2174 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2175 			error = EBUSY;
2176 	if (rule->match_tagname[0])
2177 		if ((rule->match_tag =
2178 		    pf_tagname2tag(rule->match_tagname)) == 0)
2179 			error = EBUSY;
2180 	if (rule->rt && !rule->direction)
2181 		error = EINVAL;
2182 	if (!rule->log)
2183 		rule->logif = 0;
2184 	if (rule->logif >= PFLOGIFS_MAX)
2185 		error = EINVAL;
2186 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2187 		error = ENOMEM;
2188 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2189 		error = ENOMEM;
2190 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
2191 		error = EINVAL;
2192 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
2193 	    (rule->set_prio[0] > PF_PRIO_MAX ||
2194 	    rule->set_prio[1] > PF_PRIO_MAX))
2195 		error = EINVAL;
2196 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2197 		if (pa->addr.type == PF_ADDR_TABLE) {
2198 			pa->addr.p.tbl = pfr_attach_table(ruleset,
2199 			    pa->addr.v.tblname);
2200 			if (pa->addr.p.tbl == NULL)
2201 				error = ENOMEM;
2202 		}
2203 
2204 	rule->overload_tbl = NULL;
2205 	if (rule->overload_tblname[0]) {
2206 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
2207 		    rule->overload_tblname)) == NULL)
2208 			error = EINVAL;
2209 		else
2210 			rule->overload_tbl->pfrkt_flags |=
2211 			    PFR_TFLAG_ACTIVE;
2212 	}
2213 
2214 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2215 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2216 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2217 	    (rule->rt > PF_NOPFROUTE)) &&
2218 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
2219 		error = EINVAL;
2220 
2221 	if (rule->action == PF_PASS && rule->rpool.opts & PF_POOL_STICKYADDR &&
2222 	    !rule->keep_state) {
2223 		error = EINVAL;
2224 	}
2225 
2226 	if (error) {
2227 		pf_free_rule(rule);
2228 		rule = NULL;
2229 		ERROUT(error);
2230 	}
2231 
2232 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2233 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2234 	    rule, entries);
2235 	ruleset->rules[rs_num].inactive.rcount++;
2236 
2237 	PF_RULES_WUNLOCK();
2238 	pf_hash_rule(rule);
2239 	if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2240 		PF_RULES_WLOCK();
2241 		TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2242 		ruleset->rules[rs_num].inactive.rcount--;
2243 		pf_free_rule(rule);
2244 		rule = NULL;
2245 		ERROUT(EEXIST);
2246 	}
2247 	PF_CONFIG_UNLOCK();
2248 
2249 	return (0);
2250 
2251 #undef ERROUT
2252 errout:
2253 	PF_RULES_WUNLOCK();
2254 	PF_CONFIG_UNLOCK();
2255 errout_unlocked:
2256 	pf_kkif_free(rcv_kif);
2257 	pf_kkif_free(kif);
2258 	pf_krule_free(rule);
2259 	return (error);
2260 }
2261 
2262 static bool
pf_label_match(const struct pf_krule * rule,const char * label)2263 pf_label_match(const struct pf_krule *rule, const char *label)
2264 {
2265 	int i = 0;
2266 
2267 	while (*rule->label[i]) {
2268 		if (strcmp(rule->label[i], label) == 0)
2269 			return (true);
2270 		i++;
2271 	}
2272 
2273 	return (false);
2274 }
2275 
2276 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)2277 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2278 {
2279 	struct pf_kstate *s;
2280 	int more = 0;
2281 
2282 	s = pf_find_state_all(key, dir, &more);
2283 	if (s == NULL)
2284 		return (0);
2285 
2286 	if (more) {
2287 		PF_STATE_UNLOCK(s);
2288 		return (0);
2289 	}
2290 
2291 	pf_unlink_state(s);
2292 	return (1);
2293 }
2294 
2295 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)2296 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2297 {
2298 	struct pf_kstate	*s;
2299 	struct pf_state_key	*sk;
2300 	struct pf_addr		*srcaddr, *dstaddr;
2301 	struct pf_state_key_cmp	 match_key;
2302 	int			 idx, killed = 0;
2303 	unsigned int		 dir;
2304 	u_int16_t		 srcport, dstport;
2305 	struct pfi_kkif		*kif;
2306 
2307 relock_DIOCKILLSTATES:
2308 	PF_HASHROW_LOCK(ih);
2309 	LIST_FOREACH(s, &ih->states, entry) {
2310 		/* For floating states look at the original kif. */
2311 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2312 
2313 		sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2314 		if (s->direction == PF_OUT) {
2315 			srcaddr = &sk->addr[1];
2316 			dstaddr = &sk->addr[0];
2317 			srcport = sk->port[1];
2318 			dstport = sk->port[0];
2319 		} else {
2320 			srcaddr = &sk->addr[0];
2321 			dstaddr = &sk->addr[1];
2322 			srcport = sk->port[0];
2323 			dstport = sk->port[1];
2324 		}
2325 
2326 		if (psk->psk_af && sk->af != psk->psk_af)
2327 			continue;
2328 
2329 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2330 			continue;
2331 
2332 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2333 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2334 			continue;
2335 
2336 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2337 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2338 			continue;
2339 
2340 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2341 		    &psk->psk_rt_addr.addr.v.a.addr,
2342 		    &psk->psk_rt_addr.addr.v.a.mask,
2343 		    &s->rt_addr, sk->af))
2344 			continue;
2345 
2346 		if (psk->psk_src.port_op != 0 &&
2347 		    ! pf_match_port(psk->psk_src.port_op,
2348 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2349 			continue;
2350 
2351 		if (psk->psk_dst.port_op != 0 &&
2352 		    ! pf_match_port(psk->psk_dst.port_op,
2353 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2354 			continue;
2355 
2356 		if (psk->psk_label[0] &&
2357 		    ! pf_label_match(s->rule, psk->psk_label))
2358 			continue;
2359 
2360 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2361 		    kif->pfik_name))
2362 			continue;
2363 
2364 		if (psk->psk_kill_match) {
2365 			/* Create the key to find matching states, with lock
2366 			 * held. */
2367 
2368 			bzero(&match_key, sizeof(match_key));
2369 
2370 			if (s->direction == PF_OUT) {
2371 				dir = PF_IN;
2372 				idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2373 			} else {
2374 				dir = PF_OUT;
2375 				idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2376 			}
2377 
2378 			match_key.af = s->key[idx]->af;
2379 			match_key.proto = s->key[idx]->proto;
2380 			PF_ACPY(&match_key.addr[0],
2381 			    &s->key[idx]->addr[1], match_key.af);
2382 			match_key.port[0] = s->key[idx]->port[1];
2383 			PF_ACPY(&match_key.addr[1],
2384 			    &s->key[idx]->addr[0], match_key.af);
2385 			match_key.port[1] = s->key[idx]->port[0];
2386 		}
2387 
2388 		pf_unlink_state(s);
2389 		killed++;
2390 
2391 		if (psk->psk_kill_match)
2392 			killed += pf_kill_matching_state(&match_key, dir);
2393 
2394 		goto relock_DIOCKILLSTATES;
2395 	}
2396 	PF_HASHROW_UNLOCK(ih);
2397 
2398 	return (killed);
2399 }
2400 
2401 int
pf_start(void)2402 pf_start(void)
2403 {
2404 	int error = 0;
2405 
2406 	sx_xlock(&V_pf_ioctl_lock);
2407 	if (V_pf_status.running)
2408 		error = EEXIST;
2409 	else {
2410 		hook_pf();
2411 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2412 			hook_pf_eth();
2413 		V_pf_status.running = 1;
2414 		V_pf_status.since = time_second;
2415 		new_unrhdr64(&V_pf_stateid, time_second);
2416 
2417 		DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2418 	}
2419 	sx_xunlock(&V_pf_ioctl_lock);
2420 
2421 	return (error);
2422 }
2423 
2424 int
pf_stop(void)2425 pf_stop(void)
2426 {
2427 	int error = 0;
2428 
2429 	sx_xlock(&V_pf_ioctl_lock);
2430 	if (!V_pf_status.running)
2431 		error = ENOENT;
2432 	else {
2433 		V_pf_status.running = 0;
2434 		dehook_pf();
2435 		dehook_pf_eth();
2436 		V_pf_status.since = time_second;
2437 		DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2438 	}
2439 	sx_xunlock(&V_pf_ioctl_lock);
2440 
2441 	return (error);
2442 }
2443 
2444 void
pf_ioctl_clear_status(void)2445 pf_ioctl_clear_status(void)
2446 {
2447 	PF_RULES_WLOCK();
2448 	for (int i = 0; i < PFRES_MAX; i++)
2449 		counter_u64_zero(V_pf_status.counters[i]);
2450 	for (int i = 0; i < FCNT_MAX; i++)
2451 		pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2452 	for (int i = 0; i < SCNT_MAX; i++)
2453 		counter_u64_zero(V_pf_status.scounters[i]);
2454 	for (int i = 0; i < KLCNT_MAX; i++)
2455 		counter_u64_zero(V_pf_status.lcounters[i]);
2456 	V_pf_status.since = time_second;
2457 	if (*V_pf_status.ifname)
2458 		pfi_update_status(V_pf_status.ifname, NULL);
2459 	PF_RULES_WUNLOCK();
2460 }
2461 
2462 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2463 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2464 {
2465 	uint32_t old;
2466 
2467 	if (timeout < 0 || timeout >= PFTM_MAX ||
2468 	    seconds < 0)
2469 		return (EINVAL);
2470 
2471 	PF_RULES_WLOCK();
2472 	old = V_pf_default_rule.timeout[timeout];
2473 	if (timeout == PFTM_INTERVAL && seconds == 0)
2474 		seconds = 1;
2475 	V_pf_default_rule.timeout[timeout] = seconds;
2476 	if (timeout == PFTM_INTERVAL && seconds < old)
2477 		wakeup(pf_purge_thread);
2478 
2479 	if (prev_seconds != NULL)
2480 		*prev_seconds = old;
2481 
2482 	PF_RULES_WUNLOCK();
2483 
2484 	return (0);
2485 }
2486 
2487 int
pf_ioctl_get_timeout(int timeout,int * seconds)2488 pf_ioctl_get_timeout(int timeout, int *seconds)
2489 {
2490 	PF_RULES_RLOCK_TRACKER;
2491 
2492 	if (timeout < 0 || timeout >= PFTM_MAX)
2493 		return (EINVAL);
2494 
2495 	PF_RULES_RLOCK();
2496 	*seconds = V_pf_default_rule.timeout[timeout];
2497 	PF_RULES_RUNLOCK();
2498 
2499 	return (0);
2500 }
2501 
2502 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2503 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2504 {
2505 
2506 	PF_RULES_WLOCK();
2507 	if (index < 0 || index >= PF_LIMIT_MAX ||
2508 	    V_pf_limits[index].zone == NULL) {
2509 		PF_RULES_WUNLOCK();
2510 		return (EINVAL);
2511 	}
2512 	uma_zone_set_max(V_pf_limits[index].zone, limit);
2513 	if (old_limit != NULL)
2514 		*old_limit = V_pf_limits[index].limit;
2515 	V_pf_limits[index].limit = limit;
2516 	PF_RULES_WUNLOCK();
2517 
2518 	return (0);
2519 }
2520 
2521 int
pf_ioctl_get_limit(int index,unsigned int * limit)2522 pf_ioctl_get_limit(int index, unsigned int *limit)
2523 {
2524 	PF_RULES_RLOCK_TRACKER;
2525 
2526 	if (index < 0 || index >= PF_LIMIT_MAX)
2527 		return (EINVAL);
2528 
2529 	PF_RULES_RLOCK();
2530 	*limit = V_pf_limits[index].limit;
2531 	PF_RULES_RUNLOCK();
2532 
2533 	return (0);
2534 }
2535 
2536 int
pf_ioctl_begin_addrs(uint32_t * ticket)2537 pf_ioctl_begin_addrs(uint32_t *ticket)
2538 {
2539 	PF_RULES_WLOCK();
2540 	pf_empty_kpool(&V_pf_pabuf);
2541 	*ticket = ++V_ticket_pabuf;
2542 	PF_RULES_WUNLOCK();
2543 
2544 	return (0);
2545 }
2546 
2547 int
pf_ioctl_add_addr(struct pfioc_pooladdr * pp)2548 pf_ioctl_add_addr(struct pfioc_pooladdr *pp)
2549 {
2550 	struct pf_kpooladdr	*pa = NULL;
2551 	struct pfi_kkif		*kif = NULL;
2552 	int error;
2553 
2554 #ifndef INET
2555 	if (pp->af == AF_INET)
2556 		return (EAFNOSUPPORT);
2557 #endif /* INET */
2558 #ifndef INET6
2559 	if (pp->af == AF_INET6)
2560 		return (EAFNOSUPPORT);
2561 #endif /* INET6 */
2562 
2563 	if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2564 	    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2565 	    pp->addr.addr.type != PF_ADDR_TABLE)
2566 		return (EINVAL);
2567 
2568 	if (pp->addr.addr.p.dyn != NULL)
2569 		return (EINVAL);
2570 
2571 	pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2572 	error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2573 	if (error != 0)
2574 		goto out;
2575 	if (pa->ifname[0])
2576 		kif = pf_kkif_create(M_WAITOK);
2577 	PF_RULES_WLOCK();
2578 	if (pp->ticket != V_ticket_pabuf) {
2579 		PF_RULES_WUNLOCK();
2580 		if (pa->ifname[0])
2581 			pf_kkif_free(kif);
2582 		error = EBUSY;
2583 		goto out;
2584 	}
2585 	if (pa->ifname[0]) {
2586 		pa->kif = pfi_kkif_attach(kif, pa->ifname);
2587 		kif = NULL;
2588 		pfi_kkif_ref(pa->kif);
2589 	} else
2590 		pa->kif = NULL;
2591 	if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2592 	    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2593 		if (pa->ifname[0])
2594 			pfi_kkif_unref(pa->kif);
2595 		PF_RULES_WUNLOCK();
2596 		goto out;
2597 	}
2598 	TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2599 	PF_RULES_WUNLOCK();
2600 
2601 	return (0);
2602 
2603 out:
2604 	free(pa, M_PFRULE);
2605 	return (error);
2606 }
2607 
2608 int
pf_ioctl_get_addrs(struct pfioc_pooladdr * pp)2609 pf_ioctl_get_addrs(struct pfioc_pooladdr *pp)
2610 {
2611 	struct pf_kpool		*pool;
2612 	struct pf_kpooladdr	*pa;
2613 
2614 	PF_RULES_RLOCK_TRACKER;
2615 
2616 	pp->anchor[sizeof(pp->anchor) - 1] = 0;
2617 	pp->nr = 0;
2618 
2619 	PF_RULES_RLOCK();
2620 	pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2621 	    pp->r_num, 0, 1, 0);
2622 	if (pool == NULL) {
2623 		PF_RULES_RUNLOCK();
2624 		return (EBUSY);
2625 	}
2626 	TAILQ_FOREACH(pa, &pool->list, entries)
2627 		pp->nr++;
2628 	PF_RULES_RUNLOCK();
2629 
2630 	return (0);
2631 }
2632 
2633 int
pf_ioctl_get_addr(struct pfioc_pooladdr * pp)2634 pf_ioctl_get_addr(struct pfioc_pooladdr *pp)
2635 {
2636 	struct pf_kpool		*pool;
2637 	struct pf_kpooladdr	*pa;
2638 	u_int32_t		 nr = 0;
2639 
2640 	PF_RULES_RLOCK_TRACKER;
2641 
2642 	pp->anchor[sizeof(pp->anchor) - 1] = 0;
2643 
2644 	PF_RULES_RLOCK();
2645 	pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2646 	    pp->r_num, 0, 1, 1);
2647 	if (pool == NULL) {
2648 		PF_RULES_RUNLOCK();
2649 		return (EBUSY);
2650 	}
2651 	pa = TAILQ_FIRST(&pool->list);
2652 	while ((pa != NULL) && (nr < pp->nr)) {
2653 		pa = TAILQ_NEXT(pa, entries);
2654 		nr++;
2655 	}
2656 	if (pa == NULL) {
2657 		PF_RULES_RUNLOCK();
2658 		return (EBUSY);
2659 	}
2660 	pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2661 	pf_addr_copyout(&pp->addr.addr);
2662 	PF_RULES_RUNLOCK();
2663 
2664 	return (0);
2665 }
2666 
2667 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)2668 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
2669 {
2670 	struct pf_kruleset	*ruleset;
2671 	struct pf_kanchor	*anchor;
2672 
2673 	PF_RULES_RLOCK_TRACKER;
2674 
2675 	pr->path[sizeof(pr->path) - 1] = 0;
2676 
2677 	PF_RULES_RLOCK();
2678 	if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2679 		PF_RULES_RUNLOCK();
2680 		return (ENOENT);
2681 	}
2682 	pr->nr = 0;
2683 	if (ruleset->anchor == NULL) {
2684 		/* XXX kludge for pf_main_ruleset */
2685 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2686 			if (anchor->parent == NULL)
2687 				pr->nr++;
2688 	} else {
2689 		RB_FOREACH(anchor, pf_kanchor_node,
2690 		    &ruleset->anchor->children)
2691 			pr->nr++;
2692 	}
2693 	PF_RULES_RUNLOCK();
2694 
2695 	return (0);
2696 }
2697 
2698 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)2699 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
2700 {
2701 	struct pf_kruleset	*ruleset;
2702 	struct pf_kanchor	*anchor;
2703 	u_int32_t		 nr = 0;
2704 	int			 error = 0;
2705 
2706 	PF_RULES_RLOCK_TRACKER;
2707 
2708 	PF_RULES_RLOCK();
2709 	if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2710 		PF_RULES_RUNLOCK();
2711 		return (ENOENT);
2712 	}
2713 
2714 	pr->name[0] = 0;
2715 	if (ruleset->anchor == NULL) {
2716 		/* XXX kludge for pf_main_ruleset */
2717 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2718 			if (anchor->parent == NULL && nr++ == pr->nr) {
2719 				strlcpy(pr->name, anchor->name,
2720 				    sizeof(pr->name));
2721 				break;
2722 			}
2723 	} else {
2724 		RB_FOREACH(anchor, pf_kanchor_node,
2725 		    &ruleset->anchor->children)
2726 			if (nr++ == pr->nr) {
2727 				strlcpy(pr->name, anchor->name,
2728 				    sizeof(pr->name));
2729 				break;
2730 			}
2731 	}
2732 	if (!pr->name[0])
2733 		error = EBUSY;
2734 	PF_RULES_RUNLOCK();
2735 
2736 	return (error);
2737 }
2738 
2739 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2740 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2741 {
2742 	int			 error = 0;
2743 	PF_RULES_RLOCK_TRACKER;
2744 
2745 #define	ERROUT_IOCTL(target, x)					\
2746     do {								\
2747 	    error = (x);						\
2748 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2749 	    goto target;						\
2750     } while (0)
2751 
2752 
2753 	/* XXX keep in sync with switch() below */
2754 	if (securelevel_gt(td->td_ucred, 2))
2755 		switch (cmd) {
2756 		case DIOCGETRULES:
2757 		case DIOCGETRULENV:
2758 		case DIOCGETADDRS:
2759 		case DIOCGETADDR:
2760 		case DIOCGETSTATE:
2761 		case DIOCGETSTATENV:
2762 		case DIOCSETSTATUSIF:
2763 		case DIOCGETSTATUSNV:
2764 		case DIOCCLRSTATUS:
2765 		case DIOCNATLOOK:
2766 		case DIOCSETDEBUG:
2767 #ifdef COMPAT_FREEBSD14
2768 		case DIOCGETSTATES:
2769 		case DIOCGETSTATESV2:
2770 #endif
2771 		case DIOCGETTIMEOUT:
2772 		case DIOCCLRRULECTRS:
2773 		case DIOCGETLIMIT:
2774 		case DIOCGETALTQSV0:
2775 		case DIOCGETALTQSV1:
2776 		case DIOCGETALTQV0:
2777 		case DIOCGETALTQV1:
2778 		case DIOCGETQSTATSV0:
2779 		case DIOCGETQSTATSV1:
2780 		case DIOCGETRULESETS:
2781 		case DIOCGETRULESET:
2782 		case DIOCRGETTABLES:
2783 		case DIOCRGETTSTATS:
2784 		case DIOCRCLRTSTATS:
2785 		case DIOCRCLRADDRS:
2786 		case DIOCRADDADDRS:
2787 		case DIOCRDELADDRS:
2788 		case DIOCRSETADDRS:
2789 		case DIOCRGETADDRS:
2790 		case DIOCRGETASTATS:
2791 		case DIOCRCLRASTATS:
2792 		case DIOCRTSTADDRS:
2793 		case DIOCOSFPGET:
2794 		case DIOCGETSRCNODES:
2795 		case DIOCCLRSRCNODES:
2796 		case DIOCGETSYNCOOKIES:
2797 		case DIOCIGETIFACES:
2798 		case DIOCGIFSPEEDV0:
2799 		case DIOCGIFSPEEDV1:
2800 		case DIOCSETIFFLAG:
2801 		case DIOCCLRIFFLAG:
2802 		case DIOCGETETHRULES:
2803 		case DIOCGETETHRULE:
2804 		case DIOCGETETHRULESETS:
2805 		case DIOCGETETHRULESET:
2806 			break;
2807 		case DIOCRCLRTABLES:
2808 		case DIOCRADDTABLES:
2809 		case DIOCRDELTABLES:
2810 		case DIOCRSETTFLAGS:
2811 			if (((struct pfioc_table *)addr)->pfrio_flags &
2812 			    PFR_FLAG_DUMMY)
2813 				break; /* dummy operation ok */
2814 			return (EPERM);
2815 		default:
2816 			return (EPERM);
2817 		}
2818 
2819 	if (!(flags & FWRITE))
2820 		switch (cmd) {
2821 		case DIOCGETRULES:
2822 		case DIOCGETADDRS:
2823 		case DIOCGETADDR:
2824 		case DIOCGETSTATE:
2825 		case DIOCGETSTATENV:
2826 		case DIOCGETSTATUSNV:
2827 #ifdef COMPAT_FREEBSD14
2828 		case DIOCGETSTATES:
2829 		case DIOCGETSTATESV2:
2830 #endif
2831 		case DIOCGETTIMEOUT:
2832 		case DIOCGETLIMIT:
2833 		case DIOCGETALTQSV0:
2834 		case DIOCGETALTQSV1:
2835 		case DIOCGETALTQV0:
2836 		case DIOCGETALTQV1:
2837 		case DIOCGETQSTATSV0:
2838 		case DIOCGETQSTATSV1:
2839 		case DIOCGETRULESETS:
2840 		case DIOCGETRULESET:
2841 		case DIOCNATLOOK:
2842 		case DIOCRGETTABLES:
2843 		case DIOCRGETTSTATS:
2844 		case DIOCRGETADDRS:
2845 		case DIOCRGETASTATS:
2846 		case DIOCRTSTADDRS:
2847 		case DIOCOSFPGET:
2848 		case DIOCGETSRCNODES:
2849 		case DIOCGETSYNCOOKIES:
2850 		case DIOCIGETIFACES:
2851 		case DIOCGIFSPEEDV1:
2852 		case DIOCGIFSPEEDV0:
2853 		case DIOCGETRULENV:
2854 		case DIOCGETETHRULES:
2855 		case DIOCGETETHRULE:
2856 		case DIOCGETETHRULESETS:
2857 		case DIOCGETETHRULESET:
2858 			break;
2859 		case DIOCRCLRTABLES:
2860 		case DIOCRADDTABLES:
2861 		case DIOCRDELTABLES:
2862 		case DIOCRCLRTSTATS:
2863 		case DIOCRCLRADDRS:
2864 		case DIOCRADDADDRS:
2865 		case DIOCRDELADDRS:
2866 		case DIOCRSETADDRS:
2867 		case DIOCRSETTFLAGS:
2868 			if (((struct pfioc_table *)addr)->pfrio_flags &
2869 			    PFR_FLAG_DUMMY) {
2870 				flags |= FWRITE; /* need write lock for dummy */
2871 				break; /* dummy operation ok */
2872 			}
2873 			return (EACCES);
2874 		default:
2875 			return (EACCES);
2876 		}
2877 
2878 	CURVNET_SET(TD_TO_VNET(td));
2879 
2880 	switch (cmd) {
2881 #ifdef COMPAT_FREEBSD14
2882 	case DIOCSTART:
2883 		error = pf_start();
2884 		break;
2885 
2886 	case DIOCSTOP:
2887 		error = pf_stop();
2888 		break;
2889 #endif
2890 
2891 	case DIOCGETETHRULES: {
2892 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2893 		nvlist_t		*nvl;
2894 		void			*packed;
2895 		struct pf_keth_rule	*tail;
2896 		struct pf_keth_ruleset	*rs;
2897 		u_int32_t		 ticket, nr;
2898 		const char		*anchor = "";
2899 
2900 		nvl = NULL;
2901 		packed = NULL;
2902 
2903 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2904 
2905 		if (nv->len > pf_ioctl_maxcount)
2906 			ERROUT(ENOMEM);
2907 
2908 		/* Copy the request in */
2909 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2910 		error = copyin(nv->data, packed, nv->len);
2911 		if (error)
2912 			ERROUT(error);
2913 
2914 		nvl = nvlist_unpack(packed, nv->len, 0);
2915 		if (nvl == NULL)
2916 			ERROUT(EBADMSG);
2917 
2918 		if (! nvlist_exists_string(nvl, "anchor"))
2919 			ERROUT(EBADMSG);
2920 
2921 		anchor = nvlist_get_string(nvl, "anchor");
2922 
2923 		rs = pf_find_keth_ruleset(anchor);
2924 
2925 		nvlist_destroy(nvl);
2926 		nvl = NULL;
2927 		free(packed, M_NVLIST);
2928 		packed = NULL;
2929 
2930 		if (rs == NULL)
2931 			ERROUT(ENOENT);
2932 
2933 		/* Reply */
2934 		nvl = nvlist_create(0);
2935 		if (nvl == NULL)
2936 			ERROUT(ENOMEM);
2937 
2938 		PF_RULES_RLOCK();
2939 
2940 		ticket = rs->active.ticket;
2941 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2942 		if (tail)
2943 			nr = tail->nr + 1;
2944 		else
2945 			nr = 0;
2946 
2947 		PF_RULES_RUNLOCK();
2948 
2949 		nvlist_add_number(nvl, "ticket", ticket);
2950 		nvlist_add_number(nvl, "nr", nr);
2951 
2952 		packed = nvlist_pack(nvl, &nv->len);
2953 		if (packed == NULL)
2954 			ERROUT(ENOMEM);
2955 
2956 		if (nv->size == 0)
2957 			ERROUT(0);
2958 		else if (nv->size < nv->len)
2959 			ERROUT(ENOSPC);
2960 
2961 		error = copyout(packed, nv->data, nv->len);
2962 
2963 #undef ERROUT
2964 DIOCGETETHRULES_error:
2965 		free(packed, M_NVLIST);
2966 		nvlist_destroy(nvl);
2967 		break;
2968 	}
2969 
2970 	case DIOCGETETHRULE: {
2971 		struct epoch_tracker	 et;
2972 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2973 		nvlist_t		*nvl = NULL;
2974 		void			*nvlpacked = NULL;
2975 		struct pf_keth_rule	*rule = NULL;
2976 		struct pf_keth_ruleset	*rs;
2977 		u_int32_t		 ticket, nr;
2978 		bool			 clear = false;
2979 		const char		*anchor;
2980 
2981 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2982 
2983 		if (nv->len > pf_ioctl_maxcount)
2984 			ERROUT(ENOMEM);
2985 
2986 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2987 		error = copyin(nv->data, nvlpacked, nv->len);
2988 		if (error)
2989 			ERROUT(error);
2990 
2991 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2992 		if (nvl == NULL)
2993 			ERROUT(EBADMSG);
2994 		if (! nvlist_exists_number(nvl, "ticket"))
2995 			ERROUT(EBADMSG);
2996 		ticket = nvlist_get_number(nvl, "ticket");
2997 		if (! nvlist_exists_string(nvl, "anchor"))
2998 			ERROUT(EBADMSG);
2999 		anchor = nvlist_get_string(nvl, "anchor");
3000 
3001 		if (nvlist_exists_bool(nvl, "clear"))
3002 			clear = nvlist_get_bool(nvl, "clear");
3003 
3004 		if (clear && !(flags & FWRITE))
3005 			ERROUT(EACCES);
3006 
3007 		if (! nvlist_exists_number(nvl, "nr"))
3008 			ERROUT(EBADMSG);
3009 		nr = nvlist_get_number(nvl, "nr");
3010 
3011 		PF_RULES_RLOCK();
3012 		rs = pf_find_keth_ruleset(anchor);
3013 		if (rs == NULL) {
3014 			PF_RULES_RUNLOCK();
3015 			ERROUT(ENOENT);
3016 		}
3017 		if (ticket != rs->active.ticket) {
3018 			PF_RULES_RUNLOCK();
3019 			ERROUT(EBUSY);
3020 		}
3021 
3022 		nvlist_destroy(nvl);
3023 		nvl = NULL;
3024 		free(nvlpacked, M_NVLIST);
3025 		nvlpacked = NULL;
3026 
3027 		rule = TAILQ_FIRST(rs->active.rules);
3028 		while ((rule != NULL) && (rule->nr != nr))
3029 			rule = TAILQ_NEXT(rule, entries);
3030 		if (rule == NULL) {
3031 			PF_RULES_RUNLOCK();
3032 			ERROUT(ENOENT);
3033 		}
3034 		/* Make sure rule can't go away. */
3035 		NET_EPOCH_ENTER(et);
3036 		PF_RULES_RUNLOCK();
3037 		nvl = pf_keth_rule_to_nveth_rule(rule);
3038 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
3039 			NET_EPOCH_EXIT(et);
3040 			ERROUT(EBUSY);
3041 		}
3042 		NET_EPOCH_EXIT(et);
3043 		if (nvl == NULL)
3044 			ERROUT(ENOMEM);
3045 
3046 		nvlpacked = nvlist_pack(nvl, &nv->len);
3047 		if (nvlpacked == NULL)
3048 			ERROUT(ENOMEM);
3049 
3050 		if (nv->size == 0)
3051 			ERROUT(0);
3052 		else if (nv->size < nv->len)
3053 			ERROUT(ENOSPC);
3054 
3055 		error = copyout(nvlpacked, nv->data, nv->len);
3056 		if (error == 0 && clear) {
3057 			counter_u64_zero(rule->evaluations);
3058 			for (int i = 0; i < 2; i++) {
3059 				counter_u64_zero(rule->packets[i]);
3060 				counter_u64_zero(rule->bytes[i]);
3061 			}
3062 		}
3063 
3064 #undef ERROUT
3065 DIOCGETETHRULE_error:
3066 		free(nvlpacked, M_NVLIST);
3067 		nvlist_destroy(nvl);
3068 		break;
3069 	}
3070 
3071 	case DIOCADDETHRULE: {
3072 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3073 		nvlist_t		*nvl = NULL;
3074 		void			*nvlpacked = NULL;
3075 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
3076 		struct pf_keth_ruleset	*ruleset = NULL;
3077 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
3078 		const char		*anchor = "", *anchor_call = "";
3079 
3080 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3081 
3082 		if (nv->len > pf_ioctl_maxcount)
3083 			ERROUT(ENOMEM);
3084 
3085 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3086 		error = copyin(nv->data, nvlpacked, nv->len);
3087 		if (error)
3088 			ERROUT(error);
3089 
3090 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3091 		if (nvl == NULL)
3092 			ERROUT(EBADMSG);
3093 
3094 		if (! nvlist_exists_number(nvl, "ticket"))
3095 			ERROUT(EBADMSG);
3096 
3097 		if (nvlist_exists_string(nvl, "anchor"))
3098 			anchor = nvlist_get_string(nvl, "anchor");
3099 		if (nvlist_exists_string(nvl, "anchor_call"))
3100 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3101 
3102 		ruleset = pf_find_keth_ruleset(anchor);
3103 		if (ruleset == NULL)
3104 			ERROUT(EINVAL);
3105 
3106 		if (nvlist_get_number(nvl, "ticket") !=
3107 		    ruleset->inactive.ticket) {
3108 			DPFPRINTF(PF_DEBUG_MISC,
3109 			    ("ticket: %d != %d\n",
3110 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
3111 			    ruleset->inactive.ticket));
3112 			ERROUT(EBUSY);
3113 		}
3114 
3115 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3116 		rule->timestamp = NULL;
3117 
3118 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
3119 		if (error != 0)
3120 			ERROUT(error);
3121 
3122 		if (rule->ifname[0])
3123 			kif = pf_kkif_create(M_WAITOK);
3124 		if (rule->bridge_to_name[0])
3125 			bridge_to_kif = pf_kkif_create(M_WAITOK);
3126 		rule->evaluations = counter_u64_alloc(M_WAITOK);
3127 		for (int i = 0; i < 2; i++) {
3128 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
3129 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3130 		}
3131 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3132 		    M_WAITOK | M_ZERO);
3133 
3134 		PF_RULES_WLOCK();
3135 
3136 		if (rule->ifname[0]) {
3137 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
3138 			pfi_kkif_ref(rule->kif);
3139 		} else
3140 			rule->kif = NULL;
3141 		if (rule->bridge_to_name[0]) {
3142 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3143 			    rule->bridge_to_name);
3144 			pfi_kkif_ref(rule->bridge_to);
3145 		} else
3146 			rule->bridge_to = NULL;
3147 
3148 #ifdef ALTQ
3149 		/* set queue IDs */
3150 		if (rule->qname[0] != 0) {
3151 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3152 				error = EBUSY;
3153 			else
3154 				rule->qid = rule->qid;
3155 		}
3156 #endif
3157 		if (rule->tagname[0])
3158 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3159 				error = EBUSY;
3160 		if (rule->match_tagname[0])
3161 			if ((rule->match_tag = pf_tagname2tag(
3162 			    rule->match_tagname)) == 0)
3163 				error = EBUSY;
3164 
3165 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3166 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3167 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3168 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3169 
3170 		if (error) {
3171 			pf_free_eth_rule(rule);
3172 			PF_RULES_WUNLOCK();
3173 			ERROUT(error);
3174 		}
3175 
3176 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3177 			pf_free_eth_rule(rule);
3178 			PF_RULES_WUNLOCK();
3179 			ERROUT(EINVAL);
3180 		}
3181 
3182 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3183 		if (tail)
3184 			rule->nr = tail->nr + 1;
3185 		else
3186 			rule->nr = 0;
3187 
3188 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3189 
3190 		PF_RULES_WUNLOCK();
3191 
3192 #undef ERROUT
3193 DIOCADDETHRULE_error:
3194 		nvlist_destroy(nvl);
3195 		free(nvlpacked, M_NVLIST);
3196 		break;
3197 	}
3198 
3199 	case DIOCGETETHRULESETS: {
3200 		struct epoch_tracker	 et;
3201 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3202 		nvlist_t		*nvl = NULL;
3203 		void			*nvlpacked = NULL;
3204 		struct pf_keth_ruleset	*ruleset;
3205 		struct pf_keth_anchor	*anchor;
3206 		int			 nr = 0;
3207 
3208 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3209 
3210 		if (nv->len > pf_ioctl_maxcount)
3211 			ERROUT(ENOMEM);
3212 
3213 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3214 		error = copyin(nv->data, nvlpacked, nv->len);
3215 		if (error)
3216 			ERROUT(error);
3217 
3218 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3219 		if (nvl == NULL)
3220 			ERROUT(EBADMSG);
3221 		if (! nvlist_exists_string(nvl, "path"))
3222 			ERROUT(EBADMSG);
3223 
3224 		NET_EPOCH_ENTER(et);
3225 
3226 		if ((ruleset = pf_find_keth_ruleset(
3227 		    nvlist_get_string(nvl, "path"))) == NULL) {
3228 			NET_EPOCH_EXIT(et);
3229 			ERROUT(ENOENT);
3230 		}
3231 
3232 		if (ruleset->anchor == NULL) {
3233 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3234 				if (anchor->parent == NULL)
3235 					nr++;
3236 		} else {
3237 			RB_FOREACH(anchor, pf_keth_anchor_node,
3238 			    &ruleset->anchor->children)
3239 				nr++;
3240 		}
3241 
3242 		NET_EPOCH_EXIT(et);
3243 
3244 		nvlist_destroy(nvl);
3245 		nvl = NULL;
3246 		free(nvlpacked, M_NVLIST);
3247 		nvlpacked = NULL;
3248 
3249 		nvl = nvlist_create(0);
3250 		if (nvl == NULL)
3251 			ERROUT(ENOMEM);
3252 
3253 		nvlist_add_number(nvl, "nr", nr);
3254 
3255 		nvlpacked = nvlist_pack(nvl, &nv->len);
3256 		if (nvlpacked == NULL)
3257 			ERROUT(ENOMEM);
3258 
3259 		if (nv->size == 0)
3260 			ERROUT(0);
3261 		else if (nv->size < nv->len)
3262 			ERROUT(ENOSPC);
3263 
3264 		error = copyout(nvlpacked, nv->data, nv->len);
3265 
3266 #undef ERROUT
3267 DIOCGETETHRULESETS_error:
3268 		free(nvlpacked, M_NVLIST);
3269 		nvlist_destroy(nvl);
3270 		break;
3271 	}
3272 
3273 	case DIOCGETETHRULESET: {
3274 		struct epoch_tracker	 et;
3275 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3276 		nvlist_t		*nvl = NULL;
3277 		void			*nvlpacked = NULL;
3278 		struct pf_keth_ruleset	*ruleset;
3279 		struct pf_keth_anchor	*anchor;
3280 		int			 nr = 0, req_nr = 0;
3281 		bool			 found = false;
3282 
3283 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3284 
3285 		if (nv->len > pf_ioctl_maxcount)
3286 			ERROUT(ENOMEM);
3287 
3288 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3289 		error = copyin(nv->data, nvlpacked, nv->len);
3290 		if (error)
3291 			ERROUT(error);
3292 
3293 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3294 		if (nvl == NULL)
3295 			ERROUT(EBADMSG);
3296 		if (! nvlist_exists_string(nvl, "path"))
3297 			ERROUT(EBADMSG);
3298 		if (! nvlist_exists_number(nvl, "nr"))
3299 			ERROUT(EBADMSG);
3300 
3301 		req_nr = nvlist_get_number(nvl, "nr");
3302 
3303 		NET_EPOCH_ENTER(et);
3304 
3305 		if ((ruleset = pf_find_keth_ruleset(
3306 		    nvlist_get_string(nvl, "path"))) == NULL) {
3307 			NET_EPOCH_EXIT(et);
3308 			ERROUT(ENOENT);
3309 		}
3310 
3311 		nvlist_destroy(nvl);
3312 		nvl = NULL;
3313 		free(nvlpacked, M_NVLIST);
3314 		nvlpacked = NULL;
3315 
3316 		nvl = nvlist_create(0);
3317 		if (nvl == NULL) {
3318 			NET_EPOCH_EXIT(et);
3319 			ERROUT(ENOMEM);
3320 		}
3321 
3322 		if (ruleset->anchor == NULL) {
3323 			RB_FOREACH(anchor, pf_keth_anchor_global,
3324 			    &V_pf_keth_anchors) {
3325 				if (anchor->parent == NULL && nr++ == req_nr) {
3326 					found = true;
3327 					break;
3328 				}
3329 			}
3330 		} else {
3331 			RB_FOREACH(anchor, pf_keth_anchor_node,
3332 			     &ruleset->anchor->children) {
3333 				if (nr++ == req_nr) {
3334 					found = true;
3335 					break;
3336 				}
3337 			}
3338 		}
3339 
3340 		NET_EPOCH_EXIT(et);
3341 		if (found) {
3342 			nvlist_add_number(nvl, "nr", nr);
3343 			nvlist_add_string(nvl, "name", anchor->name);
3344 			if (ruleset->anchor)
3345 				nvlist_add_string(nvl, "path",
3346 				    ruleset->anchor->path);
3347 			else
3348 				nvlist_add_string(nvl, "path", "");
3349 		} else {
3350 			ERROUT(EBUSY);
3351 		}
3352 
3353 		nvlpacked = nvlist_pack(nvl, &nv->len);
3354 		if (nvlpacked == NULL)
3355 			ERROUT(ENOMEM);
3356 
3357 		if (nv->size == 0)
3358 			ERROUT(0);
3359 		else if (nv->size < nv->len)
3360 			ERROUT(ENOSPC);
3361 
3362 		error = copyout(nvlpacked, nv->data, nv->len);
3363 
3364 #undef ERROUT
3365 DIOCGETETHRULESET_error:
3366 		free(nvlpacked, M_NVLIST);
3367 		nvlist_destroy(nvl);
3368 		break;
3369 	}
3370 
3371 	case DIOCADDRULENV: {
3372 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3373 		nvlist_t	*nvl = NULL;
3374 		void		*nvlpacked = NULL;
3375 		struct pf_krule	*rule = NULL;
3376 		const char	*anchor = "", *anchor_call = "";
3377 		uint32_t	 ticket = 0, pool_ticket = 0;
3378 
3379 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3380 
3381 		if (nv->len > pf_ioctl_maxcount)
3382 			ERROUT(ENOMEM);
3383 
3384 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3385 		error = copyin(nv->data, nvlpacked, nv->len);
3386 		if (error)
3387 			ERROUT(error);
3388 
3389 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3390 		if (nvl == NULL)
3391 			ERROUT(EBADMSG);
3392 
3393 		if (! nvlist_exists_number(nvl, "ticket"))
3394 			ERROUT(EINVAL);
3395 		ticket = nvlist_get_number(nvl, "ticket");
3396 
3397 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3398 			ERROUT(EINVAL);
3399 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3400 
3401 		if (! nvlist_exists_nvlist(nvl, "rule"))
3402 			ERROUT(EINVAL);
3403 
3404 		rule = pf_krule_alloc();
3405 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3406 		    rule);
3407 		if (error)
3408 			ERROUT(error);
3409 
3410 		if (nvlist_exists_string(nvl, "anchor"))
3411 			anchor = nvlist_get_string(nvl, "anchor");
3412 		if (nvlist_exists_string(nvl, "anchor_call"))
3413 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3414 
3415 		if ((error = nvlist_error(nvl)))
3416 			ERROUT(error);
3417 
3418 		/* Frees rule on error */
3419 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3420 		    anchor_call, td->td_ucred->cr_ruid,
3421 		    td->td_proc ? td->td_proc->p_pid : 0);
3422 
3423 		nvlist_destroy(nvl);
3424 		free(nvlpacked, M_NVLIST);
3425 		break;
3426 #undef ERROUT
3427 DIOCADDRULENV_error:
3428 		pf_krule_free(rule);
3429 		nvlist_destroy(nvl);
3430 		free(nvlpacked, M_NVLIST);
3431 
3432 		break;
3433 	}
3434 	case DIOCADDRULE: {
3435 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3436 		struct pf_krule		*rule;
3437 
3438 		rule = pf_krule_alloc();
3439 		error = pf_rule_to_krule(&pr->rule, rule);
3440 		if (error != 0) {
3441 			pf_krule_free(rule);
3442 			break;
3443 		}
3444 
3445 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3446 
3447 		/* Frees rule on error */
3448 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3449 		    pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3450 		    td->td_proc ? td->td_proc->p_pid : 0);
3451 		break;
3452 	}
3453 
3454 	case DIOCGETRULES: {
3455 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3456 
3457 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3458 
3459 		error = pf_ioctl_getrules(pr);
3460 
3461 		break;
3462 	}
3463 
3464 	case DIOCGETRULENV: {
3465 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3466 		nvlist_t		*nvrule = NULL;
3467 		nvlist_t		*nvl = NULL;
3468 		struct pf_kruleset	*ruleset;
3469 		struct pf_krule		*rule;
3470 		void			*nvlpacked = NULL;
3471 		int			 rs_num, nr;
3472 		bool			 clear_counter = false;
3473 
3474 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3475 
3476 		if (nv->len > pf_ioctl_maxcount)
3477 			ERROUT(ENOMEM);
3478 
3479 		/* Copy the request in */
3480 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3481 		error = copyin(nv->data, nvlpacked, nv->len);
3482 		if (error)
3483 			ERROUT(error);
3484 
3485 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3486 		if (nvl == NULL)
3487 			ERROUT(EBADMSG);
3488 
3489 		if (! nvlist_exists_string(nvl, "anchor"))
3490 			ERROUT(EBADMSG);
3491 		if (! nvlist_exists_number(nvl, "ruleset"))
3492 			ERROUT(EBADMSG);
3493 		if (! nvlist_exists_number(nvl, "ticket"))
3494 			ERROUT(EBADMSG);
3495 		if (! nvlist_exists_number(nvl, "nr"))
3496 			ERROUT(EBADMSG);
3497 
3498 		if (nvlist_exists_bool(nvl, "clear_counter"))
3499 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3500 
3501 		if (clear_counter && !(flags & FWRITE))
3502 			ERROUT(EACCES);
3503 
3504 		nr = nvlist_get_number(nvl, "nr");
3505 
3506 		PF_RULES_WLOCK();
3507 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3508 		if (ruleset == NULL) {
3509 			PF_RULES_WUNLOCK();
3510 			ERROUT(ENOENT);
3511 		}
3512 
3513 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3514 		if (rs_num >= PF_RULESET_MAX) {
3515 			PF_RULES_WUNLOCK();
3516 			ERROUT(EINVAL);
3517 		}
3518 
3519 		if (nvlist_get_number(nvl, "ticket") !=
3520 		    ruleset->rules[rs_num].active.ticket) {
3521 			PF_RULES_WUNLOCK();
3522 			ERROUT(EBUSY);
3523 		}
3524 
3525 		if ((error = nvlist_error(nvl))) {
3526 			PF_RULES_WUNLOCK();
3527 			ERROUT(error);
3528 		}
3529 
3530 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3531 		while ((rule != NULL) && (rule->nr != nr))
3532 			rule = TAILQ_NEXT(rule, entries);
3533 		if (rule == NULL) {
3534 			PF_RULES_WUNLOCK();
3535 			ERROUT(EBUSY);
3536 		}
3537 
3538 		nvrule = pf_krule_to_nvrule(rule);
3539 
3540 		nvlist_destroy(nvl);
3541 		nvl = nvlist_create(0);
3542 		if (nvl == NULL) {
3543 			PF_RULES_WUNLOCK();
3544 			ERROUT(ENOMEM);
3545 		}
3546 		nvlist_add_number(nvl, "nr", nr);
3547 		nvlist_add_nvlist(nvl, "rule", nvrule);
3548 		nvlist_destroy(nvrule);
3549 		nvrule = NULL;
3550 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3551 			PF_RULES_WUNLOCK();
3552 			ERROUT(EBUSY);
3553 		}
3554 
3555 		free(nvlpacked, M_NVLIST);
3556 		nvlpacked = nvlist_pack(nvl, &nv->len);
3557 		if (nvlpacked == NULL) {
3558 			PF_RULES_WUNLOCK();
3559 			ERROUT(ENOMEM);
3560 		}
3561 
3562 		if (nv->size == 0) {
3563 			PF_RULES_WUNLOCK();
3564 			ERROUT(0);
3565 		}
3566 		else if (nv->size < nv->len) {
3567 			PF_RULES_WUNLOCK();
3568 			ERROUT(ENOSPC);
3569 		}
3570 
3571 		if (clear_counter)
3572 			pf_krule_clear_counters(rule);
3573 
3574 		PF_RULES_WUNLOCK();
3575 
3576 		error = copyout(nvlpacked, nv->data, nv->len);
3577 
3578 #undef ERROUT
3579 DIOCGETRULENV_error:
3580 		free(nvlpacked, M_NVLIST);
3581 		nvlist_destroy(nvrule);
3582 		nvlist_destroy(nvl);
3583 
3584 		break;
3585 	}
3586 
3587 	case DIOCCHANGERULE: {
3588 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3589 		struct pf_kruleset	*ruleset;
3590 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3591 		struct pfi_kkif		*kif = NULL;
3592 		struct pf_kpooladdr	*pa;
3593 		u_int32_t		 nr = 0;
3594 		int			 rs_num;
3595 
3596 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3597 
3598 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3599 		    pcr->action > PF_CHANGE_GET_TICKET) {
3600 			error = EINVAL;
3601 			break;
3602 		}
3603 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3604 			error = EINVAL;
3605 			break;
3606 		}
3607 
3608 		if (pcr->action != PF_CHANGE_REMOVE) {
3609 			newrule = pf_krule_alloc();
3610 			error = pf_rule_to_krule(&pcr->rule, newrule);
3611 			if (error != 0) {
3612 				pf_krule_free(newrule);
3613 				break;
3614 			}
3615 
3616 			if (newrule->ifname[0])
3617 				kif = pf_kkif_create(M_WAITOK);
3618 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3619 			for (int i = 0; i < 2; i++) {
3620 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3621 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3622 			}
3623 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3624 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3625 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3626 			newrule->cuid = td->td_ucred->cr_ruid;
3627 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3628 			TAILQ_INIT(&newrule->rpool.list);
3629 		}
3630 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3631 
3632 		PF_CONFIG_LOCK();
3633 		PF_RULES_WLOCK();
3634 #ifdef PF_WANT_32_TO_64_COUNTER
3635 		if (newrule != NULL) {
3636 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3637 			newrule->allrulelinked = true;
3638 			V_pf_allrulecount++;
3639 		}
3640 #endif
3641 
3642 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3643 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3644 		    pcr->pool_ticket != V_ticket_pabuf)
3645 			ERROUT(EBUSY);
3646 
3647 		ruleset = pf_find_kruleset(pcr->anchor);
3648 		if (ruleset == NULL)
3649 			ERROUT(EINVAL);
3650 
3651 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3652 		if (rs_num >= PF_RULESET_MAX)
3653 			ERROUT(EINVAL);
3654 
3655 		/*
3656 		 * XXXMJG: there is no guarantee that the ruleset was
3657 		 * created by the usual route of calling DIOCXBEGIN.
3658 		 * As a result it is possible the rule tree will not
3659 		 * be allocated yet. Hack around it by doing it here.
3660 		 * Note it is fine to let the tree persist in case of
3661 		 * error as it will be freed down the road on future
3662 		 * updates (if need be).
3663 		 */
3664 		if (ruleset->rules[rs_num].active.tree == NULL) {
3665 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3666 			if (ruleset->rules[rs_num].active.tree == NULL) {
3667 				ERROUT(ENOMEM);
3668 			}
3669 		}
3670 
3671 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3672 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3673 			ERROUT(0);
3674 		} else if (pcr->ticket !=
3675 			    ruleset->rules[rs_num].active.ticket)
3676 				ERROUT(EINVAL);
3677 
3678 		if (pcr->action != PF_CHANGE_REMOVE) {
3679 			if (newrule->ifname[0]) {
3680 				newrule->kif = pfi_kkif_attach(kif,
3681 				    newrule->ifname);
3682 				kif = NULL;
3683 				pfi_kkif_ref(newrule->kif);
3684 			} else
3685 				newrule->kif = NULL;
3686 
3687 			if (newrule->rtableid > 0 &&
3688 			    newrule->rtableid >= rt_numfibs)
3689 				error = EBUSY;
3690 
3691 #ifdef ALTQ
3692 			/* set queue IDs */
3693 			if (newrule->qname[0] != 0) {
3694 				if ((newrule->qid =
3695 				    pf_qname2qid(newrule->qname)) == 0)
3696 					error = EBUSY;
3697 				else if (newrule->pqname[0] != 0) {
3698 					if ((newrule->pqid =
3699 					    pf_qname2qid(newrule->pqname)) == 0)
3700 						error = EBUSY;
3701 				} else
3702 					newrule->pqid = newrule->qid;
3703 			}
3704 #endif /* ALTQ */
3705 			if (newrule->tagname[0])
3706 				if ((newrule->tag =
3707 				    pf_tagname2tag(newrule->tagname)) == 0)
3708 					error = EBUSY;
3709 			if (newrule->match_tagname[0])
3710 				if ((newrule->match_tag = pf_tagname2tag(
3711 				    newrule->match_tagname)) == 0)
3712 					error = EBUSY;
3713 			if (newrule->rt && !newrule->direction)
3714 				error = EINVAL;
3715 			if (!newrule->log)
3716 				newrule->logif = 0;
3717 			if (newrule->logif >= PFLOGIFS_MAX)
3718 				error = EINVAL;
3719 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3720 				error = ENOMEM;
3721 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3722 				error = ENOMEM;
3723 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3724 				error = EINVAL;
3725 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3726 				if (pa->addr.type == PF_ADDR_TABLE) {
3727 					pa->addr.p.tbl =
3728 					    pfr_attach_table(ruleset,
3729 					    pa->addr.v.tblname);
3730 					if (pa->addr.p.tbl == NULL)
3731 						error = ENOMEM;
3732 				}
3733 
3734 			newrule->overload_tbl = NULL;
3735 			if (newrule->overload_tblname[0]) {
3736 				if ((newrule->overload_tbl = pfr_attach_table(
3737 				    ruleset, newrule->overload_tblname)) ==
3738 				    NULL)
3739 					error = EINVAL;
3740 				else
3741 					newrule->overload_tbl->pfrkt_flags |=
3742 					    PFR_TFLAG_ACTIVE;
3743 			}
3744 
3745 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3746 			if (((((newrule->action == PF_NAT) ||
3747 			    (newrule->action == PF_RDR) ||
3748 			    (newrule->action == PF_BINAT) ||
3749 			    (newrule->rt > PF_NOPFROUTE)) &&
3750 			    !newrule->anchor)) &&
3751 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3752 				error = EINVAL;
3753 
3754 			if (error) {
3755 				pf_free_rule(newrule);
3756 				PF_RULES_WUNLOCK();
3757 				PF_CONFIG_UNLOCK();
3758 				break;
3759 			}
3760 
3761 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3762 		}
3763 		pf_empty_kpool(&V_pf_pabuf);
3764 
3765 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3766 			oldrule = TAILQ_FIRST(
3767 			    ruleset->rules[rs_num].active.ptr);
3768 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3769 			oldrule = TAILQ_LAST(
3770 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3771 		else {
3772 			oldrule = TAILQ_FIRST(
3773 			    ruleset->rules[rs_num].active.ptr);
3774 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3775 				oldrule = TAILQ_NEXT(oldrule, entries);
3776 			if (oldrule == NULL) {
3777 				if (newrule != NULL)
3778 					pf_free_rule(newrule);
3779 				PF_RULES_WUNLOCK();
3780 				PF_CONFIG_UNLOCK();
3781 				error = EINVAL;
3782 				break;
3783 			}
3784 		}
3785 
3786 		if (pcr->action == PF_CHANGE_REMOVE) {
3787 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3788 			    oldrule);
3789 			RB_REMOVE(pf_krule_global,
3790 			    ruleset->rules[rs_num].active.tree, oldrule);
3791 			ruleset->rules[rs_num].active.rcount--;
3792 		} else {
3793 			pf_hash_rule(newrule);
3794 			if (RB_INSERT(pf_krule_global,
3795 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3796 				pf_free_rule(newrule);
3797 				PF_RULES_WUNLOCK();
3798 				PF_CONFIG_UNLOCK();
3799 				error = EEXIST;
3800 				break;
3801 			}
3802 
3803 			if (oldrule == NULL)
3804 				TAILQ_INSERT_TAIL(
3805 				    ruleset->rules[rs_num].active.ptr,
3806 				    newrule, entries);
3807 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3808 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3809 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3810 			else
3811 				TAILQ_INSERT_AFTER(
3812 				    ruleset->rules[rs_num].active.ptr,
3813 				    oldrule, newrule, entries);
3814 			ruleset->rules[rs_num].active.rcount++;
3815 		}
3816 
3817 		nr = 0;
3818 		TAILQ_FOREACH(oldrule,
3819 		    ruleset->rules[rs_num].active.ptr, entries)
3820 			oldrule->nr = nr++;
3821 
3822 		ruleset->rules[rs_num].active.ticket++;
3823 
3824 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3825 		pf_remove_if_empty_kruleset(ruleset);
3826 
3827 		PF_RULES_WUNLOCK();
3828 		PF_CONFIG_UNLOCK();
3829 		break;
3830 
3831 #undef ERROUT
3832 DIOCCHANGERULE_error:
3833 		PF_RULES_WUNLOCK();
3834 		PF_CONFIG_UNLOCK();
3835 		pf_krule_free(newrule);
3836 		pf_kkif_free(kif);
3837 		break;
3838 	}
3839 
3840 	case DIOCCLRSTATESNV: {
3841 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3842 		break;
3843 	}
3844 
3845 	case DIOCKILLSTATESNV: {
3846 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3847 		break;
3848 	}
3849 
3850 	case DIOCADDSTATE: {
3851 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3852 		struct pfsync_state_1301	*sp = &ps->state;
3853 
3854 		if (sp->timeout >= PFTM_MAX) {
3855 			error = EINVAL;
3856 			break;
3857 		}
3858 		if (V_pfsync_state_import_ptr != NULL) {
3859 			PF_RULES_RLOCK();
3860 			error = V_pfsync_state_import_ptr(
3861 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3862 			    PFSYNC_MSG_VERSION_1301);
3863 			PF_RULES_RUNLOCK();
3864 		} else
3865 			error = EOPNOTSUPP;
3866 		break;
3867 	}
3868 
3869 	case DIOCGETSTATE: {
3870 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3871 		struct pf_kstate	*s;
3872 
3873 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3874 		if (s == NULL) {
3875 			error = ENOENT;
3876 			break;
3877 		}
3878 
3879 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3880 		    s, PFSYNC_MSG_VERSION_1301);
3881 		PF_STATE_UNLOCK(s);
3882 		break;
3883 	}
3884 
3885 	case DIOCGETSTATENV: {
3886 		error = pf_getstate((struct pfioc_nv *)addr);
3887 		break;
3888 	}
3889 
3890 #ifdef COMPAT_FREEBSD14
3891 	case DIOCGETSTATES: {
3892 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3893 		struct pf_kstate	*s;
3894 		struct pfsync_state_1301	*pstore, *p;
3895 		int			 i, nr;
3896 		size_t			 slice_count = 16, count;
3897 		void			*out;
3898 
3899 		if (ps->ps_len <= 0) {
3900 			nr = uma_zone_get_cur(V_pf_state_z);
3901 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3902 			break;
3903 		}
3904 
3905 		out = ps->ps_states;
3906 		pstore = mallocarray(slice_count,
3907 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3908 		nr = 0;
3909 
3910 		for (i = 0; i <= V_pf_hashmask; i++) {
3911 			struct pf_idhash *ih = &V_pf_idhash[i];
3912 
3913 DIOCGETSTATES_retry:
3914 			p = pstore;
3915 
3916 			if (LIST_EMPTY(&ih->states))
3917 				continue;
3918 
3919 			PF_HASHROW_LOCK(ih);
3920 			count = 0;
3921 			LIST_FOREACH(s, &ih->states, entry) {
3922 				if (s->timeout == PFTM_UNLINKED)
3923 					continue;
3924 				count++;
3925 			}
3926 
3927 			if (count > slice_count) {
3928 				PF_HASHROW_UNLOCK(ih);
3929 				free(pstore, M_TEMP);
3930 				slice_count = count * 2;
3931 				pstore = mallocarray(slice_count,
3932 				    sizeof(struct pfsync_state_1301), M_TEMP,
3933 				    M_WAITOK | M_ZERO);
3934 				goto DIOCGETSTATES_retry;
3935 			}
3936 
3937 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3938 				PF_HASHROW_UNLOCK(ih);
3939 				goto DIOCGETSTATES_full;
3940 			}
3941 
3942 			LIST_FOREACH(s, &ih->states, entry) {
3943 				if (s->timeout == PFTM_UNLINKED)
3944 					continue;
3945 
3946 				pfsync_state_export((union pfsync_state_union*)p,
3947 				    s, PFSYNC_MSG_VERSION_1301);
3948 				p++;
3949 				nr++;
3950 			}
3951 			PF_HASHROW_UNLOCK(ih);
3952 			error = copyout(pstore, out,
3953 			    sizeof(struct pfsync_state_1301) * count);
3954 			if (error)
3955 				break;
3956 			out = ps->ps_states + nr;
3957 		}
3958 DIOCGETSTATES_full:
3959 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3960 		free(pstore, M_TEMP);
3961 
3962 		break;
3963 	}
3964 
3965 	case DIOCGETSTATESV2: {
3966 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3967 		struct pf_kstate	*s;
3968 		struct pf_state_export	*pstore, *p;
3969 		int i, nr;
3970 		size_t slice_count = 16, count;
3971 		void *out;
3972 
3973 		if (ps->ps_req_version > PF_STATE_VERSION) {
3974 			error = ENOTSUP;
3975 			break;
3976 		}
3977 
3978 		if (ps->ps_len <= 0) {
3979 			nr = uma_zone_get_cur(V_pf_state_z);
3980 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3981 			break;
3982 		}
3983 
3984 		out = ps->ps_states;
3985 		pstore = mallocarray(slice_count,
3986 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3987 		nr = 0;
3988 
3989 		for (i = 0; i <= V_pf_hashmask; i++) {
3990 			struct pf_idhash *ih = &V_pf_idhash[i];
3991 
3992 DIOCGETSTATESV2_retry:
3993 			p = pstore;
3994 
3995 			if (LIST_EMPTY(&ih->states))
3996 				continue;
3997 
3998 			PF_HASHROW_LOCK(ih);
3999 			count = 0;
4000 			LIST_FOREACH(s, &ih->states, entry) {
4001 				if (s->timeout == PFTM_UNLINKED)
4002 					continue;
4003 				count++;
4004 			}
4005 
4006 			if (count > slice_count) {
4007 				PF_HASHROW_UNLOCK(ih);
4008 				free(pstore, M_TEMP);
4009 				slice_count = count * 2;
4010 				pstore = mallocarray(slice_count,
4011 				    sizeof(struct pf_state_export), M_TEMP,
4012 				    M_WAITOK | M_ZERO);
4013 				goto DIOCGETSTATESV2_retry;
4014 			}
4015 
4016 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
4017 				PF_HASHROW_UNLOCK(ih);
4018 				goto DIOCGETSTATESV2_full;
4019 			}
4020 
4021 			LIST_FOREACH(s, &ih->states, entry) {
4022 				if (s->timeout == PFTM_UNLINKED)
4023 					continue;
4024 
4025 				pf_state_export(p, s);
4026 				p++;
4027 				nr++;
4028 			}
4029 			PF_HASHROW_UNLOCK(ih);
4030 			error = copyout(pstore, out,
4031 			    sizeof(struct pf_state_export) * count);
4032 			if (error)
4033 				break;
4034 			out = ps->ps_states + nr;
4035 		}
4036 DIOCGETSTATESV2_full:
4037 		ps->ps_len = nr * sizeof(struct pf_state_export);
4038 		free(pstore, M_TEMP);
4039 
4040 		break;
4041 	}
4042 #endif
4043 	case DIOCGETSTATUSNV: {
4044 		error = pf_getstatus((struct pfioc_nv *)addr);
4045 		break;
4046 	}
4047 
4048 	case DIOCSETSTATUSIF: {
4049 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
4050 
4051 		if (pi->ifname[0] == 0) {
4052 			bzero(V_pf_status.ifname, IFNAMSIZ);
4053 			break;
4054 		}
4055 		PF_RULES_WLOCK();
4056 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
4057 		PF_RULES_WUNLOCK();
4058 		break;
4059 	}
4060 
4061 	case DIOCCLRSTATUS: {
4062 		pf_ioctl_clear_status();
4063 		break;
4064 	}
4065 
4066 	case DIOCNATLOOK: {
4067 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
4068 		struct pf_state_key	*sk;
4069 		struct pf_kstate	*state;
4070 		struct pf_state_key_cmp	 key;
4071 		int			 m = 0, direction = pnl->direction;
4072 		int			 sidx, didx;
4073 
4074 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
4075 		sidx = (direction == PF_IN) ? 1 : 0;
4076 		didx = (direction == PF_IN) ? 0 : 1;
4077 
4078 		if (!pnl->proto ||
4079 		    PF_AZERO(&pnl->saddr, pnl->af) ||
4080 		    PF_AZERO(&pnl->daddr, pnl->af) ||
4081 		    ((pnl->proto == IPPROTO_TCP ||
4082 		    pnl->proto == IPPROTO_UDP) &&
4083 		    (!pnl->dport || !pnl->sport)))
4084 			error = EINVAL;
4085 		else {
4086 			bzero(&key, sizeof(key));
4087 			key.af = pnl->af;
4088 			key.proto = pnl->proto;
4089 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4090 			key.port[sidx] = pnl->sport;
4091 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4092 			key.port[didx] = pnl->dport;
4093 
4094 			state = pf_find_state_all(&key, direction, &m);
4095 			if (state == NULL) {
4096 				error = ENOENT;
4097 			} else {
4098 				if (m > 1) {
4099 					PF_STATE_UNLOCK(state);
4100 					error = E2BIG;	/* more than one state */
4101 				} else {
4102 					sk = state->key[sidx];
4103 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4104 					pnl->rsport = sk->port[sidx];
4105 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4106 					pnl->rdport = sk->port[didx];
4107 					PF_STATE_UNLOCK(state);
4108 				}
4109 			}
4110 		}
4111 		break;
4112 	}
4113 
4114 	case DIOCSETTIMEOUT: {
4115 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4116 
4117 		error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4118 		    &pt->seconds);
4119 		break;
4120 	}
4121 
4122 	case DIOCGETTIMEOUT: {
4123 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4124 
4125 		error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4126 		break;
4127 	}
4128 
4129 	case DIOCGETLIMIT: {
4130 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4131 
4132 		error = pf_ioctl_get_limit(pl->index, &pl->limit);
4133 		break;
4134 	}
4135 
4136 	case DIOCSETLIMIT: {
4137 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4138 		unsigned int old_limit;
4139 
4140 		error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4141 		pl->limit = old_limit;
4142 		break;
4143 	}
4144 
4145 	case DIOCSETDEBUG: {
4146 		u_int32_t	*level = (u_int32_t *)addr;
4147 
4148 		PF_RULES_WLOCK();
4149 		V_pf_status.debug = *level;
4150 		PF_RULES_WUNLOCK();
4151 		break;
4152 	}
4153 
4154 	case DIOCCLRRULECTRS: {
4155 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4156 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
4157 		struct pf_krule		*rule;
4158 
4159 		PF_RULES_WLOCK();
4160 		TAILQ_FOREACH(rule,
4161 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4162 			pf_counter_u64_zero(&rule->evaluations);
4163 			for (int i = 0; i < 2; i++) {
4164 				pf_counter_u64_zero(&rule->packets[i]);
4165 				pf_counter_u64_zero(&rule->bytes[i]);
4166 			}
4167 		}
4168 		PF_RULES_WUNLOCK();
4169 		break;
4170 	}
4171 
4172 	case DIOCGIFSPEEDV0:
4173 	case DIOCGIFSPEEDV1: {
4174 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
4175 		struct pf_ifspeed_v1	ps;
4176 		struct ifnet		*ifp;
4177 
4178 		if (psp->ifname[0] == '\0') {
4179 			error = EINVAL;
4180 			break;
4181 		}
4182 
4183 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4184 		if (error != 0)
4185 			break;
4186 		ifp = ifunit(ps.ifname);
4187 		if (ifp != NULL) {
4188 			psp->baudrate32 =
4189 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4190 			if (cmd == DIOCGIFSPEEDV1)
4191 				psp->baudrate = ifp->if_baudrate;
4192 		} else {
4193 			error = EINVAL;
4194 		}
4195 		break;
4196 	}
4197 
4198 #ifdef ALTQ
4199 	case DIOCSTARTALTQ: {
4200 		struct pf_altq		*altq;
4201 
4202 		PF_RULES_WLOCK();
4203 		/* enable all altq interfaces on active list */
4204 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4205 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4206 				error = pf_enable_altq(altq);
4207 				if (error != 0)
4208 					break;
4209 			}
4210 		}
4211 		if (error == 0)
4212 			V_pf_altq_running = 1;
4213 		PF_RULES_WUNLOCK();
4214 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4215 		break;
4216 	}
4217 
4218 	case DIOCSTOPALTQ: {
4219 		struct pf_altq		*altq;
4220 
4221 		PF_RULES_WLOCK();
4222 		/* disable all altq interfaces on active list */
4223 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4224 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4225 				error = pf_disable_altq(altq);
4226 				if (error != 0)
4227 					break;
4228 			}
4229 		}
4230 		if (error == 0)
4231 			V_pf_altq_running = 0;
4232 		PF_RULES_WUNLOCK();
4233 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4234 		break;
4235 	}
4236 
4237 	case DIOCADDALTQV0:
4238 	case DIOCADDALTQV1: {
4239 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4240 		struct pf_altq		*altq, *a;
4241 		struct ifnet		*ifp;
4242 
4243 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4244 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4245 		if (error)
4246 			break;
4247 		altq->local_flags = 0;
4248 
4249 		PF_RULES_WLOCK();
4250 		if (pa->ticket != V_ticket_altqs_inactive) {
4251 			PF_RULES_WUNLOCK();
4252 			free(altq, M_PFALTQ);
4253 			error = EBUSY;
4254 			break;
4255 		}
4256 
4257 		/*
4258 		 * if this is for a queue, find the discipline and
4259 		 * copy the necessary fields
4260 		 */
4261 		if (altq->qname[0] != 0) {
4262 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4263 				PF_RULES_WUNLOCK();
4264 				error = EBUSY;
4265 				free(altq, M_PFALTQ);
4266 				break;
4267 			}
4268 			altq->altq_disc = NULL;
4269 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4270 				if (strncmp(a->ifname, altq->ifname,
4271 				    IFNAMSIZ) == 0) {
4272 					altq->altq_disc = a->altq_disc;
4273 					break;
4274 				}
4275 			}
4276 		}
4277 
4278 		if ((ifp = ifunit(altq->ifname)) == NULL)
4279 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4280 		else
4281 			error = altq_add(ifp, altq);
4282 
4283 		if (error) {
4284 			PF_RULES_WUNLOCK();
4285 			free(altq, M_PFALTQ);
4286 			break;
4287 		}
4288 
4289 		if (altq->qname[0] != 0)
4290 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4291 		else
4292 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4293 		/* version error check done on import above */
4294 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4295 		PF_RULES_WUNLOCK();
4296 		break;
4297 	}
4298 
4299 	case DIOCGETALTQSV0:
4300 	case DIOCGETALTQSV1: {
4301 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4302 		struct pf_altq		*altq;
4303 
4304 		PF_RULES_RLOCK();
4305 		pa->nr = 0;
4306 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4307 			pa->nr++;
4308 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4309 			pa->nr++;
4310 		pa->ticket = V_ticket_altqs_active;
4311 		PF_RULES_RUNLOCK();
4312 		break;
4313 	}
4314 
4315 	case DIOCGETALTQV0:
4316 	case DIOCGETALTQV1: {
4317 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4318 		struct pf_altq		*altq;
4319 
4320 		PF_RULES_RLOCK();
4321 		if (pa->ticket != V_ticket_altqs_active) {
4322 			PF_RULES_RUNLOCK();
4323 			error = EBUSY;
4324 			break;
4325 		}
4326 		altq = pf_altq_get_nth_active(pa->nr);
4327 		if (altq == NULL) {
4328 			PF_RULES_RUNLOCK();
4329 			error = EBUSY;
4330 			break;
4331 		}
4332 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4333 		PF_RULES_RUNLOCK();
4334 		break;
4335 	}
4336 
4337 	case DIOCCHANGEALTQV0:
4338 	case DIOCCHANGEALTQV1:
4339 		/* CHANGEALTQ not supported yet! */
4340 		error = ENODEV;
4341 		break;
4342 
4343 	case DIOCGETQSTATSV0:
4344 	case DIOCGETQSTATSV1: {
4345 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4346 		struct pf_altq		*altq;
4347 		int			 nbytes;
4348 		u_int32_t		 version;
4349 
4350 		PF_RULES_RLOCK();
4351 		if (pq->ticket != V_ticket_altqs_active) {
4352 			PF_RULES_RUNLOCK();
4353 			error = EBUSY;
4354 			break;
4355 		}
4356 		nbytes = pq->nbytes;
4357 		altq = pf_altq_get_nth_active(pq->nr);
4358 		if (altq == NULL) {
4359 			PF_RULES_RUNLOCK();
4360 			error = EBUSY;
4361 			break;
4362 		}
4363 
4364 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4365 			PF_RULES_RUNLOCK();
4366 			error = ENXIO;
4367 			break;
4368 		}
4369 		PF_RULES_RUNLOCK();
4370 		if (cmd == DIOCGETQSTATSV0)
4371 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4372 		else
4373 			version = pq->version;
4374 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4375 		if (error == 0) {
4376 			pq->scheduler = altq->scheduler;
4377 			pq->nbytes = nbytes;
4378 		}
4379 		break;
4380 	}
4381 #endif /* ALTQ */
4382 
4383 	case DIOCBEGINADDRS: {
4384 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4385 
4386 		error = pf_ioctl_begin_addrs(&pp->ticket);
4387 		break;
4388 	}
4389 
4390 	case DIOCADDADDR: {
4391 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4392 
4393 		error = pf_ioctl_add_addr(pp);
4394 		break;
4395 	}
4396 
4397 	case DIOCGETADDRS: {
4398 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4399 
4400 		error = pf_ioctl_get_addrs(pp);
4401 		break;
4402 	}
4403 
4404 	case DIOCGETADDR: {
4405 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4406 
4407 		error = pf_ioctl_get_addr(pp);
4408 		break;
4409 	}
4410 
4411 	case DIOCCHANGEADDR: {
4412 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4413 		struct pf_kpool		*pool;
4414 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4415 		struct pf_kruleset	*ruleset;
4416 		struct pfi_kkif		*kif = NULL;
4417 
4418 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4419 
4420 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4421 		    pca->action > PF_CHANGE_REMOVE) {
4422 			error = EINVAL;
4423 			break;
4424 		}
4425 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4426 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4427 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4428 			error = EINVAL;
4429 			break;
4430 		}
4431 		if (pca->addr.addr.p.dyn != NULL) {
4432 			error = EINVAL;
4433 			break;
4434 		}
4435 
4436 		if (pca->action != PF_CHANGE_REMOVE) {
4437 #ifndef INET
4438 			if (pca->af == AF_INET) {
4439 				error = EAFNOSUPPORT;
4440 				break;
4441 			}
4442 #endif /* INET */
4443 #ifndef INET6
4444 			if (pca->af == AF_INET6) {
4445 				error = EAFNOSUPPORT;
4446 				break;
4447 			}
4448 #endif /* INET6 */
4449 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4450 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4451 			if (newpa->ifname[0])
4452 				kif = pf_kkif_create(M_WAITOK);
4453 			newpa->kif = NULL;
4454 		}
4455 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4456 		PF_RULES_WLOCK();
4457 		ruleset = pf_find_kruleset(pca->anchor);
4458 		if (ruleset == NULL)
4459 			ERROUT(EBUSY);
4460 
4461 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4462 		    pca->r_num, pca->r_last, 1, 1);
4463 		if (pool == NULL)
4464 			ERROUT(EBUSY);
4465 
4466 		if (pca->action != PF_CHANGE_REMOVE) {
4467 			if (newpa->ifname[0]) {
4468 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4469 				pfi_kkif_ref(newpa->kif);
4470 				kif = NULL;
4471 			}
4472 
4473 			switch (newpa->addr.type) {
4474 			case PF_ADDR_DYNIFTL:
4475 				error = pfi_dynaddr_setup(&newpa->addr,
4476 				    pca->af);
4477 				break;
4478 			case PF_ADDR_TABLE:
4479 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4480 				    newpa->addr.v.tblname);
4481 				if (newpa->addr.p.tbl == NULL)
4482 					error = ENOMEM;
4483 				break;
4484 			}
4485 			if (error)
4486 				goto DIOCCHANGEADDR_error;
4487 		}
4488 
4489 		switch (pca->action) {
4490 		case PF_CHANGE_ADD_HEAD:
4491 			oldpa = TAILQ_FIRST(&pool->list);
4492 			break;
4493 		case PF_CHANGE_ADD_TAIL:
4494 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4495 			break;
4496 		default:
4497 			oldpa = TAILQ_FIRST(&pool->list);
4498 			for (int i = 0; oldpa && i < pca->nr; i++)
4499 				oldpa = TAILQ_NEXT(oldpa, entries);
4500 
4501 			if (oldpa == NULL)
4502 				ERROUT(EINVAL);
4503 		}
4504 
4505 		if (pca->action == PF_CHANGE_REMOVE) {
4506 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4507 			switch (oldpa->addr.type) {
4508 			case PF_ADDR_DYNIFTL:
4509 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4510 				break;
4511 			case PF_ADDR_TABLE:
4512 				pfr_detach_table(oldpa->addr.p.tbl);
4513 				break;
4514 			}
4515 			if (oldpa->kif)
4516 				pfi_kkif_unref(oldpa->kif);
4517 			free(oldpa, M_PFRULE);
4518 		} else {
4519 			if (oldpa == NULL)
4520 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4521 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4522 			    pca->action == PF_CHANGE_ADD_BEFORE)
4523 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4524 			else
4525 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4526 				    newpa, entries);
4527 		}
4528 
4529 		pool->cur = TAILQ_FIRST(&pool->list);
4530 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4531 		PF_RULES_WUNLOCK();
4532 		break;
4533 
4534 #undef ERROUT
4535 DIOCCHANGEADDR_error:
4536 		if (newpa != NULL) {
4537 			if (newpa->kif)
4538 				pfi_kkif_unref(newpa->kif);
4539 			free(newpa, M_PFRULE);
4540 		}
4541 		PF_RULES_WUNLOCK();
4542 		pf_kkif_free(kif);
4543 		break;
4544 	}
4545 
4546 	case DIOCGETRULESETS: {
4547 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4548 
4549 		pr->path[sizeof(pr->path) - 1] = 0;
4550 
4551 		error = pf_ioctl_get_rulesets(pr);
4552 		break;
4553 	}
4554 
4555 	case DIOCGETRULESET: {
4556 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4557 
4558 		pr->path[sizeof(pr->path) - 1] = 0;
4559 
4560 		error = pf_ioctl_get_ruleset(pr);
4561 		break;
4562 	}
4563 
4564 	case DIOCRCLRTABLES: {
4565 		struct pfioc_table *io = (struct pfioc_table *)addr;
4566 
4567 		if (io->pfrio_esize != 0) {
4568 			error = ENODEV;
4569 			break;
4570 		}
4571 		PF_RULES_WLOCK();
4572 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4573 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4574 		PF_RULES_WUNLOCK();
4575 		break;
4576 	}
4577 
4578 	case DIOCRADDTABLES: {
4579 		struct pfioc_table *io = (struct pfioc_table *)addr;
4580 		struct pfr_table *pfrts;
4581 		size_t totlen;
4582 
4583 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4584 			error = ENODEV;
4585 			break;
4586 		}
4587 
4588 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4589 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4590 			error = ENOMEM;
4591 			break;
4592 		}
4593 
4594 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4595 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4596 		    M_TEMP, M_WAITOK);
4597 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4598 		if (error) {
4599 			free(pfrts, M_TEMP);
4600 			break;
4601 		}
4602 		PF_RULES_WLOCK();
4603 		error = pfr_add_tables(pfrts, io->pfrio_size,
4604 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4605 		PF_RULES_WUNLOCK();
4606 		free(pfrts, M_TEMP);
4607 		break;
4608 	}
4609 
4610 	case DIOCRDELTABLES: {
4611 		struct pfioc_table *io = (struct pfioc_table *)addr;
4612 		struct pfr_table *pfrts;
4613 		size_t totlen;
4614 
4615 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4616 			error = ENODEV;
4617 			break;
4618 		}
4619 
4620 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4621 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4622 			error = ENOMEM;
4623 			break;
4624 		}
4625 
4626 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4627 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4628 		    M_TEMP, M_WAITOK);
4629 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4630 		if (error) {
4631 			free(pfrts, M_TEMP);
4632 			break;
4633 		}
4634 		PF_RULES_WLOCK();
4635 		error = pfr_del_tables(pfrts, io->pfrio_size,
4636 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4637 		PF_RULES_WUNLOCK();
4638 		free(pfrts, M_TEMP);
4639 		break;
4640 	}
4641 
4642 	case DIOCRGETTABLES: {
4643 		struct pfioc_table *io = (struct pfioc_table *)addr;
4644 		struct pfr_table *pfrts;
4645 		size_t totlen;
4646 		int n;
4647 
4648 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4649 			error = ENODEV;
4650 			break;
4651 		}
4652 		PF_RULES_RLOCK();
4653 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4654 		if (n < 0) {
4655 			PF_RULES_RUNLOCK();
4656 			error = EINVAL;
4657 			break;
4658 		}
4659 		io->pfrio_size = min(io->pfrio_size, n);
4660 
4661 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4662 
4663 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4664 		    M_TEMP, M_NOWAIT | M_ZERO);
4665 		if (pfrts == NULL) {
4666 			error = ENOMEM;
4667 			PF_RULES_RUNLOCK();
4668 			break;
4669 		}
4670 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4671 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4672 		PF_RULES_RUNLOCK();
4673 		if (error == 0)
4674 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4675 		free(pfrts, M_TEMP);
4676 		break;
4677 	}
4678 
4679 	case DIOCRGETTSTATS: {
4680 		struct pfioc_table *io = (struct pfioc_table *)addr;
4681 		struct pfr_tstats *pfrtstats;
4682 		size_t totlen;
4683 		int n;
4684 
4685 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4686 			error = ENODEV;
4687 			break;
4688 		}
4689 		PF_TABLE_STATS_LOCK();
4690 		PF_RULES_RLOCK();
4691 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4692 		if (n < 0) {
4693 			PF_RULES_RUNLOCK();
4694 			PF_TABLE_STATS_UNLOCK();
4695 			error = EINVAL;
4696 			break;
4697 		}
4698 		io->pfrio_size = min(io->pfrio_size, n);
4699 
4700 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4701 		pfrtstats = mallocarray(io->pfrio_size,
4702 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4703 		if (pfrtstats == NULL) {
4704 			error = ENOMEM;
4705 			PF_RULES_RUNLOCK();
4706 			PF_TABLE_STATS_UNLOCK();
4707 			break;
4708 		}
4709 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4710 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4711 		PF_RULES_RUNLOCK();
4712 		PF_TABLE_STATS_UNLOCK();
4713 		if (error == 0)
4714 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4715 		free(pfrtstats, M_TEMP);
4716 		break;
4717 	}
4718 
4719 	case DIOCRCLRTSTATS: {
4720 		struct pfioc_table *io = (struct pfioc_table *)addr;
4721 		struct pfr_table *pfrts;
4722 		size_t totlen;
4723 
4724 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4725 			error = ENODEV;
4726 			break;
4727 		}
4728 
4729 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4730 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4731 			/* We used to count tables and use the minimum required
4732 			 * size, so we didn't fail on overly large requests.
4733 			 * Keep doing so. */
4734 			io->pfrio_size = pf_ioctl_maxcount;
4735 			break;
4736 		}
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 
4747 		PF_TABLE_STATS_LOCK();
4748 		PF_RULES_RLOCK();
4749 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4750 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4751 		PF_RULES_RUNLOCK();
4752 		PF_TABLE_STATS_UNLOCK();
4753 		free(pfrts, M_TEMP);
4754 		break;
4755 	}
4756 
4757 	case DIOCRSETTFLAGS: {
4758 		struct pfioc_table *io = (struct pfioc_table *)addr;
4759 		struct pfr_table *pfrts;
4760 		size_t totlen;
4761 		int n;
4762 
4763 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4764 			error = ENODEV;
4765 			break;
4766 		}
4767 
4768 		PF_RULES_RLOCK();
4769 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4770 		if (n < 0) {
4771 			PF_RULES_RUNLOCK();
4772 			error = EINVAL;
4773 			break;
4774 		}
4775 
4776 		io->pfrio_size = min(io->pfrio_size, n);
4777 		PF_RULES_RUNLOCK();
4778 
4779 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4780 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4781 		    M_TEMP, M_WAITOK);
4782 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4783 		if (error) {
4784 			free(pfrts, M_TEMP);
4785 			break;
4786 		}
4787 		PF_RULES_WLOCK();
4788 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4789 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4790 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4791 		PF_RULES_WUNLOCK();
4792 		free(pfrts, M_TEMP);
4793 		break;
4794 	}
4795 
4796 	case DIOCRCLRADDRS: {
4797 		struct pfioc_table *io = (struct pfioc_table *)addr;
4798 
4799 		if (io->pfrio_esize != 0) {
4800 			error = ENODEV;
4801 			break;
4802 		}
4803 		PF_RULES_WLOCK();
4804 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4805 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4806 		PF_RULES_WUNLOCK();
4807 		break;
4808 	}
4809 
4810 	case DIOCRADDADDRS: {
4811 		struct pfioc_table *io = (struct pfioc_table *)addr;
4812 		struct pfr_addr *pfras;
4813 		size_t totlen;
4814 
4815 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4816 			error = ENODEV;
4817 			break;
4818 		}
4819 		if (io->pfrio_size < 0 ||
4820 		    io->pfrio_size > pf_ioctl_maxcount ||
4821 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4822 			error = EINVAL;
4823 			break;
4824 		}
4825 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4826 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4827 		    M_TEMP, M_WAITOK);
4828 		error = copyin(io->pfrio_buffer, pfras, totlen);
4829 		if (error) {
4830 			free(pfras, M_TEMP);
4831 			break;
4832 		}
4833 		PF_RULES_WLOCK();
4834 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4835 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4836 		    PFR_FLAG_USERIOCTL);
4837 		PF_RULES_WUNLOCK();
4838 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4839 			error = copyout(pfras, io->pfrio_buffer, totlen);
4840 		free(pfras, M_TEMP);
4841 		break;
4842 	}
4843 
4844 	case DIOCRDELADDRS: {
4845 		struct pfioc_table *io = (struct pfioc_table *)addr;
4846 		struct pfr_addr *pfras;
4847 		size_t totlen;
4848 
4849 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4850 			error = ENODEV;
4851 			break;
4852 		}
4853 		if (io->pfrio_size < 0 ||
4854 		    io->pfrio_size > pf_ioctl_maxcount ||
4855 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4856 			error = EINVAL;
4857 			break;
4858 		}
4859 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4860 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4861 		    M_TEMP, M_WAITOK);
4862 		error = copyin(io->pfrio_buffer, pfras, totlen);
4863 		if (error) {
4864 			free(pfras, M_TEMP);
4865 			break;
4866 		}
4867 		PF_RULES_WLOCK();
4868 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4869 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4870 		    PFR_FLAG_USERIOCTL);
4871 		PF_RULES_WUNLOCK();
4872 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4873 			error = copyout(pfras, io->pfrio_buffer, totlen);
4874 		free(pfras, M_TEMP);
4875 		break;
4876 	}
4877 
4878 	case DIOCRSETADDRS: {
4879 		struct pfioc_table *io = (struct pfioc_table *)addr;
4880 		struct pfr_addr *pfras;
4881 		size_t totlen, count;
4882 
4883 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4884 			error = ENODEV;
4885 			break;
4886 		}
4887 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4888 			error = EINVAL;
4889 			break;
4890 		}
4891 		count = max(io->pfrio_size, io->pfrio_size2);
4892 		if (count > pf_ioctl_maxcount ||
4893 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4894 			error = EINVAL;
4895 			break;
4896 		}
4897 		totlen = count * sizeof(struct pfr_addr);
4898 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4899 		    M_WAITOK);
4900 		error = copyin(io->pfrio_buffer, pfras, totlen);
4901 		if (error) {
4902 			free(pfras, M_TEMP);
4903 			break;
4904 		}
4905 		PF_RULES_WLOCK();
4906 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4907 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4908 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4909 		    PFR_FLAG_USERIOCTL, 0);
4910 		PF_RULES_WUNLOCK();
4911 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4912 			error = copyout(pfras, io->pfrio_buffer, totlen);
4913 		free(pfras, M_TEMP);
4914 		break;
4915 	}
4916 
4917 	case DIOCRGETADDRS: {
4918 		struct pfioc_table *io = (struct pfioc_table *)addr;
4919 		struct pfr_addr *pfras;
4920 		size_t totlen;
4921 
4922 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4923 			error = ENODEV;
4924 			break;
4925 		}
4926 		if (io->pfrio_size < 0 ||
4927 		    io->pfrio_size > pf_ioctl_maxcount ||
4928 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4929 			error = EINVAL;
4930 			break;
4931 		}
4932 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4933 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4934 		    M_TEMP, M_WAITOK | M_ZERO);
4935 		PF_RULES_RLOCK();
4936 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4937 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4938 		PF_RULES_RUNLOCK();
4939 		if (error == 0)
4940 			error = copyout(pfras, io->pfrio_buffer, totlen);
4941 		free(pfras, M_TEMP);
4942 		break;
4943 	}
4944 
4945 	case DIOCRGETASTATS: {
4946 		struct pfioc_table *io = (struct pfioc_table *)addr;
4947 		struct pfr_astats *pfrastats;
4948 		size_t totlen;
4949 
4950 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4951 			error = ENODEV;
4952 			break;
4953 		}
4954 		if (io->pfrio_size < 0 ||
4955 		    io->pfrio_size > pf_ioctl_maxcount ||
4956 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4957 			error = EINVAL;
4958 			break;
4959 		}
4960 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4961 		pfrastats = mallocarray(io->pfrio_size,
4962 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4963 		PF_RULES_RLOCK();
4964 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4965 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4966 		PF_RULES_RUNLOCK();
4967 		if (error == 0)
4968 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4969 		free(pfrastats, M_TEMP);
4970 		break;
4971 	}
4972 
4973 	case DIOCRCLRASTATS: {
4974 		struct pfioc_table *io = (struct pfioc_table *)addr;
4975 		struct pfr_addr *pfras;
4976 		size_t totlen;
4977 
4978 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4979 			error = ENODEV;
4980 			break;
4981 		}
4982 		if (io->pfrio_size < 0 ||
4983 		    io->pfrio_size > pf_ioctl_maxcount ||
4984 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4985 			error = EINVAL;
4986 			break;
4987 		}
4988 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4989 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4990 		    M_TEMP, M_WAITOK);
4991 		error = copyin(io->pfrio_buffer, pfras, totlen);
4992 		if (error) {
4993 			free(pfras, M_TEMP);
4994 			break;
4995 		}
4996 		PF_RULES_WLOCK();
4997 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4998 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4999 		    PFR_FLAG_USERIOCTL);
5000 		PF_RULES_WUNLOCK();
5001 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5002 			error = copyout(pfras, io->pfrio_buffer, totlen);
5003 		free(pfras, M_TEMP);
5004 		break;
5005 	}
5006 
5007 	case DIOCRTSTADDRS: {
5008 		struct pfioc_table *io = (struct pfioc_table *)addr;
5009 		struct pfr_addr *pfras;
5010 		size_t totlen;
5011 
5012 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5013 			error = ENODEV;
5014 			break;
5015 		}
5016 		if (io->pfrio_size < 0 ||
5017 		    io->pfrio_size > pf_ioctl_maxcount ||
5018 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5019 			error = EINVAL;
5020 			break;
5021 		}
5022 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5023 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5024 		    M_TEMP, M_WAITOK);
5025 		error = copyin(io->pfrio_buffer, pfras, totlen);
5026 		if (error) {
5027 			free(pfras, M_TEMP);
5028 			break;
5029 		}
5030 		PF_RULES_RLOCK();
5031 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
5032 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5033 		    PFR_FLAG_USERIOCTL);
5034 		PF_RULES_RUNLOCK();
5035 		if (error == 0)
5036 			error = copyout(pfras, io->pfrio_buffer, totlen);
5037 		free(pfras, M_TEMP);
5038 		break;
5039 	}
5040 
5041 	case DIOCRINADEFINE: {
5042 		struct pfioc_table *io = (struct pfioc_table *)addr;
5043 		struct pfr_addr *pfras;
5044 		size_t totlen;
5045 
5046 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5047 			error = ENODEV;
5048 			break;
5049 		}
5050 		if (io->pfrio_size < 0 ||
5051 		    io->pfrio_size > pf_ioctl_maxcount ||
5052 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5053 			error = EINVAL;
5054 			break;
5055 		}
5056 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5057 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5058 		    M_TEMP, M_WAITOK);
5059 		error = copyin(io->pfrio_buffer, pfras, totlen);
5060 		if (error) {
5061 			free(pfras, M_TEMP);
5062 			break;
5063 		}
5064 		PF_RULES_WLOCK();
5065 		error = pfr_ina_define(&io->pfrio_table, pfras,
5066 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5067 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5068 		PF_RULES_WUNLOCK();
5069 		free(pfras, M_TEMP);
5070 		break;
5071 	}
5072 
5073 	case DIOCOSFPADD: {
5074 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5075 		PF_RULES_WLOCK();
5076 		error = pf_osfp_add(io);
5077 		PF_RULES_WUNLOCK();
5078 		break;
5079 	}
5080 
5081 	case DIOCOSFPGET: {
5082 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5083 		PF_RULES_RLOCK();
5084 		error = pf_osfp_get(io);
5085 		PF_RULES_RUNLOCK();
5086 		break;
5087 	}
5088 
5089 	case DIOCXBEGIN: {
5090 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5091 		struct pfioc_trans_e	*ioes, *ioe;
5092 		size_t			 totlen;
5093 		int			 i;
5094 
5095 		if (io->esize != sizeof(*ioe)) {
5096 			error = ENODEV;
5097 			break;
5098 		}
5099 		if (io->size < 0 ||
5100 		    io->size > pf_ioctl_maxcount ||
5101 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5102 			error = EINVAL;
5103 			break;
5104 		}
5105 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5106 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5107 		    M_TEMP, M_WAITOK);
5108 		error = copyin(io->array, ioes, totlen);
5109 		if (error) {
5110 			free(ioes, M_TEMP);
5111 			break;
5112 		}
5113 		/* Ensure there's no more ethernet rules to clean up. */
5114 		NET_EPOCH_DRAIN_CALLBACKS();
5115 		PF_RULES_WLOCK();
5116 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5117 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5118 			switch (ioe->rs_num) {
5119 			case PF_RULESET_ETH:
5120 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5121 					PF_RULES_WUNLOCK();
5122 					free(ioes, M_TEMP);
5123 					goto fail;
5124 				}
5125 				break;
5126 #ifdef ALTQ
5127 			case PF_RULESET_ALTQ:
5128 				if (ioe->anchor[0]) {
5129 					PF_RULES_WUNLOCK();
5130 					free(ioes, M_TEMP);
5131 					error = EINVAL;
5132 					goto fail;
5133 				}
5134 				if ((error = pf_begin_altq(&ioe->ticket))) {
5135 					PF_RULES_WUNLOCK();
5136 					free(ioes, M_TEMP);
5137 					goto fail;
5138 				}
5139 				break;
5140 #endif /* ALTQ */
5141 			case PF_RULESET_TABLE:
5142 			    {
5143 				struct pfr_table table;
5144 
5145 				bzero(&table, sizeof(table));
5146 				strlcpy(table.pfrt_anchor, ioe->anchor,
5147 				    sizeof(table.pfrt_anchor));
5148 				if ((error = pfr_ina_begin(&table,
5149 				    &ioe->ticket, NULL, 0))) {
5150 					PF_RULES_WUNLOCK();
5151 					free(ioes, M_TEMP);
5152 					goto fail;
5153 				}
5154 				break;
5155 			    }
5156 			default:
5157 				if ((error = pf_begin_rules(&ioe->ticket,
5158 				    ioe->rs_num, ioe->anchor))) {
5159 					PF_RULES_WUNLOCK();
5160 					free(ioes, M_TEMP);
5161 					goto fail;
5162 				}
5163 				break;
5164 			}
5165 		}
5166 		PF_RULES_WUNLOCK();
5167 		error = copyout(ioes, io->array, totlen);
5168 		free(ioes, M_TEMP);
5169 		break;
5170 	}
5171 
5172 	case DIOCXROLLBACK: {
5173 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5174 		struct pfioc_trans_e	*ioe, *ioes;
5175 		size_t			 totlen;
5176 		int			 i;
5177 
5178 		if (io->esize != sizeof(*ioe)) {
5179 			error = ENODEV;
5180 			break;
5181 		}
5182 		if (io->size < 0 ||
5183 		    io->size > pf_ioctl_maxcount ||
5184 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5185 			error = EINVAL;
5186 			break;
5187 		}
5188 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5189 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5190 		    M_TEMP, M_WAITOK);
5191 		error = copyin(io->array, ioes, totlen);
5192 		if (error) {
5193 			free(ioes, M_TEMP);
5194 			break;
5195 		}
5196 		PF_RULES_WLOCK();
5197 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5198 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5199 			switch (ioe->rs_num) {
5200 			case PF_RULESET_ETH:
5201 				if ((error = pf_rollback_eth(ioe->ticket,
5202 				    ioe->anchor))) {
5203 					PF_RULES_WUNLOCK();
5204 					free(ioes, M_TEMP);
5205 					goto fail; /* really bad */
5206 				}
5207 				break;
5208 #ifdef ALTQ
5209 			case PF_RULESET_ALTQ:
5210 				if (ioe->anchor[0]) {
5211 					PF_RULES_WUNLOCK();
5212 					free(ioes, M_TEMP);
5213 					error = EINVAL;
5214 					goto fail;
5215 				}
5216 				if ((error = pf_rollback_altq(ioe->ticket))) {
5217 					PF_RULES_WUNLOCK();
5218 					free(ioes, M_TEMP);
5219 					goto fail; /* really bad */
5220 				}
5221 				break;
5222 #endif /* ALTQ */
5223 			case PF_RULESET_TABLE:
5224 			    {
5225 				struct pfr_table table;
5226 
5227 				bzero(&table, sizeof(table));
5228 				strlcpy(table.pfrt_anchor, ioe->anchor,
5229 				    sizeof(table.pfrt_anchor));
5230 				if ((error = pfr_ina_rollback(&table,
5231 				    ioe->ticket, NULL, 0))) {
5232 					PF_RULES_WUNLOCK();
5233 					free(ioes, M_TEMP);
5234 					goto fail; /* really bad */
5235 				}
5236 				break;
5237 			    }
5238 			default:
5239 				if ((error = pf_rollback_rules(ioe->ticket,
5240 				    ioe->rs_num, ioe->anchor))) {
5241 					PF_RULES_WUNLOCK();
5242 					free(ioes, M_TEMP);
5243 					goto fail; /* really bad */
5244 				}
5245 				break;
5246 			}
5247 		}
5248 		PF_RULES_WUNLOCK();
5249 		free(ioes, M_TEMP);
5250 		break;
5251 	}
5252 
5253 	case DIOCXCOMMIT: {
5254 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5255 		struct pfioc_trans_e	*ioe, *ioes;
5256 		struct pf_kruleset	*rs;
5257 		struct pf_keth_ruleset	*ers;
5258 		size_t			 totlen;
5259 		int			 i;
5260 
5261 		if (io->esize != sizeof(*ioe)) {
5262 			error = ENODEV;
5263 			break;
5264 		}
5265 
5266 		if (io->size < 0 ||
5267 		    io->size > pf_ioctl_maxcount ||
5268 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5269 			error = EINVAL;
5270 			break;
5271 		}
5272 
5273 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5274 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5275 		    M_TEMP, M_WAITOK);
5276 		error = copyin(io->array, ioes, totlen);
5277 		if (error) {
5278 			free(ioes, M_TEMP);
5279 			break;
5280 		}
5281 		PF_RULES_WLOCK();
5282 		/* First makes sure everything will succeed. */
5283 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5284 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5285 			switch (ioe->rs_num) {
5286 			case PF_RULESET_ETH:
5287 				ers = pf_find_keth_ruleset(ioe->anchor);
5288 				if (ers == NULL || ioe->ticket == 0 ||
5289 				    ioe->ticket != ers->inactive.ticket) {
5290 					PF_RULES_WUNLOCK();
5291 					free(ioes, M_TEMP);
5292 					error = EINVAL;
5293 					goto fail;
5294 				}
5295 				break;
5296 #ifdef ALTQ
5297 			case PF_RULESET_ALTQ:
5298 				if (ioe->anchor[0]) {
5299 					PF_RULES_WUNLOCK();
5300 					free(ioes, M_TEMP);
5301 					error = EINVAL;
5302 					goto fail;
5303 				}
5304 				if (!V_altqs_inactive_open || ioe->ticket !=
5305 				    V_ticket_altqs_inactive) {
5306 					PF_RULES_WUNLOCK();
5307 					free(ioes, M_TEMP);
5308 					error = EBUSY;
5309 					goto fail;
5310 				}
5311 				break;
5312 #endif /* ALTQ */
5313 			case PF_RULESET_TABLE:
5314 				rs = pf_find_kruleset(ioe->anchor);
5315 				if (rs == NULL || !rs->topen || ioe->ticket !=
5316 				    rs->tticket) {
5317 					PF_RULES_WUNLOCK();
5318 					free(ioes, M_TEMP);
5319 					error = EBUSY;
5320 					goto fail;
5321 				}
5322 				break;
5323 			default:
5324 				if (ioe->rs_num < 0 || ioe->rs_num >=
5325 				    PF_RULESET_MAX) {
5326 					PF_RULES_WUNLOCK();
5327 					free(ioes, M_TEMP);
5328 					error = EINVAL;
5329 					goto fail;
5330 				}
5331 				rs = pf_find_kruleset(ioe->anchor);
5332 				if (rs == NULL ||
5333 				    !rs->rules[ioe->rs_num].inactive.open ||
5334 				    rs->rules[ioe->rs_num].inactive.ticket !=
5335 				    ioe->ticket) {
5336 					PF_RULES_WUNLOCK();
5337 					free(ioes, M_TEMP);
5338 					error = EBUSY;
5339 					goto fail;
5340 				}
5341 				break;
5342 			}
5343 		}
5344 		/* Now do the commit - no errors should happen here. */
5345 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5346 			switch (ioe->rs_num) {
5347 			case PF_RULESET_ETH:
5348 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5349 					PF_RULES_WUNLOCK();
5350 					free(ioes, M_TEMP);
5351 					goto fail; /* really bad */
5352 				}
5353 				break;
5354 #ifdef ALTQ
5355 			case PF_RULESET_ALTQ:
5356 				if ((error = pf_commit_altq(ioe->ticket))) {
5357 					PF_RULES_WUNLOCK();
5358 					free(ioes, M_TEMP);
5359 					goto fail; /* really bad */
5360 				}
5361 				break;
5362 #endif /* ALTQ */
5363 			case PF_RULESET_TABLE:
5364 			    {
5365 				struct pfr_table table;
5366 
5367 				bzero(&table, sizeof(table));
5368 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5369 				    sizeof(table.pfrt_anchor));
5370 				if ((error = pfr_ina_commit(&table,
5371 				    ioe->ticket, NULL, NULL, 0))) {
5372 					PF_RULES_WUNLOCK();
5373 					free(ioes, M_TEMP);
5374 					goto fail; /* really bad */
5375 				}
5376 				break;
5377 			    }
5378 			default:
5379 				if ((error = pf_commit_rules(ioe->ticket,
5380 				    ioe->rs_num, ioe->anchor))) {
5381 					PF_RULES_WUNLOCK();
5382 					free(ioes, M_TEMP);
5383 					goto fail; /* really bad */
5384 				}
5385 				break;
5386 			}
5387 		}
5388 		PF_RULES_WUNLOCK();
5389 
5390 		/* Only hook into EtherNet taffic if we've got rules for it. */
5391 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5392 			hook_pf_eth();
5393 		else
5394 			dehook_pf_eth();
5395 
5396 		free(ioes, M_TEMP);
5397 		break;
5398 	}
5399 
5400 	case DIOCGETSRCNODES: {
5401 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5402 		struct pf_srchash	*sh;
5403 		struct pf_ksrc_node	*n;
5404 		struct pf_src_node	*p, *pstore;
5405 		uint32_t		 i, nr = 0;
5406 
5407 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5408 				i++, sh++) {
5409 			PF_HASHROW_LOCK(sh);
5410 			LIST_FOREACH(n, &sh->nodes, entry)
5411 				nr++;
5412 			PF_HASHROW_UNLOCK(sh);
5413 		}
5414 
5415 		psn->psn_len = min(psn->psn_len,
5416 		    sizeof(struct pf_src_node) * nr);
5417 
5418 		if (psn->psn_len == 0) {
5419 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5420 			break;
5421 		}
5422 
5423 		nr = 0;
5424 
5425 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5426 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5427 		    i++, sh++) {
5428 		    PF_HASHROW_LOCK(sh);
5429 		    LIST_FOREACH(n, &sh->nodes, entry) {
5430 
5431 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5432 				break;
5433 
5434 			pf_src_node_copy(n, p);
5435 
5436 			p++;
5437 			nr++;
5438 		    }
5439 		    PF_HASHROW_UNLOCK(sh);
5440 		}
5441 		error = copyout(pstore, psn->psn_src_nodes,
5442 		    sizeof(struct pf_src_node) * nr);
5443 		if (error) {
5444 			free(pstore, M_TEMP);
5445 			break;
5446 		}
5447 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5448 		free(pstore, M_TEMP);
5449 		break;
5450 	}
5451 
5452 	case DIOCCLRSRCNODES: {
5453 		pf_kill_srcnodes(NULL);
5454 		break;
5455 	}
5456 
5457 	case DIOCKILLSRCNODES:
5458 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5459 		break;
5460 
5461 #ifdef COMPAT_FREEBSD13
5462 	case DIOCKEEPCOUNTERS_FREEBSD13:
5463 #endif
5464 	case DIOCKEEPCOUNTERS:
5465 		error = pf_keepcounters((struct pfioc_nv *)addr);
5466 		break;
5467 
5468 	case DIOCGETSYNCOOKIES:
5469 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5470 		break;
5471 
5472 	case DIOCSETSYNCOOKIES:
5473 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5474 		break;
5475 
5476 	case DIOCSETHOSTID: {
5477 		u_int32_t	*hostid = (u_int32_t *)addr;
5478 
5479 		PF_RULES_WLOCK();
5480 		if (*hostid == 0)
5481 			V_pf_status.hostid = arc4random();
5482 		else
5483 			V_pf_status.hostid = *hostid;
5484 		PF_RULES_WUNLOCK();
5485 		break;
5486 	}
5487 
5488 	case DIOCOSFPFLUSH:
5489 		PF_RULES_WLOCK();
5490 		pf_osfp_flush();
5491 		PF_RULES_WUNLOCK();
5492 		break;
5493 
5494 	case DIOCIGETIFACES: {
5495 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5496 		struct pfi_kif *ifstore;
5497 		size_t bufsiz;
5498 
5499 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5500 			error = ENODEV;
5501 			break;
5502 		}
5503 
5504 		if (io->pfiio_size < 0 ||
5505 		    io->pfiio_size > pf_ioctl_maxcount ||
5506 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5507 			error = EINVAL;
5508 			break;
5509 		}
5510 
5511 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5512 
5513 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5514 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5515 		    M_TEMP, M_WAITOK | M_ZERO);
5516 
5517 		PF_RULES_RLOCK();
5518 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5519 		PF_RULES_RUNLOCK();
5520 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5521 		free(ifstore, M_TEMP);
5522 		break;
5523 	}
5524 
5525 	case DIOCSETIFFLAG: {
5526 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5527 
5528 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5529 
5530 		PF_RULES_WLOCK();
5531 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5532 		PF_RULES_WUNLOCK();
5533 		break;
5534 	}
5535 
5536 	case DIOCCLRIFFLAG: {
5537 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5538 
5539 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5540 
5541 		PF_RULES_WLOCK();
5542 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5543 		PF_RULES_WUNLOCK();
5544 		break;
5545 	}
5546 
5547 	case DIOCSETREASS: {
5548 		u_int32_t	*reass = (u_int32_t *)addr;
5549 
5550 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5551 		/* Removal of DF flag without reassembly enabled is not a
5552 		 * valid combination. Disable reassembly in such case. */
5553 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5554 			V_pf_status.reass = 0;
5555 		break;
5556 	}
5557 
5558 	default:
5559 		error = ENODEV;
5560 		break;
5561 	}
5562 fail:
5563 	CURVNET_RESTORE();
5564 
5565 #undef ERROUT_IOCTL
5566 
5567 	return (error);
5568 }
5569 
5570 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5571 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5572 {
5573 	bzero(sp, sizeof(union pfsync_state_union));
5574 
5575 	/* copy from state key */
5576 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5577 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5578 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5579 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5580 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5581 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5582 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5583 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5584 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5585 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5586 
5587 	/* copy from state */
5588 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5589 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5590 	sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5591 	sp->pfs_1301.expire = pf_state_expires(st);
5592 	if (sp->pfs_1301.expire <= time_uptime)
5593 		sp->pfs_1301.expire = htonl(0);
5594 	else
5595 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5596 
5597 	sp->pfs_1301.direction = st->direction;
5598 	sp->pfs_1301.log = st->act.log;
5599 	sp->pfs_1301.timeout = st->timeout;
5600 
5601 	switch (msg_version) {
5602 		case PFSYNC_MSG_VERSION_1301:
5603 			sp->pfs_1301.state_flags = st->state_flags;
5604 			break;
5605 		case PFSYNC_MSG_VERSION_1400:
5606 			sp->pfs_1400.state_flags = htons(st->state_flags);
5607 			sp->pfs_1400.qid = htons(st->act.qid);
5608 			sp->pfs_1400.pqid = htons(st->act.pqid);
5609 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5610 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5611 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5612 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5613 			sp->pfs_1400.set_tos = st->act.set_tos;
5614 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5615 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5616 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5617 			sp->pfs_1400.rt = st->rt;
5618 			if (st->rt_kif)
5619 				strlcpy(sp->pfs_1400.rt_ifname,
5620 				    st->rt_kif->pfik_name,
5621 				    sizeof(sp->pfs_1400.rt_ifname));
5622 			break;
5623 		default:
5624 			panic("%s: Unsupported pfsync_msg_version %d",
5625 			    __func__, msg_version);
5626 	}
5627 
5628 	if (st->src_node)
5629 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5630 	if (st->nat_src_node)
5631 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5632 
5633 	sp->pfs_1301.id = st->id;
5634 	sp->pfs_1301.creatorid = st->creatorid;
5635 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5636 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5637 
5638 	if (st->rule == NULL)
5639 		sp->pfs_1301.rule = htonl(-1);
5640 	else
5641 		sp->pfs_1301.rule = htonl(st->rule->nr);
5642 	if (st->anchor == NULL)
5643 		sp->pfs_1301.anchor = htonl(-1);
5644 	else
5645 		sp->pfs_1301.anchor = htonl(st->anchor->nr);
5646 	if (st->nat_rule == NULL)
5647 		sp->pfs_1301.nat_rule = htonl(-1);
5648 	else
5649 		sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
5650 
5651 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5652 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5653 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5654 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5655 }
5656 
5657 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5658 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5659 {
5660 	bzero(sp, sizeof(*sp));
5661 
5662 	sp->version = PF_STATE_VERSION;
5663 
5664 	/* copy from state key */
5665 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5666 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5667 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5668 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5669 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5670 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5671 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5672 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5673 	sp->proto = st->key[PF_SK_WIRE]->proto;
5674 	sp->af = st->key[PF_SK_WIRE]->af;
5675 
5676 	/* copy from state */
5677 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5678 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5679 	    sizeof(sp->orig_ifname));
5680 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5681 	sp->creation = htonl(time_uptime - (st->creation / 1000));
5682 	sp->expire = pf_state_expires(st);
5683 	if (sp->expire <= time_uptime)
5684 		sp->expire = htonl(0);
5685 	else
5686 		sp->expire = htonl(sp->expire - time_uptime);
5687 
5688 	sp->direction = st->direction;
5689 	sp->log = st->act.log;
5690 	sp->timeout = st->timeout;
5691 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5692 	sp->state_flags_compat = st->state_flags;
5693 	sp->state_flags = htons(st->state_flags);
5694 	if (st->src_node)
5695 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5696 	if (st->nat_src_node)
5697 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5698 
5699 	sp->id = st->id;
5700 	sp->creatorid = st->creatorid;
5701 	pf_state_peer_hton(&st->src, &sp->src);
5702 	pf_state_peer_hton(&st->dst, &sp->dst);
5703 
5704 	if (st->rule == NULL)
5705 		sp->rule = htonl(-1);
5706 	else
5707 		sp->rule = htonl(st->rule->nr);
5708 	if (st->anchor == NULL)
5709 		sp->anchor = htonl(-1);
5710 	else
5711 		sp->anchor = htonl(st->anchor->nr);
5712 	if (st->nat_rule == NULL)
5713 		sp->nat_rule = htonl(-1);
5714 	else
5715 		sp->nat_rule = htonl(st->nat_rule->nr);
5716 
5717 	sp->packets[0] = st->packets[0];
5718 	sp->packets[1] = st->packets[1];
5719 	sp->bytes[0] = st->bytes[0];
5720 	sp->bytes[1] = st->bytes[1];
5721 
5722 	sp->qid = htons(st->act.qid);
5723 	sp->pqid = htons(st->act.pqid);
5724 	sp->dnpipe = htons(st->act.dnpipe);
5725 	sp->dnrpipe = htons(st->act.dnrpipe);
5726 	sp->rtableid = htonl(st->act.rtableid);
5727 	sp->min_ttl = st->act.min_ttl;
5728 	sp->set_tos = st->act.set_tos;
5729 	sp->max_mss = htons(st->act.max_mss);
5730 	sp->rt = st->rt;
5731 	if (st->rt_kif)
5732 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5733 		    sizeof(sp->rt_ifname));
5734 	sp->set_prio[0] = st->act.set_prio[0];
5735 	sp->set_prio[1] = st->act.set_prio[1];
5736 
5737 }
5738 
5739 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5740 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5741 {
5742 	struct pfr_ktable *kt;
5743 
5744 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5745 
5746 	kt = aw->p.tbl;
5747 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5748 		kt = kt->pfrkt_root;
5749 	aw->p.tbl = NULL;
5750 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5751 		kt->pfrkt_cnt : -1;
5752 }
5753 
5754 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5755 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5756     size_t number, char **names)
5757 {
5758 	nvlist_t        *nvc;
5759 
5760 	nvc = nvlist_create(0);
5761 	if (nvc == NULL)
5762 		return (ENOMEM);
5763 
5764 	for (int i = 0; i < number; i++) {
5765 		nvlist_append_number_array(nvc, "counters",
5766 		    counter_u64_fetch(counters[i]));
5767 		nvlist_append_string_array(nvc, "names",
5768 		    names[i]);
5769 		nvlist_append_number_array(nvc, "ids",
5770 		    i);
5771 	}
5772 	nvlist_add_nvlist(nvl, name, nvc);
5773 	nvlist_destroy(nvc);
5774 
5775 	return (0);
5776 }
5777 
5778 static int
pf_getstatus(struct pfioc_nv * nv)5779 pf_getstatus(struct pfioc_nv *nv)
5780 {
5781 	nvlist_t        *nvl = NULL, *nvc = NULL;
5782 	void            *nvlpacked = NULL;
5783 	int              error;
5784 	struct pf_status s;
5785 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5786 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5787 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5788 	PF_RULES_RLOCK_TRACKER;
5789 
5790 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5791 
5792 	PF_RULES_RLOCK();
5793 
5794 	nvl = nvlist_create(0);
5795 	if (nvl == NULL)
5796 		ERROUT(ENOMEM);
5797 
5798 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5799 	nvlist_add_number(nvl, "since", V_pf_status.since);
5800 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5801 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5802 	nvlist_add_number(nvl, "states", V_pf_status.states);
5803 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5804 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5805 	nvlist_add_bool(nvl, "syncookies_active",
5806 	    V_pf_status.syncookies_active);
5807 	nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5808 
5809 	/* counters */
5810 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5811 	    PFRES_MAX, pf_reasons);
5812 	if (error != 0)
5813 		ERROUT(error);
5814 
5815 	/* lcounters */
5816 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5817 	    KLCNT_MAX, pf_lcounter);
5818 	if (error != 0)
5819 		ERROUT(error);
5820 
5821 	/* fcounters */
5822 	nvc = nvlist_create(0);
5823 	if (nvc == NULL)
5824 		ERROUT(ENOMEM);
5825 
5826 	for (int i = 0; i < FCNT_MAX; i++) {
5827 		nvlist_append_number_array(nvc, "counters",
5828 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5829 		nvlist_append_string_array(nvc, "names",
5830 		    pf_fcounter[i]);
5831 		nvlist_append_number_array(nvc, "ids",
5832 		    i);
5833 	}
5834 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5835 	nvlist_destroy(nvc);
5836 	nvc = NULL;
5837 
5838 	/* scounters */
5839 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5840 	    SCNT_MAX, pf_fcounter);
5841 	if (error != 0)
5842 		ERROUT(error);
5843 
5844 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5845 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5846 	    PF_MD5_DIGEST_LENGTH);
5847 
5848 	pfi_update_status(V_pf_status.ifname, &s);
5849 
5850 	/* pcounters / bcounters */
5851 	for (int i = 0; i < 2; i++) {
5852 		for (int j = 0; j < 2; j++) {
5853 			for (int k = 0; k < 2; k++) {
5854 				nvlist_append_number_array(nvl, "pcounters",
5855 				    s.pcounters[i][j][k]);
5856 			}
5857 			nvlist_append_number_array(nvl, "bcounters",
5858 			    s.bcounters[i][j]);
5859 		}
5860 	}
5861 
5862 	nvlpacked = nvlist_pack(nvl, &nv->len);
5863 	if (nvlpacked == NULL)
5864 		ERROUT(ENOMEM);
5865 
5866 	if (nv->size == 0)
5867 		ERROUT(0);
5868 	else if (nv->size < nv->len)
5869 		ERROUT(ENOSPC);
5870 
5871 	PF_RULES_RUNLOCK();
5872 	error = copyout(nvlpacked, nv->data, nv->len);
5873 	goto done;
5874 
5875 #undef ERROUT
5876 errout:
5877 	PF_RULES_RUNLOCK();
5878 done:
5879 	free(nvlpacked, M_NVLIST);
5880 	nvlist_destroy(nvc);
5881 	nvlist_destroy(nvl);
5882 
5883 	return (error);
5884 }
5885 
5886 /*
5887  * XXX - Check for version mismatch!!!
5888  */
5889 static void
pf_clear_all_states(void)5890 pf_clear_all_states(void)
5891 {
5892 	struct epoch_tracker	 et;
5893 	struct pf_kstate	*s;
5894 	u_int i;
5895 
5896 	NET_EPOCH_ENTER(et);
5897 	for (i = 0; i <= V_pf_hashmask; i++) {
5898 		struct pf_idhash *ih = &V_pf_idhash[i];
5899 relock:
5900 		PF_HASHROW_LOCK(ih);
5901 		LIST_FOREACH(s, &ih->states, entry) {
5902 			s->timeout = PFTM_PURGE;
5903 			/* Don't send out individual delete messages. */
5904 			s->state_flags |= PFSTATE_NOSYNC;
5905 			pf_unlink_state(s);
5906 			goto relock;
5907 		}
5908 		PF_HASHROW_UNLOCK(ih);
5909 	}
5910 	NET_EPOCH_EXIT(et);
5911 }
5912 
5913 static int
pf_clear_tables(void)5914 pf_clear_tables(void)
5915 {
5916 	struct pfioc_table io;
5917 	int error;
5918 
5919 	bzero(&io, sizeof(io));
5920 	io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5921 
5922 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5923 	    io.pfrio_flags);
5924 
5925 	return (error);
5926 }
5927 
5928 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)5929 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5930 {
5931 	struct pf_ksrc_node_list	 kill;
5932 	u_int 				 killed;
5933 
5934 	LIST_INIT(&kill);
5935 	for (int i = 0; i <= V_pf_srchashmask; i++) {
5936 		struct pf_srchash *sh = &V_pf_srchash[i];
5937 		struct pf_ksrc_node *sn, *tmp;
5938 
5939 		PF_HASHROW_LOCK(sh);
5940 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5941 			if (psnk == NULL ||
5942 			    (PF_MATCHA(psnk->psnk_src.neg,
5943 			      &psnk->psnk_src.addr.v.a.addr,
5944 			      &psnk->psnk_src.addr.v.a.mask,
5945 			      &sn->addr, sn->af) &&
5946 			    PF_MATCHA(psnk->psnk_dst.neg,
5947 			      &psnk->psnk_dst.addr.v.a.addr,
5948 			      &psnk->psnk_dst.addr.v.a.mask,
5949 			      &sn->raddr, sn->af))) {
5950 				pf_unlink_src_node(sn);
5951 				LIST_INSERT_HEAD(&kill, sn, entry);
5952 				sn->expire = 1;
5953 			}
5954 		PF_HASHROW_UNLOCK(sh);
5955 	}
5956 
5957 	for (int i = 0; i <= V_pf_hashmask; i++) {
5958 		struct pf_idhash *ih = &V_pf_idhash[i];
5959 		struct pf_kstate *s;
5960 
5961 		PF_HASHROW_LOCK(ih);
5962 		LIST_FOREACH(s, &ih->states, entry) {
5963 			if (s->src_node && s->src_node->expire == 1)
5964 				s->src_node = NULL;
5965 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5966 				s->nat_src_node = NULL;
5967 		}
5968 		PF_HASHROW_UNLOCK(ih);
5969 	}
5970 
5971 	killed = pf_free_src_nodes(&kill);
5972 
5973 	if (psnk != NULL)
5974 		psnk->psnk_killed = killed;
5975 }
5976 
5977 static int
pf_keepcounters(struct pfioc_nv * nv)5978 pf_keepcounters(struct pfioc_nv *nv)
5979 {
5980 	nvlist_t	*nvl = NULL;
5981 	void		*nvlpacked = NULL;
5982 	int		 error = 0;
5983 
5984 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5985 
5986 	if (nv->len > pf_ioctl_maxcount)
5987 		ERROUT(ENOMEM);
5988 
5989 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5990 	error = copyin(nv->data, nvlpacked, nv->len);
5991 	if (error)
5992 		ERROUT(error);
5993 
5994 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5995 	if (nvl == NULL)
5996 		ERROUT(EBADMSG);
5997 
5998 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5999 		ERROUT(EBADMSG);
6000 
6001 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6002 
6003 on_error:
6004 	nvlist_destroy(nvl);
6005 	free(nvlpacked, M_NVLIST);
6006 	return (error);
6007 }
6008 
6009 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6010 pf_clear_states(const struct pf_kstate_kill *kill)
6011 {
6012 	struct pf_state_key_cmp	 match_key;
6013 	struct pf_kstate	*s;
6014 	struct pfi_kkif	*kif;
6015 	int		 idx;
6016 	unsigned int	 killed = 0, dir;
6017 
6018 	NET_EPOCH_ASSERT();
6019 
6020 	for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6021 		struct pf_idhash *ih = &V_pf_idhash[i];
6022 
6023 relock_DIOCCLRSTATES:
6024 		PF_HASHROW_LOCK(ih);
6025 		LIST_FOREACH(s, &ih->states, entry) {
6026 			/* For floating states look at the original kif. */
6027 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6028 
6029 			if (kill->psk_ifname[0] &&
6030 			    strcmp(kill->psk_ifname,
6031 			    kif->pfik_name))
6032 				continue;
6033 
6034 			if (kill->psk_kill_match) {
6035 				bzero(&match_key, sizeof(match_key));
6036 
6037 				if (s->direction == PF_OUT) {
6038 					dir = PF_IN;
6039 					idx = PF_SK_STACK;
6040 				} else {
6041 					dir = PF_OUT;
6042 					idx = PF_SK_WIRE;
6043 				}
6044 
6045 				match_key.af = s->key[idx]->af;
6046 				match_key.proto = s->key[idx]->proto;
6047 				PF_ACPY(&match_key.addr[0],
6048 				    &s->key[idx]->addr[1], match_key.af);
6049 				match_key.port[0] = s->key[idx]->port[1];
6050 				PF_ACPY(&match_key.addr[1],
6051 				    &s->key[idx]->addr[0], match_key.af);
6052 				match_key.port[1] = s->key[idx]->port[0];
6053 			}
6054 
6055 			/*
6056 			 * Don't send out individual
6057 			 * delete messages.
6058 			 */
6059 			s->state_flags |= PFSTATE_NOSYNC;
6060 			pf_unlink_state(s);
6061 			killed++;
6062 
6063 			if (kill->psk_kill_match)
6064 				killed += pf_kill_matching_state(&match_key,
6065 				    dir);
6066 
6067 			goto relock_DIOCCLRSTATES;
6068 		}
6069 		PF_HASHROW_UNLOCK(ih);
6070 	}
6071 
6072 	if (V_pfsync_clear_states_ptr != NULL)
6073 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6074 
6075 	return (killed);
6076 }
6077 
6078 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6079 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6080 {
6081 	struct pf_kstate	*s;
6082 
6083 	NET_EPOCH_ASSERT();
6084 	if (kill->psk_pfcmp.id) {
6085 		if (kill->psk_pfcmp.creatorid == 0)
6086 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6087 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6088 		    kill->psk_pfcmp.creatorid))) {
6089 			pf_unlink_state(s);
6090 			*killed = 1;
6091 		}
6092 		return;
6093 	}
6094 
6095 	for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6096 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6097 }
6098 
6099 static int
pf_killstates_nv(struct pfioc_nv * nv)6100 pf_killstates_nv(struct pfioc_nv *nv)
6101 {
6102 	struct pf_kstate_kill	 kill;
6103 	struct epoch_tracker	 et;
6104 	nvlist_t		*nvl = NULL;
6105 	void			*nvlpacked = NULL;
6106 	int			 error = 0;
6107 	unsigned int		 killed = 0;
6108 
6109 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6110 
6111 	if (nv->len > pf_ioctl_maxcount)
6112 		ERROUT(ENOMEM);
6113 
6114 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6115 	error = copyin(nv->data, nvlpacked, nv->len);
6116 	if (error)
6117 		ERROUT(error);
6118 
6119 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6120 	if (nvl == NULL)
6121 		ERROUT(EBADMSG);
6122 
6123 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6124 	if (error)
6125 		ERROUT(error);
6126 
6127 	NET_EPOCH_ENTER(et);
6128 	pf_killstates(&kill, &killed);
6129 	NET_EPOCH_EXIT(et);
6130 
6131 	free(nvlpacked, M_NVLIST);
6132 	nvlpacked = NULL;
6133 	nvlist_destroy(nvl);
6134 	nvl = nvlist_create(0);
6135 	if (nvl == NULL)
6136 		ERROUT(ENOMEM);
6137 
6138 	nvlist_add_number(nvl, "killed", killed);
6139 
6140 	nvlpacked = nvlist_pack(nvl, &nv->len);
6141 	if (nvlpacked == NULL)
6142 		ERROUT(ENOMEM);
6143 
6144 	if (nv->size == 0)
6145 		ERROUT(0);
6146 	else if (nv->size < nv->len)
6147 		ERROUT(ENOSPC);
6148 
6149 	error = copyout(nvlpacked, nv->data, nv->len);
6150 
6151 on_error:
6152 	nvlist_destroy(nvl);
6153 	free(nvlpacked, M_NVLIST);
6154 	return (error);
6155 }
6156 
6157 static int
pf_clearstates_nv(struct pfioc_nv * nv)6158 pf_clearstates_nv(struct pfioc_nv *nv)
6159 {
6160 	struct pf_kstate_kill	 kill;
6161 	struct epoch_tracker	 et;
6162 	nvlist_t		*nvl = NULL;
6163 	void			*nvlpacked = NULL;
6164 	int			 error = 0;
6165 	unsigned int		 killed;
6166 
6167 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6168 
6169 	if (nv->len > pf_ioctl_maxcount)
6170 		ERROUT(ENOMEM);
6171 
6172 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6173 	error = copyin(nv->data, nvlpacked, nv->len);
6174 	if (error)
6175 		ERROUT(error);
6176 
6177 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6178 	if (nvl == NULL)
6179 		ERROUT(EBADMSG);
6180 
6181 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6182 	if (error)
6183 		ERROUT(error);
6184 
6185 	NET_EPOCH_ENTER(et);
6186 	killed = pf_clear_states(&kill);
6187 	NET_EPOCH_EXIT(et);
6188 
6189 	free(nvlpacked, M_NVLIST);
6190 	nvlpacked = NULL;
6191 	nvlist_destroy(nvl);
6192 	nvl = nvlist_create(0);
6193 	if (nvl == NULL)
6194 		ERROUT(ENOMEM);
6195 
6196 	nvlist_add_number(nvl, "killed", killed);
6197 
6198 	nvlpacked = nvlist_pack(nvl, &nv->len);
6199 	if (nvlpacked == NULL)
6200 		ERROUT(ENOMEM);
6201 
6202 	if (nv->size == 0)
6203 		ERROUT(0);
6204 	else if (nv->size < nv->len)
6205 		ERROUT(ENOSPC);
6206 
6207 	error = copyout(nvlpacked, nv->data, nv->len);
6208 
6209 #undef ERROUT
6210 on_error:
6211 	nvlist_destroy(nvl);
6212 	free(nvlpacked, M_NVLIST);
6213 	return (error);
6214 }
6215 
6216 static int
pf_getstate(struct pfioc_nv * nv)6217 pf_getstate(struct pfioc_nv *nv)
6218 {
6219 	nvlist_t		*nvl = NULL, *nvls;
6220 	void			*nvlpacked = NULL;
6221 	struct pf_kstate	*s = NULL;
6222 	int			 error = 0;
6223 	uint64_t		 id, creatorid;
6224 
6225 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6226 
6227 	if (nv->len > pf_ioctl_maxcount)
6228 		ERROUT(ENOMEM);
6229 
6230 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6231 	error = copyin(nv->data, nvlpacked, nv->len);
6232 	if (error)
6233 		ERROUT(error);
6234 
6235 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6236 	if (nvl == NULL)
6237 		ERROUT(EBADMSG);
6238 
6239 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6240 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6241 
6242 	s = pf_find_state_byid(id, creatorid);
6243 	if (s == NULL)
6244 		ERROUT(ENOENT);
6245 
6246 	free(nvlpacked, M_NVLIST);
6247 	nvlpacked = NULL;
6248 	nvlist_destroy(nvl);
6249 	nvl = nvlist_create(0);
6250 	if (nvl == NULL)
6251 		ERROUT(ENOMEM);
6252 
6253 	nvls = pf_state_to_nvstate(s);
6254 	if (nvls == NULL)
6255 		ERROUT(ENOMEM);
6256 
6257 	nvlist_add_nvlist(nvl, "state", nvls);
6258 	nvlist_destroy(nvls);
6259 
6260 	nvlpacked = nvlist_pack(nvl, &nv->len);
6261 	if (nvlpacked == NULL)
6262 		ERROUT(ENOMEM);
6263 
6264 	if (nv->size == 0)
6265 		ERROUT(0);
6266 	else if (nv->size < nv->len)
6267 		ERROUT(ENOSPC);
6268 
6269 	error = copyout(nvlpacked, nv->data, nv->len);
6270 
6271 #undef ERROUT
6272 errout:
6273 	if (s != NULL)
6274 		PF_STATE_UNLOCK(s);
6275 	free(nvlpacked, M_NVLIST);
6276 	nvlist_destroy(nvl);
6277 	return (error);
6278 }
6279 
6280 /*
6281  * XXX - Check for version mismatch!!!
6282  */
6283 
6284 /*
6285  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6286  */
6287 static int
shutdown_pf(void)6288 shutdown_pf(void)
6289 {
6290 	int error = 0;
6291 	u_int32_t t[5];
6292 	char nn = '\0';
6293 	struct pf_kanchor *anchor;
6294 	struct pf_keth_anchor *eth_anchor;
6295 	int rs_num;
6296 
6297 	do {
6298 		/* Unlink rules of all user defined anchors */
6299 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6300 			/* Wildcard based anchors may not have a respective
6301 			 * explicit anchor rule or they may be left empty
6302 			 * without rules. It leads to anchor.refcnt=0, and the
6303 			 * rest of the logic does not expect it. */
6304 			if (anchor->refcnt == 0)
6305 				anchor->refcnt = 1;
6306 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6307 				if ((error = pf_begin_rules(&t[rs_num], rs_num,
6308 				    anchor->path)) != 0) {
6309 					DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6310 					    "anchor.path=%s rs_num=%d\n",
6311 					    anchor->path, rs_num));
6312 					goto error;	/* XXX: rollback? */
6313 				}
6314 			}
6315 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6316 				error = pf_commit_rules(t[rs_num], rs_num,
6317 				    anchor->path);
6318 				MPASS(error == 0);
6319 			}
6320 		}
6321 
6322 		/* Unlink rules of all user defined ether anchors */
6323 		RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6324 		    &V_pf_keth_anchors) {
6325 			/* Wildcard based anchors may not have a respective
6326 			 * explicit anchor rule or they may be left empty
6327 			 * without rules. It leads to anchor.refcnt=0, and the
6328 			 * rest of the logic does not expect it. */
6329 			if (eth_anchor->refcnt == 0)
6330 				eth_anchor->refcnt = 1;
6331 			if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6332 			    != 0) {
6333 				DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6334 				    "anchor.path=%s\n", eth_anchor->path));
6335 				goto error;
6336 			}
6337 			error = pf_commit_eth(t[0], eth_anchor->path);
6338 			MPASS(error == 0);
6339 		}
6340 
6341 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6342 		    != 0) {
6343 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6344 			break;
6345 		}
6346 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6347 		    != 0) {
6348 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6349 			break;		/* XXX: rollback? */
6350 		}
6351 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6352 		    != 0) {
6353 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6354 			break;		/* XXX: rollback? */
6355 		}
6356 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6357 		    != 0) {
6358 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6359 			break;		/* XXX: rollback? */
6360 		}
6361 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6362 		    != 0) {
6363 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6364 			break;		/* XXX: rollback? */
6365 		}
6366 
6367 		error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6368 		MPASS(error == 0);
6369 		error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6370 		MPASS(error == 0);
6371 		error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6372 		MPASS(error == 0);
6373 		error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6374 		MPASS(error == 0);
6375 		error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6376 		MPASS(error == 0);
6377 
6378 		if ((error = pf_clear_tables()) != 0)
6379 			break;
6380 
6381 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6382 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6383 			break;
6384 		}
6385 		error = pf_commit_eth(t[0], &nn);
6386 		MPASS(error == 0);
6387 
6388 #ifdef ALTQ
6389 		if ((error = pf_begin_altq(&t[0])) != 0) {
6390 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6391 			break;
6392 		}
6393 		pf_commit_altq(t[0]);
6394 #endif
6395 
6396 		pf_clear_all_states();
6397 
6398 		pf_kill_srcnodes(NULL);
6399 
6400 		/* status does not use malloced mem so no need to cleanup */
6401 		/* fingerprints and interfaces have their own cleanup code */
6402 	} while(0);
6403 
6404 error:
6405 	return (error);
6406 }
6407 
6408 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6409 pf_check_return(int chk, struct mbuf **m)
6410 {
6411 
6412 	switch (chk) {
6413 	case PF_PASS:
6414 		if (*m == NULL)
6415 			return (PFIL_CONSUMED);
6416 		else
6417 			return (PFIL_PASS);
6418 		break;
6419 	default:
6420 		if (*m != NULL) {
6421 			m_freem(*m);
6422 			*m = NULL;
6423 		}
6424 		return (PFIL_DROPPED);
6425 	}
6426 }
6427 
6428 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6429 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6430     void *ruleset __unused, struct inpcb *inp)
6431 {
6432 	int chk;
6433 
6434 	CURVNET_ASSERT_SET();
6435 
6436 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6437 
6438 	return (pf_check_return(chk, m));
6439 }
6440 
6441 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6442 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6443     void *ruleset __unused, struct inpcb *inp)
6444 {
6445 	int chk;
6446 
6447 	CURVNET_ASSERT_SET();
6448 
6449 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6450 
6451 	return (pf_check_return(chk, m));
6452 }
6453 
6454 #ifdef INET
6455 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6456 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6457     void *ruleset __unused, struct inpcb *inp)
6458 {
6459 	int chk;
6460 
6461 	CURVNET_ASSERT_SET();
6462 
6463 	chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
6464 
6465 	return (pf_check_return(chk, m));
6466 }
6467 
6468 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6469 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6470     void *ruleset __unused,  struct inpcb *inp)
6471 {
6472 	int chk;
6473 
6474 	CURVNET_ASSERT_SET();
6475 
6476 	chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
6477 
6478 	return (pf_check_return(chk, m));
6479 }
6480 #endif
6481 
6482 #ifdef INET6
6483 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6484 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6485     void *ruleset __unused,  struct inpcb *inp)
6486 {
6487 	int chk;
6488 
6489 	CURVNET_ASSERT_SET();
6490 
6491 	/*
6492 	 * In case of loopback traffic IPv6 uses the real interface in
6493 	 * order to support scoped addresses. In order to support stateful
6494 	 * filtering we have change this to lo0 as it is the case in IPv4.
6495 	 */
6496 	chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6497 	    m, inp, NULL);
6498 
6499 	return (pf_check_return(chk, m));
6500 }
6501 
6502 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6503 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6504     void *ruleset __unused,  struct inpcb *inp)
6505 {
6506 	int chk;
6507 
6508 	CURVNET_ASSERT_SET();
6509 
6510 	chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL);
6511 
6512 	return (pf_check_return(chk, m));
6513 }
6514 #endif /* INET6 */
6515 
6516 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6517 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6518 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6519 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6520 
6521 #ifdef INET
6522 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6523 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6524 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6525 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6526 #endif
6527 #ifdef INET6
6528 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6529 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6530 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6531 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6532 #endif
6533 
6534 static void
hook_pf_eth(void)6535 hook_pf_eth(void)
6536 {
6537 	struct pfil_hook_args pha = {
6538 		.pa_version = PFIL_VERSION,
6539 		.pa_modname = "pf",
6540 		.pa_type = PFIL_TYPE_ETHERNET,
6541 	};
6542 	struct pfil_link_args pla = {
6543 		.pa_version = PFIL_VERSION,
6544 	};
6545 	int ret __diagused;
6546 
6547 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6548 		return;
6549 
6550 	pha.pa_mbuf_chk = pf_eth_check_in;
6551 	pha.pa_flags = PFIL_IN;
6552 	pha.pa_rulname = "eth-in";
6553 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6554 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6555 	pla.pa_head = V_link_pfil_head;
6556 	pla.pa_hook = V_pf_eth_in_hook;
6557 	ret = pfil_link(&pla);
6558 	MPASS(ret == 0);
6559 	pha.pa_mbuf_chk = pf_eth_check_out;
6560 	pha.pa_flags = PFIL_OUT;
6561 	pha.pa_rulname = "eth-out";
6562 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6563 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6564 	pla.pa_head = V_link_pfil_head;
6565 	pla.pa_hook = V_pf_eth_out_hook;
6566 	ret = pfil_link(&pla);
6567 	MPASS(ret == 0);
6568 
6569 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6570 }
6571 
6572 static void
hook_pf(void)6573 hook_pf(void)
6574 {
6575 	struct pfil_hook_args pha = {
6576 		.pa_version = PFIL_VERSION,
6577 		.pa_modname = "pf",
6578 	};
6579 	struct pfil_link_args pla = {
6580 		.pa_version = PFIL_VERSION,
6581 	};
6582 	int ret __diagused;
6583 
6584 	if (atomic_load_bool(&V_pf_pfil_hooked))
6585 		return;
6586 
6587 #ifdef INET
6588 	pha.pa_type = PFIL_TYPE_IP4;
6589 	pha.pa_mbuf_chk = pf_check_in;
6590 	pha.pa_flags = PFIL_IN;
6591 	pha.pa_rulname = "default-in";
6592 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6593 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6594 	pla.pa_head = V_inet_pfil_head;
6595 	pla.pa_hook = V_pf_ip4_in_hook;
6596 	ret = pfil_link(&pla);
6597 	MPASS(ret == 0);
6598 	pha.pa_mbuf_chk = pf_check_out;
6599 	pha.pa_flags = PFIL_OUT;
6600 	pha.pa_rulname = "default-out";
6601 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6602 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6603 	pla.pa_head = V_inet_pfil_head;
6604 	pla.pa_hook = V_pf_ip4_out_hook;
6605 	ret = pfil_link(&pla);
6606 	MPASS(ret == 0);
6607 	if (V_pf_filter_local) {
6608 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6609 		pla.pa_head = V_inet_local_pfil_head;
6610 		pla.pa_hook = V_pf_ip4_out_hook;
6611 		ret = pfil_link(&pla);
6612 		MPASS(ret == 0);
6613 	}
6614 #endif
6615 #ifdef INET6
6616 	pha.pa_type = PFIL_TYPE_IP6;
6617 	pha.pa_mbuf_chk = pf_check6_in;
6618 	pha.pa_flags = PFIL_IN;
6619 	pha.pa_rulname = "default-in6";
6620 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6621 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6622 	pla.pa_head = V_inet6_pfil_head;
6623 	pla.pa_hook = V_pf_ip6_in_hook;
6624 	ret = pfil_link(&pla);
6625 	MPASS(ret == 0);
6626 	pha.pa_mbuf_chk = pf_check6_out;
6627 	pha.pa_rulname = "default-out6";
6628 	pha.pa_flags = PFIL_OUT;
6629 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6630 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6631 	pla.pa_head = V_inet6_pfil_head;
6632 	pla.pa_hook = V_pf_ip6_out_hook;
6633 	ret = pfil_link(&pla);
6634 	MPASS(ret == 0);
6635 	if (V_pf_filter_local) {
6636 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6637 		pla.pa_head = V_inet6_local_pfil_head;
6638 		pla.pa_hook = V_pf_ip6_out_hook;
6639 		ret = pfil_link(&pla);
6640 		MPASS(ret == 0);
6641 	}
6642 #endif
6643 
6644 	atomic_store_bool(&V_pf_pfil_hooked, true);
6645 }
6646 
6647 static void
dehook_pf_eth(void)6648 dehook_pf_eth(void)
6649 {
6650 
6651 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6652 		return;
6653 
6654 	pfil_remove_hook(V_pf_eth_in_hook);
6655 	pfil_remove_hook(V_pf_eth_out_hook);
6656 
6657 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6658 }
6659 
6660 static void
dehook_pf(void)6661 dehook_pf(void)
6662 {
6663 
6664 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6665 		return;
6666 
6667 #ifdef INET
6668 	pfil_remove_hook(V_pf_ip4_in_hook);
6669 	pfil_remove_hook(V_pf_ip4_out_hook);
6670 #endif
6671 #ifdef INET6
6672 	pfil_remove_hook(V_pf_ip6_in_hook);
6673 	pfil_remove_hook(V_pf_ip6_out_hook);
6674 #endif
6675 
6676 	atomic_store_bool(&V_pf_pfil_hooked, false);
6677 }
6678 
6679 static void
pf_load_vnet(void)6680 pf_load_vnet(void)
6681 {
6682 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6683 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6684 
6685 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6686 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6687 
6688 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6689 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6690 #ifdef ALTQ
6691 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6692 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6693 #endif
6694 
6695 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6696 
6697 	pfattach_vnet();
6698 	V_pf_vnet_active = 1;
6699 }
6700 
6701 static int
pf_load(void)6702 pf_load(void)
6703 {
6704 	int error;
6705 
6706 	sx_init(&pf_end_lock, "pf end thread");
6707 
6708 	pf_mtag_initialize();
6709 
6710 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6711 	if (pf_dev == NULL)
6712 		return (ENOMEM);
6713 
6714 	pf_end_threads = 0;
6715 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6716 	if (error != 0)
6717 		return (error);
6718 
6719 	pfi_initialize();
6720 
6721 	return (0);
6722 }
6723 
6724 static void
pf_unload_vnet(void)6725 pf_unload_vnet(void)
6726 {
6727 	int ret __diagused;
6728 
6729 	V_pf_vnet_active = 0;
6730 	V_pf_status.running = 0;
6731 	dehook_pf();
6732 	dehook_pf_eth();
6733 
6734 	PF_RULES_WLOCK();
6735 	pf_syncookies_cleanup();
6736 	shutdown_pf();
6737 	PF_RULES_WUNLOCK();
6738 
6739 	/* Make sure we've cleaned up ethernet rules before we continue. */
6740 	NET_EPOCH_DRAIN_CALLBACKS();
6741 
6742 	ret = swi_remove(V_pf_swi_cookie);
6743 	MPASS(ret == 0);
6744 	ret = intr_event_destroy(V_pf_swi_ie);
6745 	MPASS(ret == 0);
6746 
6747 	pf_unload_vnet_purge();
6748 
6749 	pf_normalize_cleanup();
6750 	PF_RULES_WLOCK();
6751 	pfi_cleanup_vnet();
6752 	PF_RULES_WUNLOCK();
6753 	pfr_cleanup();
6754 	pf_osfp_flush();
6755 	pf_cleanup();
6756 	if (IS_DEFAULT_VNET(curvnet))
6757 		pf_mtag_cleanup();
6758 
6759 	pf_cleanup_tagset(&V_pf_tags);
6760 #ifdef ALTQ
6761 	pf_cleanup_tagset(&V_pf_qids);
6762 #endif
6763 	uma_zdestroy(V_pf_tag_z);
6764 
6765 #ifdef PF_WANT_32_TO_64_COUNTER
6766 	PF_RULES_WLOCK();
6767 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6768 
6769 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6770 	MPASS(V_pf_allkifcount == 0);
6771 
6772 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6773 	V_pf_allrulecount--;
6774 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6775 
6776 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6777 	MPASS(V_pf_allrulecount == 0);
6778 
6779 	PF_RULES_WUNLOCK();
6780 
6781 	free(V_pf_kifmarker, PFI_MTYPE);
6782 	free(V_pf_rulemarker, M_PFRULE);
6783 #endif
6784 
6785 	/* Free counters last as we updated them during shutdown. */
6786 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6787 	for (int i = 0; i < 2; i++) {
6788 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6789 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6790 	}
6791 	counter_u64_free(V_pf_default_rule.states_cur);
6792 	counter_u64_free(V_pf_default_rule.states_tot);
6793 	counter_u64_free(V_pf_default_rule.src_nodes);
6794 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6795 
6796 	for (int i = 0; i < PFRES_MAX; i++)
6797 		counter_u64_free(V_pf_status.counters[i]);
6798 	for (int i = 0; i < KLCNT_MAX; i++)
6799 		counter_u64_free(V_pf_status.lcounters[i]);
6800 	for (int i = 0; i < FCNT_MAX; i++)
6801 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6802 	for (int i = 0; i < SCNT_MAX; i++)
6803 		counter_u64_free(V_pf_status.scounters[i]);
6804 
6805 	rm_destroy(&V_pf_rules_lock);
6806 	sx_destroy(&V_pf_ioctl_lock);
6807 }
6808 
6809 static void
pf_unload(void)6810 pf_unload(void)
6811 {
6812 
6813 	sx_xlock(&pf_end_lock);
6814 	pf_end_threads = 1;
6815 	while (pf_end_threads < 2) {
6816 		wakeup_one(pf_purge_thread);
6817 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6818 	}
6819 	sx_xunlock(&pf_end_lock);
6820 
6821 	pf_nl_unregister();
6822 
6823 	if (pf_dev != NULL)
6824 		destroy_dev(pf_dev);
6825 
6826 	pfi_cleanup();
6827 
6828 	sx_destroy(&pf_end_lock);
6829 }
6830 
6831 static void
vnet_pf_init(void * unused __unused)6832 vnet_pf_init(void *unused __unused)
6833 {
6834 
6835 	pf_load_vnet();
6836 }
6837 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6838     vnet_pf_init, NULL);
6839 
6840 static void
vnet_pf_uninit(const void * unused __unused)6841 vnet_pf_uninit(const void *unused __unused)
6842 {
6843 
6844 	pf_unload_vnet();
6845 }
6846 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6847 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6848     vnet_pf_uninit, NULL);
6849 
6850 static int
pf_modevent(module_t mod,int type,void * data)6851 pf_modevent(module_t mod, int type, void *data)
6852 {
6853 	int error = 0;
6854 
6855 	switch(type) {
6856 	case MOD_LOAD:
6857 		error = pf_load();
6858 		pf_nl_register();
6859 		break;
6860 	case MOD_UNLOAD:
6861 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6862 		 * the vnet_pf_uninit()s */
6863 		break;
6864 	default:
6865 		error = EINVAL;
6866 		break;
6867 	}
6868 
6869 	return (error);
6870 }
6871 
6872 static moduledata_t pf_mod = {
6873 	"pf",
6874 	pf_modevent,
6875 	0
6876 };
6877 
6878 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6879 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6880 MODULE_VERSION(pf, PF_MODVER);
6881