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