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 error = EBUSY;
2159 #ifdef ALTQ
2160 /* set queue IDs */
2161 if (rule->qname[0] != 0) {
2162 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2163 error = EBUSY;
2164 else if (rule->pqname[0] != 0) {
2165 if ((rule->pqid =
2166 pf_qname2qid(rule->pqname)) == 0)
2167 error = 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 error = EBUSY;
2175 if (rule->match_tagname[0])
2176 if ((rule->match_tag =
2177 pf_tagname2tag(rule->match_tagname)) == 0)
2178 error = EBUSY;
2179 if (rule->rt && !rule->direction)
2180 error = EINVAL;
2181 if (!rule->log)
2182 rule->logif = 0;
2183 if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit,
2184 rule->pktrate.seconds))
2185 error = ENOMEM;
2186 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2187 error = ENOMEM;
2188 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2189 error = ENOMEM;
2190 if (pf_kanchor_setup(rule, ruleset, anchor_call))
2191 error = EINVAL;
2192 if (rule->scrub_flags & PFSTATE_SETPRIO &&
2193 (rule->set_prio[0] > PF_PRIO_MAX ||
2194 rule->set_prio[1] > PF_PRIO_MAX))
2195 error = EINVAL;
2196 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 error = 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 error = 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 error = EINVAL;
2234 }
2235
2236 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) {
2237 error = 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 error = EINVAL;
2243 }
2244
2245 if (error) {
2246 pf_free_rule(rule);
2247 rule = NULL;
2248 ERROUT(error);
2249 }
2250
2251 rule->nat.cur = TAILQ_FIRST(&rule->nat.list);
2252 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list);
2253 rule->route.cur = TAILQ_FIRST(&rule->route.list);
2254 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2255 rule, entries);
2256 ruleset->rules[rs_num].inactive.rcount++;
2257
2258 PF_RULES_WUNLOCK();
2259 pf_hash_rule(rule);
2260 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2261 PF_RULES_WLOCK();
2262 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2263 ruleset->rules[rs_num].inactive.rcount--;
2264 pf_free_rule(rule);
2265 rule = NULL;
2266 ERROUT(EEXIST);
2267 }
2268 PF_CONFIG_UNLOCK();
2269
2270 return (0);
2271
2272 #undef ERROUT
2273 errout:
2274 PF_RULES_WUNLOCK();
2275 PF_CONFIG_UNLOCK();
2276 errout_unlocked:
2277 pf_kkif_free(rcv_kif);
2278 pf_kkif_free(kif);
2279 pf_krule_free(rule);
2280 return (error);
2281 }
2282
2283 static bool
pf_label_match(const struct pf_krule * rule,const char * label)2284 pf_label_match(const struct pf_krule *rule, const char *label)
2285 {
2286 int i = 0;
2287
2288 while (*rule->label[i]) {
2289 if (strcmp(rule->label[i], label) == 0)
2290 return (true);
2291 i++;
2292 }
2293
2294 return (false);
2295 }
2296
2297 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)2298 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2299 {
2300 struct pf_kstate *s;
2301 int more = 0;
2302
2303 s = pf_find_state_all(key, dir, &more);
2304 if (s == NULL)
2305 return (0);
2306
2307 if (more) {
2308 PF_STATE_UNLOCK(s);
2309 return (0);
2310 }
2311
2312 pf_remove_state(s);
2313 return (1);
2314 }
2315
2316 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)2317 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2318 {
2319 struct pf_kstate *s;
2320 struct pf_state_key *sk;
2321 struct pf_addr *srcaddr, *dstaddr;
2322 struct pf_state_key_cmp match_key;
2323 int idx, killed = 0;
2324 unsigned int dir;
2325 u_int16_t srcport, dstport;
2326 struct pfi_kkif *kif;
2327
2328 relock_DIOCKILLSTATES:
2329 PF_HASHROW_LOCK(ih);
2330 LIST_FOREACH(s, &ih->states, entry) {
2331 /* For floating states look at the original kif. */
2332 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2333
2334 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2335 if (s->direction == PF_OUT) {
2336 srcaddr = &sk->addr[1];
2337 dstaddr = &sk->addr[0];
2338 srcport = sk->port[1];
2339 dstport = sk->port[0];
2340 } else {
2341 srcaddr = &sk->addr[0];
2342 dstaddr = &sk->addr[1];
2343 srcport = sk->port[0];
2344 dstport = sk->port[1];
2345 }
2346
2347 if (psk->psk_af && sk->af != psk->psk_af)
2348 continue;
2349
2350 if (psk->psk_proto && psk->psk_proto != sk->proto)
2351 continue;
2352
2353 if (! pf_match_addr(psk->psk_src.neg,
2354 &psk->psk_src.addr.v.a.addr,
2355 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2356 continue;
2357
2358 if (! pf_match_addr(psk->psk_dst.neg,
2359 &psk->psk_dst.addr.v.a.addr,
2360 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2361 continue;
2362
2363 if (! pf_match_addr(psk->psk_rt_addr.neg,
2364 &psk->psk_rt_addr.addr.v.a.addr,
2365 &psk->psk_rt_addr.addr.v.a.mask,
2366 &s->act.rt_addr, sk->af))
2367 continue;
2368
2369 if (psk->psk_src.port_op != 0 &&
2370 ! pf_match_port(psk->psk_src.port_op,
2371 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2372 continue;
2373
2374 if (psk->psk_dst.port_op != 0 &&
2375 ! pf_match_port(psk->psk_dst.port_op,
2376 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2377 continue;
2378
2379 if (psk->psk_label[0] &&
2380 ! pf_label_match(s->rule, psk->psk_label))
2381 continue;
2382
2383 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2384 kif->pfik_name))
2385 continue;
2386
2387 if (psk->psk_kill_match) {
2388 /* Create the key to find matching states, with lock
2389 * held. */
2390
2391 bzero(&match_key, sizeof(match_key));
2392
2393 if (s->direction == PF_OUT) {
2394 dir = PF_IN;
2395 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2396 } else {
2397 dir = PF_OUT;
2398 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2399 }
2400
2401 match_key.af = s->key[idx]->af;
2402 match_key.proto = s->key[idx]->proto;
2403 pf_addrcpy(&match_key.addr[0],
2404 &s->key[idx]->addr[1], match_key.af);
2405 match_key.port[0] = s->key[idx]->port[1];
2406 pf_addrcpy(&match_key.addr[1],
2407 &s->key[idx]->addr[0], match_key.af);
2408 match_key.port[1] = s->key[idx]->port[0];
2409 }
2410
2411 pf_remove_state(s);
2412 killed++;
2413
2414 if (psk->psk_kill_match)
2415 killed += pf_kill_matching_state(&match_key, dir);
2416
2417 goto relock_DIOCKILLSTATES;
2418 }
2419 PF_HASHROW_UNLOCK(ih);
2420
2421 return (killed);
2422 }
2423
2424 void
unhandled_af(int af)2425 unhandled_af(int af)
2426 {
2427 panic("unhandled af %d", af);
2428 }
2429
2430 int
pf_start(void)2431 pf_start(void)
2432 {
2433 int error = 0;
2434
2435 sx_xlock(&V_pf_ioctl_lock);
2436 if (V_pf_status.running)
2437 error = EEXIST;
2438 else {
2439 hook_pf();
2440 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2441 hook_pf_eth();
2442 V_pf_status.running = 1;
2443 V_pf_status.since = time_uptime;
2444 new_unrhdr64(&V_pf_stateid, time_second);
2445
2446 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2447 }
2448 sx_xunlock(&V_pf_ioctl_lock);
2449
2450 return (error);
2451 }
2452
2453 int
pf_stop(void)2454 pf_stop(void)
2455 {
2456 int error = 0;
2457
2458 sx_xlock(&V_pf_ioctl_lock);
2459 if (!V_pf_status.running)
2460 error = ENOENT;
2461 else {
2462 V_pf_status.running = 0;
2463 dehook_pf();
2464 dehook_pf_eth();
2465 V_pf_status.since = time_uptime;
2466 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2467 }
2468 sx_xunlock(&V_pf_ioctl_lock);
2469
2470 return (error);
2471 }
2472
2473 void
pf_ioctl_clear_status(void)2474 pf_ioctl_clear_status(void)
2475 {
2476 PF_RULES_WLOCK();
2477 for (int i = 0; i < PFRES_MAX; i++)
2478 counter_u64_zero(V_pf_status.counters[i]);
2479 for (int i = 0; i < FCNT_MAX; i++)
2480 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2481 for (int i = 0; i < SCNT_MAX; i++)
2482 counter_u64_zero(V_pf_status.scounters[i]);
2483 for (int i = 0; i < KLCNT_MAX; i++)
2484 counter_u64_zero(V_pf_status.lcounters[i]);
2485 V_pf_status.since = time_uptime;
2486 if (*V_pf_status.ifname)
2487 pfi_update_status(V_pf_status.ifname, NULL);
2488 PF_RULES_WUNLOCK();
2489 }
2490
2491 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2492 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2493 {
2494 uint32_t old;
2495
2496 if (timeout < 0 || timeout >= PFTM_MAX ||
2497 seconds < 0)
2498 return (EINVAL);
2499
2500 PF_RULES_WLOCK();
2501 old = V_pf_default_rule.timeout[timeout];
2502 if (timeout == PFTM_INTERVAL && seconds == 0)
2503 seconds = 1;
2504 V_pf_default_rule.timeout[timeout] = seconds;
2505 if (timeout == PFTM_INTERVAL && seconds < old)
2506 wakeup(pf_purge_thread);
2507
2508 if (prev_seconds != NULL)
2509 *prev_seconds = old;
2510
2511 PF_RULES_WUNLOCK();
2512
2513 return (0);
2514 }
2515
2516 int
pf_ioctl_get_timeout(int timeout,int * seconds)2517 pf_ioctl_get_timeout(int timeout, int *seconds)
2518 {
2519 PF_RULES_RLOCK_TRACKER;
2520
2521 if (timeout < 0 || timeout >= PFTM_MAX)
2522 return (EINVAL);
2523
2524 PF_RULES_RLOCK();
2525 *seconds = V_pf_default_rule.timeout[timeout];
2526 PF_RULES_RUNLOCK();
2527
2528 return (0);
2529 }
2530
2531 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2532 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2533 {
2534
2535 PF_RULES_WLOCK();
2536 if (index < 0 || index >= PF_LIMIT_MAX ||
2537 V_pf_limits[index].zone == NULL) {
2538 PF_RULES_WUNLOCK();
2539 return (EINVAL);
2540 }
2541 uma_zone_set_max(V_pf_limits[index].zone,
2542 limit == 0 ? INT_MAX : limit);
2543 if (old_limit != NULL)
2544 *old_limit = V_pf_limits[index].limit;
2545 V_pf_limits[index].limit = limit;
2546 PF_RULES_WUNLOCK();
2547
2548 return (0);
2549 }
2550
2551 int
pf_ioctl_get_limit(int index,unsigned int * limit)2552 pf_ioctl_get_limit(int index, unsigned int *limit)
2553 {
2554 PF_RULES_RLOCK_TRACKER;
2555
2556 if (index < 0 || index >= PF_LIMIT_MAX)
2557 return (EINVAL);
2558
2559 PF_RULES_RLOCK();
2560 *limit = V_pf_limits[index].limit;
2561 PF_RULES_RUNLOCK();
2562
2563 return (0);
2564 }
2565
2566 int
pf_ioctl_begin_addrs(uint32_t * ticket)2567 pf_ioctl_begin_addrs(uint32_t *ticket)
2568 {
2569 PF_RULES_WLOCK();
2570 pf_empty_kpool(&V_pf_pabuf[0]);
2571 pf_empty_kpool(&V_pf_pabuf[1]);
2572 pf_empty_kpool(&V_pf_pabuf[2]);
2573 *ticket = ++V_ticket_pabuf;
2574 PF_RULES_WUNLOCK();
2575
2576 return (0);
2577 }
2578
2579 int
pf_ioctl_add_addr(struct pf_nl_pooladdr * pp)2580 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
2581 {
2582 struct pf_kpooladdr *pa = NULL;
2583 struct pfi_kkif *kif = NULL;
2584 int error;
2585
2586 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2587 pp->which != PF_RT)
2588 return (EINVAL);
2589
2590 switch (pp->af) {
2591 #ifdef INET
2592 case AF_INET:
2593 /* FALLTHROUGH */
2594 #endif /* INET */
2595 #ifdef INET6
2596 case AF_INET6:
2597 /* FALLTHROUGH */
2598 #endif /* INET6 */
2599 case AF_UNSPEC:
2600 break;
2601 default:
2602 return (EAFNOSUPPORT);
2603 }
2604
2605 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2606 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2607 pp->addr.addr.type != PF_ADDR_TABLE)
2608 return (EINVAL);
2609
2610 if (pp->addr.addr.p.dyn != NULL)
2611 return (EINVAL);
2612
2613 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2614 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2615 if (error != 0)
2616 goto out;
2617 if (pa->ifname[0])
2618 kif = pf_kkif_create(M_WAITOK);
2619 PF_RULES_WLOCK();
2620 if (pp->ticket != V_ticket_pabuf) {
2621 PF_RULES_WUNLOCK();
2622 if (pa->ifname[0])
2623 pf_kkif_free(kif);
2624 error = EBUSY;
2625 goto out;
2626 }
2627 if (pa->ifname[0]) {
2628 pa->kif = pfi_kkif_attach(kif, pa->ifname);
2629 kif = NULL;
2630 pfi_kkif_ref(pa->kif);
2631 } else
2632 pa->kif = NULL;
2633 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2634 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2635 if (pa->ifname[0])
2636 pfi_kkif_unref(pa->kif);
2637 PF_RULES_WUNLOCK();
2638 goto out;
2639 }
2640 switch (pp->which) {
2641 case PF_NAT:
2642 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries);
2643 break;
2644 case PF_RDR:
2645 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries);
2646 break;
2647 case PF_RT:
2648 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries);
2649 break;
2650 }
2651 PF_RULES_WUNLOCK();
2652
2653 return (0);
2654
2655 out:
2656 free(pa, M_PFRULE);
2657 return (error);
2658 }
2659
2660 int
pf_ioctl_get_addrs(struct pf_nl_pooladdr * pp)2661 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
2662 {
2663 struct pf_kpool *pool;
2664 struct pf_kpooladdr *pa;
2665
2666 PF_RULES_RLOCK_TRACKER;
2667
2668 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2669 pp->which != PF_RT)
2670 return (EINVAL);
2671
2672 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2673 pp->nr = 0;
2674
2675 PF_RULES_RLOCK();
2676 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2677 pp->r_num, 0, 1, 0, pp->which);
2678 if (pool == NULL) {
2679 PF_RULES_RUNLOCK();
2680 return (EBUSY);
2681 }
2682 TAILQ_FOREACH(pa, &pool->list, entries)
2683 pp->nr++;
2684 PF_RULES_RUNLOCK();
2685
2686 return (0);
2687 }
2688
2689 int
pf_ioctl_get_addr(struct pf_nl_pooladdr * pp)2690 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
2691 {
2692 struct pf_kpool *pool;
2693 struct pf_kpooladdr *pa;
2694 u_int32_t nr = 0;
2695
2696 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2697 pp->which != PF_RT)
2698 return (EINVAL);
2699
2700 PF_RULES_RLOCK_TRACKER;
2701
2702 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2703
2704 PF_RULES_RLOCK();
2705 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2706 pp->r_num, 0, 1, 1, pp->which);
2707 if (pool == NULL) {
2708 PF_RULES_RUNLOCK();
2709 return (EBUSY);
2710 }
2711 pa = TAILQ_FIRST(&pool->list);
2712 while ((pa != NULL) && (nr < pp->nr)) {
2713 pa = TAILQ_NEXT(pa, entries);
2714 nr++;
2715 }
2716 if (pa == NULL) {
2717 PF_RULES_RUNLOCK();
2718 return (EBUSY);
2719 }
2720 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2721 pf_addr_copyout(&pp->addr.addr);
2722 PF_RULES_RUNLOCK();
2723
2724 return (0);
2725 }
2726
2727 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)2728 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
2729 {
2730 struct pf_kruleset *ruleset;
2731 struct pf_kanchor *anchor;
2732
2733 PF_RULES_RLOCK_TRACKER;
2734
2735 pr->path[sizeof(pr->path) - 1] = 0;
2736
2737 PF_RULES_RLOCK();
2738 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2739 PF_RULES_RUNLOCK();
2740 return (ENOENT);
2741 }
2742 pr->nr = 0;
2743 if (ruleset == &pf_main_ruleset) {
2744 /* XXX kludge for pf_main_ruleset */
2745 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2746 if (anchor->parent == NULL)
2747 pr->nr++;
2748 } else {
2749 RB_FOREACH(anchor, pf_kanchor_node,
2750 &ruleset->anchor->children)
2751 pr->nr++;
2752 }
2753 PF_RULES_RUNLOCK();
2754
2755 return (0);
2756 }
2757
2758 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)2759 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
2760 {
2761 struct pf_kruleset *ruleset;
2762 struct pf_kanchor *anchor;
2763 u_int32_t nr = 0;
2764 int error = 0;
2765
2766 PF_RULES_RLOCK_TRACKER;
2767
2768 PF_RULES_RLOCK();
2769 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2770 PF_RULES_RUNLOCK();
2771 return (ENOENT);
2772 }
2773
2774 pr->name[0] = 0;
2775 if (ruleset == &pf_main_ruleset) {
2776 /* XXX kludge for pf_main_ruleset */
2777 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2778 if (anchor->parent == NULL && nr++ == pr->nr) {
2779 strlcpy(pr->name, anchor->name,
2780 sizeof(pr->name));
2781 break;
2782 }
2783 } else {
2784 RB_FOREACH(anchor, pf_kanchor_node,
2785 &ruleset->anchor->children)
2786 if (nr++ == pr->nr) {
2787 strlcpy(pr->name, anchor->name,
2788 sizeof(pr->name));
2789 break;
2790 }
2791 }
2792 if (!pr->name[0])
2793 error = EBUSY;
2794 PF_RULES_RUNLOCK();
2795
2796 return (error);
2797 }
2798
2799 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2800 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2801 {
2802 int error = 0;
2803 PF_RULES_RLOCK_TRACKER;
2804
2805 #define ERROUT_IOCTL(target, x) \
2806 do { \
2807 error = (x); \
2808 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
2809 goto target; \
2810 } while (0)
2811
2812
2813 /* XXX keep in sync with switch() below */
2814 if (securelevel_gt(td->td_ucred, 2))
2815 switch (cmd) {
2816 case DIOCGETRULES:
2817 case DIOCGETRULENV:
2818 case DIOCGETADDRS:
2819 case DIOCGETADDR:
2820 case DIOCGETSTATE:
2821 case DIOCGETSTATENV:
2822 case DIOCSETSTATUSIF:
2823 case DIOCGETSTATUSNV:
2824 case DIOCCLRSTATUS:
2825 case DIOCNATLOOK:
2826 case DIOCSETDEBUG:
2827 #ifdef COMPAT_FREEBSD14
2828 case DIOCGETSTATES:
2829 case DIOCGETSTATESV2:
2830 #endif
2831 case DIOCGETTIMEOUT:
2832 case DIOCCLRRULECTRS:
2833 case DIOCGETLIMIT:
2834 case DIOCGETALTQSV0:
2835 case DIOCGETALTQSV1:
2836 case DIOCGETALTQV0:
2837 case DIOCGETALTQV1:
2838 case DIOCGETQSTATSV0:
2839 case DIOCGETQSTATSV1:
2840 case DIOCGETRULESETS:
2841 case DIOCGETRULESET:
2842 case DIOCRGETTABLES:
2843 case DIOCRGETTSTATS:
2844 case DIOCRCLRTSTATS:
2845 case DIOCRCLRADDRS:
2846 case DIOCRADDADDRS:
2847 case DIOCRDELADDRS:
2848 case DIOCRSETADDRS:
2849 case DIOCRGETADDRS:
2850 case DIOCRGETASTATS:
2851 case DIOCRCLRASTATS:
2852 case DIOCRTSTADDRS:
2853 case DIOCOSFPGET:
2854 case DIOCGETSRCNODES:
2855 case DIOCCLRSRCNODES:
2856 case DIOCGETSYNCOOKIES:
2857 case DIOCIGETIFACES:
2858 case DIOCGIFSPEEDV0:
2859 case DIOCGIFSPEEDV1:
2860 case DIOCSETIFFLAG:
2861 case DIOCCLRIFFLAG:
2862 case DIOCGETETHRULES:
2863 case DIOCGETETHRULE:
2864 case DIOCGETETHRULESETS:
2865 case DIOCGETETHRULESET:
2866 break;
2867 case DIOCRCLRTABLES:
2868 case DIOCRADDTABLES:
2869 case DIOCRDELTABLES:
2870 case DIOCRSETTFLAGS:
2871 if (((struct pfioc_table *)addr)->pfrio_flags &
2872 PFR_FLAG_DUMMY)
2873 break; /* dummy operation ok */
2874 return (EPERM);
2875 default:
2876 return (EPERM);
2877 }
2878
2879 if (!(flags & FWRITE))
2880 switch (cmd) {
2881 case DIOCGETRULES:
2882 case DIOCGETADDRS:
2883 case DIOCGETADDR:
2884 case DIOCGETSTATE:
2885 case DIOCGETSTATENV:
2886 case DIOCGETSTATUSNV:
2887 #ifdef COMPAT_FREEBSD14
2888 case DIOCGETSTATES:
2889 case DIOCGETSTATESV2:
2890 #endif
2891 case DIOCGETTIMEOUT:
2892 case DIOCGETLIMIT:
2893 case DIOCGETALTQSV0:
2894 case DIOCGETALTQSV1:
2895 case DIOCGETALTQV0:
2896 case DIOCGETALTQV1:
2897 case DIOCGETQSTATSV0:
2898 case DIOCGETQSTATSV1:
2899 case DIOCGETRULESETS:
2900 case DIOCGETRULESET:
2901 case DIOCNATLOOK:
2902 case DIOCRGETTABLES:
2903 case DIOCRGETTSTATS:
2904 case DIOCRGETADDRS:
2905 case DIOCRGETASTATS:
2906 case DIOCRTSTADDRS:
2907 case DIOCOSFPGET:
2908 case DIOCGETSRCNODES:
2909 case DIOCGETSYNCOOKIES:
2910 case DIOCIGETIFACES:
2911 case DIOCGIFSPEEDV1:
2912 case DIOCGIFSPEEDV0:
2913 case DIOCGETRULENV:
2914 case DIOCGETETHRULES:
2915 case DIOCGETETHRULE:
2916 case DIOCGETETHRULESETS:
2917 case DIOCGETETHRULESET:
2918 break;
2919 case DIOCRCLRTABLES:
2920 case DIOCRADDTABLES:
2921 case DIOCRDELTABLES:
2922 case DIOCRCLRTSTATS:
2923 case DIOCRCLRADDRS:
2924 case DIOCRADDADDRS:
2925 case DIOCRDELADDRS:
2926 case DIOCRSETADDRS:
2927 case DIOCRSETTFLAGS:
2928 if (((struct pfioc_table *)addr)->pfrio_flags &
2929 PFR_FLAG_DUMMY) {
2930 flags |= FWRITE; /* need write lock for dummy */
2931 break; /* dummy operation ok */
2932 }
2933 return (EACCES);
2934 default:
2935 return (EACCES);
2936 }
2937
2938 CURVNET_SET(TD_TO_VNET(td));
2939
2940 switch (cmd) {
2941 #ifdef COMPAT_FREEBSD14
2942 case DIOCSTART:
2943 error = pf_start();
2944 break;
2945
2946 case DIOCSTOP:
2947 error = pf_stop();
2948 break;
2949 #endif
2950
2951 case DIOCGETETHRULES: {
2952 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2953 nvlist_t *nvl;
2954 void *packed;
2955 struct pf_keth_rule *tail;
2956 struct pf_keth_ruleset *rs;
2957 u_int32_t ticket, nr;
2958 const char *anchor = "";
2959
2960 nvl = NULL;
2961 packed = NULL;
2962
2963 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2964
2965 if (nv->len > pf_ioctl_maxcount)
2966 ERROUT(ENOMEM);
2967
2968 /* Copy the request in */
2969 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2970 error = copyin(nv->data, packed, nv->len);
2971 if (error)
2972 ERROUT(error);
2973
2974 nvl = nvlist_unpack(packed, nv->len, 0);
2975 if (nvl == NULL)
2976 ERROUT(EBADMSG);
2977
2978 if (! nvlist_exists_string(nvl, "anchor"))
2979 ERROUT(EBADMSG);
2980
2981 anchor = nvlist_get_string(nvl, "anchor");
2982
2983 rs = pf_find_keth_ruleset(anchor);
2984
2985 nvlist_destroy(nvl);
2986 nvl = NULL;
2987 free(packed, M_NVLIST);
2988 packed = NULL;
2989
2990 if (rs == NULL)
2991 ERROUT(ENOENT);
2992
2993 /* Reply */
2994 nvl = nvlist_create(0);
2995 if (nvl == NULL)
2996 ERROUT(ENOMEM);
2997
2998 PF_RULES_RLOCK();
2999
3000 ticket = rs->active.ticket;
3001 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
3002 if (tail)
3003 nr = tail->nr + 1;
3004 else
3005 nr = 0;
3006
3007 PF_RULES_RUNLOCK();
3008
3009 nvlist_add_number(nvl, "ticket", ticket);
3010 nvlist_add_number(nvl, "nr", nr);
3011
3012 packed = nvlist_pack(nvl, &nv->len);
3013 if (packed == NULL)
3014 ERROUT(ENOMEM);
3015
3016 if (nv->size == 0)
3017 ERROUT(0);
3018 else if (nv->size < nv->len)
3019 ERROUT(ENOSPC);
3020
3021 error = copyout(packed, nv->data, nv->len);
3022
3023 #undef ERROUT
3024 DIOCGETETHRULES_error:
3025 free(packed, M_NVLIST);
3026 nvlist_destroy(nvl);
3027 break;
3028 }
3029
3030 case DIOCGETETHRULE: {
3031 struct epoch_tracker et;
3032 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3033 nvlist_t *nvl = NULL;
3034 void *nvlpacked = NULL;
3035 struct pf_keth_rule *rule = NULL;
3036 struct pf_keth_ruleset *rs;
3037 u_int32_t ticket, nr;
3038 bool clear = false;
3039 const char *anchor;
3040
3041 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
3042
3043 if (nv->len > pf_ioctl_maxcount)
3044 ERROUT(ENOMEM);
3045
3046 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3047 error = copyin(nv->data, nvlpacked, nv->len);
3048 if (error)
3049 ERROUT(error);
3050
3051 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3052 if (nvl == NULL)
3053 ERROUT(EBADMSG);
3054 if (! nvlist_exists_number(nvl, "ticket"))
3055 ERROUT(EBADMSG);
3056 ticket = nvlist_get_number(nvl, "ticket");
3057 if (! nvlist_exists_string(nvl, "anchor"))
3058 ERROUT(EBADMSG);
3059 anchor = nvlist_get_string(nvl, "anchor");
3060
3061 if (nvlist_exists_bool(nvl, "clear"))
3062 clear = nvlist_get_bool(nvl, "clear");
3063
3064 if (clear && !(flags & FWRITE))
3065 ERROUT(EACCES);
3066
3067 if (! nvlist_exists_number(nvl, "nr"))
3068 ERROUT(EBADMSG);
3069 nr = nvlist_get_number(nvl, "nr");
3070
3071 PF_RULES_RLOCK();
3072 rs = pf_find_keth_ruleset(anchor);
3073 if (rs == NULL) {
3074 PF_RULES_RUNLOCK();
3075 ERROUT(ENOENT);
3076 }
3077 if (ticket != rs->active.ticket) {
3078 PF_RULES_RUNLOCK();
3079 ERROUT(EBUSY);
3080 }
3081
3082 nvlist_destroy(nvl);
3083 nvl = NULL;
3084 free(nvlpacked, M_NVLIST);
3085 nvlpacked = NULL;
3086
3087 rule = TAILQ_FIRST(rs->active.rules);
3088 while ((rule != NULL) && (rule->nr != nr))
3089 rule = TAILQ_NEXT(rule, entries);
3090 if (rule == NULL) {
3091 PF_RULES_RUNLOCK();
3092 ERROUT(ENOENT);
3093 }
3094 /* Make sure rule can't go away. */
3095 NET_EPOCH_ENTER(et);
3096 PF_RULES_RUNLOCK();
3097 nvl = pf_keth_rule_to_nveth_rule(rule);
3098 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
3099 NET_EPOCH_EXIT(et);
3100 ERROUT(EBUSY);
3101 }
3102 NET_EPOCH_EXIT(et);
3103 if (nvl == NULL)
3104 ERROUT(ENOMEM);
3105
3106 nvlpacked = nvlist_pack(nvl, &nv->len);
3107 if (nvlpacked == NULL)
3108 ERROUT(ENOMEM);
3109
3110 if (nv->size == 0)
3111 ERROUT(0);
3112 else if (nv->size < nv->len)
3113 ERROUT(ENOSPC);
3114
3115 error = copyout(nvlpacked, nv->data, nv->len);
3116 if (error == 0 && clear) {
3117 counter_u64_zero(rule->evaluations);
3118 for (int i = 0; i < 2; i++) {
3119 counter_u64_zero(rule->packets[i]);
3120 counter_u64_zero(rule->bytes[i]);
3121 }
3122 }
3123
3124 #undef ERROUT
3125 DIOCGETETHRULE_error:
3126 free(nvlpacked, M_NVLIST);
3127 nvlist_destroy(nvl);
3128 break;
3129 }
3130
3131 case DIOCADDETHRULE: {
3132 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3133 nvlist_t *nvl = NULL;
3134 void *nvlpacked = NULL;
3135 struct pf_keth_rule *rule = NULL, *tail = NULL;
3136 struct pf_keth_ruleset *ruleset = NULL;
3137 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
3138 const char *anchor = "", *anchor_call = "";
3139
3140 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3141
3142 if (nv->len > pf_ioctl_maxcount)
3143 ERROUT(ENOMEM);
3144
3145 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3146 error = copyin(nv->data, nvlpacked, nv->len);
3147 if (error)
3148 ERROUT(error);
3149
3150 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3151 if (nvl == NULL)
3152 ERROUT(EBADMSG);
3153
3154 if (! nvlist_exists_number(nvl, "ticket"))
3155 ERROUT(EBADMSG);
3156
3157 if (nvlist_exists_string(nvl, "anchor"))
3158 anchor = nvlist_get_string(nvl, "anchor");
3159 if (nvlist_exists_string(nvl, "anchor_call"))
3160 anchor_call = nvlist_get_string(nvl, "anchor_call");
3161
3162 ruleset = pf_find_keth_ruleset(anchor);
3163 if (ruleset == NULL)
3164 ERROUT(EINVAL);
3165
3166 if (nvlist_get_number(nvl, "ticket") !=
3167 ruleset->inactive.ticket) {
3168 DPFPRINTF(PF_DEBUG_MISC,
3169 ("ticket: %d != %d\n",
3170 (u_int32_t)nvlist_get_number(nvl, "ticket"),
3171 ruleset->inactive.ticket));
3172 ERROUT(EBUSY);
3173 }
3174
3175 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3176 rule->timestamp = NULL;
3177
3178 error = pf_nveth_rule_to_keth_rule(nvl, rule);
3179 if (error != 0)
3180 ERROUT(error);
3181
3182 if (rule->ifname[0])
3183 kif = pf_kkif_create(M_WAITOK);
3184 if (rule->bridge_to_name[0])
3185 bridge_to_kif = pf_kkif_create(M_WAITOK);
3186 rule->evaluations = counter_u64_alloc(M_WAITOK);
3187 for (int i = 0; i < 2; i++) {
3188 rule->packets[i] = counter_u64_alloc(M_WAITOK);
3189 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3190 }
3191 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3192 M_WAITOK | M_ZERO);
3193
3194 PF_RULES_WLOCK();
3195
3196 if (rule->ifname[0]) {
3197 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3198 pfi_kkif_ref(rule->kif);
3199 } else
3200 rule->kif = NULL;
3201 if (rule->bridge_to_name[0]) {
3202 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3203 rule->bridge_to_name);
3204 pfi_kkif_ref(rule->bridge_to);
3205 } else
3206 rule->bridge_to = NULL;
3207
3208 #ifdef ALTQ
3209 /* set queue IDs */
3210 if (rule->qname[0] != 0) {
3211 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3212 error = EBUSY;
3213 else
3214 rule->qid = rule->qid;
3215 }
3216 #endif
3217 if (rule->tagname[0])
3218 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3219 error = EBUSY;
3220 if (rule->match_tagname[0])
3221 if ((rule->match_tag = pf_tagname2tag(
3222 rule->match_tagname)) == 0)
3223 error = EBUSY;
3224
3225 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3226 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3227 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3228 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3229
3230 if (error) {
3231 pf_free_eth_rule(rule);
3232 PF_RULES_WUNLOCK();
3233 ERROUT(error);
3234 }
3235
3236 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3237 pf_free_eth_rule(rule);
3238 PF_RULES_WUNLOCK();
3239 ERROUT(EINVAL);
3240 }
3241
3242 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3243 if (tail)
3244 rule->nr = tail->nr + 1;
3245 else
3246 rule->nr = 0;
3247
3248 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3249
3250 PF_RULES_WUNLOCK();
3251
3252 #undef ERROUT
3253 DIOCADDETHRULE_error:
3254 nvlist_destroy(nvl);
3255 free(nvlpacked, M_NVLIST);
3256 break;
3257 }
3258
3259 case DIOCGETETHRULESETS: {
3260 struct epoch_tracker et;
3261 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3262 nvlist_t *nvl = NULL;
3263 void *nvlpacked = NULL;
3264 struct pf_keth_ruleset *ruleset;
3265 struct pf_keth_anchor *anchor;
3266 int nr = 0;
3267
3268 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3269
3270 if (nv->len > pf_ioctl_maxcount)
3271 ERROUT(ENOMEM);
3272
3273 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3274 error = copyin(nv->data, nvlpacked, nv->len);
3275 if (error)
3276 ERROUT(error);
3277
3278 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3279 if (nvl == NULL)
3280 ERROUT(EBADMSG);
3281 if (! nvlist_exists_string(nvl, "path"))
3282 ERROUT(EBADMSG);
3283
3284 NET_EPOCH_ENTER(et);
3285
3286 if ((ruleset = pf_find_keth_ruleset(
3287 nvlist_get_string(nvl, "path"))) == NULL) {
3288 NET_EPOCH_EXIT(et);
3289 ERROUT(ENOENT);
3290 }
3291
3292 if (ruleset->anchor == NULL) {
3293 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3294 if (anchor->parent == NULL)
3295 nr++;
3296 } else {
3297 RB_FOREACH(anchor, pf_keth_anchor_node,
3298 &ruleset->anchor->children)
3299 nr++;
3300 }
3301
3302 NET_EPOCH_EXIT(et);
3303
3304 nvlist_destroy(nvl);
3305 nvl = NULL;
3306 free(nvlpacked, M_NVLIST);
3307 nvlpacked = NULL;
3308
3309 nvl = nvlist_create(0);
3310 if (nvl == NULL)
3311 ERROUT(ENOMEM);
3312
3313 nvlist_add_number(nvl, "nr", nr);
3314
3315 nvlpacked = nvlist_pack(nvl, &nv->len);
3316 if (nvlpacked == NULL)
3317 ERROUT(ENOMEM);
3318
3319 if (nv->size == 0)
3320 ERROUT(0);
3321 else if (nv->size < nv->len)
3322 ERROUT(ENOSPC);
3323
3324 error = copyout(nvlpacked, nv->data, nv->len);
3325
3326 #undef ERROUT
3327 DIOCGETETHRULESETS_error:
3328 free(nvlpacked, M_NVLIST);
3329 nvlist_destroy(nvl);
3330 break;
3331 }
3332
3333 case DIOCGETETHRULESET: {
3334 struct epoch_tracker et;
3335 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3336 nvlist_t *nvl = NULL;
3337 void *nvlpacked = NULL;
3338 struct pf_keth_ruleset *ruleset;
3339 struct pf_keth_anchor *anchor;
3340 int nr = 0, req_nr = 0;
3341 bool found = false;
3342
3343 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3344
3345 if (nv->len > pf_ioctl_maxcount)
3346 ERROUT(ENOMEM);
3347
3348 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3349 error = copyin(nv->data, nvlpacked, nv->len);
3350 if (error)
3351 ERROUT(error);
3352
3353 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3354 if (nvl == NULL)
3355 ERROUT(EBADMSG);
3356 if (! nvlist_exists_string(nvl, "path"))
3357 ERROUT(EBADMSG);
3358 if (! nvlist_exists_number(nvl, "nr"))
3359 ERROUT(EBADMSG);
3360
3361 req_nr = nvlist_get_number(nvl, "nr");
3362
3363 NET_EPOCH_ENTER(et);
3364
3365 if ((ruleset = pf_find_keth_ruleset(
3366 nvlist_get_string(nvl, "path"))) == NULL) {
3367 NET_EPOCH_EXIT(et);
3368 ERROUT(ENOENT);
3369 }
3370
3371 nvlist_destroy(nvl);
3372 nvl = NULL;
3373 free(nvlpacked, M_NVLIST);
3374 nvlpacked = NULL;
3375
3376 nvl = nvlist_create(0);
3377 if (nvl == NULL) {
3378 NET_EPOCH_EXIT(et);
3379 ERROUT(ENOMEM);
3380 }
3381
3382 if (ruleset->anchor == NULL) {
3383 RB_FOREACH(anchor, pf_keth_anchor_global,
3384 &V_pf_keth_anchors) {
3385 if (anchor->parent == NULL && nr++ == req_nr) {
3386 found = true;
3387 break;
3388 }
3389 }
3390 } else {
3391 RB_FOREACH(anchor, pf_keth_anchor_node,
3392 &ruleset->anchor->children) {
3393 if (nr++ == req_nr) {
3394 found = true;
3395 break;
3396 }
3397 }
3398 }
3399
3400 NET_EPOCH_EXIT(et);
3401 if (found) {
3402 nvlist_add_number(nvl, "nr", nr);
3403 nvlist_add_string(nvl, "name", anchor->name);
3404 if (ruleset->anchor)
3405 nvlist_add_string(nvl, "path",
3406 ruleset->anchor->path);
3407 else
3408 nvlist_add_string(nvl, "path", "");
3409 } else {
3410 ERROUT(EBUSY);
3411 }
3412
3413 nvlpacked = nvlist_pack(nvl, &nv->len);
3414 if (nvlpacked == NULL)
3415 ERROUT(ENOMEM);
3416
3417 if (nv->size == 0)
3418 ERROUT(0);
3419 else if (nv->size < nv->len)
3420 ERROUT(ENOSPC);
3421
3422 error = copyout(nvlpacked, nv->data, nv->len);
3423
3424 #undef ERROUT
3425 DIOCGETETHRULESET_error:
3426 free(nvlpacked, M_NVLIST);
3427 nvlist_destroy(nvl);
3428 break;
3429 }
3430
3431 case DIOCADDRULENV: {
3432 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3433 nvlist_t *nvl = NULL;
3434 void *nvlpacked = NULL;
3435 struct pf_krule *rule = NULL;
3436 const char *anchor = "", *anchor_call = "";
3437 uint32_t ticket = 0, pool_ticket = 0;
3438
3439 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
3440
3441 if (nv->len > pf_ioctl_maxcount)
3442 ERROUT(ENOMEM);
3443
3444 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3445 error = copyin(nv->data, nvlpacked, nv->len);
3446 if (error)
3447 ERROUT(error);
3448
3449 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3450 if (nvl == NULL)
3451 ERROUT(EBADMSG);
3452
3453 if (! nvlist_exists_number(nvl, "ticket"))
3454 ERROUT(EINVAL);
3455 ticket = nvlist_get_number(nvl, "ticket");
3456
3457 if (! nvlist_exists_number(nvl, "pool_ticket"))
3458 ERROUT(EINVAL);
3459 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3460
3461 if (! nvlist_exists_nvlist(nvl, "rule"))
3462 ERROUT(EINVAL);
3463
3464 rule = pf_krule_alloc();
3465 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3466 rule);
3467 if (error)
3468 ERROUT(error);
3469
3470 if (nvlist_exists_string(nvl, "anchor"))
3471 anchor = nvlist_get_string(nvl, "anchor");
3472 if (nvlist_exists_string(nvl, "anchor_call"))
3473 anchor_call = nvlist_get_string(nvl, "anchor_call");
3474
3475 if ((error = nvlist_error(nvl)))
3476 ERROUT(error);
3477
3478 /* Frees rule on error */
3479 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3480 anchor_call, td->td_ucred->cr_ruid,
3481 td->td_proc ? td->td_proc->p_pid : 0);
3482
3483 nvlist_destroy(nvl);
3484 free(nvlpacked, M_NVLIST);
3485 break;
3486 #undef ERROUT
3487 DIOCADDRULENV_error:
3488 pf_krule_free(rule);
3489 nvlist_destroy(nvl);
3490 free(nvlpacked, M_NVLIST);
3491
3492 break;
3493 }
3494 case DIOCADDRULE: {
3495 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3496 struct pf_krule *rule;
3497
3498 rule = pf_krule_alloc();
3499 error = pf_rule_to_krule(&pr->rule, rule);
3500 if (error != 0) {
3501 pf_krule_free(rule);
3502 break;
3503 }
3504
3505 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3506
3507 /* Frees rule on error */
3508 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3509 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3510 td->td_proc ? td->td_proc->p_pid : 0);
3511 break;
3512 }
3513
3514 case DIOCGETRULES: {
3515 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3516
3517 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3518
3519 error = pf_ioctl_getrules(pr);
3520
3521 break;
3522 }
3523
3524 case DIOCGETRULENV: {
3525 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3526 nvlist_t *nvrule = NULL;
3527 nvlist_t *nvl = NULL;
3528 struct pf_kruleset *ruleset;
3529 struct pf_krule *rule;
3530 void *nvlpacked = NULL;
3531 int rs_num, nr;
3532 bool clear_counter = false;
3533
3534 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
3535
3536 if (nv->len > pf_ioctl_maxcount)
3537 ERROUT(ENOMEM);
3538
3539 /* Copy the request in */
3540 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3541 error = copyin(nv->data, nvlpacked, nv->len);
3542 if (error)
3543 ERROUT(error);
3544
3545 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3546 if (nvl == NULL)
3547 ERROUT(EBADMSG);
3548
3549 if (! nvlist_exists_string(nvl, "anchor"))
3550 ERROUT(EBADMSG);
3551 if (! nvlist_exists_number(nvl, "ruleset"))
3552 ERROUT(EBADMSG);
3553 if (! nvlist_exists_number(nvl, "ticket"))
3554 ERROUT(EBADMSG);
3555 if (! nvlist_exists_number(nvl, "nr"))
3556 ERROUT(EBADMSG);
3557
3558 if (nvlist_exists_bool(nvl, "clear_counter"))
3559 clear_counter = nvlist_get_bool(nvl, "clear_counter");
3560
3561 if (clear_counter && !(flags & FWRITE))
3562 ERROUT(EACCES);
3563
3564 nr = nvlist_get_number(nvl, "nr");
3565
3566 PF_RULES_WLOCK();
3567 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3568 if (ruleset == NULL) {
3569 PF_RULES_WUNLOCK();
3570 ERROUT(ENOENT);
3571 }
3572
3573 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3574 if (rs_num >= PF_RULESET_MAX) {
3575 PF_RULES_WUNLOCK();
3576 ERROUT(EINVAL);
3577 }
3578
3579 if (nvlist_get_number(nvl, "ticket") !=
3580 ruleset->rules[rs_num].active.ticket) {
3581 PF_RULES_WUNLOCK();
3582 ERROUT(EBUSY);
3583 }
3584
3585 if ((error = nvlist_error(nvl))) {
3586 PF_RULES_WUNLOCK();
3587 ERROUT(error);
3588 }
3589
3590 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3591 while ((rule != NULL) && (rule->nr != nr))
3592 rule = TAILQ_NEXT(rule, entries);
3593 if (rule == NULL) {
3594 PF_RULES_WUNLOCK();
3595 ERROUT(EBUSY);
3596 }
3597
3598 nvrule = pf_krule_to_nvrule(rule);
3599
3600 nvlist_destroy(nvl);
3601 nvl = nvlist_create(0);
3602 if (nvl == NULL) {
3603 PF_RULES_WUNLOCK();
3604 ERROUT(ENOMEM);
3605 }
3606 nvlist_add_number(nvl, "nr", nr);
3607 nvlist_add_nvlist(nvl, "rule", nvrule);
3608 nvlist_destroy(nvrule);
3609 nvrule = NULL;
3610 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3611 PF_RULES_WUNLOCK();
3612 ERROUT(EBUSY);
3613 }
3614
3615 free(nvlpacked, M_NVLIST);
3616 nvlpacked = nvlist_pack(nvl, &nv->len);
3617 if (nvlpacked == NULL) {
3618 PF_RULES_WUNLOCK();
3619 ERROUT(ENOMEM);
3620 }
3621
3622 if (nv->size == 0) {
3623 PF_RULES_WUNLOCK();
3624 ERROUT(0);
3625 }
3626 else if (nv->size < nv->len) {
3627 PF_RULES_WUNLOCK();
3628 ERROUT(ENOSPC);
3629 }
3630
3631 if (clear_counter)
3632 pf_krule_clear_counters(rule);
3633
3634 PF_RULES_WUNLOCK();
3635
3636 error = copyout(nvlpacked, nv->data, nv->len);
3637
3638 #undef ERROUT
3639 DIOCGETRULENV_error:
3640 free(nvlpacked, M_NVLIST);
3641 nvlist_destroy(nvrule);
3642 nvlist_destroy(nvl);
3643
3644 break;
3645 }
3646
3647 case DIOCCHANGERULE: {
3648 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
3649 struct pf_kruleset *ruleset;
3650 struct pf_krule *oldrule = NULL, *newrule = NULL;
3651 struct pfi_kkif *kif = NULL;
3652 struct pf_kpooladdr *pa;
3653 u_int32_t nr = 0;
3654 int rs_num;
3655
3656 pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3657
3658 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3659 pcr->action > PF_CHANGE_GET_TICKET) {
3660 error = EINVAL;
3661 break;
3662 }
3663 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3664 error = EINVAL;
3665 break;
3666 }
3667
3668 if (pcr->action != PF_CHANGE_REMOVE) {
3669 newrule = pf_krule_alloc();
3670 error = pf_rule_to_krule(&pcr->rule, newrule);
3671 if (error != 0) {
3672 pf_krule_free(newrule);
3673 break;
3674 }
3675
3676 if (newrule->ifname[0])
3677 kif = pf_kkif_create(M_WAITOK);
3678 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3679 for (int i = 0; i < 2; i++) {
3680 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3681 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3682 }
3683 newrule->states_cur = counter_u64_alloc(M_WAITOK);
3684 newrule->states_tot = counter_u64_alloc(M_WAITOK);
3685 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
3686 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
3687 newrule->cuid = td->td_ucred->cr_ruid;
3688 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3689 TAILQ_INIT(&newrule->nat.list);
3690 TAILQ_INIT(&newrule->rdr.list);
3691 TAILQ_INIT(&newrule->route.list);
3692 }
3693 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3694
3695 PF_CONFIG_LOCK();
3696 PF_RULES_WLOCK();
3697 #ifdef PF_WANT_32_TO_64_COUNTER
3698 if (newrule != NULL) {
3699 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3700 newrule->allrulelinked = true;
3701 V_pf_allrulecount++;
3702 }
3703 #endif
3704
3705 if (!(pcr->action == PF_CHANGE_REMOVE ||
3706 pcr->action == PF_CHANGE_GET_TICKET) &&
3707 pcr->pool_ticket != V_ticket_pabuf)
3708 ERROUT(EBUSY);
3709
3710 ruleset = pf_find_kruleset(pcr->anchor);
3711 if (ruleset == NULL)
3712 ERROUT(EINVAL);
3713
3714 rs_num = pf_get_ruleset_number(pcr->rule.action);
3715 if (rs_num >= PF_RULESET_MAX)
3716 ERROUT(EINVAL);
3717
3718 /*
3719 * XXXMJG: there is no guarantee that the ruleset was
3720 * created by the usual route of calling DIOCXBEGIN.
3721 * As a result it is possible the rule tree will not
3722 * be allocated yet. Hack around it by doing it here.
3723 * Note it is fine to let the tree persist in case of
3724 * error as it will be freed down the road on future
3725 * updates (if need be).
3726 */
3727 if (ruleset->rules[rs_num].active.tree == NULL) {
3728 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3729 if (ruleset->rules[rs_num].active.tree == NULL) {
3730 ERROUT(ENOMEM);
3731 }
3732 }
3733
3734 if (pcr->action == PF_CHANGE_GET_TICKET) {
3735 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3736 ERROUT(0);
3737 } else if (pcr->ticket !=
3738 ruleset->rules[rs_num].active.ticket)
3739 ERROUT(EINVAL);
3740
3741 if (pcr->action != PF_CHANGE_REMOVE) {
3742 if (newrule->ifname[0]) {
3743 newrule->kif = pfi_kkif_attach(kif,
3744 newrule->ifname);
3745 kif = NULL;
3746 pfi_kkif_ref(newrule->kif);
3747 } else
3748 newrule->kif = NULL;
3749
3750 if (newrule->rtableid > 0 &&
3751 newrule->rtableid >= rt_numfibs)
3752 error = EBUSY;
3753
3754 #ifdef ALTQ
3755 /* set queue IDs */
3756 if (newrule->qname[0] != 0) {
3757 if ((newrule->qid =
3758 pf_qname2qid(newrule->qname)) == 0)
3759 error = EBUSY;
3760 else if (newrule->pqname[0] != 0) {
3761 if ((newrule->pqid =
3762 pf_qname2qid(newrule->pqname)) == 0)
3763 error = EBUSY;
3764 } else
3765 newrule->pqid = newrule->qid;
3766 }
3767 #endif /* ALTQ */
3768 if (newrule->tagname[0])
3769 if ((newrule->tag =
3770 pf_tagname2tag(newrule->tagname)) == 0)
3771 error = EBUSY;
3772 if (newrule->match_tagname[0])
3773 if ((newrule->match_tag = pf_tagname2tag(
3774 newrule->match_tagname)) == 0)
3775 error = EBUSY;
3776 if (newrule->rt && !newrule->direction)
3777 error = EINVAL;
3778 if (!newrule->log)
3779 newrule->logif = 0;
3780 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3781 error = ENOMEM;
3782 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3783 error = ENOMEM;
3784 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3785 error = EINVAL;
3786 for (int i = 0; i < 3; i++) {
3787 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
3788 if (pa->addr.type == PF_ADDR_TABLE) {
3789 pa->addr.p.tbl =
3790 pfr_attach_table(ruleset,
3791 pa->addr.v.tblname);
3792 if (pa->addr.p.tbl == NULL)
3793 error = ENOMEM;
3794 }
3795 }
3796
3797 newrule->overload_tbl = NULL;
3798 if (newrule->overload_tblname[0]) {
3799 if ((newrule->overload_tbl = pfr_attach_table(
3800 ruleset, newrule->overload_tblname)) ==
3801 NULL)
3802 error = EINVAL;
3803 else
3804 newrule->overload_tbl->pfrkt_flags |=
3805 PFR_TFLAG_ACTIVE;
3806 }
3807
3808 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
3809 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
3810 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list);
3811 if (((((newrule->action == PF_NAT) ||
3812 (newrule->action == PF_RDR) ||
3813 (newrule->action == PF_BINAT) ||
3814 (newrule->rt > PF_NOPFROUTE)) &&
3815 !newrule->anchor)) &&
3816 (TAILQ_FIRST(&newrule->rdr.list) == NULL))
3817 error = EINVAL;
3818
3819 if (error) {
3820 pf_free_rule(newrule);
3821 PF_RULES_WUNLOCK();
3822 PF_CONFIG_UNLOCK();
3823 break;
3824 }
3825
3826 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
3827 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
3828 }
3829 pf_empty_kpool(&V_pf_pabuf[0]);
3830 pf_empty_kpool(&V_pf_pabuf[1]);
3831 pf_empty_kpool(&V_pf_pabuf[2]);
3832
3833 if (pcr->action == PF_CHANGE_ADD_HEAD)
3834 oldrule = TAILQ_FIRST(
3835 ruleset->rules[rs_num].active.ptr);
3836 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3837 oldrule = TAILQ_LAST(
3838 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3839 else {
3840 oldrule = TAILQ_FIRST(
3841 ruleset->rules[rs_num].active.ptr);
3842 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3843 oldrule = TAILQ_NEXT(oldrule, entries);
3844 if (oldrule == NULL) {
3845 if (newrule != NULL)
3846 pf_free_rule(newrule);
3847 PF_RULES_WUNLOCK();
3848 PF_CONFIG_UNLOCK();
3849 error = EINVAL;
3850 break;
3851 }
3852 }
3853
3854 if (pcr->action == PF_CHANGE_REMOVE) {
3855 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3856 oldrule);
3857 RB_REMOVE(pf_krule_global,
3858 ruleset->rules[rs_num].active.tree, oldrule);
3859 ruleset->rules[rs_num].active.rcount--;
3860 } else {
3861 pf_hash_rule(newrule);
3862 if (RB_INSERT(pf_krule_global,
3863 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3864 pf_free_rule(newrule);
3865 PF_RULES_WUNLOCK();
3866 PF_CONFIG_UNLOCK();
3867 error = EEXIST;
3868 break;
3869 }
3870
3871 if (oldrule == NULL)
3872 TAILQ_INSERT_TAIL(
3873 ruleset->rules[rs_num].active.ptr,
3874 newrule, entries);
3875 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3876 pcr->action == PF_CHANGE_ADD_BEFORE)
3877 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3878 else
3879 TAILQ_INSERT_AFTER(
3880 ruleset->rules[rs_num].active.ptr,
3881 oldrule, newrule, entries);
3882 ruleset->rules[rs_num].active.rcount++;
3883 }
3884
3885 nr = 0;
3886 TAILQ_FOREACH(oldrule,
3887 ruleset->rules[rs_num].active.ptr, entries)
3888 oldrule->nr = nr++;
3889
3890 ruleset->rules[rs_num].active.ticket++;
3891
3892 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3893 pf_remove_if_empty_kruleset(ruleset);
3894
3895 PF_RULES_WUNLOCK();
3896 PF_CONFIG_UNLOCK();
3897 break;
3898
3899 #undef ERROUT
3900 DIOCCHANGERULE_error:
3901 PF_RULES_WUNLOCK();
3902 PF_CONFIG_UNLOCK();
3903 pf_krule_free(newrule);
3904 pf_kkif_free(kif);
3905 break;
3906 }
3907
3908 case DIOCCLRSTATESNV: {
3909 error = pf_clearstates_nv((struct pfioc_nv *)addr);
3910 break;
3911 }
3912
3913 case DIOCKILLSTATESNV: {
3914 error = pf_killstates_nv((struct pfioc_nv *)addr);
3915 break;
3916 }
3917
3918 case DIOCADDSTATE: {
3919 struct pfioc_state *ps = (struct pfioc_state *)addr;
3920 struct pfsync_state_1301 *sp = &ps->state;
3921
3922 if (sp->timeout >= PFTM_MAX) {
3923 error = EINVAL;
3924 break;
3925 }
3926 if (V_pfsync_state_import_ptr != NULL) {
3927 PF_RULES_RLOCK();
3928 error = V_pfsync_state_import_ptr(
3929 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3930 PFSYNC_MSG_VERSION_1301);
3931 PF_RULES_RUNLOCK();
3932 } else
3933 error = EOPNOTSUPP;
3934 break;
3935 }
3936
3937 case DIOCGETSTATE: {
3938 struct pfioc_state *ps = (struct pfioc_state *)addr;
3939 struct pf_kstate *s;
3940
3941 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3942 if (s == NULL) {
3943 error = ENOENT;
3944 break;
3945 }
3946
3947 pfsync_state_export((union pfsync_state_union*)&ps->state,
3948 s, PFSYNC_MSG_VERSION_1301);
3949 PF_STATE_UNLOCK(s);
3950 break;
3951 }
3952
3953 case DIOCGETSTATENV: {
3954 error = pf_getstate((struct pfioc_nv *)addr);
3955 break;
3956 }
3957
3958 #ifdef COMPAT_FREEBSD14
3959 case DIOCGETSTATES: {
3960 struct pfioc_states *ps = (struct pfioc_states *)addr;
3961 struct pf_kstate *s;
3962 struct pfsync_state_1301 *pstore, *p;
3963 int i, nr;
3964 size_t slice_count = 16, count;
3965 void *out;
3966
3967 if (ps->ps_len <= 0) {
3968 nr = uma_zone_get_cur(V_pf_state_z);
3969 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3970 break;
3971 }
3972
3973 out = ps->ps_states;
3974 pstore = mallocarray(slice_count,
3975 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3976 nr = 0;
3977
3978 for (i = 0; i <= V_pf_hashmask; i++) {
3979 struct pf_idhash *ih = &V_pf_idhash[i];
3980
3981 DIOCGETSTATES_retry:
3982 p = pstore;
3983
3984 if (LIST_EMPTY(&ih->states))
3985 continue;
3986
3987 PF_HASHROW_LOCK(ih);
3988 count = 0;
3989 LIST_FOREACH(s, &ih->states, entry) {
3990 if (s->timeout == PFTM_UNLINKED)
3991 continue;
3992 count++;
3993 }
3994
3995 if (count > slice_count) {
3996 PF_HASHROW_UNLOCK(ih);
3997 free(pstore, M_TEMP);
3998 slice_count = count * 2;
3999 pstore = mallocarray(slice_count,
4000 sizeof(struct pfsync_state_1301), M_TEMP,
4001 M_WAITOK | M_ZERO);
4002 goto DIOCGETSTATES_retry;
4003 }
4004
4005 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4006 PF_HASHROW_UNLOCK(ih);
4007 goto DIOCGETSTATES_full;
4008 }
4009
4010 LIST_FOREACH(s, &ih->states, entry) {
4011 if (s->timeout == PFTM_UNLINKED)
4012 continue;
4013
4014 pfsync_state_export((union pfsync_state_union*)p,
4015 s, PFSYNC_MSG_VERSION_1301);
4016 p++;
4017 nr++;
4018 }
4019 PF_HASHROW_UNLOCK(ih);
4020 error = copyout(pstore, out,
4021 sizeof(struct pfsync_state_1301) * count);
4022 if (error)
4023 break;
4024 out = ps->ps_states + nr;
4025 }
4026 DIOCGETSTATES_full:
4027 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4028 free(pstore, M_TEMP);
4029
4030 break;
4031 }
4032
4033 case DIOCGETSTATESV2: {
4034 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
4035 struct pf_kstate *s;
4036 struct pf_state_export *pstore, *p;
4037 int i, nr;
4038 size_t slice_count = 16, count;
4039 void *out;
4040
4041 if (ps->ps_req_version > PF_STATE_VERSION) {
4042 error = ENOTSUP;
4043 break;
4044 }
4045
4046 if (ps->ps_len <= 0) {
4047 nr = uma_zone_get_cur(V_pf_state_z);
4048 ps->ps_len = sizeof(struct pf_state_export) * nr;
4049 break;
4050 }
4051
4052 out = ps->ps_states;
4053 pstore = mallocarray(slice_count,
4054 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
4055 nr = 0;
4056
4057 for (i = 0; i <= V_pf_hashmask; i++) {
4058 struct pf_idhash *ih = &V_pf_idhash[i];
4059
4060 DIOCGETSTATESV2_retry:
4061 p = pstore;
4062
4063 if (LIST_EMPTY(&ih->states))
4064 continue;
4065
4066 PF_HASHROW_LOCK(ih);
4067 count = 0;
4068 LIST_FOREACH(s, &ih->states, entry) {
4069 if (s->timeout == PFTM_UNLINKED)
4070 continue;
4071 count++;
4072 }
4073
4074 if (count > slice_count) {
4075 PF_HASHROW_UNLOCK(ih);
4076 free(pstore, M_TEMP);
4077 slice_count = count * 2;
4078 pstore = mallocarray(slice_count,
4079 sizeof(struct pf_state_export), M_TEMP,
4080 M_WAITOK | M_ZERO);
4081 goto DIOCGETSTATESV2_retry;
4082 }
4083
4084 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4085 PF_HASHROW_UNLOCK(ih);
4086 goto DIOCGETSTATESV2_full;
4087 }
4088
4089 LIST_FOREACH(s, &ih->states, entry) {
4090 if (s->timeout == PFTM_UNLINKED)
4091 continue;
4092
4093 pf_state_export(p, s);
4094 p++;
4095 nr++;
4096 }
4097 PF_HASHROW_UNLOCK(ih);
4098 error = copyout(pstore, out,
4099 sizeof(struct pf_state_export) * count);
4100 if (error)
4101 break;
4102 out = ps->ps_states + nr;
4103 }
4104 DIOCGETSTATESV2_full:
4105 ps->ps_len = nr * sizeof(struct pf_state_export);
4106 free(pstore, M_TEMP);
4107
4108 break;
4109 }
4110 #endif
4111 case DIOCGETSTATUSNV: {
4112 error = pf_getstatus((struct pfioc_nv *)addr);
4113 break;
4114 }
4115
4116 case DIOCSETSTATUSIF: {
4117 struct pfioc_if *pi = (struct pfioc_if *)addr;
4118
4119 if (pi->ifname[0] == 0) {
4120 bzero(V_pf_status.ifname, IFNAMSIZ);
4121 break;
4122 }
4123 PF_RULES_WLOCK();
4124 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
4125 PF_RULES_WUNLOCK();
4126 break;
4127 }
4128
4129 case DIOCCLRSTATUS: {
4130 pf_ioctl_clear_status();
4131 break;
4132 }
4133
4134 case DIOCNATLOOK: {
4135 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
4136 struct pf_state_key *sk;
4137 struct pf_kstate *state;
4138 struct pf_state_key_cmp key;
4139 int m = 0, direction = pnl->direction;
4140 int sidx, didx;
4141
4142 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
4143 sidx = (direction == PF_IN) ? 1 : 0;
4144 didx = (direction == PF_IN) ? 0 : 1;
4145
4146 if (!pnl->proto ||
4147 PF_AZERO(&pnl->saddr, pnl->af) ||
4148 PF_AZERO(&pnl->daddr, pnl->af) ||
4149 ((pnl->proto == IPPROTO_TCP ||
4150 pnl->proto == IPPROTO_UDP) &&
4151 (!pnl->dport || !pnl->sport)))
4152 error = EINVAL;
4153 else {
4154 bzero(&key, sizeof(key));
4155 key.af = pnl->af;
4156 key.proto = pnl->proto;
4157 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af);
4158 key.port[sidx] = pnl->sport;
4159 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af);
4160 key.port[didx] = pnl->dport;
4161
4162 state = pf_find_state_all(&key, direction, &m);
4163 if (state == NULL) {
4164 error = ENOENT;
4165 } else {
4166 if (m > 1) {
4167 PF_STATE_UNLOCK(state);
4168 error = E2BIG; /* more than one state */
4169 } else {
4170 sk = state->key[sidx];
4171 pf_addrcpy(&pnl->rsaddr,
4172 &sk->addr[sidx], sk->af);
4173 pnl->rsport = sk->port[sidx];
4174 pf_addrcpy(&pnl->rdaddr,
4175 &sk->addr[didx], sk->af);
4176 pnl->rdport = sk->port[didx];
4177 PF_STATE_UNLOCK(state);
4178 }
4179 }
4180 }
4181 break;
4182 }
4183
4184 case DIOCSETTIMEOUT: {
4185 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4186
4187 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4188 &pt->seconds);
4189 break;
4190 }
4191
4192 case DIOCGETTIMEOUT: {
4193 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4194
4195 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4196 break;
4197 }
4198
4199 case DIOCGETLIMIT: {
4200 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4201
4202 error = pf_ioctl_get_limit(pl->index, &pl->limit);
4203 break;
4204 }
4205
4206 case DIOCSETLIMIT: {
4207 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4208 unsigned int old_limit;
4209
4210 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4211 pl->limit = old_limit;
4212 break;
4213 }
4214
4215 case DIOCSETDEBUG: {
4216 u_int32_t *level = (u_int32_t *)addr;
4217
4218 PF_RULES_WLOCK();
4219 V_pf_status.debug = *level;
4220 PF_RULES_WUNLOCK();
4221 break;
4222 }
4223
4224 case DIOCCLRRULECTRS: {
4225 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4226 struct pf_kruleset *ruleset = &pf_main_ruleset;
4227 struct pf_krule *rule;
4228
4229 PF_RULES_WLOCK();
4230 TAILQ_FOREACH(rule,
4231 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4232 pf_counter_u64_zero(&rule->evaluations);
4233 for (int i = 0; i < 2; i++) {
4234 pf_counter_u64_zero(&rule->packets[i]);
4235 pf_counter_u64_zero(&rule->bytes[i]);
4236 }
4237 }
4238 PF_RULES_WUNLOCK();
4239 break;
4240 }
4241
4242 case DIOCGIFSPEEDV0:
4243 case DIOCGIFSPEEDV1: {
4244 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
4245 struct pf_ifspeed_v1 ps;
4246 struct ifnet *ifp;
4247
4248 if (psp->ifname[0] == '\0') {
4249 error = EINVAL;
4250 break;
4251 }
4252
4253 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4254 if (error != 0)
4255 break;
4256 ifp = ifunit(ps.ifname);
4257 if (ifp != NULL) {
4258 psp->baudrate32 =
4259 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4260 if (cmd == DIOCGIFSPEEDV1)
4261 psp->baudrate = ifp->if_baudrate;
4262 } else {
4263 error = EINVAL;
4264 }
4265 break;
4266 }
4267
4268 #ifdef ALTQ
4269 case DIOCSTARTALTQ: {
4270 struct pf_altq *altq;
4271
4272 PF_RULES_WLOCK();
4273 /* enable all altq interfaces on active list */
4274 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4275 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4276 error = pf_enable_altq(altq);
4277 if (error != 0)
4278 break;
4279 }
4280 }
4281 if (error == 0)
4282 V_pf_altq_running = 1;
4283 PF_RULES_WUNLOCK();
4284 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4285 break;
4286 }
4287
4288 case DIOCSTOPALTQ: {
4289 struct pf_altq *altq;
4290
4291 PF_RULES_WLOCK();
4292 /* disable all altq interfaces on active list */
4293 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4294 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4295 error = pf_disable_altq(altq);
4296 if (error != 0)
4297 break;
4298 }
4299 }
4300 if (error == 0)
4301 V_pf_altq_running = 0;
4302 PF_RULES_WUNLOCK();
4303 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4304 break;
4305 }
4306
4307 case DIOCADDALTQV0:
4308 case DIOCADDALTQV1: {
4309 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4310 struct pf_altq *altq, *a;
4311 struct ifnet *ifp;
4312
4313 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4314 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4315 if (error)
4316 break;
4317 altq->local_flags = 0;
4318
4319 PF_RULES_WLOCK();
4320 if (pa->ticket != V_ticket_altqs_inactive) {
4321 PF_RULES_WUNLOCK();
4322 free(altq, M_PFALTQ);
4323 error = EBUSY;
4324 break;
4325 }
4326
4327 /*
4328 * if this is for a queue, find the discipline and
4329 * copy the necessary fields
4330 */
4331 if (altq->qname[0] != 0) {
4332 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4333 PF_RULES_WUNLOCK();
4334 error = EBUSY;
4335 free(altq, M_PFALTQ);
4336 break;
4337 }
4338 altq->altq_disc = NULL;
4339 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4340 if (strncmp(a->ifname, altq->ifname,
4341 IFNAMSIZ) == 0) {
4342 altq->altq_disc = a->altq_disc;
4343 break;
4344 }
4345 }
4346 }
4347
4348 if ((ifp = ifunit(altq->ifname)) == NULL)
4349 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4350 else
4351 error = altq_add(ifp, altq);
4352
4353 if (error) {
4354 PF_RULES_WUNLOCK();
4355 free(altq, M_PFALTQ);
4356 break;
4357 }
4358
4359 if (altq->qname[0] != 0)
4360 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4361 else
4362 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4363 /* version error check done on import above */
4364 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4365 PF_RULES_WUNLOCK();
4366 break;
4367 }
4368
4369 case DIOCGETALTQSV0:
4370 case DIOCGETALTQSV1: {
4371 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4372 struct pf_altq *altq;
4373
4374 PF_RULES_RLOCK();
4375 pa->nr = 0;
4376 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4377 pa->nr++;
4378 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4379 pa->nr++;
4380 pa->ticket = V_ticket_altqs_active;
4381 PF_RULES_RUNLOCK();
4382 break;
4383 }
4384
4385 case DIOCGETALTQV0:
4386 case DIOCGETALTQV1: {
4387 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4388 struct pf_altq *altq;
4389
4390 PF_RULES_RLOCK();
4391 if (pa->ticket != V_ticket_altqs_active) {
4392 PF_RULES_RUNLOCK();
4393 error = EBUSY;
4394 break;
4395 }
4396 altq = pf_altq_get_nth_active(pa->nr);
4397 if (altq == NULL) {
4398 PF_RULES_RUNLOCK();
4399 error = EBUSY;
4400 break;
4401 }
4402 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4403 PF_RULES_RUNLOCK();
4404 break;
4405 }
4406
4407 case DIOCCHANGEALTQV0:
4408 case DIOCCHANGEALTQV1:
4409 /* CHANGEALTQ not supported yet! */
4410 error = ENODEV;
4411 break;
4412
4413 case DIOCGETQSTATSV0:
4414 case DIOCGETQSTATSV1: {
4415 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
4416 struct pf_altq *altq;
4417 int nbytes;
4418 u_int32_t version;
4419
4420 PF_RULES_RLOCK();
4421 if (pq->ticket != V_ticket_altqs_active) {
4422 PF_RULES_RUNLOCK();
4423 error = EBUSY;
4424 break;
4425 }
4426 nbytes = pq->nbytes;
4427 altq = pf_altq_get_nth_active(pq->nr);
4428 if (altq == NULL) {
4429 PF_RULES_RUNLOCK();
4430 error = EBUSY;
4431 break;
4432 }
4433
4434 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4435 PF_RULES_RUNLOCK();
4436 error = ENXIO;
4437 break;
4438 }
4439 PF_RULES_RUNLOCK();
4440 if (cmd == DIOCGETQSTATSV0)
4441 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
4442 else
4443 version = pq->version;
4444 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4445 if (error == 0) {
4446 pq->scheduler = altq->scheduler;
4447 pq->nbytes = nbytes;
4448 }
4449 break;
4450 }
4451 #endif /* ALTQ */
4452
4453 case DIOCBEGINADDRS: {
4454 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4455
4456 error = pf_ioctl_begin_addrs(&pp->ticket);
4457 break;
4458 }
4459
4460 case DIOCADDADDR: {
4461 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4462 struct pf_nl_pooladdr npp = {};
4463
4464 npp.which = PF_RDR;
4465 memcpy(&npp, pp, sizeof(*pp));
4466 error = pf_ioctl_add_addr(&npp);
4467 break;
4468 }
4469
4470 case DIOCGETADDRS: {
4471 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4472 struct pf_nl_pooladdr npp = {};
4473
4474 npp.which = PF_RDR;
4475 memcpy(&npp, pp, sizeof(*pp));
4476 error = pf_ioctl_get_addrs(&npp);
4477 memcpy(pp, &npp, sizeof(*pp));
4478
4479 break;
4480 }
4481
4482 case DIOCGETADDR: {
4483 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4484 struct pf_nl_pooladdr npp = {};
4485
4486 npp.which = PF_RDR;
4487 memcpy(&npp, pp, sizeof(*pp));
4488 error = pf_ioctl_get_addr(&npp);
4489 memcpy(pp, &npp, sizeof(*pp));
4490
4491 break;
4492 }
4493
4494 case DIOCCHANGEADDR: {
4495 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
4496 struct pf_kpool *pool;
4497 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
4498 struct pf_kruleset *ruleset;
4499 struct pfi_kkif *kif = NULL;
4500
4501 pca->anchor[sizeof(pca->anchor) - 1] = 0;
4502
4503 if (pca->action < PF_CHANGE_ADD_HEAD ||
4504 pca->action > PF_CHANGE_REMOVE) {
4505 error = EINVAL;
4506 break;
4507 }
4508 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4509 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4510 pca->addr.addr.type != PF_ADDR_TABLE) {
4511 error = EINVAL;
4512 break;
4513 }
4514 if (pca->addr.addr.p.dyn != NULL) {
4515 error = EINVAL;
4516 break;
4517 }
4518
4519 if (pca->action != PF_CHANGE_REMOVE) {
4520 #ifndef INET
4521 if (pca->af == AF_INET) {
4522 error = EAFNOSUPPORT;
4523 break;
4524 }
4525 #endif /* INET */
4526 #ifndef INET6
4527 if (pca->af == AF_INET6) {
4528 error = EAFNOSUPPORT;
4529 break;
4530 }
4531 #endif /* INET6 */
4532 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4533 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4534 if (newpa->ifname[0])
4535 kif = pf_kkif_create(M_WAITOK);
4536 newpa->kif = NULL;
4537 }
4538 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4539 PF_RULES_WLOCK();
4540 ruleset = pf_find_kruleset(pca->anchor);
4541 if (ruleset == NULL)
4542 ERROUT(EBUSY);
4543
4544 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4545 pca->r_num, pca->r_last, 1, 1, PF_RDR);
4546 if (pool == NULL)
4547 ERROUT(EBUSY);
4548
4549 if (pca->action != PF_CHANGE_REMOVE) {
4550 if (newpa->ifname[0]) {
4551 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4552 pfi_kkif_ref(newpa->kif);
4553 kif = NULL;
4554 }
4555
4556 switch (newpa->addr.type) {
4557 case PF_ADDR_DYNIFTL:
4558 error = pfi_dynaddr_setup(&newpa->addr,
4559 pca->af);
4560 break;
4561 case PF_ADDR_TABLE:
4562 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4563 newpa->addr.v.tblname);
4564 if (newpa->addr.p.tbl == NULL)
4565 error = ENOMEM;
4566 break;
4567 }
4568 if (error)
4569 goto DIOCCHANGEADDR_error;
4570 }
4571
4572 switch (pca->action) {
4573 case PF_CHANGE_ADD_HEAD:
4574 oldpa = TAILQ_FIRST(&pool->list);
4575 break;
4576 case PF_CHANGE_ADD_TAIL:
4577 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4578 break;
4579 default:
4580 oldpa = TAILQ_FIRST(&pool->list);
4581 for (int i = 0; oldpa && i < pca->nr; i++)
4582 oldpa = TAILQ_NEXT(oldpa, entries);
4583
4584 if (oldpa == NULL)
4585 ERROUT(EINVAL);
4586 }
4587
4588 if (pca->action == PF_CHANGE_REMOVE) {
4589 TAILQ_REMOVE(&pool->list, oldpa, entries);
4590 switch (oldpa->addr.type) {
4591 case PF_ADDR_DYNIFTL:
4592 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4593 break;
4594 case PF_ADDR_TABLE:
4595 pfr_detach_table(oldpa->addr.p.tbl);
4596 break;
4597 }
4598 if (oldpa->kif)
4599 pfi_kkif_unref(oldpa->kif);
4600 free(oldpa, M_PFRULE);
4601 } else {
4602 if (oldpa == NULL)
4603 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4604 else if (pca->action == PF_CHANGE_ADD_HEAD ||
4605 pca->action == PF_CHANGE_ADD_BEFORE)
4606 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4607 else
4608 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4609 newpa, entries);
4610 }
4611
4612 pool->cur = TAILQ_FIRST(&pool->list);
4613 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4614 PF_RULES_WUNLOCK();
4615 break;
4616
4617 #undef ERROUT
4618 DIOCCHANGEADDR_error:
4619 if (newpa != NULL) {
4620 if (newpa->kif)
4621 pfi_kkif_unref(newpa->kif);
4622 free(newpa, M_PFRULE);
4623 }
4624 PF_RULES_WUNLOCK();
4625 pf_kkif_free(kif);
4626 break;
4627 }
4628
4629 case DIOCGETRULESETS: {
4630 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4631
4632 pr->path[sizeof(pr->path) - 1] = 0;
4633
4634 error = pf_ioctl_get_rulesets(pr);
4635 break;
4636 }
4637
4638 case DIOCGETRULESET: {
4639 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4640
4641 pr->path[sizeof(pr->path) - 1] = 0;
4642
4643 error = pf_ioctl_get_ruleset(pr);
4644 break;
4645 }
4646
4647 case DIOCRCLRTABLES: {
4648 struct pfioc_table *io = (struct pfioc_table *)addr;
4649
4650 if (io->pfrio_esize != 0) {
4651 error = ENODEV;
4652 break;
4653 }
4654 PF_RULES_WLOCK();
4655 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4656 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4657 PF_RULES_WUNLOCK();
4658 break;
4659 }
4660
4661 case DIOCRADDTABLES: {
4662 struct pfioc_table *io = (struct pfioc_table *)addr;
4663 struct pfr_table *pfrts;
4664 size_t totlen;
4665
4666 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4667 error = ENODEV;
4668 break;
4669 }
4670
4671 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4672 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4673 error = ENOMEM;
4674 break;
4675 }
4676
4677 totlen = io->pfrio_size * sizeof(struct pfr_table);
4678 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4679 M_TEMP, M_WAITOK);
4680 error = copyin(io->pfrio_buffer, pfrts, totlen);
4681 if (error) {
4682 free(pfrts, M_TEMP);
4683 break;
4684 }
4685 PF_RULES_WLOCK();
4686 error = pfr_add_tables(pfrts, io->pfrio_size,
4687 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4688 PF_RULES_WUNLOCK();
4689 free(pfrts, M_TEMP);
4690 break;
4691 }
4692
4693 case DIOCRDELTABLES: {
4694 struct pfioc_table *io = (struct pfioc_table *)addr;
4695 struct pfr_table *pfrts;
4696 size_t totlen;
4697
4698 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4699 error = ENODEV;
4700 break;
4701 }
4702
4703 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4704 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4705 error = ENOMEM;
4706 break;
4707 }
4708
4709 totlen = io->pfrio_size * sizeof(struct pfr_table);
4710 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4711 M_TEMP, M_WAITOK);
4712 error = copyin(io->pfrio_buffer, pfrts, totlen);
4713 if (error) {
4714 free(pfrts, M_TEMP);
4715 break;
4716 }
4717 PF_RULES_WLOCK();
4718 error = pfr_del_tables(pfrts, io->pfrio_size,
4719 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4720 PF_RULES_WUNLOCK();
4721 free(pfrts, M_TEMP);
4722 break;
4723 }
4724
4725 case DIOCRGETTABLES: {
4726 struct pfioc_table *io = (struct pfioc_table *)addr;
4727 struct pfr_table *pfrts;
4728 size_t totlen;
4729 int n;
4730
4731 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4732 error = ENODEV;
4733 break;
4734 }
4735 PF_RULES_RLOCK();
4736 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4737 if (n < 0) {
4738 PF_RULES_RUNLOCK();
4739 error = EINVAL;
4740 break;
4741 }
4742 io->pfrio_size = min(io->pfrio_size, n);
4743
4744 totlen = io->pfrio_size * sizeof(struct pfr_table);
4745
4746 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4747 M_TEMP, M_NOWAIT | M_ZERO);
4748 if (pfrts == NULL) {
4749 error = ENOMEM;
4750 PF_RULES_RUNLOCK();
4751 break;
4752 }
4753 error = pfr_get_tables(&io->pfrio_table, pfrts,
4754 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4755 PF_RULES_RUNLOCK();
4756 if (error == 0)
4757 error = copyout(pfrts, io->pfrio_buffer, totlen);
4758 free(pfrts, M_TEMP);
4759 break;
4760 }
4761
4762 case DIOCRGETTSTATS: {
4763 struct pfioc_table *io = (struct pfioc_table *)addr;
4764 struct pfr_tstats *pfrtstats;
4765 size_t totlen;
4766 int n;
4767
4768 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4769 error = ENODEV;
4770 break;
4771 }
4772 PF_TABLE_STATS_LOCK();
4773 PF_RULES_RLOCK();
4774 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4775 if (n < 0) {
4776 PF_RULES_RUNLOCK();
4777 PF_TABLE_STATS_UNLOCK();
4778 error = EINVAL;
4779 break;
4780 }
4781 io->pfrio_size = min(io->pfrio_size, n);
4782
4783 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4784 pfrtstats = mallocarray(io->pfrio_size,
4785 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4786 if (pfrtstats == NULL) {
4787 error = ENOMEM;
4788 PF_RULES_RUNLOCK();
4789 PF_TABLE_STATS_UNLOCK();
4790 break;
4791 }
4792 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4793 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4794 PF_RULES_RUNLOCK();
4795 PF_TABLE_STATS_UNLOCK();
4796 if (error == 0)
4797 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4798 free(pfrtstats, M_TEMP);
4799 break;
4800 }
4801
4802 case DIOCRCLRTSTATS: {
4803 struct pfioc_table *io = (struct pfioc_table *)addr;
4804 struct pfr_table *pfrts;
4805 size_t totlen;
4806
4807 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4808 error = ENODEV;
4809 break;
4810 }
4811
4812 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4813 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4814 /* We used to count tables and use the minimum required
4815 * size, so we didn't fail on overly large requests.
4816 * Keep doing so. */
4817 io->pfrio_size = pf_ioctl_maxcount;
4818 break;
4819 }
4820
4821 totlen = io->pfrio_size * sizeof(struct pfr_table);
4822 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4823 M_TEMP, M_WAITOK);
4824 error = copyin(io->pfrio_buffer, pfrts, totlen);
4825 if (error) {
4826 free(pfrts, M_TEMP);
4827 break;
4828 }
4829
4830 PF_TABLE_STATS_LOCK();
4831 PF_RULES_RLOCK();
4832 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4833 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4834 PF_RULES_RUNLOCK();
4835 PF_TABLE_STATS_UNLOCK();
4836 free(pfrts, M_TEMP);
4837 break;
4838 }
4839
4840 case DIOCRSETTFLAGS: {
4841 struct pfioc_table *io = (struct pfioc_table *)addr;
4842 struct pfr_table *pfrts;
4843 size_t totlen;
4844 int n;
4845
4846 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4847 error = ENODEV;
4848 break;
4849 }
4850
4851 PF_RULES_RLOCK();
4852 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4853 if (n < 0) {
4854 PF_RULES_RUNLOCK();
4855 error = EINVAL;
4856 break;
4857 }
4858
4859 io->pfrio_size = min(io->pfrio_size, n);
4860 PF_RULES_RUNLOCK();
4861
4862 totlen = io->pfrio_size * sizeof(struct pfr_table);
4863 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4864 M_TEMP, M_WAITOK);
4865 error = copyin(io->pfrio_buffer, pfrts, totlen);
4866 if (error) {
4867 free(pfrts, M_TEMP);
4868 break;
4869 }
4870 PF_RULES_WLOCK();
4871 error = pfr_set_tflags(pfrts, io->pfrio_size,
4872 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4873 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4874 PF_RULES_WUNLOCK();
4875 free(pfrts, M_TEMP);
4876 break;
4877 }
4878
4879 case DIOCRCLRADDRS: {
4880 struct pfioc_table *io = (struct pfioc_table *)addr;
4881
4882 if (io->pfrio_esize != 0) {
4883 error = ENODEV;
4884 break;
4885 }
4886 PF_RULES_WLOCK();
4887 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4888 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4889 PF_RULES_WUNLOCK();
4890 break;
4891 }
4892
4893 case DIOCRADDADDRS: {
4894 struct pfioc_table *io = (struct pfioc_table *)addr;
4895 struct pfr_addr *pfras;
4896 size_t totlen;
4897
4898 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4899 error = ENODEV;
4900 break;
4901 }
4902 if (io->pfrio_size < 0 ||
4903 io->pfrio_size > pf_ioctl_maxcount ||
4904 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4905 error = EINVAL;
4906 break;
4907 }
4908 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4909 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4910 M_TEMP, M_WAITOK);
4911 error = copyin(io->pfrio_buffer, pfras, totlen);
4912 if (error) {
4913 free(pfras, M_TEMP);
4914 break;
4915 }
4916 PF_RULES_WLOCK();
4917 error = pfr_add_addrs(&io->pfrio_table, pfras,
4918 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4919 PFR_FLAG_USERIOCTL);
4920 PF_RULES_WUNLOCK();
4921 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4922 error = copyout(pfras, io->pfrio_buffer, totlen);
4923 free(pfras, M_TEMP);
4924 break;
4925 }
4926
4927 case DIOCRDELADDRS: {
4928 struct pfioc_table *io = (struct pfioc_table *)addr;
4929 struct pfr_addr *pfras;
4930 size_t totlen;
4931
4932 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4933 error = ENODEV;
4934 break;
4935 }
4936 if (io->pfrio_size < 0 ||
4937 io->pfrio_size > pf_ioctl_maxcount ||
4938 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4939 error = EINVAL;
4940 break;
4941 }
4942 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4943 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4944 M_TEMP, M_WAITOK);
4945 error = copyin(io->pfrio_buffer, pfras, totlen);
4946 if (error) {
4947 free(pfras, M_TEMP);
4948 break;
4949 }
4950 PF_RULES_WLOCK();
4951 error = pfr_del_addrs(&io->pfrio_table, pfras,
4952 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4953 PFR_FLAG_USERIOCTL);
4954 PF_RULES_WUNLOCK();
4955 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4956 error = copyout(pfras, io->pfrio_buffer, totlen);
4957 free(pfras, M_TEMP);
4958 break;
4959 }
4960
4961 case DIOCRSETADDRS: {
4962 struct pfioc_table *io = (struct pfioc_table *)addr;
4963 struct pfr_addr *pfras;
4964 size_t totlen, count;
4965
4966 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4967 error = ENODEV;
4968 break;
4969 }
4970 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4971 error = EINVAL;
4972 break;
4973 }
4974 count = max(io->pfrio_size, io->pfrio_size2);
4975 if (count > pf_ioctl_maxcount ||
4976 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4977 error = EINVAL;
4978 break;
4979 }
4980 totlen = count * sizeof(struct pfr_addr);
4981 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4982 M_WAITOK);
4983 error = copyin(io->pfrio_buffer, pfras, totlen);
4984 if (error) {
4985 free(pfras, M_TEMP);
4986 break;
4987 }
4988 PF_RULES_WLOCK();
4989 error = pfr_set_addrs(&io->pfrio_table, pfras,
4990 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4991 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4992 PFR_FLAG_USERIOCTL, 0);
4993 PF_RULES_WUNLOCK();
4994 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4995 error = copyout(pfras, io->pfrio_buffer, totlen);
4996 free(pfras, M_TEMP);
4997 break;
4998 }
4999
5000 case DIOCRGETADDRS: {
5001 struct pfioc_table *io = (struct pfioc_table *)addr;
5002 struct pfr_addr *pfras;
5003 size_t totlen;
5004
5005 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5006 error = ENODEV;
5007 break;
5008 }
5009 if (io->pfrio_size < 0 ||
5010 io->pfrio_size > pf_ioctl_maxcount ||
5011 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5012 error = EINVAL;
5013 break;
5014 }
5015 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5016 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5017 M_TEMP, M_WAITOK | M_ZERO);
5018 PF_RULES_RLOCK();
5019 error = pfr_get_addrs(&io->pfrio_table, pfras,
5020 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5021 PF_RULES_RUNLOCK();
5022 if (error == 0)
5023 error = copyout(pfras, io->pfrio_buffer, totlen);
5024 free(pfras, M_TEMP);
5025 break;
5026 }
5027
5028 case DIOCRGETASTATS: {
5029 struct pfioc_table *io = (struct pfioc_table *)addr;
5030 struct pfr_astats *pfrastats;
5031 size_t totlen;
5032
5033 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
5034 error = ENODEV;
5035 break;
5036 }
5037 if (io->pfrio_size < 0 ||
5038 io->pfrio_size > pf_ioctl_maxcount ||
5039 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5040 error = EINVAL;
5041 break;
5042 }
5043 totlen = io->pfrio_size * sizeof(struct pfr_astats);
5044 pfrastats = mallocarray(io->pfrio_size,
5045 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
5046 PF_RULES_RLOCK();
5047 error = pfr_get_astats(&io->pfrio_table, pfrastats,
5048 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5049 PF_RULES_RUNLOCK();
5050 if (error == 0)
5051 error = copyout(pfrastats, io->pfrio_buffer, totlen);
5052 free(pfrastats, M_TEMP);
5053 break;
5054 }
5055
5056 case DIOCRCLRASTATS: {
5057 struct pfioc_table *io = (struct pfioc_table *)addr;
5058 struct pfr_addr *pfras;
5059 size_t totlen;
5060
5061 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5062 error = ENODEV;
5063 break;
5064 }
5065 if (io->pfrio_size < 0 ||
5066 io->pfrio_size > pf_ioctl_maxcount ||
5067 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5068 error = EINVAL;
5069 break;
5070 }
5071 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5072 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5073 M_TEMP, M_WAITOK);
5074 error = copyin(io->pfrio_buffer, pfras, totlen);
5075 if (error) {
5076 free(pfras, M_TEMP);
5077 break;
5078 }
5079 PF_RULES_WLOCK();
5080 error = pfr_clr_astats(&io->pfrio_table, pfras,
5081 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5082 PFR_FLAG_USERIOCTL);
5083 PF_RULES_WUNLOCK();
5084 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5085 error = copyout(pfras, io->pfrio_buffer, totlen);
5086 free(pfras, M_TEMP);
5087 break;
5088 }
5089
5090 case DIOCRTSTADDRS: {
5091 struct pfioc_table *io = (struct pfioc_table *)addr;
5092 struct pfr_addr *pfras;
5093 size_t totlen;
5094
5095 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5096 error = ENODEV;
5097 break;
5098 }
5099 if (io->pfrio_size < 0 ||
5100 io->pfrio_size > pf_ioctl_maxcount ||
5101 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5102 error = EINVAL;
5103 break;
5104 }
5105 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5106 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5107 M_TEMP, M_WAITOK);
5108 error = copyin(io->pfrio_buffer, pfras, totlen);
5109 if (error) {
5110 free(pfras, M_TEMP);
5111 break;
5112 }
5113 PF_RULES_RLOCK();
5114 error = pfr_tst_addrs(&io->pfrio_table, pfras,
5115 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5116 PFR_FLAG_USERIOCTL);
5117 PF_RULES_RUNLOCK();
5118 if (error == 0)
5119 error = copyout(pfras, io->pfrio_buffer, totlen);
5120 free(pfras, M_TEMP);
5121 break;
5122 }
5123
5124 case DIOCRINADEFINE: {
5125 struct pfioc_table *io = (struct pfioc_table *)addr;
5126 struct pfr_addr *pfras;
5127 size_t totlen;
5128
5129 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5130 error = ENODEV;
5131 break;
5132 }
5133 if (io->pfrio_size < 0 ||
5134 io->pfrio_size > pf_ioctl_maxcount ||
5135 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5136 error = EINVAL;
5137 break;
5138 }
5139 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5140 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5141 M_TEMP, M_WAITOK);
5142 error = copyin(io->pfrio_buffer, pfras, totlen);
5143 if (error) {
5144 free(pfras, M_TEMP);
5145 break;
5146 }
5147 PF_RULES_WLOCK();
5148 error = pfr_ina_define(&io->pfrio_table, pfras,
5149 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5150 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5151 PF_RULES_WUNLOCK();
5152 free(pfras, M_TEMP);
5153 break;
5154 }
5155
5156 case DIOCOSFPADD: {
5157 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5158 PF_RULES_WLOCK();
5159 error = pf_osfp_add(io);
5160 PF_RULES_WUNLOCK();
5161 break;
5162 }
5163
5164 case DIOCOSFPGET: {
5165 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5166 PF_RULES_RLOCK();
5167 error = pf_osfp_get(io);
5168 PF_RULES_RUNLOCK();
5169 break;
5170 }
5171
5172 case DIOCXBEGIN: {
5173 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5174 struct pfioc_trans_e *ioes, *ioe;
5175 size_t totlen;
5176 int i;
5177
5178 if (io->esize != sizeof(*ioe)) {
5179 error = ENODEV;
5180 break;
5181 }
5182 if (io->size < 0 ||
5183 io->size > pf_ioctl_maxcount ||
5184 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5185 error = EINVAL;
5186 break;
5187 }
5188 totlen = sizeof(struct pfioc_trans_e) * io->size;
5189 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5190 M_TEMP, M_WAITOK);
5191 error = copyin(io->array, ioes, totlen);
5192 if (error) {
5193 free(ioes, M_TEMP);
5194 break;
5195 }
5196 PF_RULES_WLOCK();
5197 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5198 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5199 switch (ioe->rs_num) {
5200 case PF_RULESET_ETH:
5201 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5202 PF_RULES_WUNLOCK();
5203 free(ioes, M_TEMP);
5204 goto fail;
5205 }
5206 break;
5207 #ifdef ALTQ
5208 case PF_RULESET_ALTQ:
5209 if (ioe->anchor[0]) {
5210 PF_RULES_WUNLOCK();
5211 free(ioes, M_TEMP);
5212 error = EINVAL;
5213 goto fail;
5214 }
5215 if ((error = pf_begin_altq(&ioe->ticket))) {
5216 PF_RULES_WUNLOCK();
5217 free(ioes, M_TEMP);
5218 goto fail;
5219 }
5220 break;
5221 #endif /* ALTQ */
5222 case PF_RULESET_TABLE:
5223 {
5224 struct pfr_table table;
5225
5226 bzero(&table, sizeof(table));
5227 strlcpy(table.pfrt_anchor, ioe->anchor,
5228 sizeof(table.pfrt_anchor));
5229 if ((error = pfr_ina_begin(&table,
5230 &ioe->ticket, NULL, 0))) {
5231 PF_RULES_WUNLOCK();
5232 free(ioes, M_TEMP);
5233 goto fail;
5234 }
5235 break;
5236 }
5237 default:
5238 if ((error = pf_begin_rules(&ioe->ticket,
5239 ioe->rs_num, ioe->anchor))) {
5240 PF_RULES_WUNLOCK();
5241 free(ioes, M_TEMP);
5242 goto fail;
5243 }
5244 break;
5245 }
5246 }
5247 PF_RULES_WUNLOCK();
5248 error = copyout(ioes, io->array, totlen);
5249 free(ioes, M_TEMP);
5250 break;
5251 }
5252
5253 case DIOCXROLLBACK: {
5254 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5255 struct pfioc_trans_e *ioe, *ioes;
5256 size_t totlen;
5257 int i;
5258
5259 if (io->esize != sizeof(*ioe)) {
5260 error = ENODEV;
5261 break;
5262 }
5263 if (io->size < 0 ||
5264 io->size > pf_ioctl_maxcount ||
5265 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5266 error = EINVAL;
5267 break;
5268 }
5269 totlen = sizeof(struct pfioc_trans_e) * io->size;
5270 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5271 M_TEMP, M_WAITOK);
5272 error = copyin(io->array, ioes, totlen);
5273 if (error) {
5274 free(ioes, M_TEMP);
5275 break;
5276 }
5277 PF_RULES_WLOCK();
5278 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5279 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5280 switch (ioe->rs_num) {
5281 case PF_RULESET_ETH:
5282 if ((error = pf_rollback_eth(ioe->ticket,
5283 ioe->anchor))) {
5284 PF_RULES_WUNLOCK();
5285 free(ioes, M_TEMP);
5286 goto fail; /* really bad */
5287 }
5288 break;
5289 #ifdef ALTQ
5290 case PF_RULESET_ALTQ:
5291 if (ioe->anchor[0]) {
5292 PF_RULES_WUNLOCK();
5293 free(ioes, M_TEMP);
5294 error = EINVAL;
5295 goto fail;
5296 }
5297 if ((error = pf_rollback_altq(ioe->ticket))) {
5298 PF_RULES_WUNLOCK();
5299 free(ioes, M_TEMP);
5300 goto fail; /* really bad */
5301 }
5302 break;
5303 #endif /* ALTQ */
5304 case PF_RULESET_TABLE:
5305 {
5306 struct pfr_table table;
5307
5308 bzero(&table, sizeof(table));
5309 strlcpy(table.pfrt_anchor, ioe->anchor,
5310 sizeof(table.pfrt_anchor));
5311 if ((error = pfr_ina_rollback(&table,
5312 ioe->ticket, NULL, 0))) {
5313 PF_RULES_WUNLOCK();
5314 free(ioes, M_TEMP);
5315 goto fail; /* really bad */
5316 }
5317 break;
5318 }
5319 default:
5320 if ((error = pf_rollback_rules(ioe->ticket,
5321 ioe->rs_num, ioe->anchor))) {
5322 PF_RULES_WUNLOCK();
5323 free(ioes, M_TEMP);
5324 goto fail; /* really bad */
5325 }
5326 break;
5327 }
5328 }
5329 PF_RULES_WUNLOCK();
5330 free(ioes, M_TEMP);
5331 break;
5332 }
5333
5334 case DIOCXCOMMIT: {
5335 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5336 struct pfioc_trans_e *ioe, *ioes;
5337 struct pf_kruleset *rs;
5338 struct pf_keth_ruleset *ers;
5339 size_t totlen;
5340 int i;
5341
5342 if (io->esize != sizeof(*ioe)) {
5343 error = ENODEV;
5344 break;
5345 }
5346
5347 if (io->size < 0 ||
5348 io->size > pf_ioctl_maxcount ||
5349 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5350 error = EINVAL;
5351 break;
5352 }
5353
5354 totlen = sizeof(struct pfioc_trans_e) * io->size;
5355 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5356 M_TEMP, M_WAITOK);
5357 error = copyin(io->array, ioes, totlen);
5358 if (error) {
5359 free(ioes, M_TEMP);
5360 break;
5361 }
5362 PF_RULES_WLOCK();
5363 /* First makes sure everything will succeed. */
5364 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5365 ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5366 switch (ioe->rs_num) {
5367 case PF_RULESET_ETH:
5368 ers = pf_find_keth_ruleset(ioe->anchor);
5369 if (ers == NULL || ioe->ticket == 0 ||
5370 ioe->ticket != ers->inactive.ticket) {
5371 PF_RULES_WUNLOCK();
5372 free(ioes, M_TEMP);
5373 error = EINVAL;
5374 goto fail;
5375 }
5376 break;
5377 #ifdef ALTQ
5378 case PF_RULESET_ALTQ:
5379 if (ioe->anchor[0]) {
5380 PF_RULES_WUNLOCK();
5381 free(ioes, M_TEMP);
5382 error = EINVAL;
5383 goto fail;
5384 }
5385 if (!V_altqs_inactive_open || ioe->ticket !=
5386 V_ticket_altqs_inactive) {
5387 PF_RULES_WUNLOCK();
5388 free(ioes, M_TEMP);
5389 error = EBUSY;
5390 goto fail;
5391 }
5392 break;
5393 #endif /* ALTQ */
5394 case PF_RULESET_TABLE:
5395 rs = pf_find_kruleset(ioe->anchor);
5396 if (rs == NULL || !rs->topen || ioe->ticket !=
5397 rs->tticket) {
5398 PF_RULES_WUNLOCK();
5399 free(ioes, M_TEMP);
5400 error = EBUSY;
5401 goto fail;
5402 }
5403 break;
5404 default:
5405 if (ioe->rs_num < 0 || ioe->rs_num >=
5406 PF_RULESET_MAX) {
5407 PF_RULES_WUNLOCK();
5408 free(ioes, M_TEMP);
5409 error = EINVAL;
5410 goto fail;
5411 }
5412 rs = pf_find_kruleset(ioe->anchor);
5413 if (rs == NULL ||
5414 !rs->rules[ioe->rs_num].inactive.open ||
5415 rs->rules[ioe->rs_num].inactive.ticket !=
5416 ioe->ticket) {
5417 PF_RULES_WUNLOCK();
5418 free(ioes, M_TEMP);
5419 error = EBUSY;
5420 goto fail;
5421 }
5422 break;
5423 }
5424 }
5425 /* Now do the commit - no errors should happen here. */
5426 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5427 switch (ioe->rs_num) {
5428 case PF_RULESET_ETH:
5429 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5430 PF_RULES_WUNLOCK();
5431 free(ioes, M_TEMP);
5432 goto fail; /* really bad */
5433 }
5434 break;
5435 #ifdef ALTQ
5436 case PF_RULESET_ALTQ:
5437 if ((error = pf_commit_altq(ioe->ticket))) {
5438 PF_RULES_WUNLOCK();
5439 free(ioes, M_TEMP);
5440 goto fail; /* really bad */
5441 }
5442 break;
5443 #endif /* ALTQ */
5444 case PF_RULESET_TABLE:
5445 {
5446 struct pfr_table table;
5447
5448 bzero(&table, sizeof(table));
5449 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
5450 sizeof(table.pfrt_anchor));
5451 if ((error = pfr_ina_commit(&table,
5452 ioe->ticket, NULL, NULL, 0))) {
5453 PF_RULES_WUNLOCK();
5454 free(ioes, M_TEMP);
5455 goto fail; /* really bad */
5456 }
5457 break;
5458 }
5459 default:
5460 if ((error = pf_commit_rules(ioe->ticket,
5461 ioe->rs_num, ioe->anchor))) {
5462 PF_RULES_WUNLOCK();
5463 free(ioes, M_TEMP);
5464 goto fail; /* really bad */
5465 }
5466 break;
5467 }
5468 }
5469 PF_RULES_WUNLOCK();
5470
5471 /* Only hook into EtherNet taffic if we've got rules for it. */
5472 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5473 hook_pf_eth();
5474 else
5475 dehook_pf_eth();
5476
5477 free(ioes, M_TEMP);
5478 break;
5479 }
5480
5481 case DIOCGETSRCNODES: {
5482 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
5483 struct pf_srchash *sh;
5484 struct pf_ksrc_node *n;
5485 struct pf_src_node *p, *pstore;
5486 uint32_t i, nr = 0;
5487
5488 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5489 i++, sh++) {
5490 PF_HASHROW_LOCK(sh);
5491 LIST_FOREACH(n, &sh->nodes, entry)
5492 nr++;
5493 PF_HASHROW_UNLOCK(sh);
5494 }
5495
5496 psn->psn_len = min(psn->psn_len,
5497 sizeof(struct pf_src_node) * nr);
5498
5499 if (psn->psn_len == 0) {
5500 psn->psn_len = sizeof(struct pf_src_node) * nr;
5501 break;
5502 }
5503
5504 nr = 0;
5505
5506 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5507 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5508 i++, sh++) {
5509 PF_HASHROW_LOCK(sh);
5510 LIST_FOREACH(n, &sh->nodes, entry) {
5511
5512 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5513 break;
5514
5515 pf_src_node_copy(n, p);
5516
5517 p++;
5518 nr++;
5519 }
5520 PF_HASHROW_UNLOCK(sh);
5521 }
5522 error = copyout(pstore, psn->psn_src_nodes,
5523 sizeof(struct pf_src_node) * nr);
5524 if (error) {
5525 free(pstore, M_TEMP);
5526 break;
5527 }
5528 psn->psn_len = sizeof(struct pf_src_node) * nr;
5529 free(pstore, M_TEMP);
5530 break;
5531 }
5532
5533 case DIOCCLRSRCNODES: {
5534 pf_kill_srcnodes(NULL);
5535 break;
5536 }
5537
5538 case DIOCKILLSRCNODES:
5539 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5540 break;
5541
5542 #ifdef COMPAT_FREEBSD13
5543 case DIOCKEEPCOUNTERS_FREEBSD13:
5544 #endif
5545 case DIOCKEEPCOUNTERS:
5546 error = pf_keepcounters((struct pfioc_nv *)addr);
5547 break;
5548
5549 case DIOCGETSYNCOOKIES:
5550 error = pf_get_syncookies((struct pfioc_nv *)addr);
5551 break;
5552
5553 case DIOCSETSYNCOOKIES:
5554 error = pf_set_syncookies((struct pfioc_nv *)addr);
5555 break;
5556
5557 case DIOCSETHOSTID: {
5558 u_int32_t *hostid = (u_int32_t *)addr;
5559
5560 PF_RULES_WLOCK();
5561 if (*hostid == 0)
5562 V_pf_status.hostid = arc4random();
5563 else
5564 V_pf_status.hostid = *hostid;
5565 PF_RULES_WUNLOCK();
5566 break;
5567 }
5568
5569 case DIOCOSFPFLUSH:
5570 PF_RULES_WLOCK();
5571 pf_osfp_flush();
5572 PF_RULES_WUNLOCK();
5573 break;
5574
5575 case DIOCIGETIFACES: {
5576 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5577 struct pfi_kif *ifstore;
5578 size_t bufsiz;
5579
5580 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5581 error = ENODEV;
5582 break;
5583 }
5584
5585 if (io->pfiio_size < 0 ||
5586 io->pfiio_size > pf_ioctl_maxcount ||
5587 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5588 error = EINVAL;
5589 break;
5590 }
5591
5592 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5593
5594 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5595 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5596 M_TEMP, M_WAITOK | M_ZERO);
5597
5598 PF_RULES_RLOCK();
5599 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5600 PF_RULES_RUNLOCK();
5601 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5602 free(ifstore, M_TEMP);
5603 break;
5604 }
5605
5606 case DIOCSETIFFLAG: {
5607 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5608
5609 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5610
5611 PF_RULES_WLOCK();
5612 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5613 PF_RULES_WUNLOCK();
5614 break;
5615 }
5616
5617 case DIOCCLRIFFLAG: {
5618 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5619
5620 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5621
5622 PF_RULES_WLOCK();
5623 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5624 PF_RULES_WUNLOCK();
5625 break;
5626 }
5627
5628 case DIOCSETREASS: {
5629 u_int32_t *reass = (u_int32_t *)addr;
5630
5631 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5632 /* Removal of DF flag without reassembly enabled is not a
5633 * valid combination. Disable reassembly in such case. */
5634 if (!(V_pf_status.reass & PF_REASS_ENABLED))
5635 V_pf_status.reass = 0;
5636 break;
5637 }
5638
5639 default:
5640 error = ENODEV;
5641 break;
5642 }
5643 fail:
5644 CURVNET_RESTORE();
5645
5646 #undef ERROUT_IOCTL
5647
5648 return (error);
5649 }
5650
5651 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5652 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5653 {
5654 bzero(sp, sizeof(union pfsync_state_union));
5655
5656 /* copy from state key */
5657 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5658 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5659 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5660 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5661 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5662 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5663 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5664 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5665 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5666 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5667
5668 /* copy from state */
5669 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5670 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5671 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5672 sp->pfs_1301.expire = pf_state_expires(st);
5673 if (sp->pfs_1301.expire <= time_uptime)
5674 sp->pfs_1301.expire = htonl(0);
5675 else
5676 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5677
5678 sp->pfs_1301.direction = st->direction;
5679 sp->pfs_1301.log = st->act.log;
5680 sp->pfs_1301.timeout = st->timeout;
5681
5682 switch (msg_version) {
5683 case PFSYNC_MSG_VERSION_1301:
5684 sp->pfs_1301.state_flags = st->state_flags;
5685 break;
5686 case PFSYNC_MSG_VERSION_1400:
5687 sp->pfs_1400.state_flags = htons(st->state_flags);
5688 sp->pfs_1400.qid = htons(st->act.qid);
5689 sp->pfs_1400.pqid = htons(st->act.pqid);
5690 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5691 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5692 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5693 sp->pfs_1400.min_ttl = st->act.min_ttl;
5694 sp->pfs_1400.set_tos = st->act.set_tos;
5695 sp->pfs_1400.max_mss = htons(st->act.max_mss);
5696 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5697 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5698 sp->pfs_1400.rt = st->act.rt;
5699 if (st->act.rt_kif)
5700 strlcpy(sp->pfs_1400.rt_ifname,
5701 st->act.rt_kif->pfik_name,
5702 sizeof(sp->pfs_1400.rt_ifname));
5703 break;
5704 default:
5705 panic("%s: Unsupported pfsync_msg_version %d",
5706 __func__, msg_version);
5707 }
5708
5709 /*
5710 * XXX Why do we bother pfsyncing source node information if source
5711 * nodes are not synced? Showing users that there is source tracking
5712 * when there is none seems useless.
5713 */
5714 if (st->sns[PF_SN_LIMIT] != NULL)
5715 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5716 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
5717 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5718
5719 sp->pfs_1301.id = st->id;
5720 sp->pfs_1301.creatorid = st->creatorid;
5721 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5722 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5723
5724 if (st->rule == NULL)
5725 sp->pfs_1301.rule = htonl(-1);
5726 else
5727 sp->pfs_1301.rule = htonl(st->rule->nr);
5728 if (st->anchor == NULL)
5729 sp->pfs_1301.anchor = htonl(-1);
5730 else
5731 sp->pfs_1301.anchor = htonl(st->anchor->nr);
5732 if (st->nat_rule == NULL)
5733 sp->pfs_1301.nat_rule = htonl(-1);
5734 else
5735 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
5736
5737 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5738 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5739 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5740 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5741 }
5742
5743 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5744 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5745 {
5746 bzero(sp, sizeof(*sp));
5747
5748 sp->version = PF_STATE_VERSION;
5749
5750 /* copy from state key */
5751 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5752 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5753 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5754 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5755 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5756 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5757 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5758 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5759 sp->proto = st->key[PF_SK_WIRE]->proto;
5760 sp->af = st->key[PF_SK_WIRE]->af;
5761
5762 /* copy from state */
5763 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5764 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5765 sizeof(sp->orig_ifname));
5766 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr));
5767 sp->creation = htonl(time_uptime - (st->creation / 1000));
5768 sp->expire = pf_state_expires(st);
5769 if (sp->expire <= time_uptime)
5770 sp->expire = htonl(0);
5771 else
5772 sp->expire = htonl(sp->expire - time_uptime);
5773
5774 sp->direction = st->direction;
5775 sp->log = st->act.log;
5776 sp->timeout = st->timeout;
5777 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5778 sp->state_flags_compat = st->state_flags;
5779 sp->state_flags = htons(st->state_flags);
5780 if (st->sns[PF_SN_LIMIT] != NULL)
5781 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5782 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL)
5783 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5784 sp->id = st->id;
5785 sp->creatorid = st->creatorid;
5786 pf_state_peer_hton(&st->src, &sp->src);
5787 pf_state_peer_hton(&st->dst, &sp->dst);
5788
5789 if (st->rule == NULL)
5790 sp->rule = htonl(-1);
5791 else
5792 sp->rule = htonl(st->rule->nr);
5793 if (st->anchor == NULL)
5794 sp->anchor = htonl(-1);
5795 else
5796 sp->anchor = htonl(st->anchor->nr);
5797 if (st->nat_rule == NULL)
5798 sp->nat_rule = htonl(-1);
5799 else
5800 sp->nat_rule = htonl(st->nat_rule->nr);
5801
5802 sp->packets[0] = st->packets[0];
5803 sp->packets[1] = st->packets[1];
5804 sp->bytes[0] = st->bytes[0];
5805 sp->bytes[1] = st->bytes[1];
5806
5807 sp->qid = htons(st->act.qid);
5808 sp->pqid = htons(st->act.pqid);
5809 sp->dnpipe = htons(st->act.dnpipe);
5810 sp->dnrpipe = htons(st->act.dnrpipe);
5811 sp->rtableid = htonl(st->act.rtableid);
5812 sp->min_ttl = st->act.min_ttl;
5813 sp->set_tos = st->act.set_tos;
5814 sp->max_mss = htons(st->act.max_mss);
5815 sp->rt = st->act.rt;
5816 if (st->act.rt_kif)
5817 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
5818 sizeof(sp->rt_ifname));
5819 sp->set_prio[0] = st->act.set_prio[0];
5820 sp->set_prio[1] = st->act.set_prio[1];
5821
5822 }
5823
5824 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5825 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5826 {
5827 struct pfr_ktable *kt;
5828
5829 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5830
5831 kt = aw->p.tbl;
5832 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5833 kt = kt->pfrkt_root;
5834 aw->p.tbl = NULL;
5835 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5836 kt->pfrkt_cnt : -1;
5837 }
5838
5839 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5840 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5841 size_t number, char **names)
5842 {
5843 nvlist_t *nvc;
5844
5845 nvc = nvlist_create(0);
5846 if (nvc == NULL)
5847 return (ENOMEM);
5848
5849 for (int i = 0; i < number; i++) {
5850 nvlist_append_number_array(nvc, "counters",
5851 counter_u64_fetch(counters[i]));
5852 nvlist_append_string_array(nvc, "names",
5853 names[i]);
5854 nvlist_append_number_array(nvc, "ids",
5855 i);
5856 }
5857 nvlist_add_nvlist(nvl, name, nvc);
5858 nvlist_destroy(nvc);
5859
5860 return (0);
5861 }
5862
5863 static int
pf_getstatus(struct pfioc_nv * nv)5864 pf_getstatus(struct pfioc_nv *nv)
5865 {
5866 nvlist_t *nvl = NULL, *nvc = NULL;
5867 void *nvlpacked = NULL;
5868 int error;
5869 struct pf_status s;
5870 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5871 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5872 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5873 time_t since;
5874
5875 PF_RULES_RLOCK_TRACKER;
5876
5877 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
5878
5879 PF_RULES_RLOCK();
5880
5881 nvl = nvlist_create(0);
5882 if (nvl == NULL)
5883 ERROUT(ENOMEM);
5884
5885 since = time_second - (time_uptime - V_pf_status.since);
5886
5887 nvlist_add_bool(nvl, "running", V_pf_status.running);
5888 nvlist_add_number(nvl, "since", since);
5889 nvlist_add_number(nvl, "debug", V_pf_status.debug);
5890 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5891 nvlist_add_number(nvl, "states", V_pf_status.states);
5892 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5893 nvlist_add_number(nvl, "reass", V_pf_status.reass);
5894 nvlist_add_bool(nvl, "syncookies_active",
5895 V_pf_status.syncookies_active);
5896 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5897
5898 /* counters */
5899 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5900 PFRES_MAX, pf_reasons);
5901 if (error != 0)
5902 ERROUT(error);
5903
5904 /* lcounters */
5905 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5906 KLCNT_MAX, pf_lcounter);
5907 if (error != 0)
5908 ERROUT(error);
5909
5910 /* fcounters */
5911 nvc = nvlist_create(0);
5912 if (nvc == NULL)
5913 ERROUT(ENOMEM);
5914
5915 for (int i = 0; i < FCNT_MAX; i++) {
5916 nvlist_append_number_array(nvc, "counters",
5917 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5918 nvlist_append_string_array(nvc, "names",
5919 pf_fcounter[i]);
5920 nvlist_append_number_array(nvc, "ids",
5921 i);
5922 }
5923 nvlist_add_nvlist(nvl, "fcounters", nvc);
5924 nvlist_destroy(nvc);
5925 nvc = NULL;
5926
5927 /* scounters */
5928 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5929 SCNT_MAX, pf_fcounter);
5930 if (error != 0)
5931 ERROUT(error);
5932
5933 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5934 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5935 PF_MD5_DIGEST_LENGTH);
5936
5937 pfi_update_status(V_pf_status.ifname, &s);
5938
5939 /* pcounters / bcounters */
5940 for (int i = 0; i < 2; i++) {
5941 for (int j = 0; j < 2; j++) {
5942 for (int k = 0; k < 2; k++) {
5943 nvlist_append_number_array(nvl, "pcounters",
5944 s.pcounters[i][j][k]);
5945 }
5946 nvlist_append_number_array(nvl, "bcounters",
5947 s.bcounters[i][j]);
5948 }
5949 }
5950
5951 nvlpacked = nvlist_pack(nvl, &nv->len);
5952 if (nvlpacked == NULL)
5953 ERROUT(ENOMEM);
5954
5955 if (nv->size == 0)
5956 ERROUT(0);
5957 else if (nv->size < nv->len)
5958 ERROUT(ENOSPC);
5959
5960 PF_RULES_RUNLOCK();
5961 error = copyout(nvlpacked, nv->data, nv->len);
5962 goto done;
5963
5964 #undef ERROUT
5965 errout:
5966 PF_RULES_RUNLOCK();
5967 done:
5968 free(nvlpacked, M_NVLIST);
5969 nvlist_destroy(nvc);
5970 nvlist_destroy(nvl);
5971
5972 return (error);
5973 }
5974
5975 /*
5976 * XXX - Check for version mismatch!!!
5977 */
5978 static void
pf_clear_all_states(void)5979 pf_clear_all_states(void)
5980 {
5981 struct epoch_tracker et;
5982 struct pf_kstate *s;
5983 u_int i;
5984
5985 NET_EPOCH_ENTER(et);
5986 for (i = 0; i <= V_pf_hashmask; i++) {
5987 struct pf_idhash *ih = &V_pf_idhash[i];
5988 relock:
5989 PF_HASHROW_LOCK(ih);
5990 LIST_FOREACH(s, &ih->states, entry) {
5991 s->timeout = PFTM_PURGE;
5992 /* Don't send out individual delete messages. */
5993 s->state_flags |= PFSTATE_NOSYNC;
5994 pf_remove_state(s);
5995 goto relock;
5996 }
5997 PF_HASHROW_UNLOCK(ih);
5998 }
5999 NET_EPOCH_EXIT(et);
6000 }
6001
6002 static int
pf_clear_tables(void)6003 pf_clear_tables(void)
6004 {
6005 struct pfioc_table io;
6006 int error;
6007
6008 bzero(&io, sizeof(io));
6009 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
6010
6011 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
6012 io.pfrio_flags);
6013
6014 return (error);
6015 }
6016
6017 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)6018 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
6019 {
6020 struct pf_ksrc_node_list kill;
6021 u_int killed;
6022
6023 LIST_INIT(&kill);
6024 for (int i = 0; i <= V_pf_srchashmask; i++) {
6025 struct pf_srchash *sh = &V_pf_srchash[i];
6026 struct pf_ksrc_node *sn, *tmp;
6027
6028 PF_HASHROW_LOCK(sh);
6029 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
6030 if (psnk == NULL ||
6031 (pf_match_addr(psnk->psnk_src.neg,
6032 &psnk->psnk_src.addr.v.a.addr,
6033 &psnk->psnk_src.addr.v.a.mask,
6034 &sn->addr, sn->af) &&
6035 pf_match_addr(psnk->psnk_dst.neg,
6036 &psnk->psnk_dst.addr.v.a.addr,
6037 &psnk->psnk_dst.addr.v.a.mask,
6038 &sn->raddr, sn->af))) {
6039 pf_unlink_src_node(sn);
6040 LIST_INSERT_HEAD(&kill, sn, entry);
6041 sn->expire = 1;
6042 }
6043 PF_HASHROW_UNLOCK(sh);
6044 }
6045
6046 for (int i = 0; i <= V_pf_hashmask; i++) {
6047 struct pf_idhash *ih = &V_pf_idhash[i];
6048 struct pf_kstate *s;
6049
6050 PF_HASHROW_LOCK(ih);
6051 LIST_FOREACH(s, &ih->states, entry) {
6052 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX;
6053 sn_type++) {
6054 if (s->sns[sn_type] &&
6055 s->sns[sn_type]->expire == 1) {
6056 s->sns[sn_type] = NULL;
6057 }
6058 }
6059 }
6060 PF_HASHROW_UNLOCK(ih);
6061 }
6062
6063 killed = pf_free_src_nodes(&kill);
6064
6065 if (psnk != NULL)
6066 psnk->psnk_killed = killed;
6067 }
6068
6069 static int
pf_keepcounters(struct pfioc_nv * nv)6070 pf_keepcounters(struct pfioc_nv *nv)
6071 {
6072 nvlist_t *nvl = NULL;
6073 void *nvlpacked = NULL;
6074 int error = 0;
6075
6076 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6077
6078 if (nv->len > pf_ioctl_maxcount)
6079 ERROUT(ENOMEM);
6080
6081 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6082 error = copyin(nv->data, nvlpacked, nv->len);
6083 if (error)
6084 ERROUT(error);
6085
6086 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6087 if (nvl == NULL)
6088 ERROUT(EBADMSG);
6089
6090 if (! nvlist_exists_bool(nvl, "keep_counters"))
6091 ERROUT(EBADMSG);
6092
6093 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6094
6095 on_error:
6096 nvlist_destroy(nvl);
6097 free(nvlpacked, M_NVLIST);
6098 return (error);
6099 }
6100
6101 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6102 pf_clear_states(const struct pf_kstate_kill *kill)
6103 {
6104 struct pf_state_key_cmp match_key;
6105 struct pf_kstate *s;
6106 struct pfi_kkif *kif;
6107 int idx;
6108 unsigned int killed = 0, dir;
6109
6110 NET_EPOCH_ASSERT();
6111
6112 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6113 struct pf_idhash *ih = &V_pf_idhash[i];
6114
6115 relock_DIOCCLRSTATES:
6116 PF_HASHROW_LOCK(ih);
6117 LIST_FOREACH(s, &ih->states, entry) {
6118 /* For floating states look at the original kif. */
6119 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6120
6121 if (kill->psk_ifname[0] &&
6122 strcmp(kill->psk_ifname,
6123 kif->pfik_name))
6124 continue;
6125
6126 if (kill->psk_kill_match) {
6127 bzero(&match_key, sizeof(match_key));
6128
6129 if (s->direction == PF_OUT) {
6130 dir = PF_IN;
6131 idx = PF_SK_STACK;
6132 } else {
6133 dir = PF_OUT;
6134 idx = PF_SK_WIRE;
6135 }
6136
6137 match_key.af = s->key[idx]->af;
6138 match_key.proto = s->key[idx]->proto;
6139 pf_addrcpy(&match_key.addr[0],
6140 &s->key[idx]->addr[1], match_key.af);
6141 match_key.port[0] = s->key[idx]->port[1];
6142 pf_addrcpy(&match_key.addr[1],
6143 &s->key[idx]->addr[0], match_key.af);
6144 match_key.port[1] = s->key[idx]->port[0];
6145 }
6146
6147 /*
6148 * Don't send out individual
6149 * delete messages.
6150 */
6151 s->state_flags |= PFSTATE_NOSYNC;
6152 pf_remove_state(s);
6153 killed++;
6154
6155 if (kill->psk_kill_match)
6156 killed += pf_kill_matching_state(&match_key,
6157 dir);
6158
6159 goto relock_DIOCCLRSTATES;
6160 }
6161 PF_HASHROW_UNLOCK(ih);
6162 }
6163
6164 if (V_pfsync_clear_states_ptr != NULL)
6165 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6166
6167 return (killed);
6168 }
6169
6170 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6171 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6172 {
6173 struct pf_kstate *s;
6174
6175 NET_EPOCH_ASSERT();
6176 if (kill->psk_pfcmp.id) {
6177 if (kill->psk_pfcmp.creatorid == 0)
6178 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6179 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6180 kill->psk_pfcmp.creatorid))) {
6181 pf_remove_state(s);
6182 *killed = 1;
6183 }
6184 return;
6185 }
6186
6187 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6188 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6189 }
6190
6191 static int
pf_killstates_nv(struct pfioc_nv * nv)6192 pf_killstates_nv(struct pfioc_nv *nv)
6193 {
6194 struct pf_kstate_kill kill;
6195 struct epoch_tracker et;
6196 nvlist_t *nvl = NULL;
6197 void *nvlpacked = NULL;
6198 int error = 0;
6199 unsigned int killed = 0;
6200
6201 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6202
6203 if (nv->len > pf_ioctl_maxcount)
6204 ERROUT(ENOMEM);
6205
6206 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6207 error = copyin(nv->data, nvlpacked, nv->len);
6208 if (error)
6209 ERROUT(error);
6210
6211 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6212 if (nvl == NULL)
6213 ERROUT(EBADMSG);
6214
6215 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6216 if (error)
6217 ERROUT(error);
6218
6219 NET_EPOCH_ENTER(et);
6220 pf_killstates(&kill, &killed);
6221 NET_EPOCH_EXIT(et);
6222
6223 free(nvlpacked, M_NVLIST);
6224 nvlpacked = NULL;
6225 nvlist_destroy(nvl);
6226 nvl = nvlist_create(0);
6227 if (nvl == NULL)
6228 ERROUT(ENOMEM);
6229
6230 nvlist_add_number(nvl, "killed", killed);
6231
6232 nvlpacked = nvlist_pack(nvl, &nv->len);
6233 if (nvlpacked == NULL)
6234 ERROUT(ENOMEM);
6235
6236 if (nv->size == 0)
6237 ERROUT(0);
6238 else if (nv->size < nv->len)
6239 ERROUT(ENOSPC);
6240
6241 error = copyout(nvlpacked, nv->data, nv->len);
6242
6243 on_error:
6244 nvlist_destroy(nvl);
6245 free(nvlpacked, M_NVLIST);
6246 return (error);
6247 }
6248
6249 static int
pf_clearstates_nv(struct pfioc_nv * nv)6250 pf_clearstates_nv(struct pfioc_nv *nv)
6251 {
6252 struct pf_kstate_kill kill;
6253 struct epoch_tracker et;
6254 nvlist_t *nvl = NULL;
6255 void *nvlpacked = NULL;
6256 int error = 0;
6257 unsigned int killed;
6258
6259 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6260
6261 if (nv->len > pf_ioctl_maxcount)
6262 ERROUT(ENOMEM);
6263
6264 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6265 error = copyin(nv->data, nvlpacked, nv->len);
6266 if (error)
6267 ERROUT(error);
6268
6269 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6270 if (nvl == NULL)
6271 ERROUT(EBADMSG);
6272
6273 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6274 if (error)
6275 ERROUT(error);
6276
6277 NET_EPOCH_ENTER(et);
6278 killed = pf_clear_states(&kill);
6279 NET_EPOCH_EXIT(et);
6280
6281 free(nvlpacked, M_NVLIST);
6282 nvlpacked = NULL;
6283 nvlist_destroy(nvl);
6284 nvl = nvlist_create(0);
6285 if (nvl == NULL)
6286 ERROUT(ENOMEM);
6287
6288 nvlist_add_number(nvl, "killed", killed);
6289
6290 nvlpacked = nvlist_pack(nvl, &nv->len);
6291 if (nvlpacked == NULL)
6292 ERROUT(ENOMEM);
6293
6294 if (nv->size == 0)
6295 ERROUT(0);
6296 else if (nv->size < nv->len)
6297 ERROUT(ENOSPC);
6298
6299 error = copyout(nvlpacked, nv->data, nv->len);
6300
6301 #undef ERROUT
6302 on_error:
6303 nvlist_destroy(nvl);
6304 free(nvlpacked, M_NVLIST);
6305 return (error);
6306 }
6307
6308 static int
pf_getstate(struct pfioc_nv * nv)6309 pf_getstate(struct pfioc_nv *nv)
6310 {
6311 nvlist_t *nvl = NULL, *nvls;
6312 void *nvlpacked = NULL;
6313 struct pf_kstate *s = NULL;
6314 int error = 0;
6315 uint64_t id, creatorid;
6316
6317 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6318
6319 if (nv->len > pf_ioctl_maxcount)
6320 ERROUT(ENOMEM);
6321
6322 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6323 error = copyin(nv->data, nvlpacked, nv->len);
6324 if (error)
6325 ERROUT(error);
6326
6327 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6328 if (nvl == NULL)
6329 ERROUT(EBADMSG);
6330
6331 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6332 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6333
6334 s = pf_find_state_byid(id, creatorid);
6335 if (s == NULL)
6336 ERROUT(ENOENT);
6337
6338 free(nvlpacked, M_NVLIST);
6339 nvlpacked = NULL;
6340 nvlist_destroy(nvl);
6341 nvl = nvlist_create(0);
6342 if (nvl == NULL)
6343 ERROUT(ENOMEM);
6344
6345 nvls = pf_state_to_nvstate(s);
6346 if (nvls == NULL)
6347 ERROUT(ENOMEM);
6348
6349 nvlist_add_nvlist(nvl, "state", nvls);
6350 nvlist_destroy(nvls);
6351
6352 nvlpacked = nvlist_pack(nvl, &nv->len);
6353 if (nvlpacked == NULL)
6354 ERROUT(ENOMEM);
6355
6356 if (nv->size == 0)
6357 ERROUT(0);
6358 else if (nv->size < nv->len)
6359 ERROUT(ENOSPC);
6360
6361 error = copyout(nvlpacked, nv->data, nv->len);
6362
6363 #undef ERROUT
6364 errout:
6365 if (s != NULL)
6366 PF_STATE_UNLOCK(s);
6367 free(nvlpacked, M_NVLIST);
6368 nvlist_destroy(nvl);
6369 return (error);
6370 }
6371
6372 /*
6373 * XXX - Check for version mismatch!!!
6374 */
6375
6376 /*
6377 * Duplicate pfctl -Fa operation to get rid of as much as we can.
6378 */
6379 static int
shutdown_pf(void)6380 shutdown_pf(void)
6381 {
6382 int error = 0;
6383 u_int32_t t[5];
6384 char nn = '\0';
6385 struct pf_kanchor *anchor;
6386 struct pf_keth_anchor *eth_anchor;
6387 int rs_num;
6388
6389 do {
6390 /* Unlink rules of all user defined anchors */
6391 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6392 /* Wildcard based anchors may not have a respective
6393 * explicit anchor rule or they may be left empty
6394 * without rules. It leads to anchor.refcnt=0, and the
6395 * rest of the logic does not expect it. */
6396 if (anchor->refcnt == 0)
6397 anchor->refcnt = 1;
6398 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6399 if ((error = pf_begin_rules(&t[rs_num], rs_num,
6400 anchor->path)) != 0) {
6401 DPFPRINTF(PF_DEBUG_MISC, ("%s: "
6402 "anchor.path=%s rs_num=%d\n",
6403 __func__, anchor->path, rs_num));
6404 goto error; /* XXX: rollback? */
6405 }
6406 }
6407 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6408 error = pf_commit_rules(t[rs_num], rs_num,
6409 anchor->path);
6410 MPASS(error == 0);
6411 }
6412 }
6413
6414 /* Unlink rules of all user defined ether anchors */
6415 RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6416 &V_pf_keth_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 (eth_anchor->refcnt == 0)
6422 eth_anchor->refcnt = 1;
6423 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6424 != 0) {
6425 DPFPRINTF(PF_DEBUG_MISC, ("%s: eth "
6426 "anchor.path=%s\n", __func__,
6427 eth_anchor->path));
6428 goto error;
6429 }
6430 error = pf_commit_eth(t[0], eth_anchor->path);
6431 MPASS(error == 0);
6432 }
6433
6434 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6435 != 0) {
6436 DPFPRINTF(PF_DEBUG_MISC, ("%s: SCRUB\n", __func__));
6437 break;
6438 }
6439 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6440 != 0) {
6441 DPFPRINTF(PF_DEBUG_MISC, ("%s: FILTER\n", __func__));
6442 break; /* XXX: rollback? */
6443 }
6444 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6445 != 0) {
6446 DPFPRINTF(PF_DEBUG_MISC, ("%s: NAT\n", __func__));
6447 break; /* XXX: rollback? */
6448 }
6449 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6450 != 0) {
6451 DPFPRINTF(PF_DEBUG_MISC, ("%s: BINAT\n", __func__));
6452 break; /* XXX: rollback? */
6453 }
6454 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6455 != 0) {
6456 DPFPRINTF(PF_DEBUG_MISC, ("%s: RDR\n", __func__));
6457 break; /* XXX: rollback? */
6458 }
6459
6460 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6461 MPASS(error == 0);
6462 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6463 MPASS(error == 0);
6464 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6465 MPASS(error == 0);
6466 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6467 MPASS(error == 0);
6468 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6469 MPASS(error == 0);
6470
6471 if ((error = pf_clear_tables()) != 0)
6472 break;
6473
6474 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6475 DPFPRINTF(PF_DEBUG_MISC, ("%s: eth\n", __func__));
6476 break;
6477 }
6478 error = pf_commit_eth(t[0], &nn);
6479 MPASS(error == 0);
6480
6481 #ifdef ALTQ
6482 if ((error = pf_begin_altq(&t[0])) != 0) {
6483 DPFPRINTF(PF_DEBUG_MISC, ("%s: ALTQ\n", __func__));
6484 break;
6485 }
6486 pf_commit_altq(t[0]);
6487 #endif
6488
6489 pf_clear_all_states();
6490
6491 pf_kill_srcnodes(NULL);
6492
6493 /* status does not use malloced mem so no need to cleanup */
6494 /* fingerprints and interfaces have their own cleanup code */
6495 } while(0);
6496
6497 error:
6498 return (error);
6499 }
6500
6501 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6502 pf_check_return(int chk, struct mbuf **m)
6503 {
6504
6505 switch (chk) {
6506 case PF_PASS:
6507 if (*m == NULL)
6508 return (PFIL_CONSUMED);
6509 else
6510 return (PFIL_PASS);
6511 break;
6512 default:
6513 if (*m != NULL) {
6514 m_freem(*m);
6515 *m = NULL;
6516 }
6517 return (PFIL_DROPPED);
6518 }
6519 }
6520
6521 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6522 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6523 void *ruleset __unused, struct inpcb *inp)
6524 {
6525 int chk;
6526
6527 CURVNET_ASSERT_SET();
6528
6529 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6530
6531 return (pf_check_return(chk, m));
6532 }
6533
6534 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6535 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6536 void *ruleset __unused, struct inpcb *inp)
6537 {
6538 int chk;
6539
6540 CURVNET_ASSERT_SET();
6541
6542 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6543
6544 return (pf_check_return(chk, m));
6545 }
6546
6547 #ifdef INET
6548 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6549 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6550 void *ruleset __unused, struct inpcb *inp)
6551 {
6552 int chk;
6553
6554 CURVNET_ASSERT_SET();
6555
6556 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
6557
6558 return (pf_check_return(chk, m));
6559 }
6560
6561 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6562 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6563 void *ruleset __unused, struct inpcb *inp)
6564 {
6565 int chk;
6566
6567 CURVNET_ASSERT_SET();
6568
6569 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
6570
6571 return (pf_check_return(chk, m));
6572 }
6573 #endif
6574
6575 #ifdef INET6
6576 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6577 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6578 void *ruleset __unused, struct inpcb *inp)
6579 {
6580 int chk;
6581
6582 CURVNET_ASSERT_SET();
6583
6584 /*
6585 * In case of loopback traffic IPv6 uses the real interface in
6586 * order to support scoped addresses. In order to support stateful
6587 * filtering we have change this to lo0 as it is the case in IPv4.
6588 */
6589 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6590 m, inp, NULL);
6591
6592 return (pf_check_return(chk, m));
6593 }
6594
6595 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6596 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6597 void *ruleset __unused, struct inpcb *inp)
6598 {
6599 int chk;
6600
6601 CURVNET_ASSERT_SET();
6602
6603 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL);
6604
6605 return (pf_check_return(chk, m));
6606 }
6607 #endif /* INET6 */
6608
6609 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6610 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6611 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
6612 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
6613
6614 #ifdef INET
6615 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6616 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6617 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
6618 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
6619 #endif
6620 #ifdef INET6
6621 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6622 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6623 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
6624 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
6625 #endif
6626
6627 static void
hook_pf_eth(void)6628 hook_pf_eth(void)
6629 {
6630 struct pfil_hook_args pha = {
6631 .pa_version = PFIL_VERSION,
6632 .pa_modname = "pf",
6633 .pa_type = PFIL_TYPE_ETHERNET,
6634 };
6635 struct pfil_link_args pla = {
6636 .pa_version = PFIL_VERSION,
6637 };
6638 int ret __diagused;
6639
6640 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6641 return;
6642
6643 pha.pa_mbuf_chk = pf_eth_check_in;
6644 pha.pa_flags = PFIL_IN;
6645 pha.pa_rulname = "eth-in";
6646 V_pf_eth_in_hook = pfil_add_hook(&pha);
6647 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6648 pla.pa_head = V_link_pfil_head;
6649 pla.pa_hook = V_pf_eth_in_hook;
6650 ret = pfil_link(&pla);
6651 MPASS(ret == 0);
6652 pha.pa_mbuf_chk = pf_eth_check_out;
6653 pha.pa_flags = PFIL_OUT;
6654 pha.pa_rulname = "eth-out";
6655 V_pf_eth_out_hook = pfil_add_hook(&pha);
6656 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6657 pla.pa_head = V_link_pfil_head;
6658 pla.pa_hook = V_pf_eth_out_hook;
6659 ret = pfil_link(&pla);
6660 MPASS(ret == 0);
6661
6662 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6663 }
6664
6665 static void
hook_pf(void)6666 hook_pf(void)
6667 {
6668 struct pfil_hook_args pha = {
6669 .pa_version = PFIL_VERSION,
6670 .pa_modname = "pf",
6671 };
6672 struct pfil_link_args pla = {
6673 .pa_version = PFIL_VERSION,
6674 };
6675 int ret __diagused;
6676
6677 if (atomic_load_bool(&V_pf_pfil_hooked))
6678 return;
6679
6680 #ifdef INET
6681 pha.pa_type = PFIL_TYPE_IP4;
6682 pha.pa_mbuf_chk = pf_check_in;
6683 pha.pa_flags = PFIL_IN;
6684 pha.pa_rulname = "default-in";
6685 V_pf_ip4_in_hook = pfil_add_hook(&pha);
6686 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6687 pla.pa_head = V_inet_pfil_head;
6688 pla.pa_hook = V_pf_ip4_in_hook;
6689 ret = pfil_link(&pla);
6690 MPASS(ret == 0);
6691 pha.pa_mbuf_chk = pf_check_out;
6692 pha.pa_flags = PFIL_OUT;
6693 pha.pa_rulname = "default-out";
6694 V_pf_ip4_out_hook = pfil_add_hook(&pha);
6695 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6696 pla.pa_head = V_inet_pfil_head;
6697 pla.pa_hook = V_pf_ip4_out_hook;
6698 ret = pfil_link(&pla);
6699 MPASS(ret == 0);
6700 if (V_pf_filter_local) {
6701 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6702 pla.pa_head = V_inet_local_pfil_head;
6703 pla.pa_hook = V_pf_ip4_out_hook;
6704 ret = pfil_link(&pla);
6705 MPASS(ret == 0);
6706 }
6707 #endif
6708 #ifdef INET6
6709 pha.pa_type = PFIL_TYPE_IP6;
6710 pha.pa_mbuf_chk = pf_check6_in;
6711 pha.pa_flags = PFIL_IN;
6712 pha.pa_rulname = "default-in6";
6713 V_pf_ip6_in_hook = pfil_add_hook(&pha);
6714 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6715 pla.pa_head = V_inet6_pfil_head;
6716 pla.pa_hook = V_pf_ip6_in_hook;
6717 ret = pfil_link(&pla);
6718 MPASS(ret == 0);
6719 pha.pa_mbuf_chk = pf_check6_out;
6720 pha.pa_rulname = "default-out6";
6721 pha.pa_flags = PFIL_OUT;
6722 V_pf_ip6_out_hook = pfil_add_hook(&pha);
6723 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6724 pla.pa_head = V_inet6_pfil_head;
6725 pla.pa_hook = V_pf_ip6_out_hook;
6726 ret = pfil_link(&pla);
6727 MPASS(ret == 0);
6728 if (V_pf_filter_local) {
6729 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6730 pla.pa_head = V_inet6_local_pfil_head;
6731 pla.pa_hook = V_pf_ip6_out_hook;
6732 ret = pfil_link(&pla);
6733 MPASS(ret == 0);
6734 }
6735 #endif
6736
6737 atomic_store_bool(&V_pf_pfil_hooked, true);
6738 }
6739
6740 static void
dehook_pf_eth(void)6741 dehook_pf_eth(void)
6742 {
6743
6744 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6745 return;
6746
6747 pfil_remove_hook(V_pf_eth_in_hook);
6748 pfil_remove_hook(V_pf_eth_out_hook);
6749
6750 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6751 }
6752
6753 static void
dehook_pf(void)6754 dehook_pf(void)
6755 {
6756
6757 if (!atomic_load_bool(&V_pf_pfil_hooked))
6758 return;
6759
6760 #ifdef INET
6761 pfil_remove_hook(V_pf_ip4_in_hook);
6762 pfil_remove_hook(V_pf_ip4_out_hook);
6763 #endif
6764 #ifdef INET6
6765 pfil_remove_hook(V_pf_ip6_in_hook);
6766 pfil_remove_hook(V_pf_ip6_out_hook);
6767 #endif
6768
6769 atomic_store_bool(&V_pf_pfil_hooked, false);
6770 }
6771
6772 static void
pf_load_vnet(void)6773 pf_load_vnet(void)
6774 {
6775 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6776 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6777
6778 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6779 sx_init(&V_pf_ioctl_lock, "pf ioctl");
6780
6781 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6782 PF_RULE_TAG_HASH_SIZE_DEFAULT);
6783 #ifdef ALTQ
6784 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6785 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6786 #endif
6787
6788 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6789
6790 pfattach_vnet();
6791 V_pf_vnet_active = 1;
6792 }
6793
6794 static int
pf_load(void)6795 pf_load(void)
6796 {
6797 int error;
6798
6799 sx_init(&pf_end_lock, "pf end thread");
6800
6801 pf_mtag_initialize();
6802
6803 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6804 if (pf_dev == NULL)
6805 return (ENOMEM);
6806
6807 pf_end_threads = 0;
6808 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6809 if (error != 0)
6810 return (error);
6811
6812 pfi_initialize();
6813
6814 return (0);
6815 }
6816
6817 static void
pf_unload_vnet(void)6818 pf_unload_vnet(void)
6819 {
6820 int ret __diagused;
6821
6822 V_pf_vnet_active = 0;
6823 V_pf_status.running = 0;
6824 dehook_pf();
6825 dehook_pf_eth();
6826
6827 PF_RULES_WLOCK();
6828 pf_syncookies_cleanup();
6829 shutdown_pf();
6830 PF_RULES_WUNLOCK();
6831
6832 ret = swi_remove(V_pf_swi_cookie);
6833 MPASS(ret == 0);
6834 ret = intr_event_destroy(V_pf_swi_ie);
6835 MPASS(ret == 0);
6836
6837 pf_unload_vnet_purge();
6838
6839 pf_normalize_cleanup();
6840 PF_RULES_WLOCK();
6841 pfi_cleanup_vnet();
6842 PF_RULES_WUNLOCK();
6843 pfr_cleanup();
6844 pf_osfp_flush();
6845 pf_cleanup();
6846 if (IS_DEFAULT_VNET(curvnet))
6847 pf_mtag_cleanup();
6848
6849 pf_cleanup_tagset(&V_pf_tags);
6850 #ifdef ALTQ
6851 pf_cleanup_tagset(&V_pf_qids);
6852 #endif
6853 uma_zdestroy(V_pf_tag_z);
6854
6855 #ifdef PF_WANT_32_TO_64_COUNTER
6856 PF_RULES_WLOCK();
6857 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6858
6859 MPASS(LIST_EMPTY(&V_pf_allkiflist));
6860 MPASS(V_pf_allkifcount == 0);
6861
6862 LIST_REMOVE(&V_pf_default_rule, allrulelist);
6863 V_pf_allrulecount--;
6864 LIST_REMOVE(V_pf_rulemarker, allrulelist);
6865
6866 MPASS(LIST_EMPTY(&V_pf_allrulelist));
6867 MPASS(V_pf_allrulecount == 0);
6868
6869 PF_RULES_WUNLOCK();
6870
6871 free(V_pf_kifmarker, PFI_MTYPE);
6872 free(V_pf_rulemarker, M_PFRULE);
6873 #endif
6874
6875 /* Free counters last as we updated them during shutdown. */
6876 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6877 for (int i = 0; i < 2; i++) {
6878 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6879 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6880 }
6881 counter_u64_free(V_pf_default_rule.states_cur);
6882 counter_u64_free(V_pf_default_rule.states_tot);
6883 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
6884 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]);
6885 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6886
6887 for (int i = 0; i < PFRES_MAX; i++)
6888 counter_u64_free(V_pf_status.counters[i]);
6889 for (int i = 0; i < KLCNT_MAX; i++)
6890 counter_u64_free(V_pf_status.lcounters[i]);
6891 for (int i = 0; i < FCNT_MAX; i++)
6892 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6893 for (int i = 0; i < SCNT_MAX; i++)
6894 counter_u64_free(V_pf_status.scounters[i]);
6895
6896 rm_destroy(&V_pf_rules_lock);
6897 sx_destroy(&V_pf_ioctl_lock);
6898 }
6899
6900 static void
pf_unload(void)6901 pf_unload(void)
6902 {
6903
6904 sx_xlock(&pf_end_lock);
6905 pf_end_threads = 1;
6906 while (pf_end_threads < 2) {
6907 wakeup_one(pf_purge_thread);
6908 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6909 }
6910 sx_xunlock(&pf_end_lock);
6911
6912 pf_nl_unregister();
6913
6914 if (pf_dev != NULL)
6915 destroy_dev(pf_dev);
6916
6917 pfi_cleanup();
6918
6919 sx_destroy(&pf_end_lock);
6920 }
6921
6922 static void
vnet_pf_init(void * unused __unused)6923 vnet_pf_init(void *unused __unused)
6924 {
6925
6926 pf_load_vnet();
6927 }
6928 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6929 vnet_pf_init, NULL);
6930
6931 static void
vnet_pf_uninit(const void * unused __unused)6932 vnet_pf_uninit(const void *unused __unused)
6933 {
6934
6935 pf_unload_vnet();
6936 }
6937 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6938 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6939 vnet_pf_uninit, NULL);
6940
6941 static int
pf_modevent(module_t mod,int type,void * data)6942 pf_modevent(module_t mod, int type, void *data)
6943 {
6944 int error = 0;
6945
6946 switch(type) {
6947 case MOD_LOAD:
6948 error = pf_load();
6949 pf_nl_register();
6950 break;
6951 case MOD_UNLOAD:
6952 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6953 * the vnet_pf_uninit()s */
6954 break;
6955 default:
6956 error = EINVAL;
6957 break;
6958 }
6959
6960 return (error);
6961 }
6962
6963 static moduledata_t pf_mod = {
6964 "pf",
6965 pf_modevent,
6966 0
6967 };
6968
6969 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6970 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6971 MODULE_DEPEND(pf, crypto, 1, 1, 1);
6972 MODULE_VERSION(pf, PF_MODVER);
6973