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 = -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 V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
356
357 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
358 M_WAITOK | M_ZERO);
359
360 #ifdef PF_WANT_32_TO_64_COUNTER
361 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
362 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
363 PF_RULES_WLOCK();
364 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
365 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
366 V_pf_allrulecount++;
367 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
368 PF_RULES_WUNLOCK();
369 #endif
370
371 /* initialize default timeouts */
372 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
373 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
374 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
375 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
376 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
377 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
378 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
379 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
380 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
381 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
382 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
383 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
384 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
385 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
386 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
387 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
388 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
389 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
390 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
391 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
392 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
393 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
394 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
395 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
396 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
397
398 V_pf_status.debug = PF_DEBUG_URGENT;
399 /*
400 * XXX This is different than in OpenBSD where reassembly is enabled by
401 * defult. In FreeBSD we expect people to still use scrub rules and
402 * switch to the new syntax later. Only when they switch they must
403 * explicitly enable reassemle. We could change the default once the
404 * scrub rule functionality is hopefully removed some day in future.
405 */
406 V_pf_status.reass = 0;
407
408 V_pf_pfil_hooked = false;
409 V_pf_pfil_eth_hooked = false;
410
411 /* XXX do our best to avoid a conflict */
412 V_pf_status.hostid = arc4random();
413
414 for (int i = 0; i < PFRES_MAX; i++)
415 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
416 for (int i = 0; i < KLCNT_MAX; i++)
417 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
418 for (int i = 0; i < FCNT_MAX; i++)
419 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
420 for (int i = 0; i < SCNT_MAX; i++)
421 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
422
423 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
424 INTR_MPSAFE, &V_pf_swi_cookie) != 0)
425 /* XXXGL: leaked all above. */
426 return;
427 }
428
429 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)430 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
431 u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
432 u_int8_t check_ticket, int which)
433 {
434 struct pf_kruleset *ruleset;
435 struct pf_krule *rule;
436 int rs_num;
437
438 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT);
439
440 ruleset = pf_find_kruleset(anchor);
441 if (ruleset == NULL)
442 return (NULL);
443 rs_num = pf_get_ruleset_number(rule_action);
444 if (rs_num >= PF_RULESET_MAX)
445 return (NULL);
446 if (active) {
447 if (check_ticket && ticket !=
448 ruleset->rules[rs_num].active.ticket)
449 return (NULL);
450 if (r_last)
451 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
452 pf_krulequeue);
453 else
454 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
455 } else {
456 if (check_ticket && ticket !=
457 ruleset->rules[rs_num].inactive.ticket)
458 return (NULL);
459 if (r_last)
460 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
461 pf_krulequeue);
462 else
463 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
464 }
465 if (!r_last) {
466 while ((rule != NULL) && (rule->nr != rule_number))
467 rule = TAILQ_NEXT(rule, entries);
468 }
469 if (rule == NULL)
470 return (NULL);
471
472 switch (which) {
473 case PF_RDR:
474 return (&rule->rdr);
475 case PF_NAT:
476 return (&rule->nat);
477 case PF_RT:
478 return (&rule->route);
479 default:
480 panic("Unknow pool type %d", which);
481 }
482 }
483
484 static void
pf_mv_kpool(struct pf_kpalist * poola,struct pf_kpalist * poolb)485 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
486 {
487 struct pf_kpooladdr *mv_pool_pa;
488
489 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
490 TAILQ_REMOVE(poola, mv_pool_pa, entries);
491 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
492 }
493 }
494
495 static void
pf_empty_kpool(struct pf_kpalist * poola)496 pf_empty_kpool(struct pf_kpalist *poola)
497 {
498 struct pf_kpooladdr *pa;
499
500 while ((pa = TAILQ_FIRST(poola)) != NULL) {
501 switch (pa->addr.type) {
502 case PF_ADDR_DYNIFTL:
503 pfi_dynaddr_remove(pa->addr.p.dyn);
504 break;
505 case PF_ADDR_TABLE:
506 /* XXX: this could be unfinished pooladdr on pabuf */
507 if (pa->addr.p.tbl != NULL)
508 pfr_detach_table(pa->addr.p.tbl);
509 break;
510 }
511 if (pa->kif)
512 pfi_kkif_unref(pa->kif);
513 TAILQ_REMOVE(poola, pa, entries);
514 free(pa, M_PFRULE);
515 }
516 }
517
518 static void
pf_unlink_rule_locked(struct pf_krulequeue * rulequeue,struct pf_krule * rule)519 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
520 {
521
522 PF_RULES_WASSERT();
523 PF_UNLNKDRULES_ASSERT();
524
525 TAILQ_REMOVE(rulequeue, rule, entries);
526
527 rule->rule_ref |= PFRULE_REFS;
528 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
529 }
530
531 static void
pf_unlink_rule(struct pf_krulequeue * rulequeue,struct pf_krule * rule)532 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
533 {
534
535 PF_RULES_WASSERT();
536
537 PF_UNLNKDRULES_LOCK();
538 pf_unlink_rule_locked(rulequeue, rule);
539 PF_UNLNKDRULES_UNLOCK();
540 }
541
542 static void
pf_free_eth_rule(struct pf_keth_rule * rule)543 pf_free_eth_rule(struct pf_keth_rule *rule)
544 {
545 PF_RULES_WASSERT();
546
547 if (rule == NULL)
548 return;
549
550 if (rule->tag)
551 tag_unref(&V_pf_tags, rule->tag);
552 if (rule->match_tag)
553 tag_unref(&V_pf_tags, rule->match_tag);
554 #ifdef ALTQ
555 pf_qid_unref(rule->qid);
556 #endif
557
558 if (rule->bridge_to)
559 pfi_kkif_unref(rule->bridge_to);
560 if (rule->kif)
561 pfi_kkif_unref(rule->kif);
562
563 if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
564 pfr_detach_table(rule->ipsrc.addr.p.tbl);
565 if (rule->ipdst.addr.type == PF_ADDR_TABLE)
566 pfr_detach_table(rule->ipdst.addr.p.tbl);
567
568 counter_u64_free(rule->evaluations);
569 for (int i = 0; i < 2; i++) {
570 counter_u64_free(rule->packets[i]);
571 counter_u64_free(rule->bytes[i]);
572 }
573 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
574 pf_keth_anchor_remove(rule);
575
576 free(rule, M_PFRULE);
577 }
578
579 void
pf_free_rule(struct pf_krule * rule)580 pf_free_rule(struct pf_krule *rule)
581 {
582
583 PF_RULES_WASSERT();
584 PF_CONFIG_ASSERT();
585
586 if (rule->tag)
587 tag_unref(&V_pf_tags, rule->tag);
588 if (rule->match_tag)
589 tag_unref(&V_pf_tags, rule->match_tag);
590 #ifdef ALTQ
591 if (rule->pqid != rule->qid)
592 pf_qid_unref(rule->pqid);
593 pf_qid_unref(rule->qid);
594 #endif
595 switch (rule->src.addr.type) {
596 case PF_ADDR_DYNIFTL:
597 pfi_dynaddr_remove(rule->src.addr.p.dyn);
598 break;
599 case PF_ADDR_TABLE:
600 pfr_detach_table(rule->src.addr.p.tbl);
601 break;
602 }
603 switch (rule->dst.addr.type) {
604 case PF_ADDR_DYNIFTL:
605 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
606 break;
607 case PF_ADDR_TABLE:
608 pfr_detach_table(rule->dst.addr.p.tbl);
609 break;
610 }
611 if (rule->overload_tbl)
612 pfr_detach_table(rule->overload_tbl);
613 if (rule->kif)
614 pfi_kkif_unref(rule->kif);
615 if (rule->rcv_kif)
616 pfi_kkif_unref(rule->rcv_kif);
617 pf_kanchor_remove(rule);
618 pf_empty_kpool(&rule->rdr.list);
619 pf_empty_kpool(&rule->nat.list);
620 pf_empty_kpool(&rule->route.list);
621
622 pf_krule_free(rule);
623 }
624
625 static void
pf_init_tagset(struct pf_tagset * ts,unsigned int * tunable_size,unsigned int default_size)626 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
627 unsigned int default_size)
628 {
629 unsigned int i;
630 unsigned int hashsize;
631
632 if (*tunable_size == 0 || !powerof2(*tunable_size))
633 *tunable_size = default_size;
634
635 hashsize = *tunable_size;
636 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
637 M_WAITOK);
638 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
639 M_WAITOK);
640 ts->mask = hashsize - 1;
641 ts->seed = arc4random();
642 for (i = 0; i < hashsize; i++) {
643 TAILQ_INIT(&ts->namehash[i]);
644 TAILQ_INIT(&ts->taghash[i]);
645 }
646 BIT_FILL(TAGID_MAX, &ts->avail);
647 }
648
649 static void
pf_cleanup_tagset(struct pf_tagset * ts)650 pf_cleanup_tagset(struct pf_tagset *ts)
651 {
652 unsigned int i;
653 unsigned int hashsize;
654 struct pf_tagname *t, *tmp;
655
656 /*
657 * Only need to clean up one of the hashes as each tag is hashed
658 * into each table.
659 */
660 hashsize = ts->mask + 1;
661 for (i = 0; i < hashsize; i++)
662 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
663 uma_zfree(V_pf_tag_z, t);
664
665 free(ts->namehash, M_PFHASH);
666 free(ts->taghash, M_PFHASH);
667 }
668
669 static uint16_t
tagname2hashindex(const struct pf_tagset * ts,const char * tagname)670 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
671 {
672 size_t len;
673
674 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
675 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
676 }
677
678 static uint16_t
tag2hashindex(const struct pf_tagset * ts,uint16_t tag)679 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
680 {
681
682 return (tag & ts->mask);
683 }
684
685 static u_int16_t
tagname2tag(struct pf_tagset * ts,const char * tagname)686 tagname2tag(struct pf_tagset *ts, const char *tagname)
687 {
688 struct pf_tagname *tag;
689 u_int32_t index;
690 u_int16_t new_tagid;
691
692 PF_RULES_WASSERT();
693
694 index = tagname2hashindex(ts, tagname);
695 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
696 if (strcmp(tagname, tag->name) == 0) {
697 tag->ref++;
698 return (tag->tag);
699 }
700
701 /*
702 * new entry
703 *
704 * to avoid fragmentation, we do a linear search from the beginning
705 * and take the first free slot we find.
706 */
707 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
708 /*
709 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
710 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
711 * set. It may also return a bit number greater than TAGID_MAX due
712 * to rounding of the number of bits in the vector up to a multiple
713 * of the vector word size at declaration/allocation time.
714 */
715 if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
716 return (0);
717
718 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */
719 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
720
721 /* allocate and fill new struct pf_tagname */
722 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
723 if (tag == NULL)
724 return (0);
725 strlcpy(tag->name, tagname, sizeof(tag->name));
726 tag->tag = new_tagid;
727 tag->ref = 1;
728
729 /* Insert into namehash */
730 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
731
732 /* Insert into taghash */
733 index = tag2hashindex(ts, new_tagid);
734 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
735
736 return (tag->tag);
737 }
738
739 static void
tag_unref(struct pf_tagset * ts,u_int16_t tag)740 tag_unref(struct pf_tagset *ts, u_int16_t tag)
741 {
742 struct pf_tagname *t;
743 uint16_t index;
744
745 PF_RULES_WASSERT();
746
747 index = tag2hashindex(ts, tag);
748 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
749 if (tag == t->tag) {
750 if (--t->ref == 0) {
751 TAILQ_REMOVE(&ts->taghash[index], t,
752 taghash_entries);
753 index = tagname2hashindex(ts, t->name);
754 TAILQ_REMOVE(&ts->namehash[index], t,
755 namehash_entries);
756 /* Bits are 0-based for BIT_SET() */
757 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
758 uma_zfree(V_pf_tag_z, t);
759 }
760 break;
761 }
762 }
763
764 static uint16_t
pf_tagname2tag(const char * tagname)765 pf_tagname2tag(const char *tagname)
766 {
767 return (tagname2tag(&V_pf_tags, tagname));
768 }
769
770 static int
pf_begin_eth(uint32_t * ticket,const char * anchor)771 pf_begin_eth(uint32_t *ticket, const char *anchor)
772 {
773 struct pf_keth_rule *rule, *tmp;
774 struct pf_keth_ruleset *rs;
775
776 PF_RULES_WASSERT();
777
778 rs = pf_find_or_create_keth_ruleset(anchor);
779 if (rs == NULL)
780 return (EINVAL);
781
782 /* Purge old inactive rules. */
783 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
784 tmp) {
785 TAILQ_REMOVE(rs->inactive.rules, rule,
786 entries);
787 pf_free_eth_rule(rule);
788 }
789
790 *ticket = ++rs->inactive.ticket;
791 rs->inactive.open = 1;
792
793 return (0);
794 }
795
796 static int
pf_rollback_eth(uint32_t ticket,const char * anchor)797 pf_rollback_eth(uint32_t ticket, const char *anchor)
798 {
799 struct pf_keth_rule *rule, *tmp;
800 struct pf_keth_ruleset *rs;
801
802 PF_RULES_WASSERT();
803
804 rs = pf_find_keth_ruleset(anchor);
805 if (rs == NULL)
806 return (EINVAL);
807
808 if (!rs->inactive.open ||
809 ticket != rs->inactive.ticket)
810 return (0);
811
812 /* Purge old inactive rules. */
813 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
814 tmp) {
815 TAILQ_REMOVE(rs->inactive.rules, rule, entries);
816 pf_free_eth_rule(rule);
817 }
818
819 rs->inactive.open = 0;
820
821 pf_remove_if_empty_keth_ruleset(rs);
822
823 return (0);
824 }
825
826 #define PF_SET_SKIP_STEPS(i) \
827 do { \
828 while (head[i] != cur) { \
829 head[i]->skip[i].ptr = cur; \
830 head[i] = TAILQ_NEXT(head[i], entries); \
831 } \
832 } while (0)
833
834 static void
pf_eth_calc_skip_steps(struct pf_keth_ruleq * rules)835 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
836 {
837 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
838 int i;
839
840 cur = TAILQ_FIRST(rules);
841 prev = cur;
842 for (i = 0; i < PFE_SKIP_COUNT; ++i)
843 head[i] = cur;
844 while (cur != NULL) {
845 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
846 PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
847 if (cur->direction != prev->direction)
848 PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
849 if (cur->proto != prev->proto)
850 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
851 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
852 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
853 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
854 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
855 if (cur->ipsrc.neg != prev->ipsrc.neg ||
856 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
857 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
858 if (cur->ipdst.neg != prev->ipdst.neg ||
859 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
860 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
861
862 prev = cur;
863 cur = TAILQ_NEXT(cur, entries);
864 }
865 for (i = 0; i < PFE_SKIP_COUNT; ++i)
866 PF_SET_SKIP_STEPS(i);
867 }
868
869 static int
pf_commit_eth(uint32_t ticket,const char * anchor)870 pf_commit_eth(uint32_t ticket, const char *anchor)
871 {
872 struct pf_keth_ruleq *rules;
873 struct pf_keth_ruleset *rs;
874
875 rs = pf_find_keth_ruleset(anchor);
876 if (rs == NULL) {
877 return (EINVAL);
878 }
879
880 if (!rs->inactive.open ||
881 ticket != rs->inactive.ticket)
882 return (EBUSY);
883
884 PF_RULES_WASSERT();
885
886 pf_eth_calc_skip_steps(rs->inactive.rules);
887
888 rules = rs->active.rules;
889 atomic_store_ptr(&rs->active.rules, rs->inactive.rules);
890 rs->inactive.rules = rules;
891 rs->inactive.ticket = rs->active.ticket;
892
893 return (pf_rollback_eth(rs->inactive.ticket,
894 rs->anchor ? rs->anchor->path : ""));
895 }
896
897 #ifdef ALTQ
898 static uint16_t
pf_qname2qid(const char * qname)899 pf_qname2qid(const char *qname)
900 {
901 return (tagname2tag(&V_pf_qids, qname));
902 }
903
904 static void
pf_qid_unref(uint16_t qid)905 pf_qid_unref(uint16_t qid)
906 {
907 tag_unref(&V_pf_qids, qid);
908 }
909
910 static int
pf_begin_altq(u_int32_t * ticket)911 pf_begin_altq(u_int32_t *ticket)
912 {
913 struct pf_altq *altq, *tmp;
914 int error = 0;
915
916 PF_RULES_WASSERT();
917
918 /* Purge the old altq lists */
919 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
920 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
921 /* detach and destroy the discipline */
922 error = altq_remove(altq);
923 }
924 free(altq, M_PFALTQ);
925 }
926 TAILQ_INIT(V_pf_altq_ifs_inactive);
927 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
928 pf_qid_unref(altq->qid);
929 free(altq, M_PFALTQ);
930 }
931 TAILQ_INIT(V_pf_altqs_inactive);
932 if (error)
933 return (error);
934 *ticket = ++V_ticket_altqs_inactive;
935 V_altqs_inactive_open = 1;
936 return (0);
937 }
938
939 static int
pf_rollback_altq(u_int32_t ticket)940 pf_rollback_altq(u_int32_t ticket)
941 {
942 struct pf_altq *altq, *tmp;
943 int error = 0;
944
945 PF_RULES_WASSERT();
946
947 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
948 return (0);
949 /* Purge the old altq lists */
950 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
951 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
952 /* detach and destroy the discipline */
953 error = altq_remove(altq);
954 }
955 free(altq, M_PFALTQ);
956 }
957 TAILQ_INIT(V_pf_altq_ifs_inactive);
958 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
959 pf_qid_unref(altq->qid);
960 free(altq, M_PFALTQ);
961 }
962 TAILQ_INIT(V_pf_altqs_inactive);
963 V_altqs_inactive_open = 0;
964 return (error);
965 }
966
967 static int
pf_commit_altq(u_int32_t ticket)968 pf_commit_altq(u_int32_t ticket)
969 {
970 struct pf_altqqueue *old_altqs, *old_altq_ifs;
971 struct pf_altq *altq, *tmp;
972 int err, error = 0;
973
974 PF_RULES_WASSERT();
975
976 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
977 return (EBUSY);
978
979 /* swap altqs, keep the old. */
980 old_altqs = V_pf_altqs_active;
981 old_altq_ifs = V_pf_altq_ifs_active;
982 V_pf_altqs_active = V_pf_altqs_inactive;
983 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
984 V_pf_altqs_inactive = old_altqs;
985 V_pf_altq_ifs_inactive = old_altq_ifs;
986 V_ticket_altqs_active = V_ticket_altqs_inactive;
987
988 /* Attach new disciplines */
989 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
990 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
991 /* attach the discipline */
992 error = altq_pfattach(altq);
993 if (error == 0 && V_pf_altq_running)
994 error = pf_enable_altq(altq);
995 if (error != 0)
996 return (error);
997 }
998 }
999
1000 /* Purge the old altq lists */
1001 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1002 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1003 /* detach and destroy the discipline */
1004 if (V_pf_altq_running)
1005 error = pf_disable_altq(altq);
1006 err = altq_pfdetach(altq);
1007 if (err != 0 && error == 0)
1008 error = err;
1009 err = altq_remove(altq);
1010 if (err != 0 && error == 0)
1011 error = err;
1012 }
1013 free(altq, M_PFALTQ);
1014 }
1015 TAILQ_INIT(V_pf_altq_ifs_inactive);
1016 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1017 pf_qid_unref(altq->qid);
1018 free(altq, M_PFALTQ);
1019 }
1020 TAILQ_INIT(V_pf_altqs_inactive);
1021
1022 V_altqs_inactive_open = 0;
1023 return (error);
1024 }
1025
1026 static int
pf_enable_altq(struct pf_altq * altq)1027 pf_enable_altq(struct pf_altq *altq)
1028 {
1029 struct ifnet *ifp;
1030 struct tb_profile tb;
1031 int error = 0;
1032
1033 if ((ifp = ifunit(altq->ifname)) == NULL)
1034 return (EINVAL);
1035
1036 if (ifp->if_snd.altq_type != ALTQT_NONE)
1037 error = altq_enable(&ifp->if_snd);
1038
1039 /* set tokenbucket regulator */
1040 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1041 tb.rate = altq->ifbandwidth;
1042 tb.depth = altq->tbrsize;
1043 error = tbr_set(&ifp->if_snd, &tb);
1044 }
1045
1046 return (error);
1047 }
1048
1049 static int
pf_disable_altq(struct pf_altq * altq)1050 pf_disable_altq(struct pf_altq *altq)
1051 {
1052 struct ifnet *ifp;
1053 struct tb_profile tb;
1054 int error;
1055
1056 if ((ifp = ifunit(altq->ifname)) == NULL)
1057 return (EINVAL);
1058
1059 /*
1060 * when the discipline is no longer referenced, it was overridden
1061 * by a new one. if so, just return.
1062 */
1063 if (altq->altq_disc != ifp->if_snd.altq_disc)
1064 return (0);
1065
1066 error = altq_disable(&ifp->if_snd);
1067
1068 if (error == 0) {
1069 /* clear tokenbucket regulator */
1070 tb.rate = 0;
1071 error = tbr_set(&ifp->if_snd, &tb);
1072 }
1073
1074 return (error);
1075 }
1076
1077 static int
pf_altq_ifnet_event_add(struct ifnet * ifp,int remove,u_int32_t ticket,struct pf_altq * altq)1078 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1079 struct pf_altq *altq)
1080 {
1081 struct ifnet *ifp1;
1082 int error = 0;
1083
1084 /* Deactivate the interface in question */
1085 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1086 if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1087 (remove && ifp1 == ifp)) {
1088 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1089 } else {
1090 error = altq_add(ifp1, altq);
1091
1092 if (ticket != V_ticket_altqs_inactive)
1093 error = EBUSY;
1094
1095 if (error)
1096 free(altq, M_PFALTQ);
1097 }
1098
1099 return (error);
1100 }
1101
1102 void
pf_altq_ifnet_event(struct ifnet * ifp,int remove)1103 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1104 {
1105 struct pf_altq *a1, *a2, *a3;
1106 u_int32_t ticket;
1107 int error = 0;
1108
1109 /*
1110 * No need to re-evaluate the configuration for events on interfaces
1111 * that do not support ALTQ, as it's not possible for such
1112 * interfaces to be part of the configuration.
1113 */
1114 if (!ALTQ_IS_READY(&ifp->if_snd))
1115 return;
1116
1117 /* Interrupt userland queue modifications */
1118 if (V_altqs_inactive_open)
1119 pf_rollback_altq(V_ticket_altqs_inactive);
1120
1121 /* Start new altq ruleset */
1122 if (pf_begin_altq(&ticket))
1123 return;
1124
1125 /* Copy the current active set */
1126 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1127 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1128 if (a2 == NULL) {
1129 error = ENOMEM;
1130 break;
1131 }
1132 bcopy(a1, a2, sizeof(struct pf_altq));
1133
1134 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1135 if (error)
1136 break;
1137
1138 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1139 }
1140 if (error)
1141 goto out;
1142 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1143 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1144 if (a2 == NULL) {
1145 error = ENOMEM;
1146 break;
1147 }
1148 bcopy(a1, a2, sizeof(struct pf_altq));
1149
1150 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1151 error = EBUSY;
1152 free(a2, M_PFALTQ);
1153 break;
1154 }
1155 a2->altq_disc = NULL;
1156 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1157 if (strncmp(a3->ifname, a2->ifname,
1158 IFNAMSIZ) == 0) {
1159 a2->altq_disc = a3->altq_disc;
1160 break;
1161 }
1162 }
1163 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1164 if (error)
1165 break;
1166
1167 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1168 }
1169
1170 out:
1171 if (error != 0)
1172 pf_rollback_altq(ticket);
1173 else
1174 pf_commit_altq(ticket);
1175 }
1176 #endif /* ALTQ */
1177
1178 static struct pf_krule_global *
pf_rule_tree_alloc(int flags)1179 pf_rule_tree_alloc(int flags)
1180 {
1181 struct pf_krule_global *tree;
1182
1183 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1184 if (tree == NULL)
1185 return (NULL);
1186 RB_INIT(tree);
1187 return (tree);
1188 }
1189
1190 static void
pf_rule_tree_free(struct pf_krule_global * tree)1191 pf_rule_tree_free(struct pf_krule_global *tree)
1192 {
1193
1194 free(tree, M_TEMP);
1195 }
1196
1197 static int
pf_begin_rules(u_int32_t * ticket,int rs_num,const char * anchor)1198 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1199 {
1200 struct pf_krule_global *tree;
1201 struct pf_kruleset *rs;
1202 struct pf_krule *rule;
1203
1204 PF_RULES_WASSERT();
1205
1206 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1207 return (EINVAL);
1208 tree = pf_rule_tree_alloc(M_NOWAIT);
1209 if (tree == NULL)
1210 return (ENOMEM);
1211 rs = pf_find_or_create_kruleset(anchor);
1212 if (rs == NULL) {
1213 free(tree, M_TEMP);
1214 return (EINVAL);
1215 }
1216 pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1217 rs->rules[rs_num].inactive.tree = tree;
1218
1219 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1220 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1221 rs->rules[rs_num].inactive.rcount--;
1222 }
1223 *ticket = ++rs->rules[rs_num].inactive.ticket;
1224 rs->rules[rs_num].inactive.open = 1;
1225 return (0);
1226 }
1227
1228 static int
pf_rollback_rules(u_int32_t ticket,int rs_num,char * anchor)1229 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1230 {
1231 struct pf_kruleset *rs;
1232 struct pf_krule *rule;
1233
1234 PF_RULES_WASSERT();
1235
1236 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1237 return (EINVAL);
1238 rs = pf_find_kruleset(anchor);
1239 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1240 rs->rules[rs_num].inactive.ticket != ticket)
1241 return (0);
1242 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1243 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1244 rs->rules[rs_num].inactive.rcount--;
1245 }
1246 rs->rules[rs_num].inactive.open = 0;
1247 return (0);
1248 }
1249
1250 #define PF_MD5_UPD(st, elm) \
1251 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1252
1253 #define PF_MD5_UPD_STR(st, elm) \
1254 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1255
1256 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \
1257 (stor) = htonl((st)->elm); \
1258 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1259 } while (0)
1260
1261 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \
1262 (stor) = htons((st)->elm); \
1263 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1264 } while (0)
1265
1266 static void
pf_hash_rule_addr(MD5_CTX * ctx,struct pf_rule_addr * pfr)1267 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1268 {
1269 PF_MD5_UPD(pfr, addr.type);
1270 switch (pfr->addr.type) {
1271 case PF_ADDR_DYNIFTL:
1272 PF_MD5_UPD(pfr, addr.v.ifname);
1273 PF_MD5_UPD(pfr, addr.iflags);
1274 break;
1275 case PF_ADDR_TABLE:
1276 PF_MD5_UPD(pfr, addr.v.tblname);
1277 break;
1278 case PF_ADDR_ADDRMASK:
1279 /* XXX ignore af? */
1280 PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1281 PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1282 break;
1283 }
1284
1285 PF_MD5_UPD(pfr, port[0]);
1286 PF_MD5_UPD(pfr, port[1]);
1287 PF_MD5_UPD(pfr, neg);
1288 PF_MD5_UPD(pfr, port_op);
1289 }
1290
1291 static void
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)1292 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1293 {
1294 u_int16_t x;
1295 u_int32_t y;
1296
1297 pf_hash_rule_addr(ctx, &rule->src);
1298 pf_hash_rule_addr(ctx, &rule->dst);
1299 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1300 PF_MD5_UPD_STR(rule, label[i]);
1301 PF_MD5_UPD_STR(rule, ifname);
1302 PF_MD5_UPD_STR(rule, rcv_ifname);
1303 PF_MD5_UPD_STR(rule, match_tagname);
1304 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1305 PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1306 PF_MD5_UPD_HTONL(rule, prob, y);
1307 PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1308 PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1309 PF_MD5_UPD(rule, uid.op);
1310 PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1311 PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1312 PF_MD5_UPD(rule, gid.op);
1313 PF_MD5_UPD_HTONL(rule, rule_flag, y);
1314 PF_MD5_UPD(rule, action);
1315 PF_MD5_UPD(rule, direction);
1316 PF_MD5_UPD(rule, af);
1317 PF_MD5_UPD(rule, quick);
1318 PF_MD5_UPD(rule, ifnot);
1319 PF_MD5_UPD(rule, match_tag_not);
1320 PF_MD5_UPD(rule, natpass);
1321 PF_MD5_UPD(rule, keep_state);
1322 PF_MD5_UPD(rule, proto);
1323 PF_MD5_UPD(rule, type);
1324 PF_MD5_UPD(rule, code);
1325 PF_MD5_UPD(rule, flags);
1326 PF_MD5_UPD(rule, flagset);
1327 PF_MD5_UPD(rule, allow_opts);
1328 PF_MD5_UPD(rule, rt);
1329 PF_MD5_UPD(rule, tos);
1330 PF_MD5_UPD(rule, scrub_flags);
1331 PF_MD5_UPD(rule, min_ttl);
1332 PF_MD5_UPD(rule, set_tos);
1333 if (rule->anchor != NULL)
1334 PF_MD5_UPD_STR(rule, anchor->path);
1335 }
1336
1337 static void
pf_hash_rule(struct pf_krule * rule)1338 pf_hash_rule(struct pf_krule *rule)
1339 {
1340 MD5_CTX ctx;
1341
1342 MD5Init(&ctx);
1343 pf_hash_rule_rolling(&ctx, rule);
1344 MD5Final(rule->md5sum, &ctx);
1345 }
1346
1347 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1348 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1349 {
1350
1351 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1352 }
1353
1354 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1355 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1356 {
1357 struct pf_kruleset *rs;
1358 struct pf_krule *rule, **old_array, *old_rule;
1359 struct pf_krulequeue *old_rules;
1360 struct pf_krule_global *old_tree;
1361 int error;
1362 u_int32_t old_rcount;
1363
1364 PF_RULES_WASSERT();
1365
1366 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1367 return (EINVAL);
1368 rs = pf_find_kruleset(anchor);
1369 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1370 ticket != rs->rules[rs_num].inactive.ticket)
1371 return (EBUSY);
1372
1373 /* Calculate checksum for the main ruleset */
1374 if (rs == &pf_main_ruleset) {
1375 error = pf_setup_pfsync_matching(rs);
1376 if (error != 0)
1377 return (error);
1378 }
1379
1380 /* Swap rules, keep the old. */
1381 old_rules = rs->rules[rs_num].active.ptr;
1382 old_rcount = rs->rules[rs_num].active.rcount;
1383 old_array = rs->rules[rs_num].active.ptr_array;
1384 old_tree = rs->rules[rs_num].active.tree;
1385
1386 rs->rules[rs_num].active.ptr =
1387 rs->rules[rs_num].inactive.ptr;
1388 rs->rules[rs_num].active.ptr_array =
1389 rs->rules[rs_num].inactive.ptr_array;
1390 rs->rules[rs_num].active.tree =
1391 rs->rules[rs_num].inactive.tree;
1392 rs->rules[rs_num].active.rcount =
1393 rs->rules[rs_num].inactive.rcount;
1394
1395 /* Attempt to preserve counter information. */
1396 if (V_pf_status.keep_counters && old_tree != NULL) {
1397 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1398 entries) {
1399 old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1400 if (old_rule == NULL) {
1401 continue;
1402 }
1403 pf_counter_u64_critical_enter();
1404 pf_counter_u64_rollup_protected(&rule->evaluations,
1405 pf_counter_u64_fetch(&old_rule->evaluations));
1406 pf_counter_u64_rollup_protected(&rule->packets[0],
1407 pf_counter_u64_fetch(&old_rule->packets[0]));
1408 pf_counter_u64_rollup_protected(&rule->packets[1],
1409 pf_counter_u64_fetch(&old_rule->packets[1]));
1410 pf_counter_u64_rollup_protected(&rule->bytes[0],
1411 pf_counter_u64_fetch(&old_rule->bytes[0]));
1412 pf_counter_u64_rollup_protected(&rule->bytes[1],
1413 pf_counter_u64_fetch(&old_rule->bytes[1]));
1414 pf_counter_u64_critical_exit();
1415 }
1416 }
1417
1418 rs->rules[rs_num].inactive.ptr = old_rules;
1419 rs->rules[rs_num].inactive.ptr_array = old_array;
1420 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1421 rs->rules[rs_num].inactive.rcount = old_rcount;
1422
1423 rs->rules[rs_num].active.ticket =
1424 rs->rules[rs_num].inactive.ticket;
1425 pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1426
1427 /* Purge the old rule list. */
1428 PF_UNLNKDRULES_LOCK();
1429 while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1430 pf_unlink_rule_locked(old_rules, rule);
1431 PF_UNLNKDRULES_UNLOCK();
1432 if (rs->rules[rs_num].inactive.ptr_array)
1433 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1434 rs->rules[rs_num].inactive.ptr_array = NULL;
1435 rs->rules[rs_num].inactive.rcount = 0;
1436 rs->rules[rs_num].inactive.open = 0;
1437 pf_remove_if_empty_kruleset(rs);
1438 free(old_tree, M_TEMP);
1439
1440 return (0);
1441 }
1442
1443 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1444 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1445 {
1446 MD5_CTX ctx;
1447 struct pf_krule *rule;
1448 int rs_cnt;
1449 u_int8_t digest[PF_MD5_DIGEST_LENGTH];
1450
1451 MD5Init(&ctx);
1452 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1453 /* XXX PF_RULESET_SCRUB as well? */
1454 if (rs_cnt == PF_RULESET_SCRUB)
1455 continue;
1456
1457 if (rs->rules[rs_cnt].inactive.ptr_array)
1458 free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1459 rs->rules[rs_cnt].inactive.ptr_array = NULL;
1460
1461 if (rs->rules[rs_cnt].inactive.rcount) {
1462 rs->rules[rs_cnt].inactive.ptr_array =
1463 mallocarray(rs->rules[rs_cnt].inactive.rcount,
1464 sizeof(struct pf_rule **),
1465 M_TEMP, M_NOWAIT);
1466
1467 if (!rs->rules[rs_cnt].inactive.ptr_array)
1468 return (ENOMEM);
1469 }
1470
1471 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1472 entries) {
1473 pf_hash_rule_rolling(&ctx, rule);
1474 (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1475 }
1476 }
1477
1478 MD5Final(digest, &ctx);
1479 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1480 return (0);
1481 }
1482
1483 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1484 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1485 {
1486 int error = 0;
1487
1488 switch (addr->type) {
1489 case PF_ADDR_TABLE:
1490 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1491 if (addr->p.tbl == NULL)
1492 error = ENOMEM;
1493 break;
1494 default:
1495 error = EINVAL;
1496 }
1497
1498 return (error);
1499 }
1500
1501 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1502 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1503 sa_family_t af)
1504 {
1505 int error = 0;
1506
1507 switch (addr->type) {
1508 case PF_ADDR_TABLE:
1509 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1510 if (addr->p.tbl == NULL)
1511 error = ENOMEM;
1512 break;
1513 case PF_ADDR_DYNIFTL:
1514 error = pfi_dynaddr_setup(addr, af);
1515 break;
1516 }
1517
1518 return (error);
1519 }
1520
1521 void
pf_addr_copyout(struct pf_addr_wrap * addr)1522 pf_addr_copyout(struct pf_addr_wrap *addr)
1523 {
1524
1525 switch (addr->type) {
1526 case PF_ADDR_DYNIFTL:
1527 pfi_dynaddr_copyout(addr);
1528 break;
1529 case PF_ADDR_TABLE:
1530 pf_tbladdr_copyout(addr);
1531 break;
1532 }
1533 }
1534
1535 static void
pf_src_node_copy(const struct pf_ksrc_node * in,struct pf_src_node * out)1536 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1537 {
1538 int secs = time_uptime, diff;
1539
1540 bzero(out, sizeof(struct pf_src_node));
1541
1542 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1543 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1544
1545 if (in->rule != NULL)
1546 out->rule.nr = in->rule->nr;
1547
1548 for (int i = 0; i < 2; i++) {
1549 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1550 out->packets[i] = counter_u64_fetch(in->packets[i]);
1551 }
1552
1553 out->states = in->states;
1554 out->conn = in->conn;
1555 out->af = in->af;
1556 out->ruletype = in->ruletype;
1557
1558 out->creation = secs - in->creation;
1559 if (out->expire > secs)
1560 out->expire -= secs;
1561 else
1562 out->expire = 0;
1563
1564 /* Adjust the connection rate estimate. */
1565 out->conn_rate = in->conn_rate;
1566 diff = secs - in->conn_rate.last;
1567 if (diff >= in->conn_rate.seconds)
1568 out->conn_rate.count = 0;
1569 else
1570 out->conn_rate.count -=
1571 in->conn_rate.count * diff /
1572 in->conn_rate.seconds;
1573 }
1574
1575 #ifdef ALTQ
1576 /*
1577 * Handle export of struct pf_kaltq to user binaries that may be using any
1578 * version of struct pf_altq.
1579 */
1580 static int
pf_export_kaltq(struct pf_altq * q,struct pfioc_altq_v1 * pa,size_t ioc_size)1581 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1582 {
1583 u_int32_t version;
1584
1585 if (ioc_size == sizeof(struct pfioc_altq_v0))
1586 version = 0;
1587 else
1588 version = pa->version;
1589
1590 if (version > PFIOC_ALTQ_VERSION)
1591 return (EINVAL);
1592
1593 #define ASSIGN(x) exported_q->x = q->x
1594 #define COPY(x) \
1595 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1596 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1597 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1598
1599 switch (version) {
1600 case 0: {
1601 struct pf_altq_v0 *exported_q =
1602 &((struct pfioc_altq_v0 *)pa)->altq;
1603
1604 COPY(ifname);
1605
1606 ASSIGN(scheduler);
1607 ASSIGN(tbrsize);
1608 exported_q->tbrsize = SATU16(q->tbrsize);
1609 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1610
1611 COPY(qname);
1612 COPY(parent);
1613 ASSIGN(parent_qid);
1614 exported_q->bandwidth = SATU32(q->bandwidth);
1615 ASSIGN(priority);
1616 ASSIGN(local_flags);
1617
1618 ASSIGN(qlimit);
1619 ASSIGN(flags);
1620
1621 if (q->scheduler == ALTQT_HFSC) {
1622 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1623 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1624 SATU32(q->pq_u.hfsc_opts.x)
1625
1626 ASSIGN_OPT_SATU32(rtsc_m1);
1627 ASSIGN_OPT(rtsc_d);
1628 ASSIGN_OPT_SATU32(rtsc_m2);
1629
1630 ASSIGN_OPT_SATU32(lssc_m1);
1631 ASSIGN_OPT(lssc_d);
1632 ASSIGN_OPT_SATU32(lssc_m2);
1633
1634 ASSIGN_OPT_SATU32(ulsc_m1);
1635 ASSIGN_OPT(ulsc_d);
1636 ASSIGN_OPT_SATU32(ulsc_m2);
1637
1638 ASSIGN_OPT(flags);
1639
1640 #undef ASSIGN_OPT
1641 #undef ASSIGN_OPT_SATU32
1642 } else
1643 COPY(pq_u);
1644
1645 ASSIGN(qid);
1646 break;
1647 }
1648 case 1: {
1649 struct pf_altq_v1 *exported_q =
1650 &((struct pfioc_altq_v1 *)pa)->altq;
1651
1652 COPY(ifname);
1653
1654 ASSIGN(scheduler);
1655 ASSIGN(tbrsize);
1656 ASSIGN(ifbandwidth);
1657
1658 COPY(qname);
1659 COPY(parent);
1660 ASSIGN(parent_qid);
1661 ASSIGN(bandwidth);
1662 ASSIGN(priority);
1663 ASSIGN(local_flags);
1664
1665 ASSIGN(qlimit);
1666 ASSIGN(flags);
1667 COPY(pq_u);
1668
1669 ASSIGN(qid);
1670 break;
1671 }
1672 default:
1673 panic("%s: unhandled struct pfioc_altq version", __func__);
1674 break;
1675 }
1676
1677 #undef ASSIGN
1678 #undef COPY
1679 #undef SATU16
1680 #undef SATU32
1681
1682 return (0);
1683 }
1684
1685 /*
1686 * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1687 * that may be using any version of it.
1688 */
1689 static int
pf_import_kaltq(struct pfioc_altq_v1 * pa,struct pf_altq * q,size_t ioc_size)1690 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1691 {
1692 u_int32_t version;
1693
1694 if (ioc_size == sizeof(struct pfioc_altq_v0))
1695 version = 0;
1696 else
1697 version = pa->version;
1698
1699 if (version > PFIOC_ALTQ_VERSION)
1700 return (EINVAL);
1701
1702 #define ASSIGN(x) q->x = imported_q->x
1703 #define COPY(x) \
1704 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1705
1706 switch (version) {
1707 case 0: {
1708 struct pf_altq_v0 *imported_q =
1709 &((struct pfioc_altq_v0 *)pa)->altq;
1710
1711 COPY(ifname);
1712
1713 ASSIGN(scheduler);
1714 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1715 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1716
1717 COPY(qname);
1718 COPY(parent);
1719 ASSIGN(parent_qid);
1720 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1721 ASSIGN(priority);
1722 ASSIGN(local_flags);
1723
1724 ASSIGN(qlimit);
1725 ASSIGN(flags);
1726
1727 if (imported_q->scheduler == ALTQT_HFSC) {
1728 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1729
1730 /*
1731 * The m1 and m2 parameters are being copied from
1732 * 32-bit to 64-bit.
1733 */
1734 ASSIGN_OPT(rtsc_m1);
1735 ASSIGN_OPT(rtsc_d);
1736 ASSIGN_OPT(rtsc_m2);
1737
1738 ASSIGN_OPT(lssc_m1);
1739 ASSIGN_OPT(lssc_d);
1740 ASSIGN_OPT(lssc_m2);
1741
1742 ASSIGN_OPT(ulsc_m1);
1743 ASSIGN_OPT(ulsc_d);
1744 ASSIGN_OPT(ulsc_m2);
1745
1746 ASSIGN_OPT(flags);
1747
1748 #undef ASSIGN_OPT
1749 } else
1750 COPY(pq_u);
1751
1752 ASSIGN(qid);
1753 break;
1754 }
1755 case 1: {
1756 struct pf_altq_v1 *imported_q =
1757 &((struct pfioc_altq_v1 *)pa)->altq;
1758
1759 COPY(ifname);
1760
1761 ASSIGN(scheduler);
1762 ASSIGN(tbrsize);
1763 ASSIGN(ifbandwidth);
1764
1765 COPY(qname);
1766 COPY(parent);
1767 ASSIGN(parent_qid);
1768 ASSIGN(bandwidth);
1769 ASSIGN(priority);
1770 ASSIGN(local_flags);
1771
1772 ASSIGN(qlimit);
1773 ASSIGN(flags);
1774 COPY(pq_u);
1775
1776 ASSIGN(qid);
1777 break;
1778 }
1779 default:
1780 panic("%s: unhandled struct pfioc_altq version", __func__);
1781 break;
1782 }
1783
1784 #undef ASSIGN
1785 #undef COPY
1786
1787 return (0);
1788 }
1789
1790 static struct pf_altq *
pf_altq_get_nth_active(u_int32_t n)1791 pf_altq_get_nth_active(u_int32_t n)
1792 {
1793 struct pf_altq *altq;
1794 u_int32_t nr;
1795
1796 nr = 0;
1797 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1798 if (nr == n)
1799 return (altq);
1800 nr++;
1801 }
1802
1803 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1804 if (nr == n)
1805 return (altq);
1806 nr++;
1807 }
1808
1809 return (NULL);
1810 }
1811 #endif /* ALTQ */
1812
1813 struct pf_krule *
pf_krule_alloc(void)1814 pf_krule_alloc(void)
1815 {
1816 struct pf_krule *rule;
1817
1818 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1819 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF);
1820 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF);
1821 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF);
1822 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1823 M_WAITOK | M_ZERO);
1824 return (rule);
1825 }
1826
1827 void
pf_krule_free(struct pf_krule * rule)1828 pf_krule_free(struct pf_krule *rule)
1829 {
1830 #ifdef PF_WANT_32_TO_64_COUNTER
1831 bool wowned;
1832 #endif
1833
1834 if (rule == NULL)
1835 return;
1836
1837 #ifdef PF_WANT_32_TO_64_COUNTER
1838 if (rule->allrulelinked) {
1839 wowned = PF_RULES_WOWNED();
1840 if (!wowned)
1841 PF_RULES_WLOCK();
1842 LIST_REMOVE(rule, allrulelist);
1843 V_pf_allrulecount--;
1844 if (!wowned)
1845 PF_RULES_WUNLOCK();
1846 }
1847 #endif
1848
1849 pf_counter_u64_deinit(&rule->evaluations);
1850 for (int i = 0; i < 2; i++) {
1851 pf_counter_u64_deinit(&rule->packets[i]);
1852 pf_counter_u64_deinit(&rule->bytes[i]);
1853 }
1854 counter_u64_free(rule->states_cur);
1855 counter_u64_free(rule->states_tot);
1856 counter_u64_free(rule->src_nodes);
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 rule->src_nodes = counter_u64_alloc(M_WAITOK);
2093 rule->cuid = uid;
2094 rule->cpid = pid;
2095 TAILQ_INIT(&rule->rdr.list);
2096 TAILQ_INIT(&rule->nat.list);
2097 TAILQ_INIT(&rule->route.list);
2098
2099 PF_CONFIG_LOCK();
2100 PF_RULES_WLOCK();
2101 #ifdef PF_WANT_32_TO_64_COUNTER
2102 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2103 MPASS(!rule->allrulelinked);
2104 rule->allrulelinked = true;
2105 V_pf_allrulecount++;
2106 #endif
2107 ruleset = pf_find_kruleset(anchor);
2108 if (ruleset == NULL)
2109 ERROUT(EINVAL);
2110 rs_num = pf_get_ruleset_number(rule->action);
2111 if (rs_num >= PF_RULESET_MAX)
2112 ERROUT(EINVAL);
2113 if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2114 DPFPRINTF(PF_DEBUG_MISC,
2115 ("ticket: %d != [%d]%d\n", ticket, rs_num,
2116 ruleset->rules[rs_num].inactive.ticket));
2117 ERROUT(EBUSY);
2118 }
2119 if (pool_ticket != V_ticket_pabuf) {
2120 DPFPRINTF(PF_DEBUG_MISC,
2121 ("pool_ticket: %d != %d\n", pool_ticket,
2122 V_ticket_pabuf));
2123 ERROUT(EBUSY);
2124 }
2125 /*
2126 * XXXMJG hack: there is no mechanism to ensure they started the
2127 * transaction. Ticket checked above may happen to match by accident,
2128 * even if nobody called DIOCXBEGIN, let alone this process.
2129 * Partially work around it by checking if the RB tree got allocated,
2130 * see pf_begin_rules.
2131 */
2132 if (ruleset->rules[rs_num].inactive.tree == NULL) {
2133 ERROUT(EINVAL);
2134 }
2135
2136 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2137 pf_krulequeue);
2138 if (tail)
2139 rule->nr = tail->nr + 1;
2140 else
2141 rule->nr = 0;
2142 if (rule->ifname[0]) {
2143 rule->kif = pfi_kkif_attach(kif, rule->ifname);
2144 kif = NULL;
2145 pfi_kkif_ref(rule->kif);
2146 } else
2147 rule->kif = NULL;
2148
2149 if (rule->rcv_ifname[0]) {
2150 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname);
2151 rcv_kif = NULL;
2152 pfi_kkif_ref(rule->rcv_kif);
2153 } else
2154 rule->rcv_kif = NULL;
2155
2156 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2157 error = EBUSY;
2158
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_addr_setup(ruleset, &rule->src.addr, rule->af))
2184 error = ENOMEM;
2185 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2186 error = ENOMEM;
2187 if (pf_kanchor_setup(rule, ruleset, anchor_call))
2188 error = EINVAL;
2189 if (rule->scrub_flags & PFSTATE_SETPRIO &&
2190 (rule->set_prio[0] > PF_PRIO_MAX ||
2191 rule->set_prio[1] > PF_PRIO_MAX))
2192 error = EINVAL;
2193 for (int i = 0; i < 3; i++) {
2194 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
2195 if (pa->addr.type == PF_ADDR_TABLE) {
2196 pa->addr.p.tbl = pfr_attach_table(ruleset,
2197 pa->addr.v.tblname);
2198 if (pa->addr.p.tbl == NULL)
2199 error = ENOMEM;
2200 }
2201 }
2202
2203 rule->overload_tbl = NULL;
2204 if (rule->overload_tblname[0]) {
2205 if ((rule->overload_tbl = pfr_attach_table(ruleset,
2206 rule->overload_tblname)) == NULL)
2207 error = EINVAL;
2208 else
2209 rule->overload_tbl->pfrkt_flags |=
2210 PFR_TFLAG_ACTIVE;
2211 }
2212
2213 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list);
2214 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list);
2215 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list);
2216 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2217 (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2218 (rule->rt > PF_NOPFROUTE)) &&
2219 (TAILQ_FIRST(&rule->rdr.list) == NULL &&
2220 TAILQ_FIRST(&rule->route.list) == NULL))
2221 error = EINVAL;
2222
2223 if (rule->action == PF_PASS && rule->rdr.opts & PF_POOL_STICKYADDR &&
2224 !rule->keep_state) {
2225 error = EINVAL;
2226 }
2227
2228 if (error) {
2229 pf_free_rule(rule);
2230 rule = NULL;
2231 ERROUT(error);
2232 }
2233
2234 rule->nat.cur = TAILQ_FIRST(&rule->nat.list);
2235 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list);
2236 rule->route.cur = TAILQ_FIRST(&rule->route.list);
2237 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2238 rule, entries);
2239 ruleset->rules[rs_num].inactive.rcount++;
2240
2241 PF_RULES_WUNLOCK();
2242 pf_hash_rule(rule);
2243 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2244 PF_RULES_WLOCK();
2245 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2246 ruleset->rules[rs_num].inactive.rcount--;
2247 pf_free_rule(rule);
2248 rule = NULL;
2249 ERROUT(EEXIST);
2250 }
2251 PF_CONFIG_UNLOCK();
2252
2253 return (0);
2254
2255 #undef ERROUT
2256 errout:
2257 PF_RULES_WUNLOCK();
2258 PF_CONFIG_UNLOCK();
2259 errout_unlocked:
2260 pf_kkif_free(rcv_kif);
2261 pf_kkif_free(kif);
2262 pf_krule_free(rule);
2263 return (error);
2264 }
2265
2266 static bool
pf_label_match(const struct pf_krule * rule,const char * label)2267 pf_label_match(const struct pf_krule *rule, const char *label)
2268 {
2269 int i = 0;
2270
2271 while (*rule->label[i]) {
2272 if (strcmp(rule->label[i], label) == 0)
2273 return (true);
2274 i++;
2275 }
2276
2277 return (false);
2278 }
2279
2280 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)2281 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2282 {
2283 struct pf_kstate *s;
2284 int more = 0;
2285
2286 s = pf_find_state_all(key, dir, &more);
2287 if (s == NULL)
2288 return (0);
2289
2290 if (more) {
2291 PF_STATE_UNLOCK(s);
2292 return (0);
2293 }
2294
2295 pf_unlink_state(s);
2296 return (1);
2297 }
2298
2299 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)2300 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2301 {
2302 struct pf_kstate *s;
2303 struct pf_state_key *sk;
2304 struct pf_addr *srcaddr, *dstaddr;
2305 struct pf_state_key_cmp match_key;
2306 int idx, killed = 0;
2307 unsigned int dir;
2308 u_int16_t srcport, dstport;
2309 struct pfi_kkif *kif;
2310
2311 relock_DIOCKILLSTATES:
2312 PF_HASHROW_LOCK(ih);
2313 LIST_FOREACH(s, &ih->states, entry) {
2314 /* For floating states look at the original kif. */
2315 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2316
2317 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2318 if (s->direction == PF_OUT) {
2319 srcaddr = &sk->addr[1];
2320 dstaddr = &sk->addr[0];
2321 srcport = sk->port[1];
2322 dstport = sk->port[0];
2323 } else {
2324 srcaddr = &sk->addr[0];
2325 dstaddr = &sk->addr[1];
2326 srcport = sk->port[0];
2327 dstport = sk->port[1];
2328 }
2329
2330 if (psk->psk_af && sk->af != psk->psk_af)
2331 continue;
2332
2333 if (psk->psk_proto && psk->psk_proto != sk->proto)
2334 continue;
2335
2336 if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2337 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2338 continue;
2339
2340 if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2341 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2342 continue;
2343
2344 if (! PF_MATCHA(psk->psk_rt_addr.neg,
2345 &psk->psk_rt_addr.addr.v.a.addr,
2346 &psk->psk_rt_addr.addr.v.a.mask,
2347 &s->act.rt_addr, sk->af))
2348 continue;
2349
2350 if (psk->psk_src.port_op != 0 &&
2351 ! pf_match_port(psk->psk_src.port_op,
2352 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2353 continue;
2354
2355 if (psk->psk_dst.port_op != 0 &&
2356 ! pf_match_port(psk->psk_dst.port_op,
2357 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2358 continue;
2359
2360 if (psk->psk_label[0] &&
2361 ! pf_label_match(s->rule, psk->psk_label))
2362 continue;
2363
2364 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2365 kif->pfik_name))
2366 continue;
2367
2368 if (psk->psk_kill_match) {
2369 /* Create the key to find matching states, with lock
2370 * held. */
2371
2372 bzero(&match_key, sizeof(match_key));
2373
2374 if (s->direction == PF_OUT) {
2375 dir = PF_IN;
2376 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2377 } else {
2378 dir = PF_OUT;
2379 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2380 }
2381
2382 match_key.af = s->key[idx]->af;
2383 match_key.proto = s->key[idx]->proto;
2384 PF_ACPY(&match_key.addr[0],
2385 &s->key[idx]->addr[1], match_key.af);
2386 match_key.port[0] = s->key[idx]->port[1];
2387 PF_ACPY(&match_key.addr[1],
2388 &s->key[idx]->addr[0], match_key.af);
2389 match_key.port[1] = s->key[idx]->port[0];
2390 }
2391
2392 pf_unlink_state(s);
2393 killed++;
2394
2395 if (psk->psk_kill_match)
2396 killed += pf_kill_matching_state(&match_key, dir);
2397
2398 goto relock_DIOCKILLSTATES;
2399 }
2400 PF_HASHROW_UNLOCK(ih);
2401
2402 return (killed);
2403 }
2404
2405 int
pf_start(void)2406 pf_start(void)
2407 {
2408 int error = 0;
2409
2410 sx_xlock(&V_pf_ioctl_lock);
2411 if (V_pf_status.running)
2412 error = EEXIST;
2413 else {
2414 hook_pf();
2415 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2416 hook_pf_eth();
2417 V_pf_status.running = 1;
2418 V_pf_status.since = time_second;
2419 new_unrhdr64(&V_pf_stateid, time_second);
2420
2421 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2422 }
2423 sx_xunlock(&V_pf_ioctl_lock);
2424
2425 return (error);
2426 }
2427
2428 int
pf_stop(void)2429 pf_stop(void)
2430 {
2431 int error = 0;
2432
2433 sx_xlock(&V_pf_ioctl_lock);
2434 if (!V_pf_status.running)
2435 error = ENOENT;
2436 else {
2437 V_pf_status.running = 0;
2438 dehook_pf();
2439 dehook_pf_eth();
2440 V_pf_status.since = time_second;
2441 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2442 }
2443 sx_xunlock(&V_pf_ioctl_lock);
2444
2445 return (error);
2446 }
2447
2448 void
pf_ioctl_clear_status(void)2449 pf_ioctl_clear_status(void)
2450 {
2451 PF_RULES_WLOCK();
2452 for (int i = 0; i < PFRES_MAX; i++)
2453 counter_u64_zero(V_pf_status.counters[i]);
2454 for (int i = 0; i < FCNT_MAX; i++)
2455 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2456 for (int i = 0; i < SCNT_MAX; i++)
2457 counter_u64_zero(V_pf_status.scounters[i]);
2458 for (int i = 0; i < KLCNT_MAX; i++)
2459 counter_u64_zero(V_pf_status.lcounters[i]);
2460 V_pf_status.since = time_second;
2461 if (*V_pf_status.ifname)
2462 pfi_update_status(V_pf_status.ifname, NULL);
2463 PF_RULES_WUNLOCK();
2464 }
2465
2466 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2467 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2468 {
2469 uint32_t old;
2470
2471 if (timeout < 0 || timeout >= PFTM_MAX ||
2472 seconds < 0)
2473 return (EINVAL);
2474
2475 PF_RULES_WLOCK();
2476 old = V_pf_default_rule.timeout[timeout];
2477 if (timeout == PFTM_INTERVAL && seconds == 0)
2478 seconds = 1;
2479 V_pf_default_rule.timeout[timeout] = seconds;
2480 if (timeout == PFTM_INTERVAL && seconds < old)
2481 wakeup(pf_purge_thread);
2482
2483 if (prev_seconds != NULL)
2484 *prev_seconds = old;
2485
2486 PF_RULES_WUNLOCK();
2487
2488 return (0);
2489 }
2490
2491 int
pf_ioctl_get_timeout(int timeout,int * seconds)2492 pf_ioctl_get_timeout(int timeout, int *seconds)
2493 {
2494 PF_RULES_RLOCK_TRACKER;
2495
2496 if (timeout < 0 || timeout >= PFTM_MAX)
2497 return (EINVAL);
2498
2499 PF_RULES_RLOCK();
2500 *seconds = V_pf_default_rule.timeout[timeout];
2501 PF_RULES_RUNLOCK();
2502
2503 return (0);
2504 }
2505
2506 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2507 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2508 {
2509
2510 PF_RULES_WLOCK();
2511 if (index < 0 || index >= PF_LIMIT_MAX ||
2512 V_pf_limits[index].zone == NULL) {
2513 PF_RULES_WUNLOCK();
2514 return (EINVAL);
2515 }
2516 uma_zone_set_max(V_pf_limits[index].zone, limit);
2517 if (old_limit != NULL)
2518 *old_limit = V_pf_limits[index].limit;
2519 V_pf_limits[index].limit = limit;
2520 PF_RULES_WUNLOCK();
2521
2522 return (0);
2523 }
2524
2525 int
pf_ioctl_get_limit(int index,unsigned int * limit)2526 pf_ioctl_get_limit(int index, unsigned int *limit)
2527 {
2528 PF_RULES_RLOCK_TRACKER;
2529
2530 if (index < 0 || index >= PF_LIMIT_MAX)
2531 return (EINVAL);
2532
2533 PF_RULES_RLOCK();
2534 *limit = V_pf_limits[index].limit;
2535 PF_RULES_RUNLOCK();
2536
2537 return (0);
2538 }
2539
2540 int
pf_ioctl_begin_addrs(uint32_t * ticket)2541 pf_ioctl_begin_addrs(uint32_t *ticket)
2542 {
2543 PF_RULES_WLOCK();
2544 pf_empty_kpool(&V_pf_pabuf[0]);
2545 pf_empty_kpool(&V_pf_pabuf[1]);
2546 pf_empty_kpool(&V_pf_pabuf[2]);
2547 *ticket = ++V_ticket_pabuf;
2548 PF_RULES_WUNLOCK();
2549
2550 return (0);
2551 }
2552
2553 int
pf_ioctl_add_addr(struct pf_nl_pooladdr * pp)2554 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
2555 {
2556 struct pf_kpooladdr *pa = NULL;
2557 struct pfi_kkif *kif = NULL;
2558 int error;
2559
2560 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2561 pp->which != PF_RT)
2562 return (EINVAL);
2563
2564 #ifndef INET
2565 if (pp->af == AF_INET)
2566 return (EAFNOSUPPORT);
2567 #endif /* INET */
2568 #ifndef INET6
2569 if (pp->af == AF_INET6)
2570 return (EAFNOSUPPORT);
2571 #endif /* INET6 */
2572
2573 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2574 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2575 pp->addr.addr.type != PF_ADDR_TABLE)
2576 return (EINVAL);
2577
2578 if (pp->addr.addr.p.dyn != NULL)
2579 return (EINVAL);
2580
2581 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2582 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2583 if (error != 0)
2584 goto out;
2585 if (pa->ifname[0])
2586 kif = pf_kkif_create(M_WAITOK);
2587 PF_RULES_WLOCK();
2588 if (pp->ticket != V_ticket_pabuf) {
2589 PF_RULES_WUNLOCK();
2590 if (pa->ifname[0])
2591 pf_kkif_free(kif);
2592 error = EBUSY;
2593 goto out;
2594 }
2595 if (pa->ifname[0]) {
2596 pa->kif = pfi_kkif_attach(kif, pa->ifname);
2597 kif = NULL;
2598 pfi_kkif_ref(pa->kif);
2599 } else
2600 pa->kif = NULL;
2601 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2602 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2603 if (pa->ifname[0])
2604 pfi_kkif_unref(pa->kif);
2605 PF_RULES_WUNLOCK();
2606 goto out;
2607 }
2608 switch (pp->which) {
2609 case PF_NAT:
2610 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries);
2611 break;
2612 case PF_RDR:
2613 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries);
2614 break;
2615 case PF_RT:
2616 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries);
2617 break;
2618 }
2619 PF_RULES_WUNLOCK();
2620
2621 return (0);
2622
2623 out:
2624 free(pa, M_PFRULE);
2625 return (error);
2626 }
2627
2628 int
pf_ioctl_get_addrs(struct pf_nl_pooladdr * pp)2629 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
2630 {
2631 struct pf_kpool *pool;
2632 struct pf_kpooladdr *pa;
2633
2634 PF_RULES_RLOCK_TRACKER;
2635
2636 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2637 pp->which != PF_RT)
2638 return (EINVAL);
2639
2640 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2641 pp->nr = 0;
2642
2643 PF_RULES_RLOCK();
2644 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2645 pp->r_num, 0, 1, 0, pp->which);
2646 if (pool == NULL) {
2647 PF_RULES_RUNLOCK();
2648 return (EBUSY);
2649 }
2650 TAILQ_FOREACH(pa, &pool->list, entries)
2651 pp->nr++;
2652 PF_RULES_RUNLOCK();
2653
2654 return (0);
2655 }
2656
2657 int
pf_ioctl_get_addr(struct pf_nl_pooladdr * pp)2658 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
2659 {
2660 struct pf_kpool *pool;
2661 struct pf_kpooladdr *pa;
2662 u_int32_t nr = 0;
2663
2664 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2665 pp->which != PF_RT)
2666 return (EINVAL);
2667
2668 PF_RULES_RLOCK_TRACKER;
2669
2670 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2671
2672 PF_RULES_RLOCK();
2673 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2674 pp->r_num, 0, 1, 1, pp->which);
2675 if (pool == NULL) {
2676 PF_RULES_RUNLOCK();
2677 return (EBUSY);
2678 }
2679 pa = TAILQ_FIRST(&pool->list);
2680 while ((pa != NULL) && (nr < pp->nr)) {
2681 pa = TAILQ_NEXT(pa, entries);
2682 nr++;
2683 }
2684 if (pa == NULL) {
2685 PF_RULES_RUNLOCK();
2686 return (EBUSY);
2687 }
2688 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2689 pf_addr_copyout(&pp->addr.addr);
2690 PF_RULES_RUNLOCK();
2691
2692 return (0);
2693 }
2694
2695 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)2696 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
2697 {
2698 struct pf_kruleset *ruleset;
2699 struct pf_kanchor *anchor;
2700
2701 PF_RULES_RLOCK_TRACKER;
2702
2703 pr->path[sizeof(pr->path) - 1] = 0;
2704
2705 PF_RULES_RLOCK();
2706 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2707 PF_RULES_RUNLOCK();
2708 return (ENOENT);
2709 }
2710 pr->nr = 0;
2711 if (ruleset->anchor == NULL) {
2712 /* XXX kludge for pf_main_ruleset */
2713 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2714 if (anchor->parent == NULL)
2715 pr->nr++;
2716 } else {
2717 RB_FOREACH(anchor, pf_kanchor_node,
2718 &ruleset->anchor->children)
2719 pr->nr++;
2720 }
2721 PF_RULES_RUNLOCK();
2722
2723 return (0);
2724 }
2725
2726 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)2727 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
2728 {
2729 struct pf_kruleset *ruleset;
2730 struct pf_kanchor *anchor;
2731 u_int32_t nr = 0;
2732 int error = 0;
2733
2734 PF_RULES_RLOCK_TRACKER;
2735
2736 PF_RULES_RLOCK();
2737 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2738 PF_RULES_RUNLOCK();
2739 return (ENOENT);
2740 }
2741
2742 pr->name[0] = 0;
2743 if (ruleset->anchor == NULL) {
2744 /* XXX kludge for pf_main_ruleset */
2745 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2746 if (anchor->parent == NULL && nr++ == pr->nr) {
2747 strlcpy(pr->name, anchor->name,
2748 sizeof(pr->name));
2749 break;
2750 }
2751 } else {
2752 RB_FOREACH(anchor, pf_kanchor_node,
2753 &ruleset->anchor->children)
2754 if (nr++ == pr->nr) {
2755 strlcpy(pr->name, anchor->name,
2756 sizeof(pr->name));
2757 break;
2758 }
2759 }
2760 if (!pr->name[0])
2761 error = EBUSY;
2762 PF_RULES_RUNLOCK();
2763
2764 return (error);
2765 }
2766
2767 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2768 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2769 {
2770 int error = 0;
2771 PF_RULES_RLOCK_TRACKER;
2772
2773 #define ERROUT_IOCTL(target, x) \
2774 do { \
2775 error = (x); \
2776 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
2777 goto target; \
2778 } while (0)
2779
2780
2781 /* XXX keep in sync with switch() below */
2782 if (securelevel_gt(td->td_ucred, 2))
2783 switch (cmd) {
2784 case DIOCGETRULES:
2785 case DIOCGETRULENV:
2786 case DIOCGETADDRS:
2787 case DIOCGETADDR:
2788 case DIOCGETSTATE:
2789 case DIOCGETSTATENV:
2790 case DIOCSETSTATUSIF:
2791 case DIOCGETSTATUSNV:
2792 case DIOCCLRSTATUS:
2793 case DIOCNATLOOK:
2794 case DIOCSETDEBUG:
2795 #ifdef COMPAT_FREEBSD14
2796 case DIOCGETSTATES:
2797 case DIOCGETSTATESV2:
2798 #endif
2799 case DIOCGETTIMEOUT:
2800 case DIOCCLRRULECTRS:
2801 case DIOCGETLIMIT:
2802 case DIOCGETALTQSV0:
2803 case DIOCGETALTQSV1:
2804 case DIOCGETALTQV0:
2805 case DIOCGETALTQV1:
2806 case DIOCGETQSTATSV0:
2807 case DIOCGETQSTATSV1:
2808 case DIOCGETRULESETS:
2809 case DIOCGETRULESET:
2810 case DIOCRGETTABLES:
2811 case DIOCRGETTSTATS:
2812 case DIOCRCLRTSTATS:
2813 case DIOCRCLRADDRS:
2814 case DIOCRADDADDRS:
2815 case DIOCRDELADDRS:
2816 case DIOCRSETADDRS:
2817 case DIOCRGETADDRS:
2818 case DIOCRGETASTATS:
2819 case DIOCRCLRASTATS:
2820 case DIOCRTSTADDRS:
2821 case DIOCOSFPGET:
2822 case DIOCGETSRCNODES:
2823 case DIOCCLRSRCNODES:
2824 case DIOCGETSYNCOOKIES:
2825 case DIOCIGETIFACES:
2826 case DIOCGIFSPEEDV0:
2827 case DIOCGIFSPEEDV1:
2828 case DIOCSETIFFLAG:
2829 case DIOCCLRIFFLAG:
2830 case DIOCGETETHRULES:
2831 case DIOCGETETHRULE:
2832 case DIOCGETETHRULESETS:
2833 case DIOCGETETHRULESET:
2834 break;
2835 case DIOCRCLRTABLES:
2836 case DIOCRADDTABLES:
2837 case DIOCRDELTABLES:
2838 case DIOCRSETTFLAGS:
2839 if (((struct pfioc_table *)addr)->pfrio_flags &
2840 PFR_FLAG_DUMMY)
2841 break; /* dummy operation ok */
2842 return (EPERM);
2843 default:
2844 return (EPERM);
2845 }
2846
2847 if (!(flags & FWRITE))
2848 switch (cmd) {
2849 case DIOCGETRULES:
2850 case DIOCGETADDRS:
2851 case DIOCGETADDR:
2852 case DIOCGETSTATE:
2853 case DIOCGETSTATENV:
2854 case DIOCGETSTATUSNV:
2855 #ifdef COMPAT_FREEBSD14
2856 case DIOCGETSTATES:
2857 case DIOCGETSTATESV2:
2858 #endif
2859 case DIOCGETTIMEOUT:
2860 case DIOCGETLIMIT:
2861 case DIOCGETALTQSV0:
2862 case DIOCGETALTQSV1:
2863 case DIOCGETALTQV0:
2864 case DIOCGETALTQV1:
2865 case DIOCGETQSTATSV0:
2866 case DIOCGETQSTATSV1:
2867 case DIOCGETRULESETS:
2868 case DIOCGETRULESET:
2869 case DIOCNATLOOK:
2870 case DIOCRGETTABLES:
2871 case DIOCRGETTSTATS:
2872 case DIOCRGETADDRS:
2873 case DIOCRGETASTATS:
2874 case DIOCRTSTADDRS:
2875 case DIOCOSFPGET:
2876 case DIOCGETSRCNODES:
2877 case DIOCGETSYNCOOKIES:
2878 case DIOCIGETIFACES:
2879 case DIOCGIFSPEEDV1:
2880 case DIOCGIFSPEEDV0:
2881 case DIOCGETRULENV:
2882 case DIOCGETETHRULES:
2883 case DIOCGETETHRULE:
2884 case DIOCGETETHRULESETS:
2885 case DIOCGETETHRULESET:
2886 break;
2887 case DIOCRCLRTABLES:
2888 case DIOCRADDTABLES:
2889 case DIOCRDELTABLES:
2890 case DIOCRCLRTSTATS:
2891 case DIOCRCLRADDRS:
2892 case DIOCRADDADDRS:
2893 case DIOCRDELADDRS:
2894 case DIOCRSETADDRS:
2895 case DIOCRSETTFLAGS:
2896 if (((struct pfioc_table *)addr)->pfrio_flags &
2897 PFR_FLAG_DUMMY) {
2898 flags |= FWRITE; /* need write lock for dummy */
2899 break; /* dummy operation ok */
2900 }
2901 return (EACCES);
2902 default:
2903 return (EACCES);
2904 }
2905
2906 CURVNET_SET(TD_TO_VNET(td));
2907
2908 switch (cmd) {
2909 #ifdef COMPAT_FREEBSD14
2910 case DIOCSTART:
2911 error = pf_start();
2912 break;
2913
2914 case DIOCSTOP:
2915 error = pf_stop();
2916 break;
2917 #endif
2918
2919 case DIOCGETETHRULES: {
2920 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2921 nvlist_t *nvl;
2922 void *packed;
2923 struct pf_keth_rule *tail;
2924 struct pf_keth_ruleset *rs;
2925 u_int32_t ticket, nr;
2926 const char *anchor = "";
2927
2928 nvl = NULL;
2929 packed = NULL;
2930
2931 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2932
2933 if (nv->len > pf_ioctl_maxcount)
2934 ERROUT(ENOMEM);
2935
2936 /* Copy the request in */
2937 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2938 error = copyin(nv->data, packed, nv->len);
2939 if (error)
2940 ERROUT(error);
2941
2942 nvl = nvlist_unpack(packed, nv->len, 0);
2943 if (nvl == NULL)
2944 ERROUT(EBADMSG);
2945
2946 if (! nvlist_exists_string(nvl, "anchor"))
2947 ERROUT(EBADMSG);
2948
2949 anchor = nvlist_get_string(nvl, "anchor");
2950
2951 rs = pf_find_keth_ruleset(anchor);
2952
2953 nvlist_destroy(nvl);
2954 nvl = NULL;
2955 free(packed, M_NVLIST);
2956 packed = NULL;
2957
2958 if (rs == NULL)
2959 ERROUT(ENOENT);
2960
2961 /* Reply */
2962 nvl = nvlist_create(0);
2963 if (nvl == NULL)
2964 ERROUT(ENOMEM);
2965
2966 PF_RULES_RLOCK();
2967
2968 ticket = rs->active.ticket;
2969 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2970 if (tail)
2971 nr = tail->nr + 1;
2972 else
2973 nr = 0;
2974
2975 PF_RULES_RUNLOCK();
2976
2977 nvlist_add_number(nvl, "ticket", ticket);
2978 nvlist_add_number(nvl, "nr", nr);
2979
2980 packed = nvlist_pack(nvl, &nv->len);
2981 if (packed == NULL)
2982 ERROUT(ENOMEM);
2983
2984 if (nv->size == 0)
2985 ERROUT(0);
2986 else if (nv->size < nv->len)
2987 ERROUT(ENOSPC);
2988
2989 error = copyout(packed, nv->data, nv->len);
2990
2991 #undef ERROUT
2992 DIOCGETETHRULES_error:
2993 free(packed, M_NVLIST);
2994 nvlist_destroy(nvl);
2995 break;
2996 }
2997
2998 case DIOCGETETHRULE: {
2999 struct epoch_tracker et;
3000 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3001 nvlist_t *nvl = NULL;
3002 void *nvlpacked = NULL;
3003 struct pf_keth_rule *rule = NULL;
3004 struct pf_keth_ruleset *rs;
3005 u_int32_t ticket, nr;
3006 bool clear = false;
3007 const char *anchor;
3008
3009 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
3010
3011 if (nv->len > pf_ioctl_maxcount)
3012 ERROUT(ENOMEM);
3013
3014 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3015 error = copyin(nv->data, nvlpacked, nv->len);
3016 if (error)
3017 ERROUT(error);
3018
3019 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3020 if (nvl == NULL)
3021 ERROUT(EBADMSG);
3022 if (! nvlist_exists_number(nvl, "ticket"))
3023 ERROUT(EBADMSG);
3024 ticket = nvlist_get_number(nvl, "ticket");
3025 if (! nvlist_exists_string(nvl, "anchor"))
3026 ERROUT(EBADMSG);
3027 anchor = nvlist_get_string(nvl, "anchor");
3028
3029 if (nvlist_exists_bool(nvl, "clear"))
3030 clear = nvlist_get_bool(nvl, "clear");
3031
3032 if (clear && !(flags & FWRITE))
3033 ERROUT(EACCES);
3034
3035 if (! nvlist_exists_number(nvl, "nr"))
3036 ERROUT(EBADMSG);
3037 nr = nvlist_get_number(nvl, "nr");
3038
3039 PF_RULES_RLOCK();
3040 rs = pf_find_keth_ruleset(anchor);
3041 if (rs == NULL) {
3042 PF_RULES_RUNLOCK();
3043 ERROUT(ENOENT);
3044 }
3045 if (ticket != rs->active.ticket) {
3046 PF_RULES_RUNLOCK();
3047 ERROUT(EBUSY);
3048 }
3049
3050 nvlist_destroy(nvl);
3051 nvl = NULL;
3052 free(nvlpacked, M_NVLIST);
3053 nvlpacked = NULL;
3054
3055 rule = TAILQ_FIRST(rs->active.rules);
3056 while ((rule != NULL) && (rule->nr != nr))
3057 rule = TAILQ_NEXT(rule, entries);
3058 if (rule == NULL) {
3059 PF_RULES_RUNLOCK();
3060 ERROUT(ENOENT);
3061 }
3062 /* Make sure rule can't go away. */
3063 NET_EPOCH_ENTER(et);
3064 PF_RULES_RUNLOCK();
3065 nvl = pf_keth_rule_to_nveth_rule(rule);
3066 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
3067 NET_EPOCH_EXIT(et);
3068 ERROUT(EBUSY);
3069 }
3070 NET_EPOCH_EXIT(et);
3071 if (nvl == NULL)
3072 ERROUT(ENOMEM);
3073
3074 nvlpacked = nvlist_pack(nvl, &nv->len);
3075 if (nvlpacked == NULL)
3076 ERROUT(ENOMEM);
3077
3078 if (nv->size == 0)
3079 ERROUT(0);
3080 else if (nv->size < nv->len)
3081 ERROUT(ENOSPC);
3082
3083 error = copyout(nvlpacked, nv->data, nv->len);
3084 if (error == 0 && clear) {
3085 counter_u64_zero(rule->evaluations);
3086 for (int i = 0; i < 2; i++) {
3087 counter_u64_zero(rule->packets[i]);
3088 counter_u64_zero(rule->bytes[i]);
3089 }
3090 }
3091
3092 #undef ERROUT
3093 DIOCGETETHRULE_error:
3094 free(nvlpacked, M_NVLIST);
3095 nvlist_destroy(nvl);
3096 break;
3097 }
3098
3099 case DIOCADDETHRULE: {
3100 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3101 nvlist_t *nvl = NULL;
3102 void *nvlpacked = NULL;
3103 struct pf_keth_rule *rule = NULL, *tail = NULL;
3104 struct pf_keth_ruleset *ruleset = NULL;
3105 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
3106 const char *anchor = "", *anchor_call = "";
3107
3108 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3109
3110 if (nv->len > pf_ioctl_maxcount)
3111 ERROUT(ENOMEM);
3112
3113 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3114 error = copyin(nv->data, nvlpacked, nv->len);
3115 if (error)
3116 ERROUT(error);
3117
3118 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3119 if (nvl == NULL)
3120 ERROUT(EBADMSG);
3121
3122 if (! nvlist_exists_number(nvl, "ticket"))
3123 ERROUT(EBADMSG);
3124
3125 if (nvlist_exists_string(nvl, "anchor"))
3126 anchor = nvlist_get_string(nvl, "anchor");
3127 if (nvlist_exists_string(nvl, "anchor_call"))
3128 anchor_call = nvlist_get_string(nvl, "anchor_call");
3129
3130 ruleset = pf_find_keth_ruleset(anchor);
3131 if (ruleset == NULL)
3132 ERROUT(EINVAL);
3133
3134 if (nvlist_get_number(nvl, "ticket") !=
3135 ruleset->inactive.ticket) {
3136 DPFPRINTF(PF_DEBUG_MISC,
3137 ("ticket: %d != %d\n",
3138 (u_int32_t)nvlist_get_number(nvl, "ticket"),
3139 ruleset->inactive.ticket));
3140 ERROUT(EBUSY);
3141 }
3142
3143 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3144 rule->timestamp = NULL;
3145
3146 error = pf_nveth_rule_to_keth_rule(nvl, rule);
3147 if (error != 0)
3148 ERROUT(error);
3149
3150 if (rule->ifname[0])
3151 kif = pf_kkif_create(M_WAITOK);
3152 if (rule->bridge_to_name[0])
3153 bridge_to_kif = pf_kkif_create(M_WAITOK);
3154 rule->evaluations = counter_u64_alloc(M_WAITOK);
3155 for (int i = 0; i < 2; i++) {
3156 rule->packets[i] = counter_u64_alloc(M_WAITOK);
3157 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3158 }
3159 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3160 M_WAITOK | M_ZERO);
3161
3162 PF_RULES_WLOCK();
3163
3164 if (rule->ifname[0]) {
3165 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3166 pfi_kkif_ref(rule->kif);
3167 } else
3168 rule->kif = NULL;
3169 if (rule->bridge_to_name[0]) {
3170 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3171 rule->bridge_to_name);
3172 pfi_kkif_ref(rule->bridge_to);
3173 } else
3174 rule->bridge_to = NULL;
3175
3176 #ifdef ALTQ
3177 /* set queue IDs */
3178 if (rule->qname[0] != 0) {
3179 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3180 error = EBUSY;
3181 else
3182 rule->qid = rule->qid;
3183 }
3184 #endif
3185 if (rule->tagname[0])
3186 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3187 error = EBUSY;
3188 if (rule->match_tagname[0])
3189 if ((rule->match_tag = pf_tagname2tag(
3190 rule->match_tagname)) == 0)
3191 error = EBUSY;
3192
3193 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3194 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3195 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3196 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3197
3198 if (error) {
3199 pf_free_eth_rule(rule);
3200 PF_RULES_WUNLOCK();
3201 ERROUT(error);
3202 }
3203
3204 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3205 pf_free_eth_rule(rule);
3206 PF_RULES_WUNLOCK();
3207 ERROUT(EINVAL);
3208 }
3209
3210 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3211 if (tail)
3212 rule->nr = tail->nr + 1;
3213 else
3214 rule->nr = 0;
3215
3216 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3217
3218 PF_RULES_WUNLOCK();
3219
3220 #undef ERROUT
3221 DIOCADDETHRULE_error:
3222 nvlist_destroy(nvl);
3223 free(nvlpacked, M_NVLIST);
3224 break;
3225 }
3226
3227 case DIOCGETETHRULESETS: {
3228 struct epoch_tracker et;
3229 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3230 nvlist_t *nvl = NULL;
3231 void *nvlpacked = NULL;
3232 struct pf_keth_ruleset *ruleset;
3233 struct pf_keth_anchor *anchor;
3234 int nr = 0;
3235
3236 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3237
3238 if (nv->len > pf_ioctl_maxcount)
3239 ERROUT(ENOMEM);
3240
3241 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3242 error = copyin(nv->data, nvlpacked, nv->len);
3243 if (error)
3244 ERROUT(error);
3245
3246 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3247 if (nvl == NULL)
3248 ERROUT(EBADMSG);
3249 if (! nvlist_exists_string(nvl, "path"))
3250 ERROUT(EBADMSG);
3251
3252 NET_EPOCH_ENTER(et);
3253
3254 if ((ruleset = pf_find_keth_ruleset(
3255 nvlist_get_string(nvl, "path"))) == NULL) {
3256 NET_EPOCH_EXIT(et);
3257 ERROUT(ENOENT);
3258 }
3259
3260 if (ruleset->anchor == NULL) {
3261 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3262 if (anchor->parent == NULL)
3263 nr++;
3264 } else {
3265 RB_FOREACH(anchor, pf_keth_anchor_node,
3266 &ruleset->anchor->children)
3267 nr++;
3268 }
3269
3270 NET_EPOCH_EXIT(et);
3271
3272 nvlist_destroy(nvl);
3273 nvl = NULL;
3274 free(nvlpacked, M_NVLIST);
3275 nvlpacked = NULL;
3276
3277 nvl = nvlist_create(0);
3278 if (nvl == NULL)
3279 ERROUT(ENOMEM);
3280
3281 nvlist_add_number(nvl, "nr", nr);
3282
3283 nvlpacked = nvlist_pack(nvl, &nv->len);
3284 if (nvlpacked == NULL)
3285 ERROUT(ENOMEM);
3286
3287 if (nv->size == 0)
3288 ERROUT(0);
3289 else if (nv->size < nv->len)
3290 ERROUT(ENOSPC);
3291
3292 error = copyout(nvlpacked, nv->data, nv->len);
3293
3294 #undef ERROUT
3295 DIOCGETETHRULESETS_error:
3296 free(nvlpacked, M_NVLIST);
3297 nvlist_destroy(nvl);
3298 break;
3299 }
3300
3301 case DIOCGETETHRULESET: {
3302 struct epoch_tracker et;
3303 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3304 nvlist_t *nvl = NULL;
3305 void *nvlpacked = NULL;
3306 struct pf_keth_ruleset *ruleset;
3307 struct pf_keth_anchor *anchor;
3308 int nr = 0, req_nr = 0;
3309 bool found = false;
3310
3311 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3312
3313 if (nv->len > pf_ioctl_maxcount)
3314 ERROUT(ENOMEM);
3315
3316 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3317 error = copyin(nv->data, nvlpacked, nv->len);
3318 if (error)
3319 ERROUT(error);
3320
3321 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3322 if (nvl == NULL)
3323 ERROUT(EBADMSG);
3324 if (! nvlist_exists_string(nvl, "path"))
3325 ERROUT(EBADMSG);
3326 if (! nvlist_exists_number(nvl, "nr"))
3327 ERROUT(EBADMSG);
3328
3329 req_nr = nvlist_get_number(nvl, "nr");
3330
3331 NET_EPOCH_ENTER(et);
3332
3333 if ((ruleset = pf_find_keth_ruleset(
3334 nvlist_get_string(nvl, "path"))) == NULL) {
3335 NET_EPOCH_EXIT(et);
3336 ERROUT(ENOENT);
3337 }
3338
3339 nvlist_destroy(nvl);
3340 nvl = NULL;
3341 free(nvlpacked, M_NVLIST);
3342 nvlpacked = NULL;
3343
3344 nvl = nvlist_create(0);
3345 if (nvl == NULL) {
3346 NET_EPOCH_EXIT(et);
3347 ERROUT(ENOMEM);
3348 }
3349
3350 if (ruleset->anchor == NULL) {
3351 RB_FOREACH(anchor, pf_keth_anchor_global,
3352 &V_pf_keth_anchors) {
3353 if (anchor->parent == NULL && nr++ == req_nr) {
3354 found = true;
3355 break;
3356 }
3357 }
3358 } else {
3359 RB_FOREACH(anchor, pf_keth_anchor_node,
3360 &ruleset->anchor->children) {
3361 if (nr++ == req_nr) {
3362 found = true;
3363 break;
3364 }
3365 }
3366 }
3367
3368 NET_EPOCH_EXIT(et);
3369 if (found) {
3370 nvlist_add_number(nvl, "nr", nr);
3371 nvlist_add_string(nvl, "name", anchor->name);
3372 if (ruleset->anchor)
3373 nvlist_add_string(nvl, "path",
3374 ruleset->anchor->path);
3375 else
3376 nvlist_add_string(nvl, "path", "");
3377 } else {
3378 ERROUT(EBUSY);
3379 }
3380
3381 nvlpacked = nvlist_pack(nvl, &nv->len);
3382 if (nvlpacked == NULL)
3383 ERROUT(ENOMEM);
3384
3385 if (nv->size == 0)
3386 ERROUT(0);
3387 else if (nv->size < nv->len)
3388 ERROUT(ENOSPC);
3389
3390 error = copyout(nvlpacked, nv->data, nv->len);
3391
3392 #undef ERROUT
3393 DIOCGETETHRULESET_error:
3394 free(nvlpacked, M_NVLIST);
3395 nvlist_destroy(nvl);
3396 break;
3397 }
3398
3399 case DIOCADDRULENV: {
3400 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3401 nvlist_t *nvl = NULL;
3402 void *nvlpacked = NULL;
3403 struct pf_krule *rule = NULL;
3404 const char *anchor = "", *anchor_call = "";
3405 uint32_t ticket = 0, pool_ticket = 0;
3406
3407 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
3408
3409 if (nv->len > pf_ioctl_maxcount)
3410 ERROUT(ENOMEM);
3411
3412 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3413 error = copyin(nv->data, nvlpacked, nv->len);
3414 if (error)
3415 ERROUT(error);
3416
3417 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3418 if (nvl == NULL)
3419 ERROUT(EBADMSG);
3420
3421 if (! nvlist_exists_number(nvl, "ticket"))
3422 ERROUT(EINVAL);
3423 ticket = nvlist_get_number(nvl, "ticket");
3424
3425 if (! nvlist_exists_number(nvl, "pool_ticket"))
3426 ERROUT(EINVAL);
3427 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3428
3429 if (! nvlist_exists_nvlist(nvl, "rule"))
3430 ERROUT(EINVAL);
3431
3432 rule = pf_krule_alloc();
3433 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3434 rule);
3435 if (error)
3436 ERROUT(error);
3437
3438 if (nvlist_exists_string(nvl, "anchor"))
3439 anchor = nvlist_get_string(nvl, "anchor");
3440 if (nvlist_exists_string(nvl, "anchor_call"))
3441 anchor_call = nvlist_get_string(nvl, "anchor_call");
3442
3443 if ((error = nvlist_error(nvl)))
3444 ERROUT(error);
3445
3446 /* Frees rule on error */
3447 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3448 anchor_call, td->td_ucred->cr_ruid,
3449 td->td_proc ? td->td_proc->p_pid : 0);
3450
3451 nvlist_destroy(nvl);
3452 free(nvlpacked, M_NVLIST);
3453 break;
3454 #undef ERROUT
3455 DIOCADDRULENV_error:
3456 pf_krule_free(rule);
3457 nvlist_destroy(nvl);
3458 free(nvlpacked, M_NVLIST);
3459
3460 break;
3461 }
3462 case DIOCADDRULE: {
3463 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3464 struct pf_krule *rule;
3465
3466 rule = pf_krule_alloc();
3467 error = pf_rule_to_krule(&pr->rule, rule);
3468 if (error != 0) {
3469 pf_krule_free(rule);
3470 break;
3471 }
3472
3473 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3474
3475 /* Frees rule on error */
3476 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3477 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3478 td->td_proc ? td->td_proc->p_pid : 0);
3479 break;
3480 }
3481
3482 case DIOCGETRULES: {
3483 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3484
3485 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3486
3487 error = pf_ioctl_getrules(pr);
3488
3489 break;
3490 }
3491
3492 case DIOCGETRULENV: {
3493 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3494 nvlist_t *nvrule = NULL;
3495 nvlist_t *nvl = NULL;
3496 struct pf_kruleset *ruleset;
3497 struct pf_krule *rule;
3498 void *nvlpacked = NULL;
3499 int rs_num, nr;
3500 bool clear_counter = false;
3501
3502 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
3503
3504 if (nv->len > pf_ioctl_maxcount)
3505 ERROUT(ENOMEM);
3506
3507 /* Copy the request in */
3508 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3509 error = copyin(nv->data, nvlpacked, nv->len);
3510 if (error)
3511 ERROUT(error);
3512
3513 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3514 if (nvl == NULL)
3515 ERROUT(EBADMSG);
3516
3517 if (! nvlist_exists_string(nvl, "anchor"))
3518 ERROUT(EBADMSG);
3519 if (! nvlist_exists_number(nvl, "ruleset"))
3520 ERROUT(EBADMSG);
3521 if (! nvlist_exists_number(nvl, "ticket"))
3522 ERROUT(EBADMSG);
3523 if (! nvlist_exists_number(nvl, "nr"))
3524 ERROUT(EBADMSG);
3525
3526 if (nvlist_exists_bool(nvl, "clear_counter"))
3527 clear_counter = nvlist_get_bool(nvl, "clear_counter");
3528
3529 if (clear_counter && !(flags & FWRITE))
3530 ERROUT(EACCES);
3531
3532 nr = nvlist_get_number(nvl, "nr");
3533
3534 PF_RULES_WLOCK();
3535 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3536 if (ruleset == NULL) {
3537 PF_RULES_WUNLOCK();
3538 ERROUT(ENOENT);
3539 }
3540
3541 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3542 if (rs_num >= PF_RULESET_MAX) {
3543 PF_RULES_WUNLOCK();
3544 ERROUT(EINVAL);
3545 }
3546
3547 if (nvlist_get_number(nvl, "ticket") !=
3548 ruleset->rules[rs_num].active.ticket) {
3549 PF_RULES_WUNLOCK();
3550 ERROUT(EBUSY);
3551 }
3552
3553 if ((error = nvlist_error(nvl))) {
3554 PF_RULES_WUNLOCK();
3555 ERROUT(error);
3556 }
3557
3558 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3559 while ((rule != NULL) && (rule->nr != nr))
3560 rule = TAILQ_NEXT(rule, entries);
3561 if (rule == NULL) {
3562 PF_RULES_WUNLOCK();
3563 ERROUT(EBUSY);
3564 }
3565
3566 nvrule = pf_krule_to_nvrule(rule);
3567
3568 nvlist_destroy(nvl);
3569 nvl = nvlist_create(0);
3570 if (nvl == NULL) {
3571 PF_RULES_WUNLOCK();
3572 ERROUT(ENOMEM);
3573 }
3574 nvlist_add_number(nvl, "nr", nr);
3575 nvlist_add_nvlist(nvl, "rule", nvrule);
3576 nvlist_destroy(nvrule);
3577 nvrule = NULL;
3578 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3579 PF_RULES_WUNLOCK();
3580 ERROUT(EBUSY);
3581 }
3582
3583 free(nvlpacked, M_NVLIST);
3584 nvlpacked = nvlist_pack(nvl, &nv->len);
3585 if (nvlpacked == NULL) {
3586 PF_RULES_WUNLOCK();
3587 ERROUT(ENOMEM);
3588 }
3589
3590 if (nv->size == 0) {
3591 PF_RULES_WUNLOCK();
3592 ERROUT(0);
3593 }
3594 else if (nv->size < nv->len) {
3595 PF_RULES_WUNLOCK();
3596 ERROUT(ENOSPC);
3597 }
3598
3599 if (clear_counter)
3600 pf_krule_clear_counters(rule);
3601
3602 PF_RULES_WUNLOCK();
3603
3604 error = copyout(nvlpacked, nv->data, nv->len);
3605
3606 #undef ERROUT
3607 DIOCGETRULENV_error:
3608 free(nvlpacked, M_NVLIST);
3609 nvlist_destroy(nvrule);
3610 nvlist_destroy(nvl);
3611
3612 break;
3613 }
3614
3615 case DIOCCHANGERULE: {
3616 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
3617 struct pf_kruleset *ruleset;
3618 struct pf_krule *oldrule = NULL, *newrule = NULL;
3619 struct pfi_kkif *kif = NULL;
3620 struct pf_kpooladdr *pa;
3621 u_int32_t nr = 0;
3622 int rs_num;
3623
3624 pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3625
3626 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3627 pcr->action > PF_CHANGE_GET_TICKET) {
3628 error = EINVAL;
3629 break;
3630 }
3631 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3632 error = EINVAL;
3633 break;
3634 }
3635
3636 if (pcr->action != PF_CHANGE_REMOVE) {
3637 newrule = pf_krule_alloc();
3638 error = pf_rule_to_krule(&pcr->rule, newrule);
3639 if (error != 0) {
3640 pf_krule_free(newrule);
3641 break;
3642 }
3643
3644 if (newrule->ifname[0])
3645 kif = pf_kkif_create(M_WAITOK);
3646 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3647 for (int i = 0; i < 2; i++) {
3648 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3649 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3650 }
3651 newrule->states_cur = counter_u64_alloc(M_WAITOK);
3652 newrule->states_tot = counter_u64_alloc(M_WAITOK);
3653 newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3654 newrule->cuid = td->td_ucred->cr_ruid;
3655 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3656 TAILQ_INIT(&newrule->nat.list);
3657 TAILQ_INIT(&newrule->rdr.list);
3658 TAILQ_INIT(&newrule->route.list);
3659 }
3660 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3661
3662 PF_CONFIG_LOCK();
3663 PF_RULES_WLOCK();
3664 #ifdef PF_WANT_32_TO_64_COUNTER
3665 if (newrule != NULL) {
3666 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3667 newrule->allrulelinked = true;
3668 V_pf_allrulecount++;
3669 }
3670 #endif
3671
3672 if (!(pcr->action == PF_CHANGE_REMOVE ||
3673 pcr->action == PF_CHANGE_GET_TICKET) &&
3674 pcr->pool_ticket != V_ticket_pabuf)
3675 ERROUT(EBUSY);
3676
3677 ruleset = pf_find_kruleset(pcr->anchor);
3678 if (ruleset == NULL)
3679 ERROUT(EINVAL);
3680
3681 rs_num = pf_get_ruleset_number(pcr->rule.action);
3682 if (rs_num >= PF_RULESET_MAX)
3683 ERROUT(EINVAL);
3684
3685 /*
3686 * XXXMJG: there is no guarantee that the ruleset was
3687 * created by the usual route of calling DIOCXBEGIN.
3688 * As a result it is possible the rule tree will not
3689 * be allocated yet. Hack around it by doing it here.
3690 * Note it is fine to let the tree persist in case of
3691 * error as it will be freed down the road on future
3692 * updates (if need be).
3693 */
3694 if (ruleset->rules[rs_num].active.tree == NULL) {
3695 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3696 if (ruleset->rules[rs_num].active.tree == NULL) {
3697 ERROUT(ENOMEM);
3698 }
3699 }
3700
3701 if (pcr->action == PF_CHANGE_GET_TICKET) {
3702 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3703 ERROUT(0);
3704 } else if (pcr->ticket !=
3705 ruleset->rules[rs_num].active.ticket)
3706 ERROUT(EINVAL);
3707
3708 if (pcr->action != PF_CHANGE_REMOVE) {
3709 if (newrule->ifname[0]) {
3710 newrule->kif = pfi_kkif_attach(kif,
3711 newrule->ifname);
3712 kif = NULL;
3713 pfi_kkif_ref(newrule->kif);
3714 } else
3715 newrule->kif = NULL;
3716
3717 if (newrule->rtableid > 0 &&
3718 newrule->rtableid >= rt_numfibs)
3719 error = EBUSY;
3720
3721 #ifdef ALTQ
3722 /* set queue IDs */
3723 if (newrule->qname[0] != 0) {
3724 if ((newrule->qid =
3725 pf_qname2qid(newrule->qname)) == 0)
3726 error = EBUSY;
3727 else if (newrule->pqname[0] != 0) {
3728 if ((newrule->pqid =
3729 pf_qname2qid(newrule->pqname)) == 0)
3730 error = EBUSY;
3731 } else
3732 newrule->pqid = newrule->qid;
3733 }
3734 #endif /* ALTQ */
3735 if (newrule->tagname[0])
3736 if ((newrule->tag =
3737 pf_tagname2tag(newrule->tagname)) == 0)
3738 error = EBUSY;
3739 if (newrule->match_tagname[0])
3740 if ((newrule->match_tag = pf_tagname2tag(
3741 newrule->match_tagname)) == 0)
3742 error = EBUSY;
3743 if (newrule->rt && !newrule->direction)
3744 error = EINVAL;
3745 if (!newrule->log)
3746 newrule->logif = 0;
3747 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3748 error = ENOMEM;
3749 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3750 error = ENOMEM;
3751 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3752 error = EINVAL;
3753 for (int i = 0; i < 3; i++) {
3754 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
3755 if (pa->addr.type == PF_ADDR_TABLE) {
3756 pa->addr.p.tbl =
3757 pfr_attach_table(ruleset,
3758 pa->addr.v.tblname);
3759 if (pa->addr.p.tbl == NULL)
3760 error = ENOMEM;
3761 }
3762 }
3763
3764 newrule->overload_tbl = NULL;
3765 if (newrule->overload_tblname[0]) {
3766 if ((newrule->overload_tbl = pfr_attach_table(
3767 ruleset, newrule->overload_tblname)) ==
3768 NULL)
3769 error = EINVAL;
3770 else
3771 newrule->overload_tbl->pfrkt_flags |=
3772 PFR_TFLAG_ACTIVE;
3773 }
3774
3775 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
3776 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
3777 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list);
3778 if (((((newrule->action == PF_NAT) ||
3779 (newrule->action == PF_RDR) ||
3780 (newrule->action == PF_BINAT) ||
3781 (newrule->rt > PF_NOPFROUTE)) &&
3782 !newrule->anchor)) &&
3783 (TAILQ_FIRST(&newrule->rdr.list) == NULL))
3784 error = EINVAL;
3785
3786 if (error) {
3787 pf_free_rule(newrule);
3788 PF_RULES_WUNLOCK();
3789 PF_CONFIG_UNLOCK();
3790 break;
3791 }
3792
3793 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
3794 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
3795 }
3796 pf_empty_kpool(&V_pf_pabuf[0]);
3797 pf_empty_kpool(&V_pf_pabuf[1]);
3798 pf_empty_kpool(&V_pf_pabuf[2]);
3799
3800 if (pcr->action == PF_CHANGE_ADD_HEAD)
3801 oldrule = TAILQ_FIRST(
3802 ruleset->rules[rs_num].active.ptr);
3803 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3804 oldrule = TAILQ_LAST(
3805 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3806 else {
3807 oldrule = TAILQ_FIRST(
3808 ruleset->rules[rs_num].active.ptr);
3809 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3810 oldrule = TAILQ_NEXT(oldrule, entries);
3811 if (oldrule == NULL) {
3812 if (newrule != NULL)
3813 pf_free_rule(newrule);
3814 PF_RULES_WUNLOCK();
3815 PF_CONFIG_UNLOCK();
3816 error = EINVAL;
3817 break;
3818 }
3819 }
3820
3821 if (pcr->action == PF_CHANGE_REMOVE) {
3822 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3823 oldrule);
3824 RB_REMOVE(pf_krule_global,
3825 ruleset->rules[rs_num].active.tree, oldrule);
3826 ruleset->rules[rs_num].active.rcount--;
3827 } else {
3828 pf_hash_rule(newrule);
3829 if (RB_INSERT(pf_krule_global,
3830 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3831 pf_free_rule(newrule);
3832 PF_RULES_WUNLOCK();
3833 PF_CONFIG_UNLOCK();
3834 error = EEXIST;
3835 break;
3836 }
3837
3838 if (oldrule == NULL)
3839 TAILQ_INSERT_TAIL(
3840 ruleset->rules[rs_num].active.ptr,
3841 newrule, entries);
3842 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3843 pcr->action == PF_CHANGE_ADD_BEFORE)
3844 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3845 else
3846 TAILQ_INSERT_AFTER(
3847 ruleset->rules[rs_num].active.ptr,
3848 oldrule, newrule, entries);
3849 ruleset->rules[rs_num].active.rcount++;
3850 }
3851
3852 nr = 0;
3853 TAILQ_FOREACH(oldrule,
3854 ruleset->rules[rs_num].active.ptr, entries)
3855 oldrule->nr = nr++;
3856
3857 ruleset->rules[rs_num].active.ticket++;
3858
3859 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3860 pf_remove_if_empty_kruleset(ruleset);
3861
3862 PF_RULES_WUNLOCK();
3863 PF_CONFIG_UNLOCK();
3864 break;
3865
3866 #undef ERROUT
3867 DIOCCHANGERULE_error:
3868 PF_RULES_WUNLOCK();
3869 PF_CONFIG_UNLOCK();
3870 pf_krule_free(newrule);
3871 pf_kkif_free(kif);
3872 break;
3873 }
3874
3875 case DIOCCLRSTATESNV: {
3876 error = pf_clearstates_nv((struct pfioc_nv *)addr);
3877 break;
3878 }
3879
3880 case DIOCKILLSTATESNV: {
3881 error = pf_killstates_nv((struct pfioc_nv *)addr);
3882 break;
3883 }
3884
3885 case DIOCADDSTATE: {
3886 struct pfioc_state *ps = (struct pfioc_state *)addr;
3887 struct pfsync_state_1301 *sp = &ps->state;
3888
3889 if (sp->timeout >= PFTM_MAX) {
3890 error = EINVAL;
3891 break;
3892 }
3893 if (V_pfsync_state_import_ptr != NULL) {
3894 PF_RULES_RLOCK();
3895 error = V_pfsync_state_import_ptr(
3896 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3897 PFSYNC_MSG_VERSION_1301);
3898 PF_RULES_RUNLOCK();
3899 } else
3900 error = EOPNOTSUPP;
3901 break;
3902 }
3903
3904 case DIOCGETSTATE: {
3905 struct pfioc_state *ps = (struct pfioc_state *)addr;
3906 struct pf_kstate *s;
3907
3908 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3909 if (s == NULL) {
3910 error = ENOENT;
3911 break;
3912 }
3913
3914 pfsync_state_export((union pfsync_state_union*)&ps->state,
3915 s, PFSYNC_MSG_VERSION_1301);
3916 PF_STATE_UNLOCK(s);
3917 break;
3918 }
3919
3920 case DIOCGETSTATENV: {
3921 error = pf_getstate((struct pfioc_nv *)addr);
3922 break;
3923 }
3924
3925 #ifdef COMPAT_FREEBSD14
3926 case DIOCGETSTATES: {
3927 struct pfioc_states *ps = (struct pfioc_states *)addr;
3928 struct pf_kstate *s;
3929 struct pfsync_state_1301 *pstore, *p;
3930 int i, nr;
3931 size_t slice_count = 16, count;
3932 void *out;
3933
3934 if (ps->ps_len <= 0) {
3935 nr = uma_zone_get_cur(V_pf_state_z);
3936 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3937 break;
3938 }
3939
3940 out = ps->ps_states;
3941 pstore = mallocarray(slice_count,
3942 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3943 nr = 0;
3944
3945 for (i = 0; i <= V_pf_hashmask; i++) {
3946 struct pf_idhash *ih = &V_pf_idhash[i];
3947
3948 DIOCGETSTATES_retry:
3949 p = pstore;
3950
3951 if (LIST_EMPTY(&ih->states))
3952 continue;
3953
3954 PF_HASHROW_LOCK(ih);
3955 count = 0;
3956 LIST_FOREACH(s, &ih->states, entry) {
3957 if (s->timeout == PFTM_UNLINKED)
3958 continue;
3959 count++;
3960 }
3961
3962 if (count > slice_count) {
3963 PF_HASHROW_UNLOCK(ih);
3964 free(pstore, M_TEMP);
3965 slice_count = count * 2;
3966 pstore = mallocarray(slice_count,
3967 sizeof(struct pfsync_state_1301), M_TEMP,
3968 M_WAITOK | M_ZERO);
3969 goto DIOCGETSTATES_retry;
3970 }
3971
3972 if ((nr+count) * sizeof(*p) > ps->ps_len) {
3973 PF_HASHROW_UNLOCK(ih);
3974 goto DIOCGETSTATES_full;
3975 }
3976
3977 LIST_FOREACH(s, &ih->states, entry) {
3978 if (s->timeout == PFTM_UNLINKED)
3979 continue;
3980
3981 pfsync_state_export((union pfsync_state_union*)p,
3982 s, PFSYNC_MSG_VERSION_1301);
3983 p++;
3984 nr++;
3985 }
3986 PF_HASHROW_UNLOCK(ih);
3987 error = copyout(pstore, out,
3988 sizeof(struct pfsync_state_1301) * count);
3989 if (error)
3990 break;
3991 out = ps->ps_states + nr;
3992 }
3993 DIOCGETSTATES_full:
3994 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3995 free(pstore, M_TEMP);
3996
3997 break;
3998 }
3999
4000 case DIOCGETSTATESV2: {
4001 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
4002 struct pf_kstate *s;
4003 struct pf_state_export *pstore, *p;
4004 int i, nr;
4005 size_t slice_count = 16, count;
4006 void *out;
4007
4008 if (ps->ps_req_version > PF_STATE_VERSION) {
4009 error = ENOTSUP;
4010 break;
4011 }
4012
4013 if (ps->ps_len <= 0) {
4014 nr = uma_zone_get_cur(V_pf_state_z);
4015 ps->ps_len = sizeof(struct pf_state_export) * nr;
4016 break;
4017 }
4018
4019 out = ps->ps_states;
4020 pstore = mallocarray(slice_count,
4021 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
4022 nr = 0;
4023
4024 for (i = 0; i <= V_pf_hashmask; i++) {
4025 struct pf_idhash *ih = &V_pf_idhash[i];
4026
4027 DIOCGETSTATESV2_retry:
4028 p = pstore;
4029
4030 if (LIST_EMPTY(&ih->states))
4031 continue;
4032
4033 PF_HASHROW_LOCK(ih);
4034 count = 0;
4035 LIST_FOREACH(s, &ih->states, entry) {
4036 if (s->timeout == PFTM_UNLINKED)
4037 continue;
4038 count++;
4039 }
4040
4041 if (count > slice_count) {
4042 PF_HASHROW_UNLOCK(ih);
4043 free(pstore, M_TEMP);
4044 slice_count = count * 2;
4045 pstore = mallocarray(slice_count,
4046 sizeof(struct pf_state_export), M_TEMP,
4047 M_WAITOK | M_ZERO);
4048 goto DIOCGETSTATESV2_retry;
4049 }
4050
4051 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4052 PF_HASHROW_UNLOCK(ih);
4053 goto DIOCGETSTATESV2_full;
4054 }
4055
4056 LIST_FOREACH(s, &ih->states, entry) {
4057 if (s->timeout == PFTM_UNLINKED)
4058 continue;
4059
4060 pf_state_export(p, s);
4061 p++;
4062 nr++;
4063 }
4064 PF_HASHROW_UNLOCK(ih);
4065 error = copyout(pstore, out,
4066 sizeof(struct pf_state_export) * count);
4067 if (error)
4068 break;
4069 out = ps->ps_states + nr;
4070 }
4071 DIOCGETSTATESV2_full:
4072 ps->ps_len = nr * sizeof(struct pf_state_export);
4073 free(pstore, M_TEMP);
4074
4075 break;
4076 }
4077 #endif
4078 case DIOCGETSTATUSNV: {
4079 error = pf_getstatus((struct pfioc_nv *)addr);
4080 break;
4081 }
4082
4083 case DIOCSETSTATUSIF: {
4084 struct pfioc_if *pi = (struct pfioc_if *)addr;
4085
4086 if (pi->ifname[0] == 0) {
4087 bzero(V_pf_status.ifname, IFNAMSIZ);
4088 break;
4089 }
4090 PF_RULES_WLOCK();
4091 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
4092 PF_RULES_WUNLOCK();
4093 break;
4094 }
4095
4096 case DIOCCLRSTATUS: {
4097 pf_ioctl_clear_status();
4098 break;
4099 }
4100
4101 case DIOCNATLOOK: {
4102 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
4103 struct pf_state_key *sk;
4104 struct pf_kstate *state;
4105 struct pf_state_key_cmp key;
4106 int m = 0, direction = pnl->direction;
4107 int sidx, didx;
4108
4109 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
4110 sidx = (direction == PF_IN) ? 1 : 0;
4111 didx = (direction == PF_IN) ? 0 : 1;
4112
4113 if (!pnl->proto ||
4114 PF_AZERO(&pnl->saddr, pnl->af) ||
4115 PF_AZERO(&pnl->daddr, pnl->af) ||
4116 ((pnl->proto == IPPROTO_TCP ||
4117 pnl->proto == IPPROTO_UDP) &&
4118 (!pnl->dport || !pnl->sport)))
4119 error = EINVAL;
4120 else {
4121 bzero(&key, sizeof(key));
4122 key.af = pnl->af;
4123 key.proto = pnl->proto;
4124 PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4125 key.port[sidx] = pnl->sport;
4126 PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4127 key.port[didx] = pnl->dport;
4128
4129 state = pf_find_state_all(&key, direction, &m);
4130 if (state == NULL) {
4131 error = ENOENT;
4132 } else {
4133 if (m > 1) {
4134 PF_STATE_UNLOCK(state);
4135 error = E2BIG; /* more than one state */
4136 } else {
4137 sk = state->key[sidx];
4138 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4139 pnl->rsport = sk->port[sidx];
4140 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4141 pnl->rdport = sk->port[didx];
4142 PF_STATE_UNLOCK(state);
4143 }
4144 }
4145 }
4146 break;
4147 }
4148
4149 case DIOCSETTIMEOUT: {
4150 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4151
4152 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4153 &pt->seconds);
4154 break;
4155 }
4156
4157 case DIOCGETTIMEOUT: {
4158 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4159
4160 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4161 break;
4162 }
4163
4164 case DIOCGETLIMIT: {
4165 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4166
4167 error = pf_ioctl_get_limit(pl->index, &pl->limit);
4168 break;
4169 }
4170
4171 case DIOCSETLIMIT: {
4172 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4173 unsigned int old_limit;
4174
4175 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4176 pl->limit = old_limit;
4177 break;
4178 }
4179
4180 case DIOCSETDEBUG: {
4181 u_int32_t *level = (u_int32_t *)addr;
4182
4183 PF_RULES_WLOCK();
4184 V_pf_status.debug = *level;
4185 PF_RULES_WUNLOCK();
4186 break;
4187 }
4188
4189 case DIOCCLRRULECTRS: {
4190 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4191 struct pf_kruleset *ruleset = &pf_main_ruleset;
4192 struct pf_krule *rule;
4193
4194 PF_RULES_WLOCK();
4195 TAILQ_FOREACH(rule,
4196 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4197 pf_counter_u64_zero(&rule->evaluations);
4198 for (int i = 0; i < 2; i++) {
4199 pf_counter_u64_zero(&rule->packets[i]);
4200 pf_counter_u64_zero(&rule->bytes[i]);
4201 }
4202 }
4203 PF_RULES_WUNLOCK();
4204 break;
4205 }
4206
4207 case DIOCGIFSPEEDV0:
4208 case DIOCGIFSPEEDV1: {
4209 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
4210 struct pf_ifspeed_v1 ps;
4211 struct ifnet *ifp;
4212
4213 if (psp->ifname[0] == '\0') {
4214 error = EINVAL;
4215 break;
4216 }
4217
4218 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4219 if (error != 0)
4220 break;
4221 ifp = ifunit(ps.ifname);
4222 if (ifp != NULL) {
4223 psp->baudrate32 =
4224 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4225 if (cmd == DIOCGIFSPEEDV1)
4226 psp->baudrate = ifp->if_baudrate;
4227 } else {
4228 error = EINVAL;
4229 }
4230 break;
4231 }
4232
4233 #ifdef ALTQ
4234 case DIOCSTARTALTQ: {
4235 struct pf_altq *altq;
4236
4237 PF_RULES_WLOCK();
4238 /* enable all altq interfaces on active list */
4239 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4240 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4241 error = pf_enable_altq(altq);
4242 if (error != 0)
4243 break;
4244 }
4245 }
4246 if (error == 0)
4247 V_pf_altq_running = 1;
4248 PF_RULES_WUNLOCK();
4249 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4250 break;
4251 }
4252
4253 case DIOCSTOPALTQ: {
4254 struct pf_altq *altq;
4255
4256 PF_RULES_WLOCK();
4257 /* disable all altq interfaces on active list */
4258 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4259 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4260 error = pf_disable_altq(altq);
4261 if (error != 0)
4262 break;
4263 }
4264 }
4265 if (error == 0)
4266 V_pf_altq_running = 0;
4267 PF_RULES_WUNLOCK();
4268 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4269 break;
4270 }
4271
4272 case DIOCADDALTQV0:
4273 case DIOCADDALTQV1: {
4274 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4275 struct pf_altq *altq, *a;
4276 struct ifnet *ifp;
4277
4278 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4279 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4280 if (error)
4281 break;
4282 altq->local_flags = 0;
4283
4284 PF_RULES_WLOCK();
4285 if (pa->ticket != V_ticket_altqs_inactive) {
4286 PF_RULES_WUNLOCK();
4287 free(altq, M_PFALTQ);
4288 error = EBUSY;
4289 break;
4290 }
4291
4292 /*
4293 * if this is for a queue, find the discipline and
4294 * copy the necessary fields
4295 */
4296 if (altq->qname[0] != 0) {
4297 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4298 PF_RULES_WUNLOCK();
4299 error = EBUSY;
4300 free(altq, M_PFALTQ);
4301 break;
4302 }
4303 altq->altq_disc = NULL;
4304 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4305 if (strncmp(a->ifname, altq->ifname,
4306 IFNAMSIZ) == 0) {
4307 altq->altq_disc = a->altq_disc;
4308 break;
4309 }
4310 }
4311 }
4312
4313 if ((ifp = ifunit(altq->ifname)) == NULL)
4314 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4315 else
4316 error = altq_add(ifp, altq);
4317
4318 if (error) {
4319 PF_RULES_WUNLOCK();
4320 free(altq, M_PFALTQ);
4321 break;
4322 }
4323
4324 if (altq->qname[0] != 0)
4325 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4326 else
4327 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4328 /* version error check done on import above */
4329 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4330 PF_RULES_WUNLOCK();
4331 break;
4332 }
4333
4334 case DIOCGETALTQSV0:
4335 case DIOCGETALTQSV1: {
4336 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4337 struct pf_altq *altq;
4338
4339 PF_RULES_RLOCK();
4340 pa->nr = 0;
4341 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4342 pa->nr++;
4343 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4344 pa->nr++;
4345 pa->ticket = V_ticket_altqs_active;
4346 PF_RULES_RUNLOCK();
4347 break;
4348 }
4349
4350 case DIOCGETALTQV0:
4351 case DIOCGETALTQV1: {
4352 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4353 struct pf_altq *altq;
4354
4355 PF_RULES_RLOCK();
4356 if (pa->ticket != V_ticket_altqs_active) {
4357 PF_RULES_RUNLOCK();
4358 error = EBUSY;
4359 break;
4360 }
4361 altq = pf_altq_get_nth_active(pa->nr);
4362 if (altq == NULL) {
4363 PF_RULES_RUNLOCK();
4364 error = EBUSY;
4365 break;
4366 }
4367 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4368 PF_RULES_RUNLOCK();
4369 break;
4370 }
4371
4372 case DIOCCHANGEALTQV0:
4373 case DIOCCHANGEALTQV1:
4374 /* CHANGEALTQ not supported yet! */
4375 error = ENODEV;
4376 break;
4377
4378 case DIOCGETQSTATSV0:
4379 case DIOCGETQSTATSV1: {
4380 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
4381 struct pf_altq *altq;
4382 int nbytes;
4383 u_int32_t version;
4384
4385 PF_RULES_RLOCK();
4386 if (pq->ticket != V_ticket_altqs_active) {
4387 PF_RULES_RUNLOCK();
4388 error = EBUSY;
4389 break;
4390 }
4391 nbytes = pq->nbytes;
4392 altq = pf_altq_get_nth_active(pq->nr);
4393 if (altq == NULL) {
4394 PF_RULES_RUNLOCK();
4395 error = EBUSY;
4396 break;
4397 }
4398
4399 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4400 PF_RULES_RUNLOCK();
4401 error = ENXIO;
4402 break;
4403 }
4404 PF_RULES_RUNLOCK();
4405 if (cmd == DIOCGETQSTATSV0)
4406 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
4407 else
4408 version = pq->version;
4409 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4410 if (error == 0) {
4411 pq->scheduler = altq->scheduler;
4412 pq->nbytes = nbytes;
4413 }
4414 break;
4415 }
4416 #endif /* ALTQ */
4417
4418 case DIOCBEGINADDRS: {
4419 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4420
4421 error = pf_ioctl_begin_addrs(&pp->ticket);
4422 break;
4423 }
4424
4425 case DIOCADDADDR: {
4426 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4427 struct pf_nl_pooladdr npp = {};
4428
4429 npp.which = PF_RDR;
4430 memcpy(&npp, pp, sizeof(*pp));
4431 error = pf_ioctl_add_addr(&npp);
4432 break;
4433 }
4434
4435 case DIOCGETADDRS: {
4436 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4437 struct pf_nl_pooladdr npp = {};
4438
4439 npp.which = PF_RDR;
4440 memcpy(&npp, pp, sizeof(*pp));
4441 error = pf_ioctl_get_addrs(&npp);
4442 memcpy(pp, &npp, sizeof(*pp));
4443
4444 break;
4445 }
4446
4447 case DIOCGETADDR: {
4448 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4449 struct pf_nl_pooladdr npp = {};
4450
4451 npp.which = PF_RDR;
4452 memcpy(&npp, pp, sizeof(*pp));
4453 error = pf_ioctl_get_addr(&npp);
4454 memcpy(pp, &npp, sizeof(*pp));
4455
4456 break;
4457 }
4458
4459 case DIOCCHANGEADDR: {
4460 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
4461 struct pf_kpool *pool;
4462 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
4463 struct pf_kruleset *ruleset;
4464 struct pfi_kkif *kif = NULL;
4465
4466 pca->anchor[sizeof(pca->anchor) - 1] = 0;
4467
4468 if (pca->action < PF_CHANGE_ADD_HEAD ||
4469 pca->action > PF_CHANGE_REMOVE) {
4470 error = EINVAL;
4471 break;
4472 }
4473 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4474 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4475 pca->addr.addr.type != PF_ADDR_TABLE) {
4476 error = EINVAL;
4477 break;
4478 }
4479 if (pca->addr.addr.p.dyn != NULL) {
4480 error = EINVAL;
4481 break;
4482 }
4483
4484 if (pca->action != PF_CHANGE_REMOVE) {
4485 #ifndef INET
4486 if (pca->af == AF_INET) {
4487 error = EAFNOSUPPORT;
4488 break;
4489 }
4490 #endif /* INET */
4491 #ifndef INET6
4492 if (pca->af == AF_INET6) {
4493 error = EAFNOSUPPORT;
4494 break;
4495 }
4496 #endif /* INET6 */
4497 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4498 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4499 if (newpa->ifname[0])
4500 kif = pf_kkif_create(M_WAITOK);
4501 newpa->kif = NULL;
4502 }
4503 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4504 PF_RULES_WLOCK();
4505 ruleset = pf_find_kruleset(pca->anchor);
4506 if (ruleset == NULL)
4507 ERROUT(EBUSY);
4508
4509 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4510 pca->r_num, pca->r_last, 1, 1, PF_RDR);
4511 if (pool == NULL)
4512 ERROUT(EBUSY);
4513
4514 if (pca->action != PF_CHANGE_REMOVE) {
4515 if (newpa->ifname[0]) {
4516 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4517 pfi_kkif_ref(newpa->kif);
4518 kif = NULL;
4519 }
4520
4521 switch (newpa->addr.type) {
4522 case PF_ADDR_DYNIFTL:
4523 error = pfi_dynaddr_setup(&newpa->addr,
4524 pca->af);
4525 break;
4526 case PF_ADDR_TABLE:
4527 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4528 newpa->addr.v.tblname);
4529 if (newpa->addr.p.tbl == NULL)
4530 error = ENOMEM;
4531 break;
4532 }
4533 if (error)
4534 goto DIOCCHANGEADDR_error;
4535 }
4536
4537 switch (pca->action) {
4538 case PF_CHANGE_ADD_HEAD:
4539 oldpa = TAILQ_FIRST(&pool->list);
4540 break;
4541 case PF_CHANGE_ADD_TAIL:
4542 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4543 break;
4544 default:
4545 oldpa = TAILQ_FIRST(&pool->list);
4546 for (int i = 0; oldpa && i < pca->nr; i++)
4547 oldpa = TAILQ_NEXT(oldpa, entries);
4548
4549 if (oldpa == NULL)
4550 ERROUT(EINVAL);
4551 }
4552
4553 if (pca->action == PF_CHANGE_REMOVE) {
4554 TAILQ_REMOVE(&pool->list, oldpa, entries);
4555 switch (oldpa->addr.type) {
4556 case PF_ADDR_DYNIFTL:
4557 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4558 break;
4559 case PF_ADDR_TABLE:
4560 pfr_detach_table(oldpa->addr.p.tbl);
4561 break;
4562 }
4563 if (oldpa->kif)
4564 pfi_kkif_unref(oldpa->kif);
4565 free(oldpa, M_PFRULE);
4566 } else {
4567 if (oldpa == NULL)
4568 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4569 else if (pca->action == PF_CHANGE_ADD_HEAD ||
4570 pca->action == PF_CHANGE_ADD_BEFORE)
4571 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4572 else
4573 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4574 newpa, entries);
4575 }
4576
4577 pool->cur = TAILQ_FIRST(&pool->list);
4578 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4579 PF_RULES_WUNLOCK();
4580 break;
4581
4582 #undef ERROUT
4583 DIOCCHANGEADDR_error:
4584 if (newpa != NULL) {
4585 if (newpa->kif)
4586 pfi_kkif_unref(newpa->kif);
4587 free(newpa, M_PFRULE);
4588 }
4589 PF_RULES_WUNLOCK();
4590 pf_kkif_free(kif);
4591 break;
4592 }
4593
4594 case DIOCGETRULESETS: {
4595 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4596
4597 pr->path[sizeof(pr->path) - 1] = 0;
4598
4599 error = pf_ioctl_get_rulesets(pr);
4600 break;
4601 }
4602
4603 case DIOCGETRULESET: {
4604 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4605
4606 pr->path[sizeof(pr->path) - 1] = 0;
4607
4608 error = pf_ioctl_get_ruleset(pr);
4609 break;
4610 }
4611
4612 case DIOCRCLRTABLES: {
4613 struct pfioc_table *io = (struct pfioc_table *)addr;
4614
4615 if (io->pfrio_esize != 0) {
4616 error = ENODEV;
4617 break;
4618 }
4619 PF_RULES_WLOCK();
4620 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4621 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4622 PF_RULES_WUNLOCK();
4623 break;
4624 }
4625
4626 case DIOCRADDTABLES: {
4627 struct pfioc_table *io = (struct pfioc_table *)addr;
4628 struct pfr_table *pfrts;
4629 size_t totlen;
4630
4631 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4632 error = ENODEV;
4633 break;
4634 }
4635
4636 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4637 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4638 error = ENOMEM;
4639 break;
4640 }
4641
4642 totlen = io->pfrio_size * sizeof(struct pfr_table);
4643 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4644 M_TEMP, M_WAITOK);
4645 error = copyin(io->pfrio_buffer, pfrts, totlen);
4646 if (error) {
4647 free(pfrts, M_TEMP);
4648 break;
4649 }
4650 PF_RULES_WLOCK();
4651 error = pfr_add_tables(pfrts, io->pfrio_size,
4652 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4653 PF_RULES_WUNLOCK();
4654 free(pfrts, M_TEMP);
4655 break;
4656 }
4657
4658 case DIOCRDELTABLES: {
4659 struct pfioc_table *io = (struct pfioc_table *)addr;
4660 struct pfr_table *pfrts;
4661 size_t totlen;
4662
4663 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4664 error = ENODEV;
4665 break;
4666 }
4667
4668 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4669 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4670 error = ENOMEM;
4671 break;
4672 }
4673
4674 totlen = io->pfrio_size * sizeof(struct pfr_table);
4675 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4676 M_TEMP, M_WAITOK);
4677 error = copyin(io->pfrio_buffer, pfrts, totlen);
4678 if (error) {
4679 free(pfrts, M_TEMP);
4680 break;
4681 }
4682 PF_RULES_WLOCK();
4683 error = pfr_del_tables(pfrts, io->pfrio_size,
4684 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4685 PF_RULES_WUNLOCK();
4686 free(pfrts, M_TEMP);
4687 break;
4688 }
4689
4690 case DIOCRGETTABLES: {
4691 struct pfioc_table *io = (struct pfioc_table *)addr;
4692 struct pfr_table *pfrts;
4693 size_t totlen;
4694 int n;
4695
4696 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4697 error = ENODEV;
4698 break;
4699 }
4700 PF_RULES_RLOCK();
4701 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4702 if (n < 0) {
4703 PF_RULES_RUNLOCK();
4704 error = EINVAL;
4705 break;
4706 }
4707 io->pfrio_size = min(io->pfrio_size, n);
4708
4709 totlen = io->pfrio_size * sizeof(struct pfr_table);
4710
4711 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4712 M_TEMP, M_NOWAIT | M_ZERO);
4713 if (pfrts == NULL) {
4714 error = ENOMEM;
4715 PF_RULES_RUNLOCK();
4716 break;
4717 }
4718 error = pfr_get_tables(&io->pfrio_table, pfrts,
4719 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4720 PF_RULES_RUNLOCK();
4721 if (error == 0)
4722 error = copyout(pfrts, io->pfrio_buffer, totlen);
4723 free(pfrts, M_TEMP);
4724 break;
4725 }
4726
4727 case DIOCRGETTSTATS: {
4728 struct pfioc_table *io = (struct pfioc_table *)addr;
4729 struct pfr_tstats *pfrtstats;
4730 size_t totlen;
4731 int n;
4732
4733 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4734 error = ENODEV;
4735 break;
4736 }
4737 PF_TABLE_STATS_LOCK();
4738 PF_RULES_RLOCK();
4739 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4740 if (n < 0) {
4741 PF_RULES_RUNLOCK();
4742 PF_TABLE_STATS_UNLOCK();
4743 error = EINVAL;
4744 break;
4745 }
4746 io->pfrio_size = min(io->pfrio_size, n);
4747
4748 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4749 pfrtstats = mallocarray(io->pfrio_size,
4750 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4751 if (pfrtstats == NULL) {
4752 error = ENOMEM;
4753 PF_RULES_RUNLOCK();
4754 PF_TABLE_STATS_UNLOCK();
4755 break;
4756 }
4757 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4758 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4759 PF_RULES_RUNLOCK();
4760 PF_TABLE_STATS_UNLOCK();
4761 if (error == 0)
4762 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4763 free(pfrtstats, M_TEMP);
4764 break;
4765 }
4766
4767 case DIOCRCLRTSTATS: {
4768 struct pfioc_table *io = (struct pfioc_table *)addr;
4769 struct pfr_table *pfrts;
4770 size_t totlen;
4771
4772 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4773 error = ENODEV;
4774 break;
4775 }
4776
4777 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4778 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4779 /* We used to count tables and use the minimum required
4780 * size, so we didn't fail on overly large requests.
4781 * Keep doing so. */
4782 io->pfrio_size = pf_ioctl_maxcount;
4783 break;
4784 }
4785
4786 totlen = io->pfrio_size * sizeof(struct pfr_table);
4787 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4788 M_TEMP, M_WAITOK);
4789 error = copyin(io->pfrio_buffer, pfrts, totlen);
4790 if (error) {
4791 free(pfrts, M_TEMP);
4792 break;
4793 }
4794
4795 PF_TABLE_STATS_LOCK();
4796 PF_RULES_RLOCK();
4797 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4798 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4799 PF_RULES_RUNLOCK();
4800 PF_TABLE_STATS_UNLOCK();
4801 free(pfrts, M_TEMP);
4802 break;
4803 }
4804
4805 case DIOCRSETTFLAGS: {
4806 struct pfioc_table *io = (struct pfioc_table *)addr;
4807 struct pfr_table *pfrts;
4808 size_t totlen;
4809 int n;
4810
4811 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4812 error = ENODEV;
4813 break;
4814 }
4815
4816 PF_RULES_RLOCK();
4817 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4818 if (n < 0) {
4819 PF_RULES_RUNLOCK();
4820 error = EINVAL;
4821 break;
4822 }
4823
4824 io->pfrio_size = min(io->pfrio_size, n);
4825 PF_RULES_RUNLOCK();
4826
4827 totlen = io->pfrio_size * sizeof(struct pfr_table);
4828 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4829 M_TEMP, M_WAITOK);
4830 error = copyin(io->pfrio_buffer, pfrts, totlen);
4831 if (error) {
4832 free(pfrts, M_TEMP);
4833 break;
4834 }
4835 PF_RULES_WLOCK();
4836 error = pfr_set_tflags(pfrts, io->pfrio_size,
4837 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4838 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4839 PF_RULES_WUNLOCK();
4840 free(pfrts, M_TEMP);
4841 break;
4842 }
4843
4844 case DIOCRCLRADDRS: {
4845 struct pfioc_table *io = (struct pfioc_table *)addr;
4846
4847 if (io->pfrio_esize != 0) {
4848 error = ENODEV;
4849 break;
4850 }
4851 PF_RULES_WLOCK();
4852 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4853 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4854 PF_RULES_WUNLOCK();
4855 break;
4856 }
4857
4858 case DIOCRADDADDRS: {
4859 struct pfioc_table *io = (struct pfioc_table *)addr;
4860 struct pfr_addr *pfras;
4861 size_t totlen;
4862
4863 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4864 error = ENODEV;
4865 break;
4866 }
4867 if (io->pfrio_size < 0 ||
4868 io->pfrio_size > pf_ioctl_maxcount ||
4869 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4870 error = EINVAL;
4871 break;
4872 }
4873 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4874 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4875 M_TEMP, M_WAITOK);
4876 error = copyin(io->pfrio_buffer, pfras, totlen);
4877 if (error) {
4878 free(pfras, M_TEMP);
4879 break;
4880 }
4881 PF_RULES_WLOCK();
4882 error = pfr_add_addrs(&io->pfrio_table, pfras,
4883 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4884 PFR_FLAG_USERIOCTL);
4885 PF_RULES_WUNLOCK();
4886 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4887 error = copyout(pfras, io->pfrio_buffer, totlen);
4888 free(pfras, M_TEMP);
4889 break;
4890 }
4891
4892 case DIOCRDELADDRS: {
4893 struct pfioc_table *io = (struct pfioc_table *)addr;
4894 struct pfr_addr *pfras;
4895 size_t totlen;
4896
4897 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4898 error = ENODEV;
4899 break;
4900 }
4901 if (io->pfrio_size < 0 ||
4902 io->pfrio_size > pf_ioctl_maxcount ||
4903 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4904 error = EINVAL;
4905 break;
4906 }
4907 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4908 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4909 M_TEMP, M_WAITOK);
4910 error = copyin(io->pfrio_buffer, pfras, totlen);
4911 if (error) {
4912 free(pfras, M_TEMP);
4913 break;
4914 }
4915 PF_RULES_WLOCK();
4916 error = pfr_del_addrs(&io->pfrio_table, pfras,
4917 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4918 PFR_FLAG_USERIOCTL);
4919 PF_RULES_WUNLOCK();
4920 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4921 error = copyout(pfras, io->pfrio_buffer, totlen);
4922 free(pfras, M_TEMP);
4923 break;
4924 }
4925
4926 case DIOCRSETADDRS: {
4927 struct pfioc_table *io = (struct pfioc_table *)addr;
4928 struct pfr_addr *pfras;
4929 size_t totlen, count;
4930
4931 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4932 error = ENODEV;
4933 break;
4934 }
4935 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4936 error = EINVAL;
4937 break;
4938 }
4939 count = max(io->pfrio_size, io->pfrio_size2);
4940 if (count > pf_ioctl_maxcount ||
4941 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4942 error = EINVAL;
4943 break;
4944 }
4945 totlen = count * sizeof(struct pfr_addr);
4946 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4947 M_WAITOK);
4948 error = copyin(io->pfrio_buffer, pfras, totlen);
4949 if (error) {
4950 free(pfras, M_TEMP);
4951 break;
4952 }
4953 PF_RULES_WLOCK();
4954 error = pfr_set_addrs(&io->pfrio_table, pfras,
4955 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4956 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4957 PFR_FLAG_USERIOCTL, 0);
4958 PF_RULES_WUNLOCK();
4959 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4960 error = copyout(pfras, io->pfrio_buffer, totlen);
4961 free(pfras, M_TEMP);
4962 break;
4963 }
4964
4965 case DIOCRGETADDRS: {
4966 struct pfioc_table *io = (struct pfioc_table *)addr;
4967 struct pfr_addr *pfras;
4968 size_t totlen;
4969
4970 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4971 error = ENODEV;
4972 break;
4973 }
4974 if (io->pfrio_size < 0 ||
4975 io->pfrio_size > pf_ioctl_maxcount ||
4976 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4977 error = EINVAL;
4978 break;
4979 }
4980 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4981 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4982 M_TEMP, M_WAITOK | M_ZERO);
4983 PF_RULES_RLOCK();
4984 error = pfr_get_addrs(&io->pfrio_table, pfras,
4985 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4986 PF_RULES_RUNLOCK();
4987 if (error == 0)
4988 error = copyout(pfras, io->pfrio_buffer, totlen);
4989 free(pfras, M_TEMP);
4990 break;
4991 }
4992
4993 case DIOCRGETASTATS: {
4994 struct pfioc_table *io = (struct pfioc_table *)addr;
4995 struct pfr_astats *pfrastats;
4996 size_t totlen;
4997
4998 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4999 error = ENODEV;
5000 break;
5001 }
5002 if (io->pfrio_size < 0 ||
5003 io->pfrio_size > pf_ioctl_maxcount ||
5004 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5005 error = EINVAL;
5006 break;
5007 }
5008 totlen = io->pfrio_size * sizeof(struct pfr_astats);
5009 pfrastats = mallocarray(io->pfrio_size,
5010 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
5011 PF_RULES_RLOCK();
5012 error = pfr_get_astats(&io->pfrio_table, pfrastats,
5013 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5014 PF_RULES_RUNLOCK();
5015 if (error == 0)
5016 error = copyout(pfrastats, io->pfrio_buffer, totlen);
5017 free(pfrastats, M_TEMP);
5018 break;
5019 }
5020
5021 case DIOCRCLRASTATS: {
5022 struct pfioc_table *io = (struct pfioc_table *)addr;
5023 struct pfr_addr *pfras;
5024 size_t totlen;
5025
5026 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5027 error = ENODEV;
5028 break;
5029 }
5030 if (io->pfrio_size < 0 ||
5031 io->pfrio_size > pf_ioctl_maxcount ||
5032 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5033 error = EINVAL;
5034 break;
5035 }
5036 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5037 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5038 M_TEMP, M_WAITOK);
5039 error = copyin(io->pfrio_buffer, pfras, totlen);
5040 if (error) {
5041 free(pfras, M_TEMP);
5042 break;
5043 }
5044 PF_RULES_WLOCK();
5045 error = pfr_clr_astats(&io->pfrio_table, pfras,
5046 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5047 PFR_FLAG_USERIOCTL);
5048 PF_RULES_WUNLOCK();
5049 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5050 error = copyout(pfras, io->pfrio_buffer, totlen);
5051 free(pfras, M_TEMP);
5052 break;
5053 }
5054
5055 case DIOCRTSTADDRS: {
5056 struct pfioc_table *io = (struct pfioc_table *)addr;
5057 struct pfr_addr *pfras;
5058 size_t totlen;
5059
5060 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5061 error = ENODEV;
5062 break;
5063 }
5064 if (io->pfrio_size < 0 ||
5065 io->pfrio_size > pf_ioctl_maxcount ||
5066 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5067 error = EINVAL;
5068 break;
5069 }
5070 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5071 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5072 M_TEMP, M_WAITOK);
5073 error = copyin(io->pfrio_buffer, pfras, totlen);
5074 if (error) {
5075 free(pfras, M_TEMP);
5076 break;
5077 }
5078 PF_RULES_RLOCK();
5079 error = pfr_tst_addrs(&io->pfrio_table, pfras,
5080 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5081 PFR_FLAG_USERIOCTL);
5082 PF_RULES_RUNLOCK();
5083 if (error == 0)
5084 error = copyout(pfras, io->pfrio_buffer, totlen);
5085 free(pfras, M_TEMP);
5086 break;
5087 }
5088
5089 case DIOCRINADEFINE: {
5090 struct pfioc_table *io = (struct pfioc_table *)addr;
5091 struct pfr_addr *pfras;
5092 size_t totlen;
5093
5094 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5095 error = ENODEV;
5096 break;
5097 }
5098 if (io->pfrio_size < 0 ||
5099 io->pfrio_size > pf_ioctl_maxcount ||
5100 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5101 error = EINVAL;
5102 break;
5103 }
5104 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5105 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5106 M_TEMP, M_WAITOK);
5107 error = copyin(io->pfrio_buffer, pfras, totlen);
5108 if (error) {
5109 free(pfras, M_TEMP);
5110 break;
5111 }
5112 PF_RULES_WLOCK();
5113 error = pfr_ina_define(&io->pfrio_table, pfras,
5114 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5115 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5116 PF_RULES_WUNLOCK();
5117 free(pfras, M_TEMP);
5118 break;
5119 }
5120
5121 case DIOCOSFPADD: {
5122 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5123 PF_RULES_WLOCK();
5124 error = pf_osfp_add(io);
5125 PF_RULES_WUNLOCK();
5126 break;
5127 }
5128
5129 case DIOCOSFPGET: {
5130 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5131 PF_RULES_RLOCK();
5132 error = pf_osfp_get(io);
5133 PF_RULES_RUNLOCK();
5134 break;
5135 }
5136
5137 case DIOCXBEGIN: {
5138 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5139 struct pfioc_trans_e *ioes, *ioe;
5140 size_t totlen;
5141 int i;
5142
5143 if (io->esize != sizeof(*ioe)) {
5144 error = ENODEV;
5145 break;
5146 }
5147 if (io->size < 0 ||
5148 io->size > pf_ioctl_maxcount ||
5149 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5150 error = EINVAL;
5151 break;
5152 }
5153 totlen = sizeof(struct pfioc_trans_e) * io->size;
5154 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5155 M_TEMP, M_WAITOK);
5156 error = copyin(io->array, ioes, totlen);
5157 if (error) {
5158 free(ioes, M_TEMP);
5159 break;
5160 }
5161 PF_RULES_WLOCK();
5162 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5163 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5164 switch (ioe->rs_num) {
5165 case PF_RULESET_ETH:
5166 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5167 PF_RULES_WUNLOCK();
5168 free(ioes, M_TEMP);
5169 goto fail;
5170 }
5171 break;
5172 #ifdef ALTQ
5173 case PF_RULESET_ALTQ:
5174 if (ioe->anchor[0]) {
5175 PF_RULES_WUNLOCK();
5176 free(ioes, M_TEMP);
5177 error = EINVAL;
5178 goto fail;
5179 }
5180 if ((error = pf_begin_altq(&ioe->ticket))) {
5181 PF_RULES_WUNLOCK();
5182 free(ioes, M_TEMP);
5183 goto fail;
5184 }
5185 break;
5186 #endif /* ALTQ */
5187 case PF_RULESET_TABLE:
5188 {
5189 struct pfr_table table;
5190
5191 bzero(&table, sizeof(table));
5192 strlcpy(table.pfrt_anchor, ioe->anchor,
5193 sizeof(table.pfrt_anchor));
5194 if ((error = pfr_ina_begin(&table,
5195 &ioe->ticket, NULL, 0))) {
5196 PF_RULES_WUNLOCK();
5197 free(ioes, M_TEMP);
5198 goto fail;
5199 }
5200 break;
5201 }
5202 default:
5203 if ((error = pf_begin_rules(&ioe->ticket,
5204 ioe->rs_num, ioe->anchor))) {
5205 PF_RULES_WUNLOCK();
5206 free(ioes, M_TEMP);
5207 goto fail;
5208 }
5209 break;
5210 }
5211 }
5212 PF_RULES_WUNLOCK();
5213 error = copyout(ioes, io->array, totlen);
5214 free(ioes, M_TEMP);
5215 break;
5216 }
5217
5218 case DIOCXROLLBACK: {
5219 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5220 struct pfioc_trans_e *ioe, *ioes;
5221 size_t totlen;
5222 int i;
5223
5224 if (io->esize != sizeof(*ioe)) {
5225 error = ENODEV;
5226 break;
5227 }
5228 if (io->size < 0 ||
5229 io->size > pf_ioctl_maxcount ||
5230 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5231 error = EINVAL;
5232 break;
5233 }
5234 totlen = sizeof(struct pfioc_trans_e) * io->size;
5235 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5236 M_TEMP, M_WAITOK);
5237 error = copyin(io->array, ioes, totlen);
5238 if (error) {
5239 free(ioes, M_TEMP);
5240 break;
5241 }
5242 PF_RULES_WLOCK();
5243 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5244 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5245 switch (ioe->rs_num) {
5246 case PF_RULESET_ETH:
5247 if ((error = pf_rollback_eth(ioe->ticket,
5248 ioe->anchor))) {
5249 PF_RULES_WUNLOCK();
5250 free(ioes, M_TEMP);
5251 goto fail; /* really bad */
5252 }
5253 break;
5254 #ifdef ALTQ
5255 case PF_RULESET_ALTQ:
5256 if (ioe->anchor[0]) {
5257 PF_RULES_WUNLOCK();
5258 free(ioes, M_TEMP);
5259 error = EINVAL;
5260 goto fail;
5261 }
5262 if ((error = pf_rollback_altq(ioe->ticket))) {
5263 PF_RULES_WUNLOCK();
5264 free(ioes, M_TEMP);
5265 goto fail; /* really bad */
5266 }
5267 break;
5268 #endif /* ALTQ */
5269 case PF_RULESET_TABLE:
5270 {
5271 struct pfr_table table;
5272
5273 bzero(&table, sizeof(table));
5274 strlcpy(table.pfrt_anchor, ioe->anchor,
5275 sizeof(table.pfrt_anchor));
5276 if ((error = pfr_ina_rollback(&table,
5277 ioe->ticket, NULL, 0))) {
5278 PF_RULES_WUNLOCK();
5279 free(ioes, M_TEMP);
5280 goto fail; /* really bad */
5281 }
5282 break;
5283 }
5284 default:
5285 if ((error = pf_rollback_rules(ioe->ticket,
5286 ioe->rs_num, ioe->anchor))) {
5287 PF_RULES_WUNLOCK();
5288 free(ioes, M_TEMP);
5289 goto fail; /* really bad */
5290 }
5291 break;
5292 }
5293 }
5294 PF_RULES_WUNLOCK();
5295 free(ioes, M_TEMP);
5296 break;
5297 }
5298
5299 case DIOCXCOMMIT: {
5300 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5301 struct pfioc_trans_e *ioe, *ioes;
5302 struct pf_kruleset *rs;
5303 struct pf_keth_ruleset *ers;
5304 size_t totlen;
5305 int i;
5306
5307 if (io->esize != sizeof(*ioe)) {
5308 error = ENODEV;
5309 break;
5310 }
5311
5312 if (io->size < 0 ||
5313 io->size > pf_ioctl_maxcount ||
5314 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5315 error = EINVAL;
5316 break;
5317 }
5318
5319 totlen = sizeof(struct pfioc_trans_e) * io->size;
5320 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5321 M_TEMP, M_WAITOK);
5322 error = copyin(io->array, ioes, totlen);
5323 if (error) {
5324 free(ioes, M_TEMP);
5325 break;
5326 }
5327 PF_RULES_WLOCK();
5328 /* First makes sure everything will succeed. */
5329 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5330 ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5331 switch (ioe->rs_num) {
5332 case PF_RULESET_ETH:
5333 ers = pf_find_keth_ruleset(ioe->anchor);
5334 if (ers == NULL || ioe->ticket == 0 ||
5335 ioe->ticket != ers->inactive.ticket) {
5336 PF_RULES_WUNLOCK();
5337 free(ioes, M_TEMP);
5338 error = EINVAL;
5339 goto fail;
5340 }
5341 break;
5342 #ifdef ALTQ
5343 case PF_RULESET_ALTQ:
5344 if (ioe->anchor[0]) {
5345 PF_RULES_WUNLOCK();
5346 free(ioes, M_TEMP);
5347 error = EINVAL;
5348 goto fail;
5349 }
5350 if (!V_altqs_inactive_open || ioe->ticket !=
5351 V_ticket_altqs_inactive) {
5352 PF_RULES_WUNLOCK();
5353 free(ioes, M_TEMP);
5354 error = EBUSY;
5355 goto fail;
5356 }
5357 break;
5358 #endif /* ALTQ */
5359 case PF_RULESET_TABLE:
5360 rs = pf_find_kruleset(ioe->anchor);
5361 if (rs == NULL || !rs->topen || ioe->ticket !=
5362 rs->tticket) {
5363 PF_RULES_WUNLOCK();
5364 free(ioes, M_TEMP);
5365 error = EBUSY;
5366 goto fail;
5367 }
5368 break;
5369 default:
5370 if (ioe->rs_num < 0 || ioe->rs_num >=
5371 PF_RULESET_MAX) {
5372 PF_RULES_WUNLOCK();
5373 free(ioes, M_TEMP);
5374 error = EINVAL;
5375 goto fail;
5376 }
5377 rs = pf_find_kruleset(ioe->anchor);
5378 if (rs == NULL ||
5379 !rs->rules[ioe->rs_num].inactive.open ||
5380 rs->rules[ioe->rs_num].inactive.ticket !=
5381 ioe->ticket) {
5382 PF_RULES_WUNLOCK();
5383 free(ioes, M_TEMP);
5384 error = EBUSY;
5385 goto fail;
5386 }
5387 break;
5388 }
5389 }
5390 /* Now do the commit - no errors should happen here. */
5391 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5392 switch (ioe->rs_num) {
5393 case PF_RULESET_ETH:
5394 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5395 PF_RULES_WUNLOCK();
5396 free(ioes, M_TEMP);
5397 goto fail; /* really bad */
5398 }
5399 break;
5400 #ifdef ALTQ
5401 case PF_RULESET_ALTQ:
5402 if ((error = pf_commit_altq(ioe->ticket))) {
5403 PF_RULES_WUNLOCK();
5404 free(ioes, M_TEMP);
5405 goto fail; /* really bad */
5406 }
5407 break;
5408 #endif /* ALTQ */
5409 case PF_RULESET_TABLE:
5410 {
5411 struct pfr_table table;
5412
5413 bzero(&table, sizeof(table));
5414 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
5415 sizeof(table.pfrt_anchor));
5416 if ((error = pfr_ina_commit(&table,
5417 ioe->ticket, NULL, NULL, 0))) {
5418 PF_RULES_WUNLOCK();
5419 free(ioes, M_TEMP);
5420 goto fail; /* really bad */
5421 }
5422 break;
5423 }
5424 default:
5425 if ((error = pf_commit_rules(ioe->ticket,
5426 ioe->rs_num, ioe->anchor))) {
5427 PF_RULES_WUNLOCK();
5428 free(ioes, M_TEMP);
5429 goto fail; /* really bad */
5430 }
5431 break;
5432 }
5433 }
5434 PF_RULES_WUNLOCK();
5435
5436 /* Only hook into EtherNet taffic if we've got rules for it. */
5437 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5438 hook_pf_eth();
5439 else
5440 dehook_pf_eth();
5441
5442 free(ioes, M_TEMP);
5443 break;
5444 }
5445
5446 case DIOCGETSRCNODES: {
5447 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
5448 struct pf_srchash *sh;
5449 struct pf_ksrc_node *n;
5450 struct pf_src_node *p, *pstore;
5451 uint32_t i, nr = 0;
5452
5453 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5454 i++, sh++) {
5455 PF_HASHROW_LOCK(sh);
5456 LIST_FOREACH(n, &sh->nodes, entry)
5457 nr++;
5458 PF_HASHROW_UNLOCK(sh);
5459 }
5460
5461 psn->psn_len = min(psn->psn_len,
5462 sizeof(struct pf_src_node) * nr);
5463
5464 if (psn->psn_len == 0) {
5465 psn->psn_len = sizeof(struct pf_src_node) * nr;
5466 break;
5467 }
5468
5469 nr = 0;
5470
5471 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5472 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5473 i++, sh++) {
5474 PF_HASHROW_LOCK(sh);
5475 LIST_FOREACH(n, &sh->nodes, entry) {
5476
5477 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5478 break;
5479
5480 pf_src_node_copy(n, p);
5481
5482 p++;
5483 nr++;
5484 }
5485 PF_HASHROW_UNLOCK(sh);
5486 }
5487 error = copyout(pstore, psn->psn_src_nodes,
5488 sizeof(struct pf_src_node) * nr);
5489 if (error) {
5490 free(pstore, M_TEMP);
5491 break;
5492 }
5493 psn->psn_len = sizeof(struct pf_src_node) * nr;
5494 free(pstore, M_TEMP);
5495 break;
5496 }
5497
5498 case DIOCCLRSRCNODES: {
5499 pf_kill_srcnodes(NULL);
5500 break;
5501 }
5502
5503 case DIOCKILLSRCNODES:
5504 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5505 break;
5506
5507 #ifdef COMPAT_FREEBSD13
5508 case DIOCKEEPCOUNTERS_FREEBSD13:
5509 #endif
5510 case DIOCKEEPCOUNTERS:
5511 error = pf_keepcounters((struct pfioc_nv *)addr);
5512 break;
5513
5514 case DIOCGETSYNCOOKIES:
5515 error = pf_get_syncookies((struct pfioc_nv *)addr);
5516 break;
5517
5518 case DIOCSETSYNCOOKIES:
5519 error = pf_set_syncookies((struct pfioc_nv *)addr);
5520 break;
5521
5522 case DIOCSETHOSTID: {
5523 u_int32_t *hostid = (u_int32_t *)addr;
5524
5525 PF_RULES_WLOCK();
5526 if (*hostid == 0)
5527 V_pf_status.hostid = arc4random();
5528 else
5529 V_pf_status.hostid = *hostid;
5530 PF_RULES_WUNLOCK();
5531 break;
5532 }
5533
5534 case DIOCOSFPFLUSH:
5535 PF_RULES_WLOCK();
5536 pf_osfp_flush();
5537 PF_RULES_WUNLOCK();
5538 break;
5539
5540 case DIOCIGETIFACES: {
5541 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5542 struct pfi_kif *ifstore;
5543 size_t bufsiz;
5544
5545 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5546 error = ENODEV;
5547 break;
5548 }
5549
5550 if (io->pfiio_size < 0 ||
5551 io->pfiio_size > pf_ioctl_maxcount ||
5552 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5553 error = EINVAL;
5554 break;
5555 }
5556
5557 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5558
5559 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5560 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5561 M_TEMP, M_WAITOK | M_ZERO);
5562
5563 PF_RULES_RLOCK();
5564 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5565 PF_RULES_RUNLOCK();
5566 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5567 free(ifstore, M_TEMP);
5568 break;
5569 }
5570
5571 case DIOCSETIFFLAG: {
5572 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5573
5574 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5575
5576 PF_RULES_WLOCK();
5577 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5578 PF_RULES_WUNLOCK();
5579 break;
5580 }
5581
5582 case DIOCCLRIFFLAG: {
5583 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5584
5585 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5586
5587 PF_RULES_WLOCK();
5588 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5589 PF_RULES_WUNLOCK();
5590 break;
5591 }
5592
5593 case DIOCSETREASS: {
5594 u_int32_t *reass = (u_int32_t *)addr;
5595
5596 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5597 /* Removal of DF flag without reassembly enabled is not a
5598 * valid combination. Disable reassembly in such case. */
5599 if (!(V_pf_status.reass & PF_REASS_ENABLED))
5600 V_pf_status.reass = 0;
5601 break;
5602 }
5603
5604 default:
5605 error = ENODEV;
5606 break;
5607 }
5608 fail:
5609 CURVNET_RESTORE();
5610
5611 #undef ERROUT_IOCTL
5612
5613 return (error);
5614 }
5615
5616 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5617 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5618 {
5619 bzero(sp, sizeof(union pfsync_state_union));
5620
5621 /* copy from state key */
5622 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5623 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5624 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5625 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5626 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5627 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5628 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5629 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5630 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5631 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5632
5633 /* copy from state */
5634 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5635 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5636 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5637 sp->pfs_1301.expire = pf_state_expires(st);
5638 if (sp->pfs_1301.expire <= time_uptime)
5639 sp->pfs_1301.expire = htonl(0);
5640 else
5641 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5642
5643 sp->pfs_1301.direction = st->direction;
5644 sp->pfs_1301.log = st->act.log;
5645 sp->pfs_1301.timeout = st->timeout;
5646
5647 switch (msg_version) {
5648 case PFSYNC_MSG_VERSION_1301:
5649 sp->pfs_1301.state_flags = st->state_flags;
5650 break;
5651 case PFSYNC_MSG_VERSION_1400:
5652 sp->pfs_1400.state_flags = htons(st->state_flags);
5653 sp->pfs_1400.qid = htons(st->act.qid);
5654 sp->pfs_1400.pqid = htons(st->act.pqid);
5655 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5656 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5657 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5658 sp->pfs_1400.min_ttl = st->act.min_ttl;
5659 sp->pfs_1400.set_tos = st->act.set_tos;
5660 sp->pfs_1400.max_mss = htons(st->act.max_mss);
5661 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5662 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5663 sp->pfs_1400.rt = st->act.rt;
5664 if (st->act.rt_kif)
5665 strlcpy(sp->pfs_1400.rt_ifname,
5666 st->act.rt_kif->pfik_name,
5667 sizeof(sp->pfs_1400.rt_ifname));
5668 break;
5669 default:
5670 panic("%s: Unsupported pfsync_msg_version %d",
5671 __func__, msg_version);
5672 }
5673
5674 if (st->src_node)
5675 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5676 if (st->nat_src_node)
5677 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5678
5679 sp->pfs_1301.id = st->id;
5680 sp->pfs_1301.creatorid = st->creatorid;
5681 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5682 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5683
5684 if (st->rule == NULL)
5685 sp->pfs_1301.rule = htonl(-1);
5686 else
5687 sp->pfs_1301.rule = htonl(st->rule->nr);
5688 if (st->anchor == NULL)
5689 sp->pfs_1301.anchor = htonl(-1);
5690 else
5691 sp->pfs_1301.anchor = htonl(st->anchor->nr);
5692 if (st->nat_rule == NULL)
5693 sp->pfs_1301.nat_rule = htonl(-1);
5694 else
5695 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
5696
5697 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5698 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5699 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5700 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5701 }
5702
5703 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5704 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5705 {
5706 bzero(sp, sizeof(*sp));
5707
5708 sp->version = PF_STATE_VERSION;
5709
5710 /* copy from state key */
5711 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5712 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5713 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5714 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5715 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5716 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5717 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5718 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5719 sp->proto = st->key[PF_SK_WIRE]->proto;
5720 sp->af = st->key[PF_SK_WIRE]->af;
5721
5722 /* copy from state */
5723 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5724 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5725 sizeof(sp->orig_ifname));
5726 bcopy(&st->act.rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5727 sp->creation = htonl(time_uptime - (st->creation / 1000));
5728 sp->expire = pf_state_expires(st);
5729 if (sp->expire <= time_uptime)
5730 sp->expire = htonl(0);
5731 else
5732 sp->expire = htonl(sp->expire - time_uptime);
5733
5734 sp->direction = st->direction;
5735 sp->log = st->act.log;
5736 sp->timeout = st->timeout;
5737 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5738 sp->state_flags_compat = st->state_flags;
5739 sp->state_flags = htons(st->state_flags);
5740 if (st->src_node)
5741 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5742 if (st->nat_src_node)
5743 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5744
5745 sp->id = st->id;
5746 sp->creatorid = st->creatorid;
5747 pf_state_peer_hton(&st->src, &sp->src);
5748 pf_state_peer_hton(&st->dst, &sp->dst);
5749
5750 if (st->rule == NULL)
5751 sp->rule = htonl(-1);
5752 else
5753 sp->rule = htonl(st->rule->nr);
5754 if (st->anchor == NULL)
5755 sp->anchor = htonl(-1);
5756 else
5757 sp->anchor = htonl(st->anchor->nr);
5758 if (st->nat_rule == NULL)
5759 sp->nat_rule = htonl(-1);
5760 else
5761 sp->nat_rule = htonl(st->nat_rule->nr);
5762
5763 sp->packets[0] = st->packets[0];
5764 sp->packets[1] = st->packets[1];
5765 sp->bytes[0] = st->bytes[0];
5766 sp->bytes[1] = st->bytes[1];
5767
5768 sp->qid = htons(st->act.qid);
5769 sp->pqid = htons(st->act.pqid);
5770 sp->dnpipe = htons(st->act.dnpipe);
5771 sp->dnrpipe = htons(st->act.dnrpipe);
5772 sp->rtableid = htonl(st->act.rtableid);
5773 sp->min_ttl = st->act.min_ttl;
5774 sp->set_tos = st->act.set_tos;
5775 sp->max_mss = htons(st->act.max_mss);
5776 sp->rt = st->act.rt;
5777 if (st->act.rt_kif)
5778 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
5779 sizeof(sp->rt_ifname));
5780 sp->set_prio[0] = st->act.set_prio[0];
5781 sp->set_prio[1] = st->act.set_prio[1];
5782
5783 }
5784
5785 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5786 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5787 {
5788 struct pfr_ktable *kt;
5789
5790 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5791
5792 kt = aw->p.tbl;
5793 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5794 kt = kt->pfrkt_root;
5795 aw->p.tbl = NULL;
5796 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5797 kt->pfrkt_cnt : -1;
5798 }
5799
5800 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5801 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5802 size_t number, char **names)
5803 {
5804 nvlist_t *nvc;
5805
5806 nvc = nvlist_create(0);
5807 if (nvc == NULL)
5808 return (ENOMEM);
5809
5810 for (int i = 0; i < number; i++) {
5811 nvlist_append_number_array(nvc, "counters",
5812 counter_u64_fetch(counters[i]));
5813 nvlist_append_string_array(nvc, "names",
5814 names[i]);
5815 nvlist_append_number_array(nvc, "ids",
5816 i);
5817 }
5818 nvlist_add_nvlist(nvl, name, nvc);
5819 nvlist_destroy(nvc);
5820
5821 return (0);
5822 }
5823
5824 static int
pf_getstatus(struct pfioc_nv * nv)5825 pf_getstatus(struct pfioc_nv *nv)
5826 {
5827 nvlist_t *nvl = NULL, *nvc = NULL;
5828 void *nvlpacked = NULL;
5829 int error;
5830 struct pf_status s;
5831 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5832 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5833 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5834 PF_RULES_RLOCK_TRACKER;
5835
5836 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
5837
5838 PF_RULES_RLOCK();
5839
5840 nvl = nvlist_create(0);
5841 if (nvl == NULL)
5842 ERROUT(ENOMEM);
5843
5844 nvlist_add_bool(nvl, "running", V_pf_status.running);
5845 nvlist_add_number(nvl, "since", V_pf_status.since);
5846 nvlist_add_number(nvl, "debug", V_pf_status.debug);
5847 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5848 nvlist_add_number(nvl, "states", V_pf_status.states);
5849 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5850 nvlist_add_number(nvl, "reass", V_pf_status.reass);
5851 nvlist_add_bool(nvl, "syncookies_active",
5852 V_pf_status.syncookies_active);
5853 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5854
5855 /* counters */
5856 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5857 PFRES_MAX, pf_reasons);
5858 if (error != 0)
5859 ERROUT(error);
5860
5861 /* lcounters */
5862 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5863 KLCNT_MAX, pf_lcounter);
5864 if (error != 0)
5865 ERROUT(error);
5866
5867 /* fcounters */
5868 nvc = nvlist_create(0);
5869 if (nvc == NULL)
5870 ERROUT(ENOMEM);
5871
5872 for (int i = 0; i < FCNT_MAX; i++) {
5873 nvlist_append_number_array(nvc, "counters",
5874 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5875 nvlist_append_string_array(nvc, "names",
5876 pf_fcounter[i]);
5877 nvlist_append_number_array(nvc, "ids",
5878 i);
5879 }
5880 nvlist_add_nvlist(nvl, "fcounters", nvc);
5881 nvlist_destroy(nvc);
5882 nvc = NULL;
5883
5884 /* scounters */
5885 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5886 SCNT_MAX, pf_fcounter);
5887 if (error != 0)
5888 ERROUT(error);
5889
5890 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5891 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5892 PF_MD5_DIGEST_LENGTH);
5893
5894 pfi_update_status(V_pf_status.ifname, &s);
5895
5896 /* pcounters / bcounters */
5897 for (int i = 0; i < 2; i++) {
5898 for (int j = 0; j < 2; j++) {
5899 for (int k = 0; k < 2; k++) {
5900 nvlist_append_number_array(nvl, "pcounters",
5901 s.pcounters[i][j][k]);
5902 }
5903 nvlist_append_number_array(nvl, "bcounters",
5904 s.bcounters[i][j]);
5905 }
5906 }
5907
5908 nvlpacked = nvlist_pack(nvl, &nv->len);
5909 if (nvlpacked == NULL)
5910 ERROUT(ENOMEM);
5911
5912 if (nv->size == 0)
5913 ERROUT(0);
5914 else if (nv->size < nv->len)
5915 ERROUT(ENOSPC);
5916
5917 PF_RULES_RUNLOCK();
5918 error = copyout(nvlpacked, nv->data, nv->len);
5919 goto done;
5920
5921 #undef ERROUT
5922 errout:
5923 PF_RULES_RUNLOCK();
5924 done:
5925 free(nvlpacked, M_NVLIST);
5926 nvlist_destroy(nvc);
5927 nvlist_destroy(nvl);
5928
5929 return (error);
5930 }
5931
5932 /*
5933 * XXX - Check for version mismatch!!!
5934 */
5935 static void
pf_clear_all_states(void)5936 pf_clear_all_states(void)
5937 {
5938 struct epoch_tracker et;
5939 struct pf_kstate *s;
5940 u_int i;
5941
5942 NET_EPOCH_ENTER(et);
5943 for (i = 0; i <= V_pf_hashmask; i++) {
5944 struct pf_idhash *ih = &V_pf_idhash[i];
5945 relock:
5946 PF_HASHROW_LOCK(ih);
5947 LIST_FOREACH(s, &ih->states, entry) {
5948 s->timeout = PFTM_PURGE;
5949 /* Don't send out individual delete messages. */
5950 s->state_flags |= PFSTATE_NOSYNC;
5951 pf_unlink_state(s);
5952 goto relock;
5953 }
5954 PF_HASHROW_UNLOCK(ih);
5955 }
5956 NET_EPOCH_EXIT(et);
5957 }
5958
5959 static int
pf_clear_tables(void)5960 pf_clear_tables(void)
5961 {
5962 struct pfioc_table io;
5963 int error;
5964
5965 bzero(&io, sizeof(io));
5966 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5967
5968 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5969 io.pfrio_flags);
5970
5971 return (error);
5972 }
5973
5974 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)5975 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5976 {
5977 struct pf_ksrc_node_list kill;
5978 u_int killed;
5979
5980 LIST_INIT(&kill);
5981 for (int i = 0; i <= V_pf_srchashmask; i++) {
5982 struct pf_srchash *sh = &V_pf_srchash[i];
5983 struct pf_ksrc_node *sn, *tmp;
5984
5985 PF_HASHROW_LOCK(sh);
5986 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5987 if (psnk == NULL ||
5988 (PF_MATCHA(psnk->psnk_src.neg,
5989 &psnk->psnk_src.addr.v.a.addr,
5990 &psnk->psnk_src.addr.v.a.mask,
5991 &sn->addr, sn->af) &&
5992 PF_MATCHA(psnk->psnk_dst.neg,
5993 &psnk->psnk_dst.addr.v.a.addr,
5994 &psnk->psnk_dst.addr.v.a.mask,
5995 &sn->raddr, sn->af))) {
5996 pf_unlink_src_node(sn);
5997 LIST_INSERT_HEAD(&kill, sn, entry);
5998 sn->expire = 1;
5999 }
6000 PF_HASHROW_UNLOCK(sh);
6001 }
6002
6003 for (int i = 0; i <= V_pf_hashmask; i++) {
6004 struct pf_idhash *ih = &V_pf_idhash[i];
6005 struct pf_kstate *s;
6006
6007 PF_HASHROW_LOCK(ih);
6008 LIST_FOREACH(s, &ih->states, entry) {
6009 if (s->src_node && s->src_node->expire == 1)
6010 s->src_node = NULL;
6011 if (s->nat_src_node && s->nat_src_node->expire == 1)
6012 s->nat_src_node = NULL;
6013 }
6014 PF_HASHROW_UNLOCK(ih);
6015 }
6016
6017 killed = pf_free_src_nodes(&kill);
6018
6019 if (psnk != NULL)
6020 psnk->psnk_killed = killed;
6021 }
6022
6023 static int
pf_keepcounters(struct pfioc_nv * nv)6024 pf_keepcounters(struct pfioc_nv *nv)
6025 {
6026 nvlist_t *nvl = NULL;
6027 void *nvlpacked = NULL;
6028 int error = 0;
6029
6030 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6031
6032 if (nv->len > pf_ioctl_maxcount)
6033 ERROUT(ENOMEM);
6034
6035 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6036 error = copyin(nv->data, nvlpacked, nv->len);
6037 if (error)
6038 ERROUT(error);
6039
6040 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6041 if (nvl == NULL)
6042 ERROUT(EBADMSG);
6043
6044 if (! nvlist_exists_bool(nvl, "keep_counters"))
6045 ERROUT(EBADMSG);
6046
6047 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6048
6049 on_error:
6050 nvlist_destroy(nvl);
6051 free(nvlpacked, M_NVLIST);
6052 return (error);
6053 }
6054
6055 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6056 pf_clear_states(const struct pf_kstate_kill *kill)
6057 {
6058 struct pf_state_key_cmp match_key;
6059 struct pf_kstate *s;
6060 struct pfi_kkif *kif;
6061 int idx;
6062 unsigned int killed = 0, dir;
6063
6064 NET_EPOCH_ASSERT();
6065
6066 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6067 struct pf_idhash *ih = &V_pf_idhash[i];
6068
6069 relock_DIOCCLRSTATES:
6070 PF_HASHROW_LOCK(ih);
6071 LIST_FOREACH(s, &ih->states, entry) {
6072 /* For floating states look at the original kif. */
6073 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6074
6075 if (kill->psk_ifname[0] &&
6076 strcmp(kill->psk_ifname,
6077 kif->pfik_name))
6078 continue;
6079
6080 if (kill->psk_kill_match) {
6081 bzero(&match_key, sizeof(match_key));
6082
6083 if (s->direction == PF_OUT) {
6084 dir = PF_IN;
6085 idx = PF_SK_STACK;
6086 } else {
6087 dir = PF_OUT;
6088 idx = PF_SK_WIRE;
6089 }
6090
6091 match_key.af = s->key[idx]->af;
6092 match_key.proto = s->key[idx]->proto;
6093 PF_ACPY(&match_key.addr[0],
6094 &s->key[idx]->addr[1], match_key.af);
6095 match_key.port[0] = s->key[idx]->port[1];
6096 PF_ACPY(&match_key.addr[1],
6097 &s->key[idx]->addr[0], match_key.af);
6098 match_key.port[1] = s->key[idx]->port[0];
6099 }
6100
6101 /*
6102 * Don't send out individual
6103 * delete messages.
6104 */
6105 s->state_flags |= PFSTATE_NOSYNC;
6106 pf_unlink_state(s);
6107 killed++;
6108
6109 if (kill->psk_kill_match)
6110 killed += pf_kill_matching_state(&match_key,
6111 dir);
6112
6113 goto relock_DIOCCLRSTATES;
6114 }
6115 PF_HASHROW_UNLOCK(ih);
6116 }
6117
6118 if (V_pfsync_clear_states_ptr != NULL)
6119 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6120
6121 return (killed);
6122 }
6123
6124 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6125 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6126 {
6127 struct pf_kstate *s;
6128
6129 NET_EPOCH_ASSERT();
6130 if (kill->psk_pfcmp.id) {
6131 if (kill->psk_pfcmp.creatorid == 0)
6132 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6133 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6134 kill->psk_pfcmp.creatorid))) {
6135 pf_unlink_state(s);
6136 *killed = 1;
6137 }
6138 return;
6139 }
6140
6141 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6142 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6143 }
6144
6145 static int
pf_killstates_nv(struct pfioc_nv * nv)6146 pf_killstates_nv(struct pfioc_nv *nv)
6147 {
6148 struct pf_kstate_kill kill;
6149 struct epoch_tracker et;
6150 nvlist_t *nvl = NULL;
6151 void *nvlpacked = NULL;
6152 int error = 0;
6153 unsigned int killed = 0;
6154
6155 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6156
6157 if (nv->len > pf_ioctl_maxcount)
6158 ERROUT(ENOMEM);
6159
6160 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6161 error = copyin(nv->data, nvlpacked, nv->len);
6162 if (error)
6163 ERROUT(error);
6164
6165 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6166 if (nvl == NULL)
6167 ERROUT(EBADMSG);
6168
6169 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6170 if (error)
6171 ERROUT(error);
6172
6173 NET_EPOCH_ENTER(et);
6174 pf_killstates(&kill, &killed);
6175 NET_EPOCH_EXIT(et);
6176
6177 free(nvlpacked, M_NVLIST);
6178 nvlpacked = NULL;
6179 nvlist_destroy(nvl);
6180 nvl = nvlist_create(0);
6181 if (nvl == NULL)
6182 ERROUT(ENOMEM);
6183
6184 nvlist_add_number(nvl, "killed", killed);
6185
6186 nvlpacked = nvlist_pack(nvl, &nv->len);
6187 if (nvlpacked == NULL)
6188 ERROUT(ENOMEM);
6189
6190 if (nv->size == 0)
6191 ERROUT(0);
6192 else if (nv->size < nv->len)
6193 ERROUT(ENOSPC);
6194
6195 error = copyout(nvlpacked, nv->data, nv->len);
6196
6197 on_error:
6198 nvlist_destroy(nvl);
6199 free(nvlpacked, M_NVLIST);
6200 return (error);
6201 }
6202
6203 static int
pf_clearstates_nv(struct pfioc_nv * nv)6204 pf_clearstates_nv(struct pfioc_nv *nv)
6205 {
6206 struct pf_kstate_kill kill;
6207 struct epoch_tracker et;
6208 nvlist_t *nvl = NULL;
6209 void *nvlpacked = NULL;
6210 int error = 0;
6211 unsigned int killed;
6212
6213 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6214
6215 if (nv->len > pf_ioctl_maxcount)
6216 ERROUT(ENOMEM);
6217
6218 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6219 error = copyin(nv->data, nvlpacked, nv->len);
6220 if (error)
6221 ERROUT(error);
6222
6223 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6224 if (nvl == NULL)
6225 ERROUT(EBADMSG);
6226
6227 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6228 if (error)
6229 ERROUT(error);
6230
6231 NET_EPOCH_ENTER(et);
6232 killed = pf_clear_states(&kill);
6233 NET_EPOCH_EXIT(et);
6234
6235 free(nvlpacked, M_NVLIST);
6236 nvlpacked = NULL;
6237 nvlist_destroy(nvl);
6238 nvl = nvlist_create(0);
6239 if (nvl == NULL)
6240 ERROUT(ENOMEM);
6241
6242 nvlist_add_number(nvl, "killed", killed);
6243
6244 nvlpacked = nvlist_pack(nvl, &nv->len);
6245 if (nvlpacked == NULL)
6246 ERROUT(ENOMEM);
6247
6248 if (nv->size == 0)
6249 ERROUT(0);
6250 else if (nv->size < nv->len)
6251 ERROUT(ENOSPC);
6252
6253 error = copyout(nvlpacked, nv->data, nv->len);
6254
6255 #undef ERROUT
6256 on_error:
6257 nvlist_destroy(nvl);
6258 free(nvlpacked, M_NVLIST);
6259 return (error);
6260 }
6261
6262 static int
pf_getstate(struct pfioc_nv * nv)6263 pf_getstate(struct pfioc_nv *nv)
6264 {
6265 nvlist_t *nvl = NULL, *nvls;
6266 void *nvlpacked = NULL;
6267 struct pf_kstate *s = NULL;
6268 int error = 0;
6269 uint64_t id, creatorid;
6270
6271 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6272
6273 if (nv->len > pf_ioctl_maxcount)
6274 ERROUT(ENOMEM);
6275
6276 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6277 error = copyin(nv->data, nvlpacked, nv->len);
6278 if (error)
6279 ERROUT(error);
6280
6281 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6282 if (nvl == NULL)
6283 ERROUT(EBADMSG);
6284
6285 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6286 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6287
6288 s = pf_find_state_byid(id, creatorid);
6289 if (s == NULL)
6290 ERROUT(ENOENT);
6291
6292 free(nvlpacked, M_NVLIST);
6293 nvlpacked = NULL;
6294 nvlist_destroy(nvl);
6295 nvl = nvlist_create(0);
6296 if (nvl == NULL)
6297 ERROUT(ENOMEM);
6298
6299 nvls = pf_state_to_nvstate(s);
6300 if (nvls == NULL)
6301 ERROUT(ENOMEM);
6302
6303 nvlist_add_nvlist(nvl, "state", nvls);
6304 nvlist_destroy(nvls);
6305
6306 nvlpacked = nvlist_pack(nvl, &nv->len);
6307 if (nvlpacked == NULL)
6308 ERROUT(ENOMEM);
6309
6310 if (nv->size == 0)
6311 ERROUT(0);
6312 else if (nv->size < nv->len)
6313 ERROUT(ENOSPC);
6314
6315 error = copyout(nvlpacked, nv->data, nv->len);
6316
6317 #undef ERROUT
6318 errout:
6319 if (s != NULL)
6320 PF_STATE_UNLOCK(s);
6321 free(nvlpacked, M_NVLIST);
6322 nvlist_destroy(nvl);
6323 return (error);
6324 }
6325
6326 /*
6327 * XXX - Check for version mismatch!!!
6328 */
6329
6330 /*
6331 * Duplicate pfctl -Fa operation to get rid of as much as we can.
6332 */
6333 static int
shutdown_pf(void)6334 shutdown_pf(void)
6335 {
6336 int error = 0;
6337 u_int32_t t[5];
6338 char nn = '\0';
6339 struct pf_kanchor *anchor;
6340 struct pf_keth_anchor *eth_anchor;
6341 int rs_num;
6342
6343 do {
6344 /* Unlink rules of all user defined anchors */
6345 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6346 /* Wildcard based anchors may not have a respective
6347 * explicit anchor rule or they may be left empty
6348 * without rules. It leads to anchor.refcnt=0, and the
6349 * rest of the logic does not expect it. */
6350 if (anchor->refcnt == 0)
6351 anchor->refcnt = 1;
6352 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6353 if ((error = pf_begin_rules(&t[rs_num], rs_num,
6354 anchor->path)) != 0) {
6355 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6356 "anchor.path=%s rs_num=%d\n",
6357 anchor->path, rs_num));
6358 goto error; /* XXX: rollback? */
6359 }
6360 }
6361 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6362 error = pf_commit_rules(t[rs_num], rs_num,
6363 anchor->path);
6364 MPASS(error == 0);
6365 }
6366 }
6367
6368 /* Unlink rules of all user defined ether anchors */
6369 RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6370 &V_pf_keth_anchors) {
6371 /* Wildcard based anchors may not have a respective
6372 * explicit anchor rule or they may be left empty
6373 * without rules. It leads to anchor.refcnt=0, and the
6374 * rest of the logic does not expect it. */
6375 if (eth_anchor->refcnt == 0)
6376 eth_anchor->refcnt = 1;
6377 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6378 != 0) {
6379 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6380 "anchor.path=%s\n", eth_anchor->path));
6381 goto error;
6382 }
6383 error = pf_commit_eth(t[0], eth_anchor->path);
6384 MPASS(error == 0);
6385 }
6386
6387 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6388 != 0) {
6389 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6390 break;
6391 }
6392 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6393 != 0) {
6394 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6395 break; /* XXX: rollback? */
6396 }
6397 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6398 != 0) {
6399 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6400 break; /* XXX: rollback? */
6401 }
6402 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6403 != 0) {
6404 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6405 break; /* XXX: rollback? */
6406 }
6407 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6408 != 0) {
6409 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6410 break; /* XXX: rollback? */
6411 }
6412
6413 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6414 MPASS(error == 0);
6415 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6416 MPASS(error == 0);
6417 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6418 MPASS(error == 0);
6419 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6420 MPASS(error == 0);
6421 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6422 MPASS(error == 0);
6423
6424 if ((error = pf_clear_tables()) != 0)
6425 break;
6426
6427 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6428 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6429 break;
6430 }
6431 error = pf_commit_eth(t[0], &nn);
6432 MPASS(error == 0);
6433
6434 #ifdef ALTQ
6435 if ((error = pf_begin_altq(&t[0])) != 0) {
6436 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6437 break;
6438 }
6439 pf_commit_altq(t[0]);
6440 #endif
6441
6442 pf_clear_all_states();
6443
6444 pf_kill_srcnodes(NULL);
6445
6446 /* status does not use malloced mem so no need to cleanup */
6447 /* fingerprints and interfaces have their own cleanup code */
6448 } while(0);
6449
6450 error:
6451 return (error);
6452 }
6453
6454 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6455 pf_check_return(int chk, struct mbuf **m)
6456 {
6457
6458 switch (chk) {
6459 case PF_PASS:
6460 if (*m == NULL)
6461 return (PFIL_CONSUMED);
6462 else
6463 return (PFIL_PASS);
6464 break;
6465 default:
6466 if (*m != NULL) {
6467 m_freem(*m);
6468 *m = NULL;
6469 }
6470 return (PFIL_DROPPED);
6471 }
6472 }
6473
6474 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6475 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6476 void *ruleset __unused, struct inpcb *inp)
6477 {
6478 int chk;
6479
6480 CURVNET_ASSERT_SET();
6481
6482 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6483
6484 return (pf_check_return(chk, m));
6485 }
6486
6487 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6488 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6489 void *ruleset __unused, struct inpcb *inp)
6490 {
6491 int chk;
6492
6493 CURVNET_ASSERT_SET();
6494
6495 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6496
6497 return (pf_check_return(chk, m));
6498 }
6499
6500 #ifdef INET
6501 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6502 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6503 void *ruleset __unused, struct inpcb *inp)
6504 {
6505 int chk;
6506
6507 CURVNET_ASSERT_SET();
6508
6509 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
6510
6511 return (pf_check_return(chk, m));
6512 }
6513
6514 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6515 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6516 void *ruleset __unused, struct inpcb *inp)
6517 {
6518 int chk;
6519
6520 CURVNET_ASSERT_SET();
6521
6522 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
6523
6524 return (pf_check_return(chk, m));
6525 }
6526 #endif
6527
6528 #ifdef INET6
6529 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6530 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6531 void *ruleset __unused, struct inpcb *inp)
6532 {
6533 int chk;
6534
6535 CURVNET_ASSERT_SET();
6536
6537 /*
6538 * In case of loopback traffic IPv6 uses the real interface in
6539 * order to support scoped addresses. In order to support stateful
6540 * filtering we have change this to lo0 as it is the case in IPv4.
6541 */
6542 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6543 m, inp, NULL);
6544
6545 return (pf_check_return(chk, m));
6546 }
6547
6548 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6549 pf_check6_out(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_INET6, PF_OUT, flags, ifp, m, inp, NULL);
6557
6558 return (pf_check_return(chk, m));
6559 }
6560 #endif /* INET6 */
6561
6562 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6563 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6564 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
6565 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
6566
6567 #ifdef INET
6568 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6569 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6570 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
6571 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
6572 #endif
6573 #ifdef INET6
6574 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6575 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6576 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
6577 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
6578 #endif
6579
6580 static void
hook_pf_eth(void)6581 hook_pf_eth(void)
6582 {
6583 struct pfil_hook_args pha = {
6584 .pa_version = PFIL_VERSION,
6585 .pa_modname = "pf",
6586 .pa_type = PFIL_TYPE_ETHERNET,
6587 };
6588 struct pfil_link_args pla = {
6589 .pa_version = PFIL_VERSION,
6590 };
6591 int ret __diagused;
6592
6593 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6594 return;
6595
6596 pha.pa_mbuf_chk = pf_eth_check_in;
6597 pha.pa_flags = PFIL_IN;
6598 pha.pa_rulname = "eth-in";
6599 V_pf_eth_in_hook = pfil_add_hook(&pha);
6600 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6601 pla.pa_head = V_link_pfil_head;
6602 pla.pa_hook = V_pf_eth_in_hook;
6603 ret = pfil_link(&pla);
6604 MPASS(ret == 0);
6605 pha.pa_mbuf_chk = pf_eth_check_out;
6606 pha.pa_flags = PFIL_OUT;
6607 pha.pa_rulname = "eth-out";
6608 V_pf_eth_out_hook = pfil_add_hook(&pha);
6609 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6610 pla.pa_head = V_link_pfil_head;
6611 pla.pa_hook = V_pf_eth_out_hook;
6612 ret = pfil_link(&pla);
6613 MPASS(ret == 0);
6614
6615 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6616 }
6617
6618 static void
hook_pf(void)6619 hook_pf(void)
6620 {
6621 struct pfil_hook_args pha = {
6622 .pa_version = PFIL_VERSION,
6623 .pa_modname = "pf",
6624 };
6625 struct pfil_link_args pla = {
6626 .pa_version = PFIL_VERSION,
6627 };
6628 int ret __diagused;
6629
6630 if (atomic_load_bool(&V_pf_pfil_hooked))
6631 return;
6632
6633 #ifdef INET
6634 pha.pa_type = PFIL_TYPE_IP4;
6635 pha.pa_mbuf_chk = pf_check_in;
6636 pha.pa_flags = PFIL_IN;
6637 pha.pa_rulname = "default-in";
6638 V_pf_ip4_in_hook = pfil_add_hook(&pha);
6639 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6640 pla.pa_head = V_inet_pfil_head;
6641 pla.pa_hook = V_pf_ip4_in_hook;
6642 ret = pfil_link(&pla);
6643 MPASS(ret == 0);
6644 pha.pa_mbuf_chk = pf_check_out;
6645 pha.pa_flags = PFIL_OUT;
6646 pha.pa_rulname = "default-out";
6647 V_pf_ip4_out_hook = pfil_add_hook(&pha);
6648 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6649 pla.pa_head = V_inet_pfil_head;
6650 pla.pa_hook = V_pf_ip4_out_hook;
6651 ret = pfil_link(&pla);
6652 MPASS(ret == 0);
6653 if (V_pf_filter_local) {
6654 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6655 pla.pa_head = V_inet_local_pfil_head;
6656 pla.pa_hook = V_pf_ip4_out_hook;
6657 ret = pfil_link(&pla);
6658 MPASS(ret == 0);
6659 }
6660 #endif
6661 #ifdef INET6
6662 pha.pa_type = PFIL_TYPE_IP6;
6663 pha.pa_mbuf_chk = pf_check6_in;
6664 pha.pa_flags = PFIL_IN;
6665 pha.pa_rulname = "default-in6";
6666 V_pf_ip6_in_hook = pfil_add_hook(&pha);
6667 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6668 pla.pa_head = V_inet6_pfil_head;
6669 pla.pa_hook = V_pf_ip6_in_hook;
6670 ret = pfil_link(&pla);
6671 MPASS(ret == 0);
6672 pha.pa_mbuf_chk = pf_check6_out;
6673 pha.pa_rulname = "default-out6";
6674 pha.pa_flags = PFIL_OUT;
6675 V_pf_ip6_out_hook = pfil_add_hook(&pha);
6676 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6677 pla.pa_head = V_inet6_pfil_head;
6678 pla.pa_hook = V_pf_ip6_out_hook;
6679 ret = pfil_link(&pla);
6680 MPASS(ret == 0);
6681 if (V_pf_filter_local) {
6682 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6683 pla.pa_head = V_inet6_local_pfil_head;
6684 pla.pa_hook = V_pf_ip6_out_hook;
6685 ret = pfil_link(&pla);
6686 MPASS(ret == 0);
6687 }
6688 #endif
6689
6690 atomic_store_bool(&V_pf_pfil_hooked, true);
6691 }
6692
6693 static void
dehook_pf_eth(void)6694 dehook_pf_eth(void)
6695 {
6696
6697 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6698 return;
6699
6700 pfil_remove_hook(V_pf_eth_in_hook);
6701 pfil_remove_hook(V_pf_eth_out_hook);
6702
6703 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6704 }
6705
6706 static void
dehook_pf(void)6707 dehook_pf(void)
6708 {
6709
6710 if (!atomic_load_bool(&V_pf_pfil_hooked))
6711 return;
6712
6713 #ifdef INET
6714 pfil_remove_hook(V_pf_ip4_in_hook);
6715 pfil_remove_hook(V_pf_ip4_out_hook);
6716 #endif
6717 #ifdef INET6
6718 pfil_remove_hook(V_pf_ip6_in_hook);
6719 pfil_remove_hook(V_pf_ip6_out_hook);
6720 #endif
6721
6722 atomic_store_bool(&V_pf_pfil_hooked, false);
6723 }
6724
6725 static void
pf_load_vnet(void)6726 pf_load_vnet(void)
6727 {
6728 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6729 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6730
6731 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6732 sx_init(&V_pf_ioctl_lock, "pf ioctl");
6733
6734 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6735 PF_RULE_TAG_HASH_SIZE_DEFAULT);
6736 #ifdef ALTQ
6737 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6738 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6739 #endif
6740
6741 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6742
6743 pfattach_vnet();
6744 V_pf_vnet_active = 1;
6745 }
6746
6747 static int
pf_load(void)6748 pf_load(void)
6749 {
6750 int error;
6751
6752 sx_init(&pf_end_lock, "pf end thread");
6753
6754 pf_mtag_initialize();
6755
6756 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6757 if (pf_dev == NULL)
6758 return (ENOMEM);
6759
6760 pf_end_threads = 0;
6761 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6762 if (error != 0)
6763 return (error);
6764
6765 pfi_initialize();
6766
6767 return (0);
6768 }
6769
6770 static void
pf_unload_vnet(void)6771 pf_unload_vnet(void)
6772 {
6773 int ret __diagused;
6774
6775 V_pf_vnet_active = 0;
6776 V_pf_status.running = 0;
6777 dehook_pf();
6778 dehook_pf_eth();
6779
6780 PF_RULES_WLOCK();
6781 pf_syncookies_cleanup();
6782 shutdown_pf();
6783 PF_RULES_WUNLOCK();
6784
6785 ret = swi_remove(V_pf_swi_cookie);
6786 MPASS(ret == 0);
6787 ret = intr_event_destroy(V_pf_swi_ie);
6788 MPASS(ret == 0);
6789
6790 pf_unload_vnet_purge();
6791
6792 pf_normalize_cleanup();
6793 PF_RULES_WLOCK();
6794 pfi_cleanup_vnet();
6795 PF_RULES_WUNLOCK();
6796 pfr_cleanup();
6797 pf_osfp_flush();
6798 pf_cleanup();
6799 if (IS_DEFAULT_VNET(curvnet))
6800 pf_mtag_cleanup();
6801
6802 pf_cleanup_tagset(&V_pf_tags);
6803 #ifdef ALTQ
6804 pf_cleanup_tagset(&V_pf_qids);
6805 #endif
6806 uma_zdestroy(V_pf_tag_z);
6807
6808 #ifdef PF_WANT_32_TO_64_COUNTER
6809 PF_RULES_WLOCK();
6810 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6811
6812 MPASS(LIST_EMPTY(&V_pf_allkiflist));
6813 MPASS(V_pf_allkifcount == 0);
6814
6815 LIST_REMOVE(&V_pf_default_rule, allrulelist);
6816 V_pf_allrulecount--;
6817 LIST_REMOVE(V_pf_rulemarker, allrulelist);
6818
6819 MPASS(LIST_EMPTY(&V_pf_allrulelist));
6820 MPASS(V_pf_allrulecount == 0);
6821
6822 PF_RULES_WUNLOCK();
6823
6824 free(V_pf_kifmarker, PFI_MTYPE);
6825 free(V_pf_rulemarker, M_PFRULE);
6826 #endif
6827
6828 /* Free counters last as we updated them during shutdown. */
6829 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6830 for (int i = 0; i < 2; i++) {
6831 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6832 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6833 }
6834 counter_u64_free(V_pf_default_rule.states_cur);
6835 counter_u64_free(V_pf_default_rule.states_tot);
6836 counter_u64_free(V_pf_default_rule.src_nodes);
6837 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6838
6839 for (int i = 0; i < PFRES_MAX; i++)
6840 counter_u64_free(V_pf_status.counters[i]);
6841 for (int i = 0; i < KLCNT_MAX; i++)
6842 counter_u64_free(V_pf_status.lcounters[i]);
6843 for (int i = 0; i < FCNT_MAX; i++)
6844 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6845 for (int i = 0; i < SCNT_MAX; i++)
6846 counter_u64_free(V_pf_status.scounters[i]);
6847
6848 rm_destroy(&V_pf_rules_lock);
6849 sx_destroy(&V_pf_ioctl_lock);
6850 }
6851
6852 static void
pf_unload(void)6853 pf_unload(void)
6854 {
6855
6856 sx_xlock(&pf_end_lock);
6857 pf_end_threads = 1;
6858 while (pf_end_threads < 2) {
6859 wakeup_one(pf_purge_thread);
6860 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6861 }
6862 sx_xunlock(&pf_end_lock);
6863
6864 pf_nl_unregister();
6865
6866 if (pf_dev != NULL)
6867 destroy_dev(pf_dev);
6868
6869 pfi_cleanup();
6870
6871 sx_destroy(&pf_end_lock);
6872 }
6873
6874 static void
vnet_pf_init(void * unused __unused)6875 vnet_pf_init(void *unused __unused)
6876 {
6877
6878 pf_load_vnet();
6879 }
6880 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6881 vnet_pf_init, NULL);
6882
6883 static void
vnet_pf_uninit(const void * unused __unused)6884 vnet_pf_uninit(const void *unused __unused)
6885 {
6886
6887 pf_unload_vnet();
6888 }
6889 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6890 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6891 vnet_pf_uninit, NULL);
6892
6893 static int
pf_modevent(module_t mod,int type,void * data)6894 pf_modevent(module_t mod, int type, void *data)
6895 {
6896 int error = 0;
6897
6898 switch(type) {
6899 case MOD_LOAD:
6900 error = pf_load();
6901 pf_nl_register();
6902 break;
6903 case MOD_UNLOAD:
6904 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6905 * the vnet_pf_uninit()s */
6906 break;
6907 default:
6908 error = EINVAL;
6909 break;
6910 }
6911
6912 return (error);
6913 }
6914
6915 static moduledata_t pf_mod = {
6916 "pf",
6917 pf_modevent,
6918 0
6919 };
6920
6921 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6922 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6923 MODULE_VERSION(pf, PF_MODVER);
6924