1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2001 Daniel Hartmeier
5 * Copyright (c) 2002,2003 Henning Brauer
6 * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
13 * - Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * - Redistributions in binary form must reproduce the above
16 * copyright notice, this list of conditions and the following
17 * disclaimer in the documentation and/or other materials provided
18 * with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 * POSSIBILITY OF SUCH DAMAGE.
32 *
33 * Effort sponsored in part by the Defense Advanced Research Projects
34 * Agency (DARPA) and Air Force Research Laboratory, Air Force
35 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36 *
37 * $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38 */
39
40 #include <sys/cdefs.h>
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45
46 #include <sys/param.h>
47 #include <sys/_bitset.h>
48 #include <sys/bitset.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/hash.h>
55 #include <sys/interrupt.h>
56 #include <sys/jail.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/lock.h>
60 #include <sys/mbuf.h>
61 #include <sys/module.h>
62 #include <sys/nv.h>
63 #include <sys/proc.h>
64 #include <sys/sdt.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/if_private.h>
74 #include <net/vnet.h>
75 #include <net/route.h>
76 #include <net/pfil.h>
77 #include <net/pfvar.h>
78 #include <net/if_pfsync.h>
79 #include <net/if_pflog.h>
80
81 #include <netinet/in.h>
82 #include <netinet/ip.h>
83 #include <netinet/ip_var.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet/ip_icmp.h>
86 #include <netpfil/pf/pf_nl.h>
87 #include <netpfil/pf/pf_nv.h>
88
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101
102 static struct pf_kpool *pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 u_int32_t, u_int8_t, u_int8_t, u_int8_t, int);
104
105 static void pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void pf_empty_kpool(struct pf_kpalist *);
107 static int pfioctl(struct cdev *, u_long, caddr_t, int,
108 struct thread *);
109 static int pf_begin_eth(uint32_t *, const char *);
110 static int pf_rollback_eth(uint32_t, const char *);
111 static int pf_commit_eth(uint32_t, const char *);
112 static void pf_free_eth_rule(struct pf_keth_rule *);
113 #ifdef ALTQ
114 static int pf_begin_altq(u_int32_t *);
115 static int pf_rollback_altq(u_int32_t);
116 static int pf_commit_altq(u_int32_t);
117 static int pf_enable_altq(struct pf_altq *);
118 static int pf_disable_altq(struct pf_altq *);
119 static uint16_t pf_qname2qid(const char *);
120 static void pf_qid_unref(uint16_t);
121 #endif /* ALTQ */
122 static int pf_begin_rules(u_int32_t *, int, const char *);
123 static int pf_rollback_rules(u_int32_t, int, char *);
124 static int pf_setup_pfsync_matching(struct pf_kruleset *);
125 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
126 static void pf_hash_rule(struct pf_krule *);
127 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
128 static int pf_commit_rules(u_int32_t, int, char *);
129 static int pf_addr_setup(struct pf_kruleset *,
130 struct pf_addr_wrap *, sa_family_t);
131 static void pf_src_node_copy(const struct pf_ksrc_node *,
132 struct pf_src_node *);
133 #ifdef ALTQ
134 static int pf_export_kaltq(struct pf_altq *,
135 struct pfioc_altq_v1 *, size_t);
136 static int pf_import_kaltq(struct pfioc_altq_v1 *,
137 struct pf_altq *, size_t);
138 #endif /* ALTQ */
139
140 VNET_DEFINE(struct pf_krule, pf_default_rule);
141
142 static __inline int pf_krule_compare(struct pf_krule *,
143 struct pf_krule *);
144
145 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
146
147 #ifdef ALTQ
148 VNET_DEFINE_STATIC(int, pf_altq_running);
149 #define V_pf_altq_running VNET(pf_altq_running)
150 #endif
151
152 #define TAGID_MAX 50000
153 struct pf_tagname {
154 TAILQ_ENTRY(pf_tagname) namehash_entries;
155 TAILQ_ENTRY(pf_tagname) taghash_entries;
156 char name[PF_TAG_NAME_SIZE];
157 uint16_t tag;
158 int ref;
159 };
160
161 struct pf_tagset {
162 TAILQ_HEAD(, pf_tagname) *namehash;
163 TAILQ_HEAD(, pf_tagname) *taghash;
164 unsigned int mask;
165 uint32_t seed;
166 BITSET_DEFINE(, TAGID_MAX) avail;
167 };
168
169 VNET_DEFINE(struct pf_tagset, pf_tags);
170 #define V_pf_tags VNET(pf_tags)
171 static unsigned int pf_rule_tag_hashsize;
172 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128
173 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
174 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
175 "Size of pf(4) rule tag hashtable");
176
177 #ifdef ALTQ
178 VNET_DEFINE(struct pf_tagset, pf_qids);
179 #define V_pf_qids VNET(pf_qids)
180 static unsigned int pf_queue_tag_hashsize;
181 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128
182 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
183 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
184 "Size of pf(4) queue tag hashtable");
185 #endif
186 VNET_DEFINE(uma_zone_t, pf_tag_z);
187 #define V_pf_tag_z VNET(pf_tag_z)
188 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
189 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
190
191 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
192 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
193 #endif
194
195 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
196 #define V_pf_filter_local VNET(pf_filter_local)
197 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
198 &VNET_NAME(pf_filter_local), false,
199 "Enable filtering for packets delivered to local network stack");
200
201 #ifdef PF_DEFAULT_TO_DROP
202 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
203 #else
204 VNET_DEFINE_STATIC(bool, default_to_drop);
205 #endif
206 #define V_default_to_drop VNET(default_to_drop)
207 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
208 &VNET_NAME(default_to_drop), false,
209 "Make the default rule drop all packets.");
210
211 static void pf_init_tagset(struct pf_tagset *, unsigned int *,
212 unsigned int);
213 static void pf_cleanup_tagset(struct pf_tagset *);
214 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *);
215 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t);
216 static u_int16_t tagname2tag(struct pf_tagset *, const char *);
217 static u_int16_t pf_tagname2tag(const char *);
218 static void tag_unref(struct pf_tagset *, u_int16_t);
219
220 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
221
222 struct cdev *pf_dev;
223
224 /*
225 * XXX - These are new and need to be checked when moveing to a new version
226 */
227 static void pf_clear_all_states(void);
228 static int pf_killstates_row(struct pf_kstate_kill *,
229 struct pf_idhash *);
230 static int pf_killstates_nv(struct pfioc_nv *);
231 static int pf_clearstates_nv(struct pfioc_nv *);
232 static int pf_getstate(struct pfioc_nv *);
233 static int pf_getstatus(struct pfioc_nv *);
234 static int pf_clear_tables(void);
235 static void pf_kill_srcnodes(struct pfioc_src_node_kill *);
236 static int pf_keepcounters(struct pfioc_nv *);
237 static void pf_tbladdr_copyout(struct pf_addr_wrap *);
238
239 /*
240 * Wrapper functions for pfil(9) hooks
241 */
242 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
243 int flags, void *ruleset __unused, struct inpcb *inp);
244 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
245 int flags, void *ruleset __unused, struct inpcb *inp);
246 #ifdef INET
247 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
248 int flags, void *ruleset __unused, struct inpcb *inp);
249 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
250 int flags, void *ruleset __unused, struct inpcb *inp);
251 #endif
252 #ifdef INET6
253 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
254 int flags, void *ruleset __unused, struct inpcb *inp);
255 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
256 int flags, void *ruleset __unused, struct inpcb *inp);
257 #endif
258
259 static void hook_pf_eth(void);
260 static void hook_pf(void);
261 static void dehook_pf_eth(void);
262 static void dehook_pf(void);
263 static int shutdown_pf(void);
264 static int pf_load(void);
265 static void pf_unload(void);
266
267 static struct cdevsw pf_cdevsw = {
268 .d_ioctl = pfioctl,
269 .d_name = PF_NAME,
270 .d_version = D_VERSION,
271 };
272
273 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
274 #define V_pf_pfil_hooked VNET(pf_pfil_hooked)
275 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
276 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked)
277
278 /*
279 * We need a flag that is neither hooked nor running to know when
280 * the VNET is "valid". We primarily need this to control (global)
281 * external event, e.g., eventhandlers.
282 */
283 VNET_DEFINE(int, pf_vnet_active);
284 #define V_pf_vnet_active VNET(pf_vnet_active)
285
286 int pf_end_threads;
287 struct proc *pf_purge_proc;
288
289 VNET_DEFINE(struct rmlock, pf_rules_lock);
290 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
291 #define V_pf_ioctl_lock VNET(pf_ioctl_lock)
292 struct sx pf_end_lock;
293
294 /* pfsync */
295 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
296 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
297 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
298 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
299 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
300 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
301 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
302 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
303
304 /* pflog */
305 pflog_packet_t *pflog_packet_ptr = NULL;
306
307 /*
308 * Copy a user-provided string, returning an error if truncation would occur.
309 * Avoid scanning past "sz" bytes in the source string since there's no
310 * guarantee that it's nul-terminated.
311 */
312 static int
pf_user_strcpy(char * dst,const char * src,size_t sz)313 pf_user_strcpy(char *dst, const char *src, size_t sz)
314 {
315 if (strnlen(src, sz) == sz)
316 return (EINVAL);
317 (void)strlcpy(dst, src, sz);
318 return (0);
319 }
320
321 static void
pfattach_vnet(void)322 pfattach_vnet(void)
323 {
324 u_int32_t *my_timeout = V_pf_default_rule.timeout;
325
326 bzero(&V_pf_status, sizeof(V_pf_status));
327
328 pf_initialize();
329 pfr_initialize();
330 pfi_initialize_vnet();
331 pf_normalize_init();
332 pf_syncookies_init();
333
334 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
335 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
336
337 RB_INIT(&V_pf_anchors);
338 pf_init_kruleset(&pf_main_ruleset);
339
340 pf_init_keth(V_pf_keth);
341
342 /* default rule should never be garbage collected */
343 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
344 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
345 V_pf_default_rule.nr = (uint32_t)-1;
346 V_pf_default_rule.rtableid = -1;
347
348 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
349 for (int i = 0; i < 2; i++) {
350 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
351 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
352 }
353 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
354 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
355 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++)
356 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
357
358 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
359 M_WAITOK | M_ZERO);
360
361 #ifdef PF_WANT_32_TO_64_COUNTER
362 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
363 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
364 PF_RULES_WLOCK();
365 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
366 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
367 V_pf_allrulecount++;
368 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
369 PF_RULES_WUNLOCK();
370 #endif
371
372 /* initialize default timeouts */
373 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
374 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
375 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
376 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
377 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
378 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
379 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
380 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
381 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
382 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
383 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
384 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
385 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
386 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
387 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
388 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
389 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
390 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
391 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
392 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
393 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
394 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
395 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
396 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
397 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
398
399 V_pf_status.debug = PF_DEBUG_URGENT;
400 /*
401 * XXX This is different than in OpenBSD where reassembly is enabled by
402 * defult. In FreeBSD we expect people to still use scrub rules and
403 * switch to the new syntax later. Only when they switch they must
404 * explicitly enable reassemle. We could change the default once the
405 * scrub rule functionality is hopefully removed some day in future.
406 */
407 V_pf_status.reass = 0;
408
409 V_pf_pfil_hooked = false;
410 V_pf_pfil_eth_hooked = false;
411
412 /* XXX do our best to avoid a conflict */
413 V_pf_status.hostid = arc4random();
414
415 for (int i = 0; i < PFRES_MAX; i++)
416 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
417 for (int i = 0; i < KLCNT_MAX; i++)
418 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
419 for (int i = 0; i < FCNT_MAX; i++)
420 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
421 for (int i = 0; i < SCNT_MAX; i++)
422 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
423
424 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
425 INTR_MPSAFE, &V_pf_swi_cookie) != 0)
426 /* XXXGL: leaked all above. */
427 return;
428 }
429
430 static struct pf_kpool *
pf_get_kpool(const char * anchor,u_int32_t ticket,u_int8_t rule_action,u_int32_t rule_number,u_int8_t r_last,u_int8_t active,u_int8_t check_ticket,int which)431 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
432 u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
433 u_int8_t check_ticket, int which)
434 {
435 struct pf_kruleset *ruleset;
436 struct pf_krule *rule;
437 int rs_num;
438
439 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT);
440
441 ruleset = pf_find_kruleset(anchor);
442 if (ruleset == NULL)
443 return (NULL);
444 rs_num = pf_get_ruleset_number(rule_action);
445 if (rs_num >= PF_RULESET_MAX)
446 return (NULL);
447 if (active) {
448 if (check_ticket && ticket !=
449 ruleset->rules[rs_num].active.ticket)
450 return (NULL);
451 if (r_last)
452 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
453 pf_krulequeue);
454 else
455 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
456 } else {
457 if (check_ticket && ticket !=
458 ruleset->rules[rs_num].inactive.ticket)
459 return (NULL);
460 if (r_last)
461 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
462 pf_krulequeue);
463 else
464 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
465 }
466 if (!r_last) {
467 while ((rule != NULL) && (rule->nr != rule_number))
468 rule = TAILQ_NEXT(rule, entries);
469 }
470 if (rule == NULL)
471 return (NULL);
472
473 switch (which) {
474 case PF_RDR:
475 return (&rule->rdr);
476 case PF_NAT:
477 return (&rule->nat);
478 case PF_RT:
479 return (&rule->route);
480 default:
481 panic("Unknow pool type %d", which);
482 }
483 }
484
485 static void
pf_mv_kpool(struct pf_kpalist * poola,struct pf_kpalist * poolb)486 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
487 {
488 struct pf_kpooladdr *mv_pool_pa;
489
490 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
491 TAILQ_REMOVE(poola, mv_pool_pa, entries);
492 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
493 }
494 }
495
496 static void
pf_empty_kpool(struct pf_kpalist * poola)497 pf_empty_kpool(struct pf_kpalist *poola)
498 {
499 struct pf_kpooladdr *pa;
500
501 while ((pa = TAILQ_FIRST(poola)) != NULL) {
502 switch (pa->addr.type) {
503 case PF_ADDR_DYNIFTL:
504 pfi_dynaddr_remove(pa->addr.p.dyn);
505 break;
506 case PF_ADDR_TABLE:
507 /* XXX: this could be unfinished pooladdr on pabuf */
508 if (pa->addr.p.tbl != NULL)
509 pfr_detach_table(pa->addr.p.tbl);
510 break;
511 }
512 if (pa->kif)
513 pfi_kkif_unref(pa->kif);
514 TAILQ_REMOVE(poola, pa, entries);
515 free(pa, M_PFRULE);
516 }
517 }
518
519 static void
pf_unlink_rule_locked(struct pf_krulequeue * rulequeue,struct pf_krule * rule)520 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
521 {
522
523 PF_RULES_WASSERT();
524 PF_UNLNKDRULES_ASSERT();
525
526 TAILQ_REMOVE(rulequeue, rule, entries);
527
528 rule->rule_ref |= PFRULE_REFS;
529 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
530 }
531
532 static void
pf_unlink_rule(struct pf_krulequeue * rulequeue,struct pf_krule * rule)533 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
534 {
535
536 PF_RULES_WASSERT();
537
538 PF_UNLNKDRULES_LOCK();
539 pf_unlink_rule_locked(rulequeue, rule);
540 PF_UNLNKDRULES_UNLOCK();
541 }
542
543 static void
pf_free_eth_rule(struct pf_keth_rule * rule)544 pf_free_eth_rule(struct pf_keth_rule *rule)
545 {
546 PF_RULES_WASSERT();
547
548 if (rule == NULL)
549 return;
550
551 if (rule->tag)
552 tag_unref(&V_pf_tags, rule->tag);
553 if (rule->match_tag)
554 tag_unref(&V_pf_tags, rule->match_tag);
555 #ifdef ALTQ
556 pf_qid_unref(rule->qid);
557 #endif
558
559 if (rule->bridge_to)
560 pfi_kkif_unref(rule->bridge_to);
561 if (rule->kif)
562 pfi_kkif_unref(rule->kif);
563
564 if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
565 pfr_detach_table(rule->ipsrc.addr.p.tbl);
566 if (rule->ipdst.addr.type == PF_ADDR_TABLE)
567 pfr_detach_table(rule->ipdst.addr.p.tbl);
568
569 counter_u64_free(rule->evaluations);
570 for (int i = 0; i < 2; i++) {
571 counter_u64_free(rule->packets[i]);
572 counter_u64_free(rule->bytes[i]);
573 }
574 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
575 pf_keth_anchor_remove(rule);
576
577 free(rule, M_PFRULE);
578 }
579
580 void
pf_free_rule(struct pf_krule * rule)581 pf_free_rule(struct pf_krule *rule)
582 {
583
584 PF_RULES_WASSERT();
585 PF_CONFIG_ASSERT();
586
587 if (rule->tag)
588 tag_unref(&V_pf_tags, rule->tag);
589 if (rule->match_tag)
590 tag_unref(&V_pf_tags, rule->match_tag);
591 #ifdef ALTQ
592 if (rule->pqid != rule->qid)
593 pf_qid_unref(rule->pqid);
594 pf_qid_unref(rule->qid);
595 #endif
596 switch (rule->src.addr.type) {
597 case PF_ADDR_DYNIFTL:
598 pfi_dynaddr_remove(rule->src.addr.p.dyn);
599 break;
600 case PF_ADDR_TABLE:
601 pfr_detach_table(rule->src.addr.p.tbl);
602 break;
603 }
604 switch (rule->dst.addr.type) {
605 case PF_ADDR_DYNIFTL:
606 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
607 break;
608 case PF_ADDR_TABLE:
609 pfr_detach_table(rule->dst.addr.p.tbl);
610 break;
611 }
612 if (rule->overload_tbl)
613 pfr_detach_table(rule->overload_tbl);
614 if (rule->kif)
615 pfi_kkif_unref(rule->kif);
616 if (rule->rcv_kif)
617 pfi_kkif_unref(rule->rcv_kif);
618 pf_kanchor_remove(rule);
619 pf_empty_kpool(&rule->rdr.list);
620 pf_empty_kpool(&rule->nat.list);
621 pf_empty_kpool(&rule->route.list);
622
623 pf_krule_free(rule);
624 }
625
626 static void
pf_init_tagset(struct pf_tagset * ts,unsigned int * tunable_size,unsigned int default_size)627 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
628 unsigned int default_size)
629 {
630 unsigned int i;
631 unsigned int hashsize;
632
633 if (*tunable_size == 0 || !powerof2(*tunable_size))
634 *tunable_size = default_size;
635
636 hashsize = *tunable_size;
637 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
638 M_WAITOK);
639 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
640 M_WAITOK);
641 ts->mask = hashsize - 1;
642 ts->seed = arc4random();
643 for (i = 0; i < hashsize; i++) {
644 TAILQ_INIT(&ts->namehash[i]);
645 TAILQ_INIT(&ts->taghash[i]);
646 }
647 BIT_FILL(TAGID_MAX, &ts->avail);
648 }
649
650 static void
pf_cleanup_tagset(struct pf_tagset * ts)651 pf_cleanup_tagset(struct pf_tagset *ts)
652 {
653 unsigned int i;
654 unsigned int hashsize;
655 struct pf_tagname *t, *tmp;
656
657 /*
658 * Only need to clean up one of the hashes as each tag is hashed
659 * into each table.
660 */
661 hashsize = ts->mask + 1;
662 for (i = 0; i < hashsize; i++)
663 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
664 uma_zfree(V_pf_tag_z, t);
665
666 free(ts->namehash, M_PFHASH);
667 free(ts->taghash, M_PFHASH);
668 }
669
670 static uint16_t
tagname2hashindex(const struct pf_tagset * ts,const char * tagname)671 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
672 {
673 size_t len;
674
675 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
676 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
677 }
678
679 static uint16_t
tag2hashindex(const struct pf_tagset * ts,uint16_t tag)680 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
681 {
682
683 return (tag & ts->mask);
684 }
685
686 static u_int16_t
tagname2tag(struct pf_tagset * ts,const char * tagname)687 tagname2tag(struct pf_tagset *ts, const char *tagname)
688 {
689 struct pf_tagname *tag;
690 u_int32_t index;
691 u_int16_t new_tagid;
692
693 PF_RULES_WASSERT();
694
695 index = tagname2hashindex(ts, tagname);
696 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
697 if (strcmp(tagname, tag->name) == 0) {
698 tag->ref++;
699 return (tag->tag);
700 }
701
702 /*
703 * new entry
704 *
705 * to avoid fragmentation, we do a linear search from the beginning
706 * and take the first free slot we find.
707 */
708 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
709 /*
710 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
711 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
712 * set. It may also return a bit number greater than TAGID_MAX due
713 * to rounding of the number of bits in the vector up to a multiple
714 * of the vector word size at declaration/allocation time.
715 */
716 if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
717 return (0);
718
719 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */
720 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
721
722 /* allocate and fill new struct pf_tagname */
723 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
724 if (tag == NULL)
725 return (0);
726 strlcpy(tag->name, tagname, sizeof(tag->name));
727 tag->tag = new_tagid;
728 tag->ref = 1;
729
730 /* Insert into namehash */
731 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
732
733 /* Insert into taghash */
734 index = tag2hashindex(ts, new_tagid);
735 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
736
737 return (tag->tag);
738 }
739
740 static void
tag_unref(struct pf_tagset * ts,u_int16_t tag)741 tag_unref(struct pf_tagset *ts, u_int16_t tag)
742 {
743 struct pf_tagname *t;
744 uint16_t index;
745
746 PF_RULES_WASSERT();
747
748 index = tag2hashindex(ts, tag);
749 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
750 if (tag == t->tag) {
751 if (--t->ref == 0) {
752 TAILQ_REMOVE(&ts->taghash[index], t,
753 taghash_entries);
754 index = tagname2hashindex(ts, t->name);
755 TAILQ_REMOVE(&ts->namehash[index], t,
756 namehash_entries);
757 /* Bits are 0-based for BIT_SET() */
758 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
759 uma_zfree(V_pf_tag_z, t);
760 }
761 break;
762 }
763 }
764
765 static uint16_t
pf_tagname2tag(const char * tagname)766 pf_tagname2tag(const char *tagname)
767 {
768 return (tagname2tag(&V_pf_tags, tagname));
769 }
770
771 static int
pf_begin_eth(uint32_t * ticket,const char * anchor)772 pf_begin_eth(uint32_t *ticket, const char *anchor)
773 {
774 struct pf_keth_rule *rule, *tmp;
775 struct pf_keth_ruleset *rs;
776
777 PF_RULES_WASSERT();
778
779 rs = pf_find_or_create_keth_ruleset(anchor);
780 if (rs == NULL)
781 return (EINVAL);
782
783 /* Purge old inactive rules. */
784 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
785 tmp) {
786 TAILQ_REMOVE(rs->inactive.rules, rule,
787 entries);
788 pf_free_eth_rule(rule);
789 }
790
791 *ticket = ++rs->inactive.ticket;
792 rs->inactive.open = 1;
793
794 return (0);
795 }
796
797 static int
pf_rollback_eth(uint32_t ticket,const char * anchor)798 pf_rollback_eth(uint32_t ticket, const char *anchor)
799 {
800 struct pf_keth_rule *rule, *tmp;
801 struct pf_keth_ruleset *rs;
802
803 PF_RULES_WASSERT();
804
805 rs = pf_find_keth_ruleset(anchor);
806 if (rs == NULL)
807 return (EINVAL);
808
809 if (!rs->inactive.open ||
810 ticket != rs->inactive.ticket)
811 return (0);
812
813 /* Purge old inactive rules. */
814 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
815 tmp) {
816 TAILQ_REMOVE(rs->inactive.rules, rule, entries);
817 pf_free_eth_rule(rule);
818 }
819
820 rs->inactive.open = 0;
821
822 pf_remove_if_empty_keth_ruleset(rs);
823
824 return (0);
825 }
826
827 #define PF_SET_SKIP_STEPS(i) \
828 do { \
829 while (head[i] != cur) { \
830 head[i]->skip[i].ptr = cur; \
831 head[i] = TAILQ_NEXT(head[i], entries); \
832 } \
833 } while (0)
834
835 static void
pf_eth_calc_skip_steps(struct pf_keth_ruleq * rules)836 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
837 {
838 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
839 int i;
840
841 cur = TAILQ_FIRST(rules);
842 prev = cur;
843 for (i = 0; i < PFE_SKIP_COUNT; ++i)
844 head[i] = cur;
845 while (cur != NULL) {
846 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
847 PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
848 if (cur->direction != prev->direction)
849 PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
850 if (cur->proto != prev->proto)
851 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
852 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
853 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
854 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
855 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
856 if (cur->ipsrc.neg != prev->ipsrc.neg ||
857 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
858 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
859 if (cur->ipdst.neg != prev->ipdst.neg ||
860 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
861 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
862
863 prev = cur;
864 cur = TAILQ_NEXT(cur, entries);
865 }
866 for (i = 0; i < PFE_SKIP_COUNT; ++i)
867 PF_SET_SKIP_STEPS(i);
868 }
869
870 static int
pf_commit_eth(uint32_t ticket,const char * anchor)871 pf_commit_eth(uint32_t ticket, const char *anchor)
872 {
873 struct pf_keth_ruleq *rules;
874 struct pf_keth_ruleset *rs;
875
876 rs = pf_find_keth_ruleset(anchor);
877 if (rs == NULL) {
878 return (EINVAL);
879 }
880
881 if (!rs->inactive.open ||
882 ticket != rs->inactive.ticket)
883 return (EBUSY);
884
885 PF_RULES_WASSERT();
886
887 pf_eth_calc_skip_steps(rs->inactive.rules);
888
889 rules = rs->active.rules;
890 atomic_store_ptr(&rs->active.rules, rs->inactive.rules);
891 rs->inactive.rules = rules;
892 rs->inactive.ticket = rs->active.ticket;
893
894 return (pf_rollback_eth(rs->inactive.ticket,
895 rs->anchor ? rs->anchor->path : ""));
896 }
897
898 #ifdef ALTQ
899 static uint16_t
pf_qname2qid(const char * qname)900 pf_qname2qid(const char *qname)
901 {
902 return (tagname2tag(&V_pf_qids, qname));
903 }
904
905 static void
pf_qid_unref(uint16_t qid)906 pf_qid_unref(uint16_t qid)
907 {
908 tag_unref(&V_pf_qids, qid);
909 }
910
911 static int
pf_begin_altq(u_int32_t * ticket)912 pf_begin_altq(u_int32_t *ticket)
913 {
914 struct pf_altq *altq, *tmp;
915 int error = 0;
916
917 PF_RULES_WASSERT();
918
919 /* Purge the old altq lists */
920 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
921 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
922 /* detach and destroy the discipline */
923 error = altq_remove(altq);
924 }
925 free(altq, M_PFALTQ);
926 }
927 TAILQ_INIT(V_pf_altq_ifs_inactive);
928 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
929 pf_qid_unref(altq->qid);
930 free(altq, M_PFALTQ);
931 }
932 TAILQ_INIT(V_pf_altqs_inactive);
933 if (error)
934 return (error);
935 *ticket = ++V_ticket_altqs_inactive;
936 V_altqs_inactive_open = 1;
937 return (0);
938 }
939
940 static int
pf_rollback_altq(u_int32_t ticket)941 pf_rollback_altq(u_int32_t ticket)
942 {
943 struct pf_altq *altq, *tmp;
944 int error = 0;
945
946 PF_RULES_WASSERT();
947
948 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
949 return (0);
950 /* Purge the old altq lists */
951 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
952 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
953 /* detach and destroy the discipline */
954 error = altq_remove(altq);
955 }
956 free(altq, M_PFALTQ);
957 }
958 TAILQ_INIT(V_pf_altq_ifs_inactive);
959 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
960 pf_qid_unref(altq->qid);
961 free(altq, M_PFALTQ);
962 }
963 TAILQ_INIT(V_pf_altqs_inactive);
964 V_altqs_inactive_open = 0;
965 return (error);
966 }
967
968 static int
pf_commit_altq(u_int32_t ticket)969 pf_commit_altq(u_int32_t ticket)
970 {
971 struct pf_altqqueue *old_altqs, *old_altq_ifs;
972 struct pf_altq *altq, *tmp;
973 int err, error = 0;
974
975 PF_RULES_WASSERT();
976
977 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
978 return (EBUSY);
979
980 /* swap altqs, keep the old. */
981 old_altqs = V_pf_altqs_active;
982 old_altq_ifs = V_pf_altq_ifs_active;
983 V_pf_altqs_active = V_pf_altqs_inactive;
984 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
985 V_pf_altqs_inactive = old_altqs;
986 V_pf_altq_ifs_inactive = old_altq_ifs;
987 V_ticket_altqs_active = V_ticket_altqs_inactive;
988
989 /* Attach new disciplines */
990 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
991 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
992 /* attach the discipline */
993 error = altq_pfattach(altq);
994 if (error == 0 && V_pf_altq_running)
995 error = pf_enable_altq(altq);
996 if (error != 0)
997 return (error);
998 }
999 }
1000
1001 /* Purge the old altq lists */
1002 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1003 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1004 /* detach and destroy the discipline */
1005 if (V_pf_altq_running)
1006 error = pf_disable_altq(altq);
1007 err = altq_pfdetach(altq);
1008 if (err != 0 && error == 0)
1009 error = err;
1010 err = altq_remove(altq);
1011 if (err != 0 && error == 0)
1012 error = err;
1013 }
1014 free(altq, M_PFALTQ);
1015 }
1016 TAILQ_INIT(V_pf_altq_ifs_inactive);
1017 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1018 pf_qid_unref(altq->qid);
1019 free(altq, M_PFALTQ);
1020 }
1021 TAILQ_INIT(V_pf_altqs_inactive);
1022
1023 V_altqs_inactive_open = 0;
1024 return (error);
1025 }
1026
1027 static int
pf_enable_altq(struct pf_altq * altq)1028 pf_enable_altq(struct pf_altq *altq)
1029 {
1030 struct ifnet *ifp;
1031 struct tb_profile tb;
1032 int error = 0;
1033
1034 if ((ifp = ifunit(altq->ifname)) == NULL)
1035 return (EINVAL);
1036
1037 if (ifp->if_snd.altq_type != ALTQT_NONE)
1038 error = altq_enable(&ifp->if_snd);
1039
1040 /* set tokenbucket regulator */
1041 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1042 tb.rate = altq->ifbandwidth;
1043 tb.depth = altq->tbrsize;
1044 error = tbr_set(&ifp->if_snd, &tb);
1045 }
1046
1047 return (error);
1048 }
1049
1050 static int
pf_disable_altq(struct pf_altq * altq)1051 pf_disable_altq(struct pf_altq *altq)
1052 {
1053 struct ifnet *ifp;
1054 struct tb_profile tb;
1055 int error;
1056
1057 if ((ifp = ifunit(altq->ifname)) == NULL)
1058 return (EINVAL);
1059
1060 /*
1061 * when the discipline is no longer referenced, it was overridden
1062 * by a new one. if so, just return.
1063 */
1064 if (altq->altq_disc != ifp->if_snd.altq_disc)
1065 return (0);
1066
1067 error = altq_disable(&ifp->if_snd);
1068
1069 if (error == 0) {
1070 /* clear tokenbucket regulator */
1071 tb.rate = 0;
1072 error = tbr_set(&ifp->if_snd, &tb);
1073 }
1074
1075 return (error);
1076 }
1077
1078 static int
pf_altq_ifnet_event_add(struct ifnet * ifp,int remove,u_int32_t ticket,struct pf_altq * altq)1079 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1080 struct pf_altq *altq)
1081 {
1082 struct ifnet *ifp1;
1083 int error = 0;
1084
1085 /* Deactivate the interface in question */
1086 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1087 if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1088 (remove && ifp1 == ifp)) {
1089 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1090 } else {
1091 error = altq_add(ifp1, altq);
1092
1093 if (ticket != V_ticket_altqs_inactive)
1094 error = EBUSY;
1095
1096 if (error)
1097 free(altq, M_PFALTQ);
1098 }
1099
1100 return (error);
1101 }
1102
1103 void
pf_altq_ifnet_event(struct ifnet * ifp,int remove)1104 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1105 {
1106 struct pf_altq *a1, *a2, *a3;
1107 u_int32_t ticket;
1108 int error = 0;
1109
1110 /*
1111 * No need to re-evaluate the configuration for events on interfaces
1112 * that do not support ALTQ, as it's not possible for such
1113 * interfaces to be part of the configuration.
1114 */
1115 if (!ALTQ_IS_READY(&ifp->if_snd))
1116 return;
1117
1118 /* Interrupt userland queue modifications */
1119 if (V_altqs_inactive_open)
1120 pf_rollback_altq(V_ticket_altqs_inactive);
1121
1122 /* Start new altq ruleset */
1123 if (pf_begin_altq(&ticket))
1124 return;
1125
1126 /* Copy the current active set */
1127 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1128 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1129 if (a2 == NULL) {
1130 error = ENOMEM;
1131 break;
1132 }
1133 bcopy(a1, a2, sizeof(struct pf_altq));
1134
1135 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1136 if (error)
1137 break;
1138
1139 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1140 }
1141 if (error)
1142 goto out;
1143 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1144 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1145 if (a2 == NULL) {
1146 error = ENOMEM;
1147 break;
1148 }
1149 bcopy(a1, a2, sizeof(struct pf_altq));
1150
1151 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1152 error = EBUSY;
1153 free(a2, M_PFALTQ);
1154 break;
1155 }
1156 a2->altq_disc = NULL;
1157 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1158 if (strncmp(a3->ifname, a2->ifname,
1159 IFNAMSIZ) == 0) {
1160 a2->altq_disc = a3->altq_disc;
1161 break;
1162 }
1163 }
1164 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1165 if (error)
1166 break;
1167
1168 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1169 }
1170
1171 out:
1172 if (error != 0)
1173 pf_rollback_altq(ticket);
1174 else
1175 pf_commit_altq(ticket);
1176 }
1177 #endif /* ALTQ */
1178
1179 static struct pf_krule_global *
pf_rule_tree_alloc(int flags)1180 pf_rule_tree_alloc(int flags)
1181 {
1182 struct pf_krule_global *tree;
1183
1184 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1185 if (tree == NULL)
1186 return (NULL);
1187 RB_INIT(tree);
1188 return (tree);
1189 }
1190
1191 static void
pf_rule_tree_free(struct pf_krule_global * tree)1192 pf_rule_tree_free(struct pf_krule_global *tree)
1193 {
1194
1195 free(tree, M_TEMP);
1196 }
1197
1198 static int
pf_begin_rules(u_int32_t * ticket,int rs_num,const char * anchor)1199 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1200 {
1201 struct pf_krule_global *tree;
1202 struct pf_kruleset *rs;
1203 struct pf_krule *rule;
1204
1205 PF_RULES_WASSERT();
1206
1207 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1208 return (EINVAL);
1209 tree = pf_rule_tree_alloc(M_NOWAIT);
1210 if (tree == NULL)
1211 return (ENOMEM);
1212 rs = pf_find_or_create_kruleset(anchor);
1213 if (rs == NULL) {
1214 free(tree, M_TEMP);
1215 return (EINVAL);
1216 }
1217 pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1218 rs->rules[rs_num].inactive.tree = tree;
1219
1220 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1221 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1222 rs->rules[rs_num].inactive.rcount--;
1223 }
1224 *ticket = ++rs->rules[rs_num].inactive.ticket;
1225 rs->rules[rs_num].inactive.open = 1;
1226 return (0);
1227 }
1228
1229 static int
pf_rollback_rules(u_int32_t ticket,int rs_num,char * anchor)1230 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1231 {
1232 struct pf_kruleset *rs;
1233 struct pf_krule *rule;
1234
1235 PF_RULES_WASSERT();
1236
1237 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1238 return (EINVAL);
1239 rs = pf_find_kruleset(anchor);
1240 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1241 rs->rules[rs_num].inactive.ticket != ticket)
1242 return (0);
1243 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1244 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1245 rs->rules[rs_num].inactive.rcount--;
1246 }
1247 rs->rules[rs_num].inactive.open = 0;
1248 return (0);
1249 }
1250
1251 #define PF_MD5_UPD(st, elm) \
1252 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1253
1254 #define PF_MD5_UPD_STR(st, elm) \
1255 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1256
1257 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \
1258 (stor) = htonl((st)->elm); \
1259 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1260 } while (0)
1261
1262 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \
1263 (stor) = htons((st)->elm); \
1264 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1265 } while (0)
1266
1267 static void
pf_hash_rule_addr(MD5_CTX * ctx,struct pf_rule_addr * pfr)1268 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1269 {
1270 PF_MD5_UPD(pfr, addr.type);
1271 switch (pfr->addr.type) {
1272 case PF_ADDR_DYNIFTL:
1273 PF_MD5_UPD(pfr, addr.v.ifname);
1274 PF_MD5_UPD(pfr, addr.iflags);
1275 break;
1276 case PF_ADDR_TABLE:
1277 PF_MD5_UPD(pfr, addr.v.tblname);
1278 break;
1279 case PF_ADDR_ADDRMASK:
1280 /* XXX ignore af? */
1281 PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1282 PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1283 break;
1284 }
1285
1286 PF_MD5_UPD(pfr, port[0]);
1287 PF_MD5_UPD(pfr, port[1]);
1288 PF_MD5_UPD(pfr, neg);
1289 PF_MD5_UPD(pfr, port_op);
1290 }
1291
1292 static void
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)1293 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1294 {
1295 u_int16_t x;
1296 u_int32_t y;
1297
1298 pf_hash_rule_addr(ctx, &rule->src);
1299 pf_hash_rule_addr(ctx, &rule->dst);
1300 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1301 PF_MD5_UPD_STR(rule, label[i]);
1302 PF_MD5_UPD_STR(rule, ifname);
1303 PF_MD5_UPD_STR(rule, rcv_ifname);
1304 PF_MD5_UPD_STR(rule, match_tagname);
1305 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1306 PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1307 PF_MD5_UPD_HTONL(rule, prob, y);
1308 PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1309 PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1310 PF_MD5_UPD(rule, uid.op);
1311 PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1312 PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1313 PF_MD5_UPD(rule, gid.op);
1314 PF_MD5_UPD_HTONL(rule, rule_flag, y);
1315 PF_MD5_UPD(rule, action);
1316 PF_MD5_UPD(rule, direction);
1317 PF_MD5_UPD(rule, af);
1318 PF_MD5_UPD(rule, quick);
1319 PF_MD5_UPD(rule, ifnot);
1320 PF_MD5_UPD(rule, rcvifnot);
1321 PF_MD5_UPD(rule, match_tag_not);
1322 PF_MD5_UPD(rule, natpass);
1323 PF_MD5_UPD(rule, keep_state);
1324 PF_MD5_UPD(rule, proto);
1325 PF_MD5_UPD(rule, type);
1326 PF_MD5_UPD(rule, code);
1327 PF_MD5_UPD(rule, flags);
1328 PF_MD5_UPD(rule, flagset);
1329 PF_MD5_UPD(rule, allow_opts);
1330 PF_MD5_UPD(rule, rt);
1331 PF_MD5_UPD(rule, tos);
1332 PF_MD5_UPD(rule, scrub_flags);
1333 PF_MD5_UPD(rule, min_ttl);
1334 PF_MD5_UPD(rule, set_tos);
1335 if (rule->anchor != NULL)
1336 PF_MD5_UPD_STR(rule, anchor->path);
1337 }
1338
1339 static void
pf_hash_rule(struct pf_krule * rule)1340 pf_hash_rule(struct pf_krule *rule)
1341 {
1342 MD5_CTX ctx;
1343
1344 MD5Init(&ctx);
1345 pf_hash_rule_rolling(&ctx, rule);
1346 MD5Final(rule->md5sum, &ctx);
1347 }
1348
1349 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1350 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1351 {
1352
1353 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1354 }
1355
1356 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1357 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1358 {
1359 struct pf_kruleset *rs;
1360 struct pf_krule *rule, **old_array, *old_rule;
1361 struct pf_krulequeue *old_rules;
1362 struct pf_krule_global *old_tree;
1363 int error;
1364 u_int32_t old_rcount;
1365
1366 PF_RULES_WASSERT();
1367
1368 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1369 return (EINVAL);
1370 rs = pf_find_kruleset(anchor);
1371 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1372 ticket != rs->rules[rs_num].inactive.ticket)
1373 return (EBUSY);
1374
1375 /* Calculate checksum for the main ruleset */
1376 if (rs == &pf_main_ruleset) {
1377 error = pf_setup_pfsync_matching(rs);
1378 if (error != 0)
1379 return (error);
1380 }
1381
1382 /* Swap rules, keep the old. */
1383 old_rules = rs->rules[rs_num].active.ptr;
1384 old_rcount = rs->rules[rs_num].active.rcount;
1385 old_array = rs->rules[rs_num].active.ptr_array;
1386 old_tree = rs->rules[rs_num].active.tree;
1387
1388 rs->rules[rs_num].active.ptr =
1389 rs->rules[rs_num].inactive.ptr;
1390 rs->rules[rs_num].active.ptr_array =
1391 rs->rules[rs_num].inactive.ptr_array;
1392 rs->rules[rs_num].active.tree =
1393 rs->rules[rs_num].inactive.tree;
1394 rs->rules[rs_num].active.rcount =
1395 rs->rules[rs_num].inactive.rcount;
1396
1397 /* Attempt to preserve counter information. */
1398 if (V_pf_status.keep_counters && old_tree != NULL) {
1399 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1400 entries) {
1401 old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1402 if (old_rule == NULL) {
1403 continue;
1404 }
1405 pf_counter_u64_critical_enter();
1406 pf_counter_u64_rollup_protected(&rule->evaluations,
1407 pf_counter_u64_fetch(&old_rule->evaluations));
1408 pf_counter_u64_rollup_protected(&rule->packets[0],
1409 pf_counter_u64_fetch(&old_rule->packets[0]));
1410 pf_counter_u64_rollup_protected(&rule->packets[1],
1411 pf_counter_u64_fetch(&old_rule->packets[1]));
1412 pf_counter_u64_rollup_protected(&rule->bytes[0],
1413 pf_counter_u64_fetch(&old_rule->bytes[0]));
1414 pf_counter_u64_rollup_protected(&rule->bytes[1],
1415 pf_counter_u64_fetch(&old_rule->bytes[1]));
1416 pf_counter_u64_critical_exit();
1417 }
1418 }
1419
1420 rs->rules[rs_num].inactive.ptr = old_rules;
1421 rs->rules[rs_num].inactive.ptr_array = old_array;
1422 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1423 rs->rules[rs_num].inactive.rcount = old_rcount;
1424
1425 rs->rules[rs_num].active.ticket =
1426 rs->rules[rs_num].inactive.ticket;
1427 pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1428
1429 /* Purge the old rule list. */
1430 PF_UNLNKDRULES_LOCK();
1431 while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1432 pf_unlink_rule_locked(old_rules, rule);
1433 PF_UNLNKDRULES_UNLOCK();
1434 if (rs->rules[rs_num].inactive.ptr_array)
1435 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1436 rs->rules[rs_num].inactive.ptr_array = NULL;
1437 rs->rules[rs_num].inactive.rcount = 0;
1438 rs->rules[rs_num].inactive.open = 0;
1439 pf_remove_if_empty_kruleset(rs);
1440 free(old_tree, M_TEMP);
1441
1442 return (0);
1443 }
1444
1445 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1446 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1447 {
1448 MD5_CTX ctx;
1449 struct pf_krule *rule;
1450 int rs_cnt;
1451 u_int8_t digest[PF_MD5_DIGEST_LENGTH];
1452
1453 MD5Init(&ctx);
1454 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1455 /* XXX PF_RULESET_SCRUB as well? */
1456 if (rs_cnt == PF_RULESET_SCRUB)
1457 continue;
1458
1459 if (rs->rules[rs_cnt].inactive.ptr_array)
1460 free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1461 rs->rules[rs_cnt].inactive.ptr_array = NULL;
1462
1463 if (rs->rules[rs_cnt].inactive.rcount) {
1464 rs->rules[rs_cnt].inactive.ptr_array =
1465 mallocarray(rs->rules[rs_cnt].inactive.rcount,
1466 sizeof(struct pf_rule **),
1467 M_TEMP, M_NOWAIT);
1468
1469 if (!rs->rules[rs_cnt].inactive.ptr_array)
1470 return (ENOMEM);
1471 }
1472
1473 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1474 entries) {
1475 pf_hash_rule_rolling(&ctx, rule);
1476 (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1477 }
1478 }
1479
1480 MD5Final(digest, &ctx);
1481 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1482 return (0);
1483 }
1484
1485 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1486 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1487 {
1488 int error = 0;
1489
1490 switch (addr->type) {
1491 case PF_ADDR_TABLE:
1492 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1493 if (addr->p.tbl == NULL)
1494 error = ENOMEM;
1495 break;
1496 default:
1497 error = EINVAL;
1498 }
1499
1500 return (error);
1501 }
1502
1503 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1504 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1505 sa_family_t af)
1506 {
1507 int error = 0;
1508
1509 switch (addr->type) {
1510 case PF_ADDR_TABLE:
1511 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1512 if (addr->p.tbl == NULL)
1513 error = ENOMEM;
1514 break;
1515 case PF_ADDR_DYNIFTL:
1516 error = pfi_dynaddr_setup(addr, af);
1517 break;
1518 }
1519
1520 return (error);
1521 }
1522
1523 void
pf_addr_copyout(struct pf_addr_wrap * addr)1524 pf_addr_copyout(struct pf_addr_wrap *addr)
1525 {
1526
1527 switch (addr->type) {
1528 case PF_ADDR_DYNIFTL:
1529 pfi_dynaddr_copyout(addr);
1530 break;
1531 case PF_ADDR_TABLE:
1532 pf_tbladdr_copyout(addr);
1533 break;
1534 }
1535 }
1536
1537 static void
pf_src_node_copy(const struct pf_ksrc_node * in,struct pf_src_node * out)1538 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1539 {
1540 int secs = time_uptime, diff;
1541
1542 bzero(out, sizeof(struct pf_src_node));
1543
1544 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1545 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1546
1547 if (in->rule != NULL)
1548 out->rule.nr = in->rule->nr;
1549
1550 for (int i = 0; i < 2; i++) {
1551 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1552 out->packets[i] = counter_u64_fetch(in->packets[i]);
1553 }
1554
1555 out->states = in->states;
1556 out->conn = in->conn;
1557 out->af = in->af;
1558 out->ruletype = in->ruletype;
1559
1560 out->creation = secs - in->creation;
1561 if (out->expire > secs)
1562 out->expire -= secs;
1563 else
1564 out->expire = 0;
1565
1566 /* Adjust the connection rate estimate. */
1567 out->conn_rate = in->conn_rate;
1568 diff = secs - in->conn_rate.last;
1569 if (diff >= in->conn_rate.seconds)
1570 out->conn_rate.count = 0;
1571 else
1572 out->conn_rate.count -=
1573 in->conn_rate.count * diff /
1574 in->conn_rate.seconds;
1575 }
1576
1577 #ifdef ALTQ
1578 /*
1579 * Handle export of struct pf_kaltq to user binaries that may be using any
1580 * version of struct pf_altq.
1581 */
1582 static int
pf_export_kaltq(struct pf_altq * q,struct pfioc_altq_v1 * pa,size_t ioc_size)1583 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1584 {
1585 u_int32_t version;
1586
1587 if (ioc_size == sizeof(struct pfioc_altq_v0))
1588 version = 0;
1589 else
1590 version = pa->version;
1591
1592 if (version > PFIOC_ALTQ_VERSION)
1593 return (EINVAL);
1594
1595 #define ASSIGN(x) exported_q->x = q->x
1596 #define COPY(x) \
1597 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1598 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1599 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1600
1601 switch (version) {
1602 case 0: {
1603 struct pf_altq_v0 *exported_q =
1604 &((struct pfioc_altq_v0 *)pa)->altq;
1605
1606 COPY(ifname);
1607
1608 ASSIGN(scheduler);
1609 ASSIGN(tbrsize);
1610 exported_q->tbrsize = SATU16(q->tbrsize);
1611 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1612
1613 COPY(qname);
1614 COPY(parent);
1615 ASSIGN(parent_qid);
1616 exported_q->bandwidth = SATU32(q->bandwidth);
1617 ASSIGN(priority);
1618 ASSIGN(local_flags);
1619
1620 ASSIGN(qlimit);
1621 ASSIGN(flags);
1622
1623 if (q->scheduler == ALTQT_HFSC) {
1624 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1625 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1626 SATU32(q->pq_u.hfsc_opts.x)
1627
1628 ASSIGN_OPT_SATU32(rtsc_m1);
1629 ASSIGN_OPT(rtsc_d);
1630 ASSIGN_OPT_SATU32(rtsc_m2);
1631
1632 ASSIGN_OPT_SATU32(lssc_m1);
1633 ASSIGN_OPT(lssc_d);
1634 ASSIGN_OPT_SATU32(lssc_m2);
1635
1636 ASSIGN_OPT_SATU32(ulsc_m1);
1637 ASSIGN_OPT(ulsc_d);
1638 ASSIGN_OPT_SATU32(ulsc_m2);
1639
1640 ASSIGN_OPT(flags);
1641
1642 #undef ASSIGN_OPT
1643 #undef ASSIGN_OPT_SATU32
1644 } else
1645 COPY(pq_u);
1646
1647 ASSIGN(qid);
1648 break;
1649 }
1650 case 1: {
1651 struct pf_altq_v1 *exported_q =
1652 &((struct pfioc_altq_v1 *)pa)->altq;
1653
1654 COPY(ifname);
1655
1656 ASSIGN(scheduler);
1657 ASSIGN(tbrsize);
1658 ASSIGN(ifbandwidth);
1659
1660 COPY(qname);
1661 COPY(parent);
1662 ASSIGN(parent_qid);
1663 ASSIGN(bandwidth);
1664 ASSIGN(priority);
1665 ASSIGN(local_flags);
1666
1667 ASSIGN(qlimit);
1668 ASSIGN(flags);
1669 COPY(pq_u);
1670
1671 ASSIGN(qid);
1672 break;
1673 }
1674 default:
1675 panic("%s: unhandled struct pfioc_altq version", __func__);
1676 break;
1677 }
1678
1679 #undef ASSIGN
1680 #undef COPY
1681 #undef SATU16
1682 #undef SATU32
1683
1684 return (0);
1685 }
1686
1687 /*
1688 * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1689 * that may be using any version of it.
1690 */
1691 static int
pf_import_kaltq(struct pfioc_altq_v1 * pa,struct pf_altq * q,size_t ioc_size)1692 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1693 {
1694 u_int32_t version;
1695
1696 if (ioc_size == sizeof(struct pfioc_altq_v0))
1697 version = 0;
1698 else
1699 version = pa->version;
1700
1701 if (version > PFIOC_ALTQ_VERSION)
1702 return (EINVAL);
1703
1704 #define ASSIGN(x) q->x = imported_q->x
1705 #define COPY(x) \
1706 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1707
1708 switch (version) {
1709 case 0: {
1710 struct pf_altq_v0 *imported_q =
1711 &((struct pfioc_altq_v0 *)pa)->altq;
1712
1713 COPY(ifname);
1714
1715 ASSIGN(scheduler);
1716 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1717 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1718
1719 COPY(qname);
1720 COPY(parent);
1721 ASSIGN(parent_qid);
1722 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1723 ASSIGN(priority);
1724 ASSIGN(local_flags);
1725
1726 ASSIGN(qlimit);
1727 ASSIGN(flags);
1728
1729 if (imported_q->scheduler == ALTQT_HFSC) {
1730 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1731
1732 /*
1733 * The m1 and m2 parameters are being copied from
1734 * 32-bit to 64-bit.
1735 */
1736 ASSIGN_OPT(rtsc_m1);
1737 ASSIGN_OPT(rtsc_d);
1738 ASSIGN_OPT(rtsc_m2);
1739
1740 ASSIGN_OPT(lssc_m1);
1741 ASSIGN_OPT(lssc_d);
1742 ASSIGN_OPT(lssc_m2);
1743
1744 ASSIGN_OPT(ulsc_m1);
1745 ASSIGN_OPT(ulsc_d);
1746 ASSIGN_OPT(ulsc_m2);
1747
1748 ASSIGN_OPT(flags);
1749
1750 #undef ASSIGN_OPT
1751 } else
1752 COPY(pq_u);
1753
1754 ASSIGN(qid);
1755 break;
1756 }
1757 case 1: {
1758 struct pf_altq_v1 *imported_q =
1759 &((struct pfioc_altq_v1 *)pa)->altq;
1760
1761 COPY(ifname);
1762
1763 ASSIGN(scheduler);
1764 ASSIGN(tbrsize);
1765 ASSIGN(ifbandwidth);
1766
1767 COPY(qname);
1768 COPY(parent);
1769 ASSIGN(parent_qid);
1770 ASSIGN(bandwidth);
1771 ASSIGN(priority);
1772 ASSIGN(local_flags);
1773
1774 ASSIGN(qlimit);
1775 ASSIGN(flags);
1776 COPY(pq_u);
1777
1778 ASSIGN(qid);
1779 break;
1780 }
1781 default:
1782 panic("%s: unhandled struct pfioc_altq version", __func__);
1783 break;
1784 }
1785
1786 #undef ASSIGN
1787 #undef COPY
1788
1789 return (0);
1790 }
1791
1792 static struct pf_altq *
pf_altq_get_nth_active(u_int32_t n)1793 pf_altq_get_nth_active(u_int32_t n)
1794 {
1795 struct pf_altq *altq;
1796 u_int32_t nr;
1797
1798 nr = 0;
1799 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1800 if (nr == n)
1801 return (altq);
1802 nr++;
1803 }
1804
1805 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1806 if (nr == n)
1807 return (altq);
1808 nr++;
1809 }
1810
1811 return (NULL);
1812 }
1813 #endif /* ALTQ */
1814
1815 struct pf_krule *
pf_krule_alloc(void)1816 pf_krule_alloc(void)
1817 {
1818 struct pf_krule *rule;
1819
1820 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1821 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF);
1822 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF);
1823 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF);
1824 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1825 M_WAITOK | M_ZERO);
1826 return (rule);
1827 }
1828
1829 void
pf_krule_free(struct pf_krule * rule)1830 pf_krule_free(struct pf_krule *rule)
1831 {
1832 #ifdef PF_WANT_32_TO_64_COUNTER
1833 bool wowned;
1834 #endif
1835
1836 if (rule == NULL)
1837 return;
1838
1839 #ifdef PF_WANT_32_TO_64_COUNTER
1840 if (rule->allrulelinked) {
1841 wowned = PF_RULES_WOWNED();
1842 if (!wowned)
1843 PF_RULES_WLOCK();
1844 LIST_REMOVE(rule, allrulelist);
1845 V_pf_allrulecount--;
1846 if (!wowned)
1847 PF_RULES_WUNLOCK();
1848 }
1849 #endif
1850
1851 pf_counter_u64_deinit(&rule->evaluations);
1852 for (int i = 0; i < 2; i++) {
1853 pf_counter_u64_deinit(&rule->packets[i]);
1854 pf_counter_u64_deinit(&rule->bytes[i]);
1855 }
1856 counter_u64_free(rule->states_cur);
1857 counter_u64_free(rule->states_tot);
1858 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
1859 counter_u64_free(rule->src_nodes[sn_type]);
1860 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1861
1862 mtx_destroy(&rule->nat.mtx);
1863 mtx_destroy(&rule->rdr.mtx);
1864 mtx_destroy(&rule->route.mtx);
1865 free(rule, M_PFRULE);
1866 }
1867
1868 void
pf_krule_clear_counters(struct pf_krule * rule)1869 pf_krule_clear_counters(struct pf_krule *rule)
1870 {
1871 pf_counter_u64_zero(&rule->evaluations);
1872 for (int i = 0; i < 2; i++) {
1873 pf_counter_u64_zero(&rule->packets[i]);
1874 pf_counter_u64_zero(&rule->bytes[i]);
1875 }
1876 counter_u64_zero(rule->states_tot);
1877 }
1878
1879 static void
pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr * kpool,struct pf_pooladdr * pool)1880 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1881 struct pf_pooladdr *pool)
1882 {
1883
1884 bzero(pool, sizeof(*pool));
1885 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1886 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1887 }
1888
1889 static int
pf_pooladdr_to_kpooladdr(const struct pf_pooladdr * pool,struct pf_kpooladdr * kpool)1890 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1891 struct pf_kpooladdr *kpool)
1892 {
1893 int ret;
1894
1895 bzero(kpool, sizeof(*kpool));
1896 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1897 ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1898 sizeof(kpool->ifname));
1899 return (ret);
1900 }
1901
1902 static void
pf_pool_to_kpool(const struct pf_pool * pool,struct pf_kpool * kpool)1903 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1904 {
1905 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1906 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1907
1908 bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1909 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1910
1911 kpool->tblidx = pool->tblidx;
1912 kpool->proxy_port[0] = pool->proxy_port[0];
1913 kpool->proxy_port[1] = pool->proxy_port[1];
1914 kpool->opts = pool->opts;
1915 }
1916
1917 static int
pf_rule_to_krule(const struct pf_rule * rule,struct pf_krule * krule)1918 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1919 {
1920 int ret;
1921
1922 #ifndef INET
1923 if (rule->af == AF_INET) {
1924 return (EAFNOSUPPORT);
1925 }
1926 #endif /* INET */
1927 #ifndef INET6
1928 if (rule->af == AF_INET6) {
1929 return (EAFNOSUPPORT);
1930 }
1931 #endif /* INET6 */
1932
1933 ret = pf_check_rule_addr(&rule->src);
1934 if (ret != 0)
1935 return (ret);
1936 ret = pf_check_rule_addr(&rule->dst);
1937 if (ret != 0)
1938 return (ret);
1939
1940 bcopy(&rule->src, &krule->src, sizeof(rule->src));
1941 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1942
1943 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1944 if (ret != 0)
1945 return (ret);
1946 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1947 if (ret != 0)
1948 return (ret);
1949 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1950 if (ret != 0)
1951 return (ret);
1952 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1953 if (ret != 0)
1954 return (ret);
1955 ret = pf_user_strcpy(krule->tagname, rule->tagname,
1956 sizeof(rule->tagname));
1957 if (ret != 0)
1958 return (ret);
1959 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1960 sizeof(rule->match_tagname));
1961 if (ret != 0)
1962 return (ret);
1963 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1964 sizeof(rule->overload_tblname));
1965 if (ret != 0)
1966 return (ret);
1967
1968 pf_pool_to_kpool(&rule->rpool, &krule->rdr);
1969
1970 /* Don't allow userspace to set evaluations, packets or bytes. */
1971 /* kif, anchor, overload_tbl are not copied over. */
1972
1973 krule->os_fingerprint = rule->os_fingerprint;
1974
1975 krule->rtableid = rule->rtableid;
1976 /* pf_rule->timeout is smaller than pf_krule->timeout */
1977 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1978 krule->max_states = rule->max_states;
1979 krule->max_src_nodes = rule->max_src_nodes;
1980 krule->max_src_states = rule->max_src_states;
1981 krule->max_src_conn = rule->max_src_conn;
1982 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1983 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1984 krule->qid = rule->qid;
1985 krule->pqid = rule->pqid;
1986 krule->nr = rule->nr;
1987 krule->prob = rule->prob;
1988 krule->cuid = rule->cuid;
1989 krule->cpid = rule->cpid;
1990
1991 krule->return_icmp = rule->return_icmp;
1992 krule->return_icmp6 = rule->return_icmp6;
1993 krule->max_mss = rule->max_mss;
1994 krule->tag = rule->tag;
1995 krule->match_tag = rule->match_tag;
1996 krule->scrub_flags = rule->scrub_flags;
1997
1998 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1999 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2000
2001 krule->rule_flag = rule->rule_flag;
2002 krule->action = rule->action;
2003 krule->direction = rule->direction;
2004 krule->log = rule->log;
2005 krule->logif = rule->logif;
2006 krule->quick = rule->quick;
2007 krule->ifnot = rule->ifnot;
2008 krule->match_tag_not = rule->match_tag_not;
2009 krule->natpass = rule->natpass;
2010
2011 krule->keep_state = rule->keep_state;
2012 krule->af = rule->af;
2013 krule->proto = rule->proto;
2014 krule->type = rule->type;
2015 krule->code = rule->code;
2016 krule->flags = rule->flags;
2017 krule->flagset = rule->flagset;
2018 krule->min_ttl = rule->min_ttl;
2019 krule->allow_opts = rule->allow_opts;
2020 krule->rt = rule->rt;
2021 krule->return_ttl = rule->return_ttl;
2022 krule->tos = rule->tos;
2023 krule->set_tos = rule->set_tos;
2024
2025 krule->flush = rule->flush;
2026 krule->prio = rule->prio;
2027 krule->set_prio[0] = rule->set_prio[0];
2028 krule->set_prio[1] = rule->set_prio[1];
2029
2030 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2031
2032 return (0);
2033 }
2034
2035 int
pf_ioctl_getrules(struct pfioc_rule * pr)2036 pf_ioctl_getrules(struct pfioc_rule *pr)
2037 {
2038 struct pf_kruleset *ruleset;
2039 struct pf_krule *tail;
2040 int rs_num;
2041
2042 PF_RULES_WLOCK();
2043 ruleset = pf_find_kruleset(pr->anchor);
2044 if (ruleset == NULL) {
2045 PF_RULES_WUNLOCK();
2046 return (EINVAL);
2047 }
2048 rs_num = pf_get_ruleset_number(pr->rule.action);
2049 if (rs_num >= PF_RULESET_MAX) {
2050 PF_RULES_WUNLOCK();
2051 return (EINVAL);
2052 }
2053 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2054 pf_krulequeue);
2055 if (tail)
2056 pr->nr = tail->nr + 1;
2057 else
2058 pr->nr = 0;
2059 pr->ticket = ruleset->rules[rs_num].active.ticket;
2060 PF_RULES_WUNLOCK();
2061
2062 return (0);
2063 }
2064
2065 int
pf_ioctl_addrule(struct pf_krule * rule,uint32_t ticket,uint32_t pool_ticket,const char * anchor,const char * anchor_call,uid_t uid,pid_t pid)2066 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2067 uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2068 uid_t uid, pid_t pid)
2069 {
2070 struct pf_kruleset *ruleset;
2071 struct pf_krule *tail;
2072 struct pf_kpooladdr *pa;
2073 struct pfi_kkif *kif = NULL, *rcv_kif = NULL;
2074 int rs_num;
2075 int error = 0;
2076
2077 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2078 error = EINVAL;
2079 goto errout_unlocked;
2080 }
2081
2082 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
2083
2084 if (rule->ifname[0])
2085 kif = pf_kkif_create(M_WAITOK);
2086 if (rule->rcv_ifname[0])
2087 rcv_kif = pf_kkif_create(M_WAITOK);
2088 pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2089 for (int i = 0; i < 2; i++) {
2090 pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2091 pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2092 }
2093 rule->states_cur = counter_u64_alloc(M_WAITOK);
2094 rule->states_tot = counter_u64_alloc(M_WAITOK);
2095 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
2096 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
2097 rule->cuid = uid;
2098 rule->cpid = pid;
2099 TAILQ_INIT(&rule->rdr.list);
2100 TAILQ_INIT(&rule->nat.list);
2101 TAILQ_INIT(&rule->route.list);
2102
2103 PF_CONFIG_LOCK();
2104 PF_RULES_WLOCK();
2105 #ifdef PF_WANT_32_TO_64_COUNTER
2106 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2107 MPASS(!rule->allrulelinked);
2108 rule->allrulelinked = true;
2109 V_pf_allrulecount++;
2110 #endif
2111 ruleset = pf_find_kruleset(anchor);
2112 if (ruleset == NULL)
2113 ERROUT(EINVAL);
2114 rs_num = pf_get_ruleset_number(rule->action);
2115 if (rs_num >= PF_RULESET_MAX)
2116 ERROUT(EINVAL);
2117 if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2118 DPFPRINTF(PF_DEBUG_MISC,
2119 ("ticket: %d != [%d]%d\n", ticket, rs_num,
2120 ruleset->rules[rs_num].inactive.ticket));
2121 ERROUT(EBUSY);
2122 }
2123 if (pool_ticket != V_ticket_pabuf) {
2124 DPFPRINTF(PF_DEBUG_MISC,
2125 ("pool_ticket: %d != %d\n", pool_ticket,
2126 V_ticket_pabuf));
2127 ERROUT(EBUSY);
2128 }
2129 /*
2130 * XXXMJG hack: there is no mechanism to ensure they started the
2131 * transaction. Ticket checked above may happen to match by accident,
2132 * even if nobody called DIOCXBEGIN, let alone this process.
2133 * Partially work around it by checking if the RB tree got allocated,
2134 * see pf_begin_rules.
2135 */
2136 if (ruleset->rules[rs_num].inactive.tree == NULL) {
2137 ERROUT(EINVAL);
2138 }
2139
2140 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2141 pf_krulequeue);
2142 if (tail)
2143 rule->nr = tail->nr + 1;
2144 else
2145 rule->nr = 0;
2146 if (rule->ifname[0]) {
2147 rule->kif = pfi_kkif_attach(kif, rule->ifname);
2148 kif = NULL;
2149 pfi_kkif_ref(rule->kif);
2150 } else
2151 rule->kif = NULL;
2152
2153 if (rule->rcv_ifname[0]) {
2154 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname);
2155 rcv_kif = NULL;
2156 pfi_kkif_ref(rule->rcv_kif);
2157 } else
2158 rule->rcv_kif = NULL;
2159
2160 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2161 error = EBUSY;
2162
2163 #ifdef ALTQ
2164 /* set queue IDs */
2165 if (rule->qname[0] != 0) {
2166 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2167 error = EBUSY;
2168 else if (rule->pqname[0] != 0) {
2169 if ((rule->pqid =
2170 pf_qname2qid(rule->pqname)) == 0)
2171 error = EBUSY;
2172 } else
2173 rule->pqid = rule->qid;
2174 }
2175 #endif
2176 if (rule->tagname[0])
2177 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2178 error = EBUSY;
2179 if (rule->match_tagname[0])
2180 if ((rule->match_tag =
2181 pf_tagname2tag(rule->match_tagname)) == 0)
2182 error = EBUSY;
2183 if (rule->rt && !rule->direction)
2184 error = EINVAL;
2185 if (!rule->log)
2186 rule->logif = 0;
2187 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2188 error = ENOMEM;
2189 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2190 error = ENOMEM;
2191 if (pf_kanchor_setup(rule, ruleset, anchor_call))
2192 error = EINVAL;
2193 if (rule->scrub_flags & PFSTATE_SETPRIO &&
2194 (rule->set_prio[0] > PF_PRIO_MAX ||
2195 rule->set_prio[1] > PF_PRIO_MAX))
2196 error = EINVAL;
2197 for (int i = 0; i < 3; i++) {
2198 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
2199 if (pa->addr.type == PF_ADDR_TABLE) {
2200 pa->addr.p.tbl = pfr_attach_table(ruleset,
2201 pa->addr.v.tblname);
2202 if (pa->addr.p.tbl == NULL)
2203 error = ENOMEM;
2204 }
2205 }
2206
2207 rule->overload_tbl = NULL;
2208 if (rule->overload_tblname[0]) {
2209 if ((rule->overload_tbl = pfr_attach_table(ruleset,
2210 rule->overload_tblname)) == NULL)
2211 error = EINVAL;
2212 else
2213 rule->overload_tbl->pfrkt_flags |=
2214 PFR_TFLAG_ACTIVE;
2215 }
2216
2217 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list);
2218
2219 /*
2220 * Old version of pfctl provide route redirection pools in single
2221 * common redirection pool rdr. New versions use rdr only for
2222 * rdr-to rules.
2223 */
2224 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) {
2225 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list);
2226 } else {
2227 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list);
2228 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list);
2229 }
2230
2231 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2232 (rule->action == PF_BINAT)) && rule->anchor == NULL &&
2233 TAILQ_FIRST(&rule->rdr.list) == NULL) {
2234 error = EINVAL;
2235 }
2236
2237 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) {
2238 error = EINVAL;
2239 }
2240
2241 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR ||
2242 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) {
2243 error = EINVAL;
2244 }
2245
2246 if (error) {
2247 pf_free_rule(rule);
2248 rule = NULL;
2249 ERROUT(error);
2250 }
2251
2252 rule->nat.cur = TAILQ_FIRST(&rule->nat.list);
2253 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list);
2254 rule->route.cur = TAILQ_FIRST(&rule->route.list);
2255 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2256 rule, entries);
2257 ruleset->rules[rs_num].inactive.rcount++;
2258
2259 PF_RULES_WUNLOCK();
2260 pf_hash_rule(rule);
2261 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2262 PF_RULES_WLOCK();
2263 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2264 ruleset->rules[rs_num].inactive.rcount--;
2265 pf_free_rule(rule);
2266 rule = NULL;
2267 ERROUT(EEXIST);
2268 }
2269 PF_CONFIG_UNLOCK();
2270
2271 return (0);
2272
2273 #undef ERROUT
2274 errout:
2275 PF_RULES_WUNLOCK();
2276 PF_CONFIG_UNLOCK();
2277 errout_unlocked:
2278 pf_kkif_free(rcv_kif);
2279 pf_kkif_free(kif);
2280 pf_krule_free(rule);
2281 return (error);
2282 }
2283
2284 static bool
pf_label_match(const struct pf_krule * rule,const char * label)2285 pf_label_match(const struct pf_krule *rule, const char *label)
2286 {
2287 int i = 0;
2288
2289 while (*rule->label[i]) {
2290 if (strcmp(rule->label[i], label) == 0)
2291 return (true);
2292 i++;
2293 }
2294
2295 return (false);
2296 }
2297
2298 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)2299 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2300 {
2301 struct pf_kstate *s;
2302 int more = 0;
2303
2304 s = pf_find_state_all(key, dir, &more);
2305 if (s == NULL)
2306 return (0);
2307
2308 if (more) {
2309 PF_STATE_UNLOCK(s);
2310 return (0);
2311 }
2312
2313 pf_remove_state(s);
2314 return (1);
2315 }
2316
2317 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)2318 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2319 {
2320 struct pf_kstate *s;
2321 struct pf_state_key *sk;
2322 struct pf_addr *srcaddr, *dstaddr;
2323 struct pf_state_key_cmp match_key;
2324 int idx, killed = 0;
2325 unsigned int dir;
2326 u_int16_t srcport, dstport;
2327 struct pfi_kkif *kif;
2328
2329 relock_DIOCKILLSTATES:
2330 PF_HASHROW_LOCK(ih);
2331 LIST_FOREACH(s, &ih->states, entry) {
2332 /* For floating states look at the original kif. */
2333 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2334
2335 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2336 if (s->direction == PF_OUT) {
2337 srcaddr = &sk->addr[1];
2338 dstaddr = &sk->addr[0];
2339 srcport = sk->port[1];
2340 dstport = sk->port[0];
2341 } else {
2342 srcaddr = &sk->addr[0];
2343 dstaddr = &sk->addr[1];
2344 srcport = sk->port[0];
2345 dstport = sk->port[1];
2346 }
2347
2348 if (psk->psk_af && sk->af != psk->psk_af)
2349 continue;
2350
2351 if (psk->psk_proto && psk->psk_proto != sk->proto)
2352 continue;
2353
2354 if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2355 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2356 continue;
2357
2358 if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2359 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2360 continue;
2361
2362 if (! PF_MATCHA(psk->psk_rt_addr.neg,
2363 &psk->psk_rt_addr.addr.v.a.addr,
2364 &psk->psk_rt_addr.addr.v.a.mask,
2365 &s->act.rt_addr, sk->af))
2366 continue;
2367
2368 if (psk->psk_src.port_op != 0 &&
2369 ! pf_match_port(psk->psk_src.port_op,
2370 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2371 continue;
2372
2373 if (psk->psk_dst.port_op != 0 &&
2374 ! pf_match_port(psk->psk_dst.port_op,
2375 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2376 continue;
2377
2378 if (psk->psk_label[0] &&
2379 ! pf_label_match(s->rule, psk->psk_label))
2380 continue;
2381
2382 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2383 kif->pfik_name))
2384 continue;
2385
2386 if (psk->psk_kill_match) {
2387 /* Create the key to find matching states, with lock
2388 * held. */
2389
2390 bzero(&match_key, sizeof(match_key));
2391
2392 if (s->direction == PF_OUT) {
2393 dir = PF_IN;
2394 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2395 } else {
2396 dir = PF_OUT;
2397 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2398 }
2399
2400 match_key.af = s->key[idx]->af;
2401 match_key.proto = s->key[idx]->proto;
2402 PF_ACPY(&match_key.addr[0],
2403 &s->key[idx]->addr[1], match_key.af);
2404 match_key.port[0] = s->key[idx]->port[1];
2405 PF_ACPY(&match_key.addr[1],
2406 &s->key[idx]->addr[0], match_key.af);
2407 match_key.port[1] = s->key[idx]->port[0];
2408 }
2409
2410 pf_remove_state(s);
2411 killed++;
2412
2413 if (psk->psk_kill_match)
2414 killed += pf_kill_matching_state(&match_key, dir);
2415
2416 goto relock_DIOCKILLSTATES;
2417 }
2418 PF_HASHROW_UNLOCK(ih);
2419
2420 return (killed);
2421 }
2422
2423 void
unhandled_af(int af)2424 unhandled_af(int af)
2425 {
2426 panic("unhandled af %d", af);
2427 }
2428
2429 int
pf_start(void)2430 pf_start(void)
2431 {
2432 int error = 0;
2433
2434 sx_xlock(&V_pf_ioctl_lock);
2435 if (V_pf_status.running)
2436 error = EEXIST;
2437 else {
2438 hook_pf();
2439 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2440 hook_pf_eth();
2441 V_pf_status.running = 1;
2442 V_pf_status.since = time_second;
2443 new_unrhdr64(&V_pf_stateid, time_second);
2444
2445 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2446 }
2447 sx_xunlock(&V_pf_ioctl_lock);
2448
2449 return (error);
2450 }
2451
2452 int
pf_stop(void)2453 pf_stop(void)
2454 {
2455 int error = 0;
2456
2457 sx_xlock(&V_pf_ioctl_lock);
2458 if (!V_pf_status.running)
2459 error = ENOENT;
2460 else {
2461 V_pf_status.running = 0;
2462 dehook_pf();
2463 dehook_pf_eth();
2464 V_pf_status.since = time_second;
2465 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2466 }
2467 sx_xunlock(&V_pf_ioctl_lock);
2468
2469 return (error);
2470 }
2471
2472 void
pf_ioctl_clear_status(void)2473 pf_ioctl_clear_status(void)
2474 {
2475 PF_RULES_WLOCK();
2476 for (int i = 0; i < PFRES_MAX; i++)
2477 counter_u64_zero(V_pf_status.counters[i]);
2478 for (int i = 0; i < FCNT_MAX; i++)
2479 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2480 for (int i = 0; i < SCNT_MAX; i++)
2481 counter_u64_zero(V_pf_status.scounters[i]);
2482 for (int i = 0; i < KLCNT_MAX; i++)
2483 counter_u64_zero(V_pf_status.lcounters[i]);
2484 V_pf_status.since = time_second;
2485 if (*V_pf_status.ifname)
2486 pfi_update_status(V_pf_status.ifname, NULL);
2487 PF_RULES_WUNLOCK();
2488 }
2489
2490 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2491 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2492 {
2493 uint32_t old;
2494
2495 if (timeout < 0 || timeout >= PFTM_MAX ||
2496 seconds < 0)
2497 return (EINVAL);
2498
2499 PF_RULES_WLOCK();
2500 old = V_pf_default_rule.timeout[timeout];
2501 if (timeout == PFTM_INTERVAL && seconds == 0)
2502 seconds = 1;
2503 V_pf_default_rule.timeout[timeout] = seconds;
2504 if (timeout == PFTM_INTERVAL && seconds < old)
2505 wakeup(pf_purge_thread);
2506
2507 if (prev_seconds != NULL)
2508 *prev_seconds = old;
2509
2510 PF_RULES_WUNLOCK();
2511
2512 return (0);
2513 }
2514
2515 int
pf_ioctl_get_timeout(int timeout,int * seconds)2516 pf_ioctl_get_timeout(int timeout, int *seconds)
2517 {
2518 PF_RULES_RLOCK_TRACKER;
2519
2520 if (timeout < 0 || timeout >= PFTM_MAX)
2521 return (EINVAL);
2522
2523 PF_RULES_RLOCK();
2524 *seconds = V_pf_default_rule.timeout[timeout];
2525 PF_RULES_RUNLOCK();
2526
2527 return (0);
2528 }
2529
2530 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2531 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2532 {
2533
2534 PF_RULES_WLOCK();
2535 if (index < 0 || index >= PF_LIMIT_MAX ||
2536 V_pf_limits[index].zone == NULL) {
2537 PF_RULES_WUNLOCK();
2538 return (EINVAL);
2539 }
2540 uma_zone_set_max(V_pf_limits[index].zone, limit);
2541 if (old_limit != NULL)
2542 *old_limit = V_pf_limits[index].limit;
2543 V_pf_limits[index].limit = limit;
2544 PF_RULES_WUNLOCK();
2545
2546 return (0);
2547 }
2548
2549 int
pf_ioctl_get_limit(int index,unsigned int * limit)2550 pf_ioctl_get_limit(int index, unsigned int *limit)
2551 {
2552 PF_RULES_RLOCK_TRACKER;
2553
2554 if (index < 0 || index >= PF_LIMIT_MAX)
2555 return (EINVAL);
2556
2557 PF_RULES_RLOCK();
2558 *limit = V_pf_limits[index].limit;
2559 PF_RULES_RUNLOCK();
2560
2561 return (0);
2562 }
2563
2564 int
pf_ioctl_begin_addrs(uint32_t * ticket)2565 pf_ioctl_begin_addrs(uint32_t *ticket)
2566 {
2567 PF_RULES_WLOCK();
2568 pf_empty_kpool(&V_pf_pabuf[0]);
2569 pf_empty_kpool(&V_pf_pabuf[1]);
2570 pf_empty_kpool(&V_pf_pabuf[2]);
2571 *ticket = ++V_ticket_pabuf;
2572 PF_RULES_WUNLOCK();
2573
2574 return (0);
2575 }
2576
2577 int
pf_ioctl_add_addr(struct pf_nl_pooladdr * pp)2578 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
2579 {
2580 struct pf_kpooladdr *pa = NULL;
2581 struct pfi_kkif *kif = NULL;
2582 int error;
2583
2584 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2585 pp->which != PF_RT)
2586 return (EINVAL);
2587
2588 switch (pp->af) {
2589 #ifdef INET
2590 case AF_INET:
2591 /* FALLTHROUGH */
2592 #endif /* INET */
2593 #ifdef INET6
2594 case AF_INET6:
2595 /* FALLTHROUGH */
2596 #endif /* INET6 */
2597 case AF_UNSPEC:
2598 break;
2599 default:
2600 return (EAFNOSUPPORT);
2601 }
2602
2603 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2604 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2605 pp->addr.addr.type != PF_ADDR_TABLE)
2606 return (EINVAL);
2607
2608 if (pp->addr.addr.p.dyn != NULL)
2609 return (EINVAL);
2610
2611 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2612 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2613 if (error != 0)
2614 goto out;
2615 if (pa->ifname[0])
2616 kif = pf_kkif_create(M_WAITOK);
2617 PF_RULES_WLOCK();
2618 if (pp->ticket != V_ticket_pabuf) {
2619 PF_RULES_WUNLOCK();
2620 if (pa->ifname[0])
2621 pf_kkif_free(kif);
2622 error = EBUSY;
2623 goto out;
2624 }
2625 if (pa->ifname[0]) {
2626 pa->kif = pfi_kkif_attach(kif, pa->ifname);
2627 kif = NULL;
2628 pfi_kkif_ref(pa->kif);
2629 } else
2630 pa->kif = NULL;
2631 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2632 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2633 if (pa->ifname[0])
2634 pfi_kkif_unref(pa->kif);
2635 PF_RULES_WUNLOCK();
2636 goto out;
2637 }
2638 switch (pp->which) {
2639 case PF_NAT:
2640 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries);
2641 break;
2642 case PF_RDR:
2643 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries);
2644 break;
2645 case PF_RT:
2646 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries);
2647 break;
2648 }
2649 PF_RULES_WUNLOCK();
2650
2651 return (0);
2652
2653 out:
2654 free(pa, M_PFRULE);
2655 return (error);
2656 }
2657
2658 int
pf_ioctl_get_addrs(struct pf_nl_pooladdr * pp)2659 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
2660 {
2661 struct pf_kpool *pool;
2662 struct pf_kpooladdr *pa;
2663
2664 PF_RULES_RLOCK_TRACKER;
2665
2666 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2667 pp->which != PF_RT)
2668 return (EINVAL);
2669
2670 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2671 pp->nr = 0;
2672
2673 PF_RULES_RLOCK();
2674 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2675 pp->r_num, 0, 1, 0, pp->which);
2676 if (pool == NULL) {
2677 PF_RULES_RUNLOCK();
2678 return (EBUSY);
2679 }
2680 TAILQ_FOREACH(pa, &pool->list, entries)
2681 pp->nr++;
2682 PF_RULES_RUNLOCK();
2683
2684 return (0);
2685 }
2686
2687 int
pf_ioctl_get_addr(struct pf_nl_pooladdr * pp)2688 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
2689 {
2690 struct pf_kpool *pool;
2691 struct pf_kpooladdr *pa;
2692 u_int32_t nr = 0;
2693
2694 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2695 pp->which != PF_RT)
2696 return (EINVAL);
2697
2698 PF_RULES_RLOCK_TRACKER;
2699
2700 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2701
2702 PF_RULES_RLOCK();
2703 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2704 pp->r_num, 0, 1, 1, pp->which);
2705 if (pool == NULL) {
2706 PF_RULES_RUNLOCK();
2707 return (EBUSY);
2708 }
2709 pa = TAILQ_FIRST(&pool->list);
2710 while ((pa != NULL) && (nr < pp->nr)) {
2711 pa = TAILQ_NEXT(pa, entries);
2712 nr++;
2713 }
2714 if (pa == NULL) {
2715 PF_RULES_RUNLOCK();
2716 return (EBUSY);
2717 }
2718 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2719 pf_addr_copyout(&pp->addr.addr);
2720 PF_RULES_RUNLOCK();
2721
2722 return (0);
2723 }
2724
2725 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)2726 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
2727 {
2728 struct pf_kruleset *ruleset;
2729 struct pf_kanchor *anchor;
2730
2731 PF_RULES_RLOCK_TRACKER;
2732
2733 pr->path[sizeof(pr->path) - 1] = 0;
2734
2735 PF_RULES_RLOCK();
2736 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2737 PF_RULES_RUNLOCK();
2738 return (ENOENT);
2739 }
2740 pr->nr = 0;
2741 if (ruleset->anchor == NULL) {
2742 /* XXX kludge for pf_main_ruleset */
2743 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2744 if (anchor->parent == NULL)
2745 pr->nr++;
2746 } else {
2747 RB_FOREACH(anchor, pf_kanchor_node,
2748 &ruleset->anchor->children)
2749 pr->nr++;
2750 }
2751 PF_RULES_RUNLOCK();
2752
2753 return (0);
2754 }
2755
2756 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)2757 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
2758 {
2759 struct pf_kruleset *ruleset;
2760 struct pf_kanchor *anchor;
2761 u_int32_t nr = 0;
2762 int error = 0;
2763
2764 PF_RULES_RLOCK_TRACKER;
2765
2766 PF_RULES_RLOCK();
2767 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2768 PF_RULES_RUNLOCK();
2769 return (ENOENT);
2770 }
2771
2772 pr->name[0] = 0;
2773 if (ruleset->anchor == NULL) {
2774 /* XXX kludge for pf_main_ruleset */
2775 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2776 if (anchor->parent == NULL && nr++ == pr->nr) {
2777 strlcpy(pr->name, anchor->name,
2778 sizeof(pr->name));
2779 break;
2780 }
2781 } else {
2782 RB_FOREACH(anchor, pf_kanchor_node,
2783 &ruleset->anchor->children)
2784 if (nr++ == pr->nr) {
2785 strlcpy(pr->name, anchor->name,
2786 sizeof(pr->name));
2787 break;
2788 }
2789 }
2790 if (!pr->name[0])
2791 error = EBUSY;
2792 PF_RULES_RUNLOCK();
2793
2794 return (error);
2795 }
2796
2797 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2798 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2799 {
2800 int error = 0;
2801 PF_RULES_RLOCK_TRACKER;
2802
2803 #define ERROUT_IOCTL(target, x) \
2804 do { \
2805 error = (x); \
2806 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
2807 goto target; \
2808 } while (0)
2809
2810
2811 /* XXX keep in sync with switch() below */
2812 if (securelevel_gt(td->td_ucred, 2))
2813 switch (cmd) {
2814 case DIOCGETRULES:
2815 case DIOCGETRULENV:
2816 case DIOCGETADDRS:
2817 case DIOCGETADDR:
2818 case DIOCGETSTATE:
2819 case DIOCGETSTATENV:
2820 case DIOCSETSTATUSIF:
2821 case DIOCGETSTATUSNV:
2822 case DIOCCLRSTATUS:
2823 case DIOCNATLOOK:
2824 case DIOCSETDEBUG:
2825 #ifdef COMPAT_FREEBSD14
2826 case DIOCGETSTATES:
2827 case DIOCGETSTATESV2:
2828 #endif
2829 case DIOCGETTIMEOUT:
2830 case DIOCCLRRULECTRS:
2831 case DIOCGETLIMIT:
2832 case DIOCGETALTQSV0:
2833 case DIOCGETALTQSV1:
2834 case DIOCGETALTQV0:
2835 case DIOCGETALTQV1:
2836 case DIOCGETQSTATSV0:
2837 case DIOCGETQSTATSV1:
2838 case DIOCGETRULESETS:
2839 case DIOCGETRULESET:
2840 case DIOCRGETTABLES:
2841 case DIOCRGETTSTATS:
2842 case DIOCRCLRTSTATS:
2843 case DIOCRCLRADDRS:
2844 case DIOCRADDADDRS:
2845 case DIOCRDELADDRS:
2846 case DIOCRSETADDRS:
2847 case DIOCRGETADDRS:
2848 case DIOCRGETASTATS:
2849 case DIOCRCLRASTATS:
2850 case DIOCRTSTADDRS:
2851 case DIOCOSFPGET:
2852 case DIOCGETSRCNODES:
2853 case DIOCCLRSRCNODES:
2854 case DIOCGETSYNCOOKIES:
2855 case DIOCIGETIFACES:
2856 case DIOCGIFSPEEDV0:
2857 case DIOCGIFSPEEDV1:
2858 case DIOCSETIFFLAG:
2859 case DIOCCLRIFFLAG:
2860 case DIOCGETETHRULES:
2861 case DIOCGETETHRULE:
2862 case DIOCGETETHRULESETS:
2863 case DIOCGETETHRULESET:
2864 break;
2865 case DIOCRCLRTABLES:
2866 case DIOCRADDTABLES:
2867 case DIOCRDELTABLES:
2868 case DIOCRSETTFLAGS:
2869 if (((struct pfioc_table *)addr)->pfrio_flags &
2870 PFR_FLAG_DUMMY)
2871 break; /* dummy operation ok */
2872 return (EPERM);
2873 default:
2874 return (EPERM);
2875 }
2876
2877 if (!(flags & FWRITE))
2878 switch (cmd) {
2879 case DIOCGETRULES:
2880 case DIOCGETADDRS:
2881 case DIOCGETADDR:
2882 case DIOCGETSTATE:
2883 case DIOCGETSTATENV:
2884 case DIOCGETSTATUSNV:
2885 #ifdef COMPAT_FREEBSD14
2886 case DIOCGETSTATES:
2887 case DIOCGETSTATESV2:
2888 #endif
2889 case DIOCGETTIMEOUT:
2890 case DIOCGETLIMIT:
2891 case DIOCGETALTQSV0:
2892 case DIOCGETALTQSV1:
2893 case DIOCGETALTQV0:
2894 case DIOCGETALTQV1:
2895 case DIOCGETQSTATSV0:
2896 case DIOCGETQSTATSV1:
2897 case DIOCGETRULESETS:
2898 case DIOCGETRULESET:
2899 case DIOCNATLOOK:
2900 case DIOCRGETTABLES:
2901 case DIOCRGETTSTATS:
2902 case DIOCRGETADDRS:
2903 case DIOCRGETASTATS:
2904 case DIOCRTSTADDRS:
2905 case DIOCOSFPGET:
2906 case DIOCGETSRCNODES:
2907 case DIOCGETSYNCOOKIES:
2908 case DIOCIGETIFACES:
2909 case DIOCGIFSPEEDV1:
2910 case DIOCGIFSPEEDV0:
2911 case DIOCGETRULENV:
2912 case DIOCGETETHRULES:
2913 case DIOCGETETHRULE:
2914 case DIOCGETETHRULESETS:
2915 case DIOCGETETHRULESET:
2916 break;
2917 case DIOCRCLRTABLES:
2918 case DIOCRADDTABLES:
2919 case DIOCRDELTABLES:
2920 case DIOCRCLRTSTATS:
2921 case DIOCRCLRADDRS:
2922 case DIOCRADDADDRS:
2923 case DIOCRDELADDRS:
2924 case DIOCRSETADDRS:
2925 case DIOCRSETTFLAGS:
2926 if (((struct pfioc_table *)addr)->pfrio_flags &
2927 PFR_FLAG_DUMMY) {
2928 flags |= FWRITE; /* need write lock for dummy */
2929 break; /* dummy operation ok */
2930 }
2931 return (EACCES);
2932 default:
2933 return (EACCES);
2934 }
2935
2936 CURVNET_SET(TD_TO_VNET(td));
2937
2938 switch (cmd) {
2939 #ifdef COMPAT_FREEBSD14
2940 case DIOCSTART:
2941 error = pf_start();
2942 break;
2943
2944 case DIOCSTOP:
2945 error = pf_stop();
2946 break;
2947 #endif
2948
2949 case DIOCGETETHRULES: {
2950 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2951 nvlist_t *nvl;
2952 void *packed;
2953 struct pf_keth_rule *tail;
2954 struct pf_keth_ruleset *rs;
2955 u_int32_t ticket, nr;
2956 const char *anchor = "";
2957
2958 nvl = NULL;
2959 packed = NULL;
2960
2961 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2962
2963 if (nv->len > pf_ioctl_maxcount)
2964 ERROUT(ENOMEM);
2965
2966 /* Copy the request in */
2967 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2968 error = copyin(nv->data, packed, nv->len);
2969 if (error)
2970 ERROUT(error);
2971
2972 nvl = nvlist_unpack(packed, nv->len, 0);
2973 if (nvl == NULL)
2974 ERROUT(EBADMSG);
2975
2976 if (! nvlist_exists_string(nvl, "anchor"))
2977 ERROUT(EBADMSG);
2978
2979 anchor = nvlist_get_string(nvl, "anchor");
2980
2981 rs = pf_find_keth_ruleset(anchor);
2982
2983 nvlist_destroy(nvl);
2984 nvl = NULL;
2985 free(packed, M_NVLIST);
2986 packed = NULL;
2987
2988 if (rs == NULL)
2989 ERROUT(ENOENT);
2990
2991 /* Reply */
2992 nvl = nvlist_create(0);
2993 if (nvl == NULL)
2994 ERROUT(ENOMEM);
2995
2996 PF_RULES_RLOCK();
2997
2998 ticket = rs->active.ticket;
2999 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
3000 if (tail)
3001 nr = tail->nr + 1;
3002 else
3003 nr = 0;
3004
3005 PF_RULES_RUNLOCK();
3006
3007 nvlist_add_number(nvl, "ticket", ticket);
3008 nvlist_add_number(nvl, "nr", nr);
3009
3010 packed = nvlist_pack(nvl, &nv->len);
3011 if (packed == NULL)
3012 ERROUT(ENOMEM);
3013
3014 if (nv->size == 0)
3015 ERROUT(0);
3016 else if (nv->size < nv->len)
3017 ERROUT(ENOSPC);
3018
3019 error = copyout(packed, nv->data, nv->len);
3020
3021 #undef ERROUT
3022 DIOCGETETHRULES_error:
3023 free(packed, M_NVLIST);
3024 nvlist_destroy(nvl);
3025 break;
3026 }
3027
3028 case DIOCGETETHRULE: {
3029 struct epoch_tracker et;
3030 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3031 nvlist_t *nvl = NULL;
3032 void *nvlpacked = NULL;
3033 struct pf_keth_rule *rule = NULL;
3034 struct pf_keth_ruleset *rs;
3035 u_int32_t ticket, nr;
3036 bool clear = false;
3037 const char *anchor;
3038
3039 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
3040
3041 if (nv->len > pf_ioctl_maxcount)
3042 ERROUT(ENOMEM);
3043
3044 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3045 error = copyin(nv->data, nvlpacked, nv->len);
3046 if (error)
3047 ERROUT(error);
3048
3049 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3050 if (nvl == NULL)
3051 ERROUT(EBADMSG);
3052 if (! nvlist_exists_number(nvl, "ticket"))
3053 ERROUT(EBADMSG);
3054 ticket = nvlist_get_number(nvl, "ticket");
3055 if (! nvlist_exists_string(nvl, "anchor"))
3056 ERROUT(EBADMSG);
3057 anchor = nvlist_get_string(nvl, "anchor");
3058
3059 if (nvlist_exists_bool(nvl, "clear"))
3060 clear = nvlist_get_bool(nvl, "clear");
3061
3062 if (clear && !(flags & FWRITE))
3063 ERROUT(EACCES);
3064
3065 if (! nvlist_exists_number(nvl, "nr"))
3066 ERROUT(EBADMSG);
3067 nr = nvlist_get_number(nvl, "nr");
3068
3069 PF_RULES_RLOCK();
3070 rs = pf_find_keth_ruleset(anchor);
3071 if (rs == NULL) {
3072 PF_RULES_RUNLOCK();
3073 ERROUT(ENOENT);
3074 }
3075 if (ticket != rs->active.ticket) {
3076 PF_RULES_RUNLOCK();
3077 ERROUT(EBUSY);
3078 }
3079
3080 nvlist_destroy(nvl);
3081 nvl = NULL;
3082 free(nvlpacked, M_NVLIST);
3083 nvlpacked = NULL;
3084
3085 rule = TAILQ_FIRST(rs->active.rules);
3086 while ((rule != NULL) && (rule->nr != nr))
3087 rule = TAILQ_NEXT(rule, entries);
3088 if (rule == NULL) {
3089 PF_RULES_RUNLOCK();
3090 ERROUT(ENOENT);
3091 }
3092 /* Make sure rule can't go away. */
3093 NET_EPOCH_ENTER(et);
3094 PF_RULES_RUNLOCK();
3095 nvl = pf_keth_rule_to_nveth_rule(rule);
3096 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
3097 NET_EPOCH_EXIT(et);
3098 ERROUT(EBUSY);
3099 }
3100 NET_EPOCH_EXIT(et);
3101 if (nvl == NULL)
3102 ERROUT(ENOMEM);
3103
3104 nvlpacked = nvlist_pack(nvl, &nv->len);
3105 if (nvlpacked == NULL)
3106 ERROUT(ENOMEM);
3107
3108 if (nv->size == 0)
3109 ERROUT(0);
3110 else if (nv->size < nv->len)
3111 ERROUT(ENOSPC);
3112
3113 error = copyout(nvlpacked, nv->data, nv->len);
3114 if (error == 0 && clear) {
3115 counter_u64_zero(rule->evaluations);
3116 for (int i = 0; i < 2; i++) {
3117 counter_u64_zero(rule->packets[i]);
3118 counter_u64_zero(rule->bytes[i]);
3119 }
3120 }
3121
3122 #undef ERROUT
3123 DIOCGETETHRULE_error:
3124 free(nvlpacked, M_NVLIST);
3125 nvlist_destroy(nvl);
3126 break;
3127 }
3128
3129 case DIOCADDETHRULE: {
3130 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3131 nvlist_t *nvl = NULL;
3132 void *nvlpacked = NULL;
3133 struct pf_keth_rule *rule = NULL, *tail = NULL;
3134 struct pf_keth_ruleset *ruleset = NULL;
3135 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
3136 const char *anchor = "", *anchor_call = "";
3137
3138 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3139
3140 if (nv->len > pf_ioctl_maxcount)
3141 ERROUT(ENOMEM);
3142
3143 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3144 error = copyin(nv->data, nvlpacked, nv->len);
3145 if (error)
3146 ERROUT(error);
3147
3148 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3149 if (nvl == NULL)
3150 ERROUT(EBADMSG);
3151
3152 if (! nvlist_exists_number(nvl, "ticket"))
3153 ERROUT(EBADMSG);
3154
3155 if (nvlist_exists_string(nvl, "anchor"))
3156 anchor = nvlist_get_string(nvl, "anchor");
3157 if (nvlist_exists_string(nvl, "anchor_call"))
3158 anchor_call = nvlist_get_string(nvl, "anchor_call");
3159
3160 ruleset = pf_find_keth_ruleset(anchor);
3161 if (ruleset == NULL)
3162 ERROUT(EINVAL);
3163
3164 if (nvlist_get_number(nvl, "ticket") !=
3165 ruleset->inactive.ticket) {
3166 DPFPRINTF(PF_DEBUG_MISC,
3167 ("ticket: %d != %d\n",
3168 (u_int32_t)nvlist_get_number(nvl, "ticket"),
3169 ruleset->inactive.ticket));
3170 ERROUT(EBUSY);
3171 }
3172
3173 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3174 rule->timestamp = NULL;
3175
3176 error = pf_nveth_rule_to_keth_rule(nvl, rule);
3177 if (error != 0)
3178 ERROUT(error);
3179
3180 if (rule->ifname[0])
3181 kif = pf_kkif_create(M_WAITOK);
3182 if (rule->bridge_to_name[0])
3183 bridge_to_kif = pf_kkif_create(M_WAITOK);
3184 rule->evaluations = counter_u64_alloc(M_WAITOK);
3185 for (int i = 0; i < 2; i++) {
3186 rule->packets[i] = counter_u64_alloc(M_WAITOK);
3187 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3188 }
3189 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3190 M_WAITOK | M_ZERO);
3191
3192 PF_RULES_WLOCK();
3193
3194 if (rule->ifname[0]) {
3195 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3196 pfi_kkif_ref(rule->kif);
3197 } else
3198 rule->kif = NULL;
3199 if (rule->bridge_to_name[0]) {
3200 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3201 rule->bridge_to_name);
3202 pfi_kkif_ref(rule->bridge_to);
3203 } else
3204 rule->bridge_to = NULL;
3205
3206 #ifdef ALTQ
3207 /* set queue IDs */
3208 if (rule->qname[0] != 0) {
3209 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3210 error = EBUSY;
3211 else
3212 rule->qid = rule->qid;
3213 }
3214 #endif
3215 if (rule->tagname[0])
3216 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3217 error = EBUSY;
3218 if (rule->match_tagname[0])
3219 if ((rule->match_tag = pf_tagname2tag(
3220 rule->match_tagname)) == 0)
3221 error = EBUSY;
3222
3223 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3224 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3225 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3226 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3227
3228 if (error) {
3229 pf_free_eth_rule(rule);
3230 PF_RULES_WUNLOCK();
3231 ERROUT(error);
3232 }
3233
3234 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3235 pf_free_eth_rule(rule);
3236 PF_RULES_WUNLOCK();
3237 ERROUT(EINVAL);
3238 }
3239
3240 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3241 if (tail)
3242 rule->nr = tail->nr + 1;
3243 else
3244 rule->nr = 0;
3245
3246 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3247
3248 PF_RULES_WUNLOCK();
3249
3250 #undef ERROUT
3251 DIOCADDETHRULE_error:
3252 nvlist_destroy(nvl);
3253 free(nvlpacked, M_NVLIST);
3254 break;
3255 }
3256
3257 case DIOCGETETHRULESETS: {
3258 struct epoch_tracker et;
3259 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3260 nvlist_t *nvl = NULL;
3261 void *nvlpacked = NULL;
3262 struct pf_keth_ruleset *ruleset;
3263 struct pf_keth_anchor *anchor;
3264 int nr = 0;
3265
3266 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3267
3268 if (nv->len > pf_ioctl_maxcount)
3269 ERROUT(ENOMEM);
3270
3271 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3272 error = copyin(nv->data, nvlpacked, nv->len);
3273 if (error)
3274 ERROUT(error);
3275
3276 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3277 if (nvl == NULL)
3278 ERROUT(EBADMSG);
3279 if (! nvlist_exists_string(nvl, "path"))
3280 ERROUT(EBADMSG);
3281
3282 NET_EPOCH_ENTER(et);
3283
3284 if ((ruleset = pf_find_keth_ruleset(
3285 nvlist_get_string(nvl, "path"))) == NULL) {
3286 NET_EPOCH_EXIT(et);
3287 ERROUT(ENOENT);
3288 }
3289
3290 if (ruleset->anchor == NULL) {
3291 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3292 if (anchor->parent == NULL)
3293 nr++;
3294 } else {
3295 RB_FOREACH(anchor, pf_keth_anchor_node,
3296 &ruleset->anchor->children)
3297 nr++;
3298 }
3299
3300 NET_EPOCH_EXIT(et);
3301
3302 nvlist_destroy(nvl);
3303 nvl = NULL;
3304 free(nvlpacked, M_NVLIST);
3305 nvlpacked = NULL;
3306
3307 nvl = nvlist_create(0);
3308 if (nvl == NULL)
3309 ERROUT(ENOMEM);
3310
3311 nvlist_add_number(nvl, "nr", nr);
3312
3313 nvlpacked = nvlist_pack(nvl, &nv->len);
3314 if (nvlpacked == NULL)
3315 ERROUT(ENOMEM);
3316
3317 if (nv->size == 0)
3318 ERROUT(0);
3319 else if (nv->size < nv->len)
3320 ERROUT(ENOSPC);
3321
3322 error = copyout(nvlpacked, nv->data, nv->len);
3323
3324 #undef ERROUT
3325 DIOCGETETHRULESETS_error:
3326 free(nvlpacked, M_NVLIST);
3327 nvlist_destroy(nvl);
3328 break;
3329 }
3330
3331 case DIOCGETETHRULESET: {
3332 struct epoch_tracker et;
3333 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3334 nvlist_t *nvl = NULL;
3335 void *nvlpacked = NULL;
3336 struct pf_keth_ruleset *ruleset;
3337 struct pf_keth_anchor *anchor;
3338 int nr = 0, req_nr = 0;
3339 bool found = false;
3340
3341 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3342
3343 if (nv->len > pf_ioctl_maxcount)
3344 ERROUT(ENOMEM);
3345
3346 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3347 error = copyin(nv->data, nvlpacked, nv->len);
3348 if (error)
3349 ERROUT(error);
3350
3351 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3352 if (nvl == NULL)
3353 ERROUT(EBADMSG);
3354 if (! nvlist_exists_string(nvl, "path"))
3355 ERROUT(EBADMSG);
3356 if (! nvlist_exists_number(nvl, "nr"))
3357 ERROUT(EBADMSG);
3358
3359 req_nr = nvlist_get_number(nvl, "nr");
3360
3361 NET_EPOCH_ENTER(et);
3362
3363 if ((ruleset = pf_find_keth_ruleset(
3364 nvlist_get_string(nvl, "path"))) == NULL) {
3365 NET_EPOCH_EXIT(et);
3366 ERROUT(ENOENT);
3367 }
3368
3369 nvlist_destroy(nvl);
3370 nvl = NULL;
3371 free(nvlpacked, M_NVLIST);
3372 nvlpacked = NULL;
3373
3374 nvl = nvlist_create(0);
3375 if (nvl == NULL) {
3376 NET_EPOCH_EXIT(et);
3377 ERROUT(ENOMEM);
3378 }
3379
3380 if (ruleset->anchor == NULL) {
3381 RB_FOREACH(anchor, pf_keth_anchor_global,
3382 &V_pf_keth_anchors) {
3383 if (anchor->parent == NULL && nr++ == req_nr) {
3384 found = true;
3385 break;
3386 }
3387 }
3388 } else {
3389 RB_FOREACH(anchor, pf_keth_anchor_node,
3390 &ruleset->anchor->children) {
3391 if (nr++ == req_nr) {
3392 found = true;
3393 break;
3394 }
3395 }
3396 }
3397
3398 NET_EPOCH_EXIT(et);
3399 if (found) {
3400 nvlist_add_number(nvl, "nr", nr);
3401 nvlist_add_string(nvl, "name", anchor->name);
3402 if (ruleset->anchor)
3403 nvlist_add_string(nvl, "path",
3404 ruleset->anchor->path);
3405 else
3406 nvlist_add_string(nvl, "path", "");
3407 } else {
3408 ERROUT(EBUSY);
3409 }
3410
3411 nvlpacked = nvlist_pack(nvl, &nv->len);
3412 if (nvlpacked == NULL)
3413 ERROUT(ENOMEM);
3414
3415 if (nv->size == 0)
3416 ERROUT(0);
3417 else if (nv->size < nv->len)
3418 ERROUT(ENOSPC);
3419
3420 error = copyout(nvlpacked, nv->data, nv->len);
3421
3422 #undef ERROUT
3423 DIOCGETETHRULESET_error:
3424 free(nvlpacked, M_NVLIST);
3425 nvlist_destroy(nvl);
3426 break;
3427 }
3428
3429 case DIOCADDRULENV: {
3430 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3431 nvlist_t *nvl = NULL;
3432 void *nvlpacked = NULL;
3433 struct pf_krule *rule = NULL;
3434 const char *anchor = "", *anchor_call = "";
3435 uint32_t ticket = 0, pool_ticket = 0;
3436
3437 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
3438
3439 if (nv->len > pf_ioctl_maxcount)
3440 ERROUT(ENOMEM);
3441
3442 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3443 error = copyin(nv->data, nvlpacked, nv->len);
3444 if (error)
3445 ERROUT(error);
3446
3447 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3448 if (nvl == NULL)
3449 ERROUT(EBADMSG);
3450
3451 if (! nvlist_exists_number(nvl, "ticket"))
3452 ERROUT(EINVAL);
3453 ticket = nvlist_get_number(nvl, "ticket");
3454
3455 if (! nvlist_exists_number(nvl, "pool_ticket"))
3456 ERROUT(EINVAL);
3457 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3458
3459 if (! nvlist_exists_nvlist(nvl, "rule"))
3460 ERROUT(EINVAL);
3461
3462 rule = pf_krule_alloc();
3463 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3464 rule);
3465 if (error)
3466 ERROUT(error);
3467
3468 if (nvlist_exists_string(nvl, "anchor"))
3469 anchor = nvlist_get_string(nvl, "anchor");
3470 if (nvlist_exists_string(nvl, "anchor_call"))
3471 anchor_call = nvlist_get_string(nvl, "anchor_call");
3472
3473 if ((error = nvlist_error(nvl)))
3474 ERROUT(error);
3475
3476 /* Frees rule on error */
3477 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3478 anchor_call, td->td_ucred->cr_ruid,
3479 td->td_proc ? td->td_proc->p_pid : 0);
3480
3481 nvlist_destroy(nvl);
3482 free(nvlpacked, M_NVLIST);
3483 break;
3484 #undef ERROUT
3485 DIOCADDRULENV_error:
3486 pf_krule_free(rule);
3487 nvlist_destroy(nvl);
3488 free(nvlpacked, M_NVLIST);
3489
3490 break;
3491 }
3492 case DIOCADDRULE: {
3493 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3494 struct pf_krule *rule;
3495
3496 rule = pf_krule_alloc();
3497 error = pf_rule_to_krule(&pr->rule, rule);
3498 if (error != 0) {
3499 pf_krule_free(rule);
3500 break;
3501 }
3502
3503 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3504
3505 /* Frees rule on error */
3506 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3507 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3508 td->td_proc ? td->td_proc->p_pid : 0);
3509 break;
3510 }
3511
3512 case DIOCGETRULES: {
3513 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3514
3515 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3516
3517 error = pf_ioctl_getrules(pr);
3518
3519 break;
3520 }
3521
3522 case DIOCGETRULENV: {
3523 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3524 nvlist_t *nvrule = NULL;
3525 nvlist_t *nvl = NULL;
3526 struct pf_kruleset *ruleset;
3527 struct pf_krule *rule;
3528 void *nvlpacked = NULL;
3529 int rs_num, nr;
3530 bool clear_counter = false;
3531
3532 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
3533
3534 if (nv->len > pf_ioctl_maxcount)
3535 ERROUT(ENOMEM);
3536
3537 /* Copy the request in */
3538 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3539 error = copyin(nv->data, nvlpacked, nv->len);
3540 if (error)
3541 ERROUT(error);
3542
3543 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3544 if (nvl == NULL)
3545 ERROUT(EBADMSG);
3546
3547 if (! nvlist_exists_string(nvl, "anchor"))
3548 ERROUT(EBADMSG);
3549 if (! nvlist_exists_number(nvl, "ruleset"))
3550 ERROUT(EBADMSG);
3551 if (! nvlist_exists_number(nvl, "ticket"))
3552 ERROUT(EBADMSG);
3553 if (! nvlist_exists_number(nvl, "nr"))
3554 ERROUT(EBADMSG);
3555
3556 if (nvlist_exists_bool(nvl, "clear_counter"))
3557 clear_counter = nvlist_get_bool(nvl, "clear_counter");
3558
3559 if (clear_counter && !(flags & FWRITE))
3560 ERROUT(EACCES);
3561
3562 nr = nvlist_get_number(nvl, "nr");
3563
3564 PF_RULES_WLOCK();
3565 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3566 if (ruleset == NULL) {
3567 PF_RULES_WUNLOCK();
3568 ERROUT(ENOENT);
3569 }
3570
3571 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3572 if (rs_num >= PF_RULESET_MAX) {
3573 PF_RULES_WUNLOCK();
3574 ERROUT(EINVAL);
3575 }
3576
3577 if (nvlist_get_number(nvl, "ticket") !=
3578 ruleset->rules[rs_num].active.ticket) {
3579 PF_RULES_WUNLOCK();
3580 ERROUT(EBUSY);
3581 }
3582
3583 if ((error = nvlist_error(nvl))) {
3584 PF_RULES_WUNLOCK();
3585 ERROUT(error);
3586 }
3587
3588 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3589 while ((rule != NULL) && (rule->nr != nr))
3590 rule = TAILQ_NEXT(rule, entries);
3591 if (rule == NULL) {
3592 PF_RULES_WUNLOCK();
3593 ERROUT(EBUSY);
3594 }
3595
3596 nvrule = pf_krule_to_nvrule(rule);
3597
3598 nvlist_destroy(nvl);
3599 nvl = nvlist_create(0);
3600 if (nvl == NULL) {
3601 PF_RULES_WUNLOCK();
3602 ERROUT(ENOMEM);
3603 }
3604 nvlist_add_number(nvl, "nr", nr);
3605 nvlist_add_nvlist(nvl, "rule", nvrule);
3606 nvlist_destroy(nvrule);
3607 nvrule = NULL;
3608 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3609 PF_RULES_WUNLOCK();
3610 ERROUT(EBUSY);
3611 }
3612
3613 free(nvlpacked, M_NVLIST);
3614 nvlpacked = nvlist_pack(nvl, &nv->len);
3615 if (nvlpacked == NULL) {
3616 PF_RULES_WUNLOCK();
3617 ERROUT(ENOMEM);
3618 }
3619
3620 if (nv->size == 0) {
3621 PF_RULES_WUNLOCK();
3622 ERROUT(0);
3623 }
3624 else if (nv->size < nv->len) {
3625 PF_RULES_WUNLOCK();
3626 ERROUT(ENOSPC);
3627 }
3628
3629 if (clear_counter)
3630 pf_krule_clear_counters(rule);
3631
3632 PF_RULES_WUNLOCK();
3633
3634 error = copyout(nvlpacked, nv->data, nv->len);
3635
3636 #undef ERROUT
3637 DIOCGETRULENV_error:
3638 free(nvlpacked, M_NVLIST);
3639 nvlist_destroy(nvrule);
3640 nvlist_destroy(nvl);
3641
3642 break;
3643 }
3644
3645 case DIOCCHANGERULE: {
3646 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
3647 struct pf_kruleset *ruleset;
3648 struct pf_krule *oldrule = NULL, *newrule = NULL;
3649 struct pfi_kkif *kif = NULL;
3650 struct pf_kpooladdr *pa;
3651 u_int32_t nr = 0;
3652 int rs_num;
3653
3654 pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3655
3656 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3657 pcr->action > PF_CHANGE_GET_TICKET) {
3658 error = EINVAL;
3659 break;
3660 }
3661 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3662 error = EINVAL;
3663 break;
3664 }
3665
3666 if (pcr->action != PF_CHANGE_REMOVE) {
3667 newrule = pf_krule_alloc();
3668 error = pf_rule_to_krule(&pcr->rule, newrule);
3669 if (error != 0) {
3670 pf_krule_free(newrule);
3671 break;
3672 }
3673
3674 if (newrule->ifname[0])
3675 kif = pf_kkif_create(M_WAITOK);
3676 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3677 for (int i = 0; i < 2; i++) {
3678 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3679 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3680 }
3681 newrule->states_cur = counter_u64_alloc(M_WAITOK);
3682 newrule->states_tot = counter_u64_alloc(M_WAITOK);
3683 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
3684 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
3685 newrule->cuid = td->td_ucred->cr_ruid;
3686 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3687 TAILQ_INIT(&newrule->nat.list);
3688 TAILQ_INIT(&newrule->rdr.list);
3689 TAILQ_INIT(&newrule->route.list);
3690 }
3691 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3692
3693 PF_CONFIG_LOCK();
3694 PF_RULES_WLOCK();
3695 #ifdef PF_WANT_32_TO_64_COUNTER
3696 if (newrule != NULL) {
3697 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3698 newrule->allrulelinked = true;
3699 V_pf_allrulecount++;
3700 }
3701 #endif
3702
3703 if (!(pcr->action == PF_CHANGE_REMOVE ||
3704 pcr->action == PF_CHANGE_GET_TICKET) &&
3705 pcr->pool_ticket != V_ticket_pabuf)
3706 ERROUT(EBUSY);
3707
3708 ruleset = pf_find_kruleset(pcr->anchor);
3709 if (ruleset == NULL)
3710 ERROUT(EINVAL);
3711
3712 rs_num = pf_get_ruleset_number(pcr->rule.action);
3713 if (rs_num >= PF_RULESET_MAX)
3714 ERROUT(EINVAL);
3715
3716 /*
3717 * XXXMJG: there is no guarantee that the ruleset was
3718 * created by the usual route of calling DIOCXBEGIN.
3719 * As a result it is possible the rule tree will not
3720 * be allocated yet. Hack around it by doing it here.
3721 * Note it is fine to let the tree persist in case of
3722 * error as it will be freed down the road on future
3723 * updates (if need be).
3724 */
3725 if (ruleset->rules[rs_num].active.tree == NULL) {
3726 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3727 if (ruleset->rules[rs_num].active.tree == NULL) {
3728 ERROUT(ENOMEM);
3729 }
3730 }
3731
3732 if (pcr->action == PF_CHANGE_GET_TICKET) {
3733 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3734 ERROUT(0);
3735 } else if (pcr->ticket !=
3736 ruleset->rules[rs_num].active.ticket)
3737 ERROUT(EINVAL);
3738
3739 if (pcr->action != PF_CHANGE_REMOVE) {
3740 if (newrule->ifname[0]) {
3741 newrule->kif = pfi_kkif_attach(kif,
3742 newrule->ifname);
3743 kif = NULL;
3744 pfi_kkif_ref(newrule->kif);
3745 } else
3746 newrule->kif = NULL;
3747
3748 if (newrule->rtableid > 0 &&
3749 newrule->rtableid >= rt_numfibs)
3750 error = EBUSY;
3751
3752 #ifdef ALTQ
3753 /* set queue IDs */
3754 if (newrule->qname[0] != 0) {
3755 if ((newrule->qid =
3756 pf_qname2qid(newrule->qname)) == 0)
3757 error = EBUSY;
3758 else if (newrule->pqname[0] != 0) {
3759 if ((newrule->pqid =
3760 pf_qname2qid(newrule->pqname)) == 0)
3761 error = EBUSY;
3762 } else
3763 newrule->pqid = newrule->qid;
3764 }
3765 #endif /* ALTQ */
3766 if (newrule->tagname[0])
3767 if ((newrule->tag =
3768 pf_tagname2tag(newrule->tagname)) == 0)
3769 error = EBUSY;
3770 if (newrule->match_tagname[0])
3771 if ((newrule->match_tag = pf_tagname2tag(
3772 newrule->match_tagname)) == 0)
3773 error = EBUSY;
3774 if (newrule->rt && !newrule->direction)
3775 error = EINVAL;
3776 if (!newrule->log)
3777 newrule->logif = 0;
3778 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3779 error = ENOMEM;
3780 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3781 error = ENOMEM;
3782 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3783 error = EINVAL;
3784 for (int i = 0; i < 3; i++) {
3785 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
3786 if (pa->addr.type == PF_ADDR_TABLE) {
3787 pa->addr.p.tbl =
3788 pfr_attach_table(ruleset,
3789 pa->addr.v.tblname);
3790 if (pa->addr.p.tbl == NULL)
3791 error = ENOMEM;
3792 }
3793 }
3794
3795 newrule->overload_tbl = NULL;
3796 if (newrule->overload_tblname[0]) {
3797 if ((newrule->overload_tbl = pfr_attach_table(
3798 ruleset, newrule->overload_tblname)) ==
3799 NULL)
3800 error = EINVAL;
3801 else
3802 newrule->overload_tbl->pfrkt_flags |=
3803 PFR_TFLAG_ACTIVE;
3804 }
3805
3806 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
3807 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
3808 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list);
3809 if (((((newrule->action == PF_NAT) ||
3810 (newrule->action == PF_RDR) ||
3811 (newrule->action == PF_BINAT) ||
3812 (newrule->rt > PF_NOPFROUTE)) &&
3813 !newrule->anchor)) &&
3814 (TAILQ_FIRST(&newrule->rdr.list) == NULL))
3815 error = EINVAL;
3816
3817 if (error) {
3818 pf_free_rule(newrule);
3819 PF_RULES_WUNLOCK();
3820 PF_CONFIG_UNLOCK();
3821 break;
3822 }
3823
3824 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
3825 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
3826 }
3827 pf_empty_kpool(&V_pf_pabuf[0]);
3828 pf_empty_kpool(&V_pf_pabuf[1]);
3829 pf_empty_kpool(&V_pf_pabuf[2]);
3830
3831 if (pcr->action == PF_CHANGE_ADD_HEAD)
3832 oldrule = TAILQ_FIRST(
3833 ruleset->rules[rs_num].active.ptr);
3834 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3835 oldrule = TAILQ_LAST(
3836 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3837 else {
3838 oldrule = TAILQ_FIRST(
3839 ruleset->rules[rs_num].active.ptr);
3840 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3841 oldrule = TAILQ_NEXT(oldrule, entries);
3842 if (oldrule == NULL) {
3843 if (newrule != NULL)
3844 pf_free_rule(newrule);
3845 PF_RULES_WUNLOCK();
3846 PF_CONFIG_UNLOCK();
3847 error = EINVAL;
3848 break;
3849 }
3850 }
3851
3852 if (pcr->action == PF_CHANGE_REMOVE) {
3853 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3854 oldrule);
3855 RB_REMOVE(pf_krule_global,
3856 ruleset->rules[rs_num].active.tree, oldrule);
3857 ruleset->rules[rs_num].active.rcount--;
3858 } else {
3859 pf_hash_rule(newrule);
3860 if (RB_INSERT(pf_krule_global,
3861 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3862 pf_free_rule(newrule);
3863 PF_RULES_WUNLOCK();
3864 PF_CONFIG_UNLOCK();
3865 error = EEXIST;
3866 break;
3867 }
3868
3869 if (oldrule == NULL)
3870 TAILQ_INSERT_TAIL(
3871 ruleset->rules[rs_num].active.ptr,
3872 newrule, entries);
3873 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3874 pcr->action == PF_CHANGE_ADD_BEFORE)
3875 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3876 else
3877 TAILQ_INSERT_AFTER(
3878 ruleset->rules[rs_num].active.ptr,
3879 oldrule, newrule, entries);
3880 ruleset->rules[rs_num].active.rcount++;
3881 }
3882
3883 nr = 0;
3884 TAILQ_FOREACH(oldrule,
3885 ruleset->rules[rs_num].active.ptr, entries)
3886 oldrule->nr = nr++;
3887
3888 ruleset->rules[rs_num].active.ticket++;
3889
3890 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3891 pf_remove_if_empty_kruleset(ruleset);
3892
3893 PF_RULES_WUNLOCK();
3894 PF_CONFIG_UNLOCK();
3895 break;
3896
3897 #undef ERROUT
3898 DIOCCHANGERULE_error:
3899 PF_RULES_WUNLOCK();
3900 PF_CONFIG_UNLOCK();
3901 pf_krule_free(newrule);
3902 pf_kkif_free(kif);
3903 break;
3904 }
3905
3906 case DIOCCLRSTATESNV: {
3907 error = pf_clearstates_nv((struct pfioc_nv *)addr);
3908 break;
3909 }
3910
3911 case DIOCKILLSTATESNV: {
3912 error = pf_killstates_nv((struct pfioc_nv *)addr);
3913 break;
3914 }
3915
3916 case DIOCADDSTATE: {
3917 struct pfioc_state *ps = (struct pfioc_state *)addr;
3918 struct pfsync_state_1301 *sp = &ps->state;
3919
3920 if (sp->timeout >= PFTM_MAX) {
3921 error = EINVAL;
3922 break;
3923 }
3924 if (V_pfsync_state_import_ptr != NULL) {
3925 PF_RULES_RLOCK();
3926 error = V_pfsync_state_import_ptr(
3927 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3928 PFSYNC_MSG_VERSION_1301);
3929 PF_RULES_RUNLOCK();
3930 } else
3931 error = EOPNOTSUPP;
3932 break;
3933 }
3934
3935 case DIOCGETSTATE: {
3936 struct pfioc_state *ps = (struct pfioc_state *)addr;
3937 struct pf_kstate *s;
3938
3939 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3940 if (s == NULL) {
3941 error = ENOENT;
3942 break;
3943 }
3944
3945 pfsync_state_export((union pfsync_state_union*)&ps->state,
3946 s, PFSYNC_MSG_VERSION_1301);
3947 PF_STATE_UNLOCK(s);
3948 break;
3949 }
3950
3951 case DIOCGETSTATENV: {
3952 error = pf_getstate((struct pfioc_nv *)addr);
3953 break;
3954 }
3955
3956 #ifdef COMPAT_FREEBSD14
3957 case DIOCGETSTATES: {
3958 struct pfioc_states *ps = (struct pfioc_states *)addr;
3959 struct pf_kstate *s;
3960 struct pfsync_state_1301 *pstore, *p;
3961 int i, nr;
3962 size_t slice_count = 16, count;
3963 void *out;
3964
3965 if (ps->ps_len <= 0) {
3966 nr = uma_zone_get_cur(V_pf_state_z);
3967 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3968 break;
3969 }
3970
3971 out = ps->ps_states;
3972 pstore = mallocarray(slice_count,
3973 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3974 nr = 0;
3975
3976 for (i = 0; i <= V_pf_hashmask; i++) {
3977 struct pf_idhash *ih = &V_pf_idhash[i];
3978
3979 DIOCGETSTATES_retry:
3980 p = pstore;
3981
3982 if (LIST_EMPTY(&ih->states))
3983 continue;
3984
3985 PF_HASHROW_LOCK(ih);
3986 count = 0;
3987 LIST_FOREACH(s, &ih->states, entry) {
3988 if (s->timeout == PFTM_UNLINKED)
3989 continue;
3990 count++;
3991 }
3992
3993 if (count > slice_count) {
3994 PF_HASHROW_UNLOCK(ih);
3995 free(pstore, M_TEMP);
3996 slice_count = count * 2;
3997 pstore = mallocarray(slice_count,
3998 sizeof(struct pfsync_state_1301), M_TEMP,
3999 M_WAITOK | M_ZERO);
4000 goto DIOCGETSTATES_retry;
4001 }
4002
4003 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4004 PF_HASHROW_UNLOCK(ih);
4005 goto DIOCGETSTATES_full;
4006 }
4007
4008 LIST_FOREACH(s, &ih->states, entry) {
4009 if (s->timeout == PFTM_UNLINKED)
4010 continue;
4011
4012 pfsync_state_export((union pfsync_state_union*)p,
4013 s, PFSYNC_MSG_VERSION_1301);
4014 p++;
4015 nr++;
4016 }
4017 PF_HASHROW_UNLOCK(ih);
4018 error = copyout(pstore, out,
4019 sizeof(struct pfsync_state_1301) * count);
4020 if (error)
4021 break;
4022 out = ps->ps_states + nr;
4023 }
4024 DIOCGETSTATES_full:
4025 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4026 free(pstore, M_TEMP);
4027
4028 break;
4029 }
4030
4031 case DIOCGETSTATESV2: {
4032 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
4033 struct pf_kstate *s;
4034 struct pf_state_export *pstore, *p;
4035 int i, nr;
4036 size_t slice_count = 16, count;
4037 void *out;
4038
4039 if (ps->ps_req_version > PF_STATE_VERSION) {
4040 error = ENOTSUP;
4041 break;
4042 }
4043
4044 if (ps->ps_len <= 0) {
4045 nr = uma_zone_get_cur(V_pf_state_z);
4046 ps->ps_len = sizeof(struct pf_state_export) * nr;
4047 break;
4048 }
4049
4050 out = ps->ps_states;
4051 pstore = mallocarray(slice_count,
4052 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
4053 nr = 0;
4054
4055 for (i = 0; i <= V_pf_hashmask; i++) {
4056 struct pf_idhash *ih = &V_pf_idhash[i];
4057
4058 DIOCGETSTATESV2_retry:
4059 p = pstore;
4060
4061 if (LIST_EMPTY(&ih->states))
4062 continue;
4063
4064 PF_HASHROW_LOCK(ih);
4065 count = 0;
4066 LIST_FOREACH(s, &ih->states, entry) {
4067 if (s->timeout == PFTM_UNLINKED)
4068 continue;
4069 count++;
4070 }
4071
4072 if (count > slice_count) {
4073 PF_HASHROW_UNLOCK(ih);
4074 free(pstore, M_TEMP);
4075 slice_count = count * 2;
4076 pstore = mallocarray(slice_count,
4077 sizeof(struct pf_state_export), M_TEMP,
4078 M_WAITOK | M_ZERO);
4079 goto DIOCGETSTATESV2_retry;
4080 }
4081
4082 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4083 PF_HASHROW_UNLOCK(ih);
4084 goto DIOCGETSTATESV2_full;
4085 }
4086
4087 LIST_FOREACH(s, &ih->states, entry) {
4088 if (s->timeout == PFTM_UNLINKED)
4089 continue;
4090
4091 pf_state_export(p, s);
4092 p++;
4093 nr++;
4094 }
4095 PF_HASHROW_UNLOCK(ih);
4096 error = copyout(pstore, out,
4097 sizeof(struct pf_state_export) * count);
4098 if (error)
4099 break;
4100 out = ps->ps_states + nr;
4101 }
4102 DIOCGETSTATESV2_full:
4103 ps->ps_len = nr * sizeof(struct pf_state_export);
4104 free(pstore, M_TEMP);
4105
4106 break;
4107 }
4108 #endif
4109 case DIOCGETSTATUSNV: {
4110 error = pf_getstatus((struct pfioc_nv *)addr);
4111 break;
4112 }
4113
4114 case DIOCSETSTATUSIF: {
4115 struct pfioc_if *pi = (struct pfioc_if *)addr;
4116
4117 if (pi->ifname[0] == 0) {
4118 bzero(V_pf_status.ifname, IFNAMSIZ);
4119 break;
4120 }
4121 PF_RULES_WLOCK();
4122 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
4123 PF_RULES_WUNLOCK();
4124 break;
4125 }
4126
4127 case DIOCCLRSTATUS: {
4128 pf_ioctl_clear_status();
4129 break;
4130 }
4131
4132 case DIOCNATLOOK: {
4133 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
4134 struct pf_state_key *sk;
4135 struct pf_kstate *state;
4136 struct pf_state_key_cmp key;
4137 int m = 0, direction = pnl->direction;
4138 int sidx, didx;
4139
4140 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
4141 sidx = (direction == PF_IN) ? 1 : 0;
4142 didx = (direction == PF_IN) ? 0 : 1;
4143
4144 if (!pnl->proto ||
4145 PF_AZERO(&pnl->saddr, pnl->af) ||
4146 PF_AZERO(&pnl->daddr, pnl->af) ||
4147 ((pnl->proto == IPPROTO_TCP ||
4148 pnl->proto == IPPROTO_UDP) &&
4149 (!pnl->dport || !pnl->sport)))
4150 error = EINVAL;
4151 else {
4152 bzero(&key, sizeof(key));
4153 key.af = pnl->af;
4154 key.proto = pnl->proto;
4155 PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4156 key.port[sidx] = pnl->sport;
4157 PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4158 key.port[didx] = pnl->dport;
4159
4160 state = pf_find_state_all(&key, direction, &m);
4161 if (state == NULL) {
4162 error = ENOENT;
4163 } else {
4164 if (m > 1) {
4165 PF_STATE_UNLOCK(state);
4166 error = E2BIG; /* more than one state */
4167 } else {
4168 sk = state->key[sidx];
4169 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4170 pnl->rsport = sk->port[sidx];
4171 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4172 pnl->rdport = sk->port[didx];
4173 PF_STATE_UNLOCK(state);
4174 }
4175 }
4176 }
4177 break;
4178 }
4179
4180 case DIOCSETTIMEOUT: {
4181 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4182
4183 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4184 &pt->seconds);
4185 break;
4186 }
4187
4188 case DIOCGETTIMEOUT: {
4189 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4190
4191 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4192 break;
4193 }
4194
4195 case DIOCGETLIMIT: {
4196 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4197
4198 error = pf_ioctl_get_limit(pl->index, &pl->limit);
4199 break;
4200 }
4201
4202 case DIOCSETLIMIT: {
4203 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4204 unsigned int old_limit;
4205
4206 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4207 pl->limit = old_limit;
4208 break;
4209 }
4210
4211 case DIOCSETDEBUG: {
4212 u_int32_t *level = (u_int32_t *)addr;
4213
4214 PF_RULES_WLOCK();
4215 V_pf_status.debug = *level;
4216 PF_RULES_WUNLOCK();
4217 break;
4218 }
4219
4220 case DIOCCLRRULECTRS: {
4221 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4222 struct pf_kruleset *ruleset = &pf_main_ruleset;
4223 struct pf_krule *rule;
4224
4225 PF_RULES_WLOCK();
4226 TAILQ_FOREACH(rule,
4227 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4228 pf_counter_u64_zero(&rule->evaluations);
4229 for (int i = 0; i < 2; i++) {
4230 pf_counter_u64_zero(&rule->packets[i]);
4231 pf_counter_u64_zero(&rule->bytes[i]);
4232 }
4233 }
4234 PF_RULES_WUNLOCK();
4235 break;
4236 }
4237
4238 case DIOCGIFSPEEDV0:
4239 case DIOCGIFSPEEDV1: {
4240 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
4241 struct pf_ifspeed_v1 ps;
4242 struct ifnet *ifp;
4243
4244 if (psp->ifname[0] == '\0') {
4245 error = EINVAL;
4246 break;
4247 }
4248
4249 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4250 if (error != 0)
4251 break;
4252 ifp = ifunit(ps.ifname);
4253 if (ifp != NULL) {
4254 psp->baudrate32 =
4255 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4256 if (cmd == DIOCGIFSPEEDV1)
4257 psp->baudrate = ifp->if_baudrate;
4258 } else {
4259 error = EINVAL;
4260 }
4261 break;
4262 }
4263
4264 #ifdef ALTQ
4265 case DIOCSTARTALTQ: {
4266 struct pf_altq *altq;
4267
4268 PF_RULES_WLOCK();
4269 /* enable all altq interfaces on active list */
4270 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4271 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4272 error = pf_enable_altq(altq);
4273 if (error != 0)
4274 break;
4275 }
4276 }
4277 if (error == 0)
4278 V_pf_altq_running = 1;
4279 PF_RULES_WUNLOCK();
4280 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4281 break;
4282 }
4283
4284 case DIOCSTOPALTQ: {
4285 struct pf_altq *altq;
4286
4287 PF_RULES_WLOCK();
4288 /* disable all altq interfaces on active list */
4289 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4290 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4291 error = pf_disable_altq(altq);
4292 if (error != 0)
4293 break;
4294 }
4295 }
4296 if (error == 0)
4297 V_pf_altq_running = 0;
4298 PF_RULES_WUNLOCK();
4299 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4300 break;
4301 }
4302
4303 case DIOCADDALTQV0:
4304 case DIOCADDALTQV1: {
4305 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4306 struct pf_altq *altq, *a;
4307 struct ifnet *ifp;
4308
4309 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4310 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4311 if (error)
4312 break;
4313 altq->local_flags = 0;
4314
4315 PF_RULES_WLOCK();
4316 if (pa->ticket != V_ticket_altqs_inactive) {
4317 PF_RULES_WUNLOCK();
4318 free(altq, M_PFALTQ);
4319 error = EBUSY;
4320 break;
4321 }
4322
4323 /*
4324 * if this is for a queue, find the discipline and
4325 * copy the necessary fields
4326 */
4327 if (altq->qname[0] != 0) {
4328 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4329 PF_RULES_WUNLOCK();
4330 error = EBUSY;
4331 free(altq, M_PFALTQ);
4332 break;
4333 }
4334 altq->altq_disc = NULL;
4335 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4336 if (strncmp(a->ifname, altq->ifname,
4337 IFNAMSIZ) == 0) {
4338 altq->altq_disc = a->altq_disc;
4339 break;
4340 }
4341 }
4342 }
4343
4344 if ((ifp = ifunit(altq->ifname)) == NULL)
4345 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4346 else
4347 error = altq_add(ifp, altq);
4348
4349 if (error) {
4350 PF_RULES_WUNLOCK();
4351 free(altq, M_PFALTQ);
4352 break;
4353 }
4354
4355 if (altq->qname[0] != 0)
4356 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4357 else
4358 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4359 /* version error check done on import above */
4360 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4361 PF_RULES_WUNLOCK();
4362 break;
4363 }
4364
4365 case DIOCGETALTQSV0:
4366 case DIOCGETALTQSV1: {
4367 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4368 struct pf_altq *altq;
4369
4370 PF_RULES_RLOCK();
4371 pa->nr = 0;
4372 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4373 pa->nr++;
4374 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4375 pa->nr++;
4376 pa->ticket = V_ticket_altqs_active;
4377 PF_RULES_RUNLOCK();
4378 break;
4379 }
4380
4381 case DIOCGETALTQV0:
4382 case DIOCGETALTQV1: {
4383 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4384 struct pf_altq *altq;
4385
4386 PF_RULES_RLOCK();
4387 if (pa->ticket != V_ticket_altqs_active) {
4388 PF_RULES_RUNLOCK();
4389 error = EBUSY;
4390 break;
4391 }
4392 altq = pf_altq_get_nth_active(pa->nr);
4393 if (altq == NULL) {
4394 PF_RULES_RUNLOCK();
4395 error = EBUSY;
4396 break;
4397 }
4398 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4399 PF_RULES_RUNLOCK();
4400 break;
4401 }
4402
4403 case DIOCCHANGEALTQV0:
4404 case DIOCCHANGEALTQV1:
4405 /* CHANGEALTQ not supported yet! */
4406 error = ENODEV;
4407 break;
4408
4409 case DIOCGETQSTATSV0:
4410 case DIOCGETQSTATSV1: {
4411 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
4412 struct pf_altq *altq;
4413 int nbytes;
4414 u_int32_t version;
4415
4416 PF_RULES_RLOCK();
4417 if (pq->ticket != V_ticket_altqs_active) {
4418 PF_RULES_RUNLOCK();
4419 error = EBUSY;
4420 break;
4421 }
4422 nbytes = pq->nbytes;
4423 altq = pf_altq_get_nth_active(pq->nr);
4424 if (altq == NULL) {
4425 PF_RULES_RUNLOCK();
4426 error = EBUSY;
4427 break;
4428 }
4429
4430 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4431 PF_RULES_RUNLOCK();
4432 error = ENXIO;
4433 break;
4434 }
4435 PF_RULES_RUNLOCK();
4436 if (cmd == DIOCGETQSTATSV0)
4437 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
4438 else
4439 version = pq->version;
4440 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4441 if (error == 0) {
4442 pq->scheduler = altq->scheduler;
4443 pq->nbytes = nbytes;
4444 }
4445 break;
4446 }
4447 #endif /* ALTQ */
4448
4449 case DIOCBEGINADDRS: {
4450 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4451
4452 error = pf_ioctl_begin_addrs(&pp->ticket);
4453 break;
4454 }
4455
4456 case DIOCADDADDR: {
4457 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4458 struct pf_nl_pooladdr npp = {};
4459
4460 npp.which = PF_RDR;
4461 memcpy(&npp, pp, sizeof(*pp));
4462 error = pf_ioctl_add_addr(&npp);
4463 break;
4464 }
4465
4466 case DIOCGETADDRS: {
4467 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4468 struct pf_nl_pooladdr npp = {};
4469
4470 npp.which = PF_RDR;
4471 memcpy(&npp, pp, sizeof(*pp));
4472 error = pf_ioctl_get_addrs(&npp);
4473 memcpy(pp, &npp, sizeof(*pp));
4474
4475 break;
4476 }
4477
4478 case DIOCGETADDR: {
4479 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4480 struct pf_nl_pooladdr npp = {};
4481
4482 npp.which = PF_RDR;
4483 memcpy(&npp, pp, sizeof(*pp));
4484 error = pf_ioctl_get_addr(&npp);
4485 memcpy(pp, &npp, sizeof(*pp));
4486
4487 break;
4488 }
4489
4490 case DIOCCHANGEADDR: {
4491 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
4492 struct pf_kpool *pool;
4493 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
4494 struct pf_kruleset *ruleset;
4495 struct pfi_kkif *kif = NULL;
4496
4497 pca->anchor[sizeof(pca->anchor) - 1] = 0;
4498
4499 if (pca->action < PF_CHANGE_ADD_HEAD ||
4500 pca->action > PF_CHANGE_REMOVE) {
4501 error = EINVAL;
4502 break;
4503 }
4504 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4505 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4506 pca->addr.addr.type != PF_ADDR_TABLE) {
4507 error = EINVAL;
4508 break;
4509 }
4510 if (pca->addr.addr.p.dyn != NULL) {
4511 error = EINVAL;
4512 break;
4513 }
4514
4515 if (pca->action != PF_CHANGE_REMOVE) {
4516 #ifndef INET
4517 if (pca->af == AF_INET) {
4518 error = EAFNOSUPPORT;
4519 break;
4520 }
4521 #endif /* INET */
4522 #ifndef INET6
4523 if (pca->af == AF_INET6) {
4524 error = EAFNOSUPPORT;
4525 break;
4526 }
4527 #endif /* INET6 */
4528 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4529 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4530 if (newpa->ifname[0])
4531 kif = pf_kkif_create(M_WAITOK);
4532 newpa->kif = NULL;
4533 }
4534 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4535 PF_RULES_WLOCK();
4536 ruleset = pf_find_kruleset(pca->anchor);
4537 if (ruleset == NULL)
4538 ERROUT(EBUSY);
4539
4540 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4541 pca->r_num, pca->r_last, 1, 1, PF_RDR);
4542 if (pool == NULL)
4543 ERROUT(EBUSY);
4544
4545 if (pca->action != PF_CHANGE_REMOVE) {
4546 if (newpa->ifname[0]) {
4547 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4548 pfi_kkif_ref(newpa->kif);
4549 kif = NULL;
4550 }
4551
4552 switch (newpa->addr.type) {
4553 case PF_ADDR_DYNIFTL:
4554 error = pfi_dynaddr_setup(&newpa->addr,
4555 pca->af);
4556 break;
4557 case PF_ADDR_TABLE:
4558 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4559 newpa->addr.v.tblname);
4560 if (newpa->addr.p.tbl == NULL)
4561 error = ENOMEM;
4562 break;
4563 }
4564 if (error)
4565 goto DIOCCHANGEADDR_error;
4566 }
4567
4568 switch (pca->action) {
4569 case PF_CHANGE_ADD_HEAD:
4570 oldpa = TAILQ_FIRST(&pool->list);
4571 break;
4572 case PF_CHANGE_ADD_TAIL:
4573 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4574 break;
4575 default:
4576 oldpa = TAILQ_FIRST(&pool->list);
4577 for (int i = 0; oldpa && i < pca->nr; i++)
4578 oldpa = TAILQ_NEXT(oldpa, entries);
4579
4580 if (oldpa == NULL)
4581 ERROUT(EINVAL);
4582 }
4583
4584 if (pca->action == PF_CHANGE_REMOVE) {
4585 TAILQ_REMOVE(&pool->list, oldpa, entries);
4586 switch (oldpa->addr.type) {
4587 case PF_ADDR_DYNIFTL:
4588 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4589 break;
4590 case PF_ADDR_TABLE:
4591 pfr_detach_table(oldpa->addr.p.tbl);
4592 break;
4593 }
4594 if (oldpa->kif)
4595 pfi_kkif_unref(oldpa->kif);
4596 free(oldpa, M_PFRULE);
4597 } else {
4598 if (oldpa == NULL)
4599 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4600 else if (pca->action == PF_CHANGE_ADD_HEAD ||
4601 pca->action == PF_CHANGE_ADD_BEFORE)
4602 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4603 else
4604 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4605 newpa, entries);
4606 }
4607
4608 pool->cur = TAILQ_FIRST(&pool->list);
4609 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4610 PF_RULES_WUNLOCK();
4611 break;
4612
4613 #undef ERROUT
4614 DIOCCHANGEADDR_error:
4615 if (newpa != NULL) {
4616 if (newpa->kif)
4617 pfi_kkif_unref(newpa->kif);
4618 free(newpa, M_PFRULE);
4619 }
4620 PF_RULES_WUNLOCK();
4621 pf_kkif_free(kif);
4622 break;
4623 }
4624
4625 case DIOCGETRULESETS: {
4626 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4627
4628 pr->path[sizeof(pr->path) - 1] = 0;
4629
4630 error = pf_ioctl_get_rulesets(pr);
4631 break;
4632 }
4633
4634 case DIOCGETRULESET: {
4635 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4636
4637 pr->path[sizeof(pr->path) - 1] = 0;
4638
4639 error = pf_ioctl_get_ruleset(pr);
4640 break;
4641 }
4642
4643 case DIOCRCLRTABLES: {
4644 struct pfioc_table *io = (struct pfioc_table *)addr;
4645
4646 if (io->pfrio_esize != 0) {
4647 error = ENODEV;
4648 break;
4649 }
4650 PF_RULES_WLOCK();
4651 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4652 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4653 PF_RULES_WUNLOCK();
4654 break;
4655 }
4656
4657 case DIOCRADDTABLES: {
4658 struct pfioc_table *io = (struct pfioc_table *)addr;
4659 struct pfr_table *pfrts;
4660 size_t totlen;
4661
4662 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4663 error = ENODEV;
4664 break;
4665 }
4666
4667 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4668 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4669 error = ENOMEM;
4670 break;
4671 }
4672
4673 totlen = io->pfrio_size * sizeof(struct pfr_table);
4674 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4675 M_TEMP, M_WAITOK);
4676 error = copyin(io->pfrio_buffer, pfrts, totlen);
4677 if (error) {
4678 free(pfrts, M_TEMP);
4679 break;
4680 }
4681 PF_RULES_WLOCK();
4682 error = pfr_add_tables(pfrts, io->pfrio_size,
4683 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4684 PF_RULES_WUNLOCK();
4685 free(pfrts, M_TEMP);
4686 break;
4687 }
4688
4689 case DIOCRDELTABLES: {
4690 struct pfioc_table *io = (struct pfioc_table *)addr;
4691 struct pfr_table *pfrts;
4692 size_t totlen;
4693
4694 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4695 error = ENODEV;
4696 break;
4697 }
4698
4699 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4700 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4701 error = ENOMEM;
4702 break;
4703 }
4704
4705 totlen = io->pfrio_size * sizeof(struct pfr_table);
4706 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4707 M_TEMP, M_WAITOK);
4708 error = copyin(io->pfrio_buffer, pfrts, totlen);
4709 if (error) {
4710 free(pfrts, M_TEMP);
4711 break;
4712 }
4713 PF_RULES_WLOCK();
4714 error = pfr_del_tables(pfrts, io->pfrio_size,
4715 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4716 PF_RULES_WUNLOCK();
4717 free(pfrts, M_TEMP);
4718 break;
4719 }
4720
4721 case DIOCRGETTABLES: {
4722 struct pfioc_table *io = (struct pfioc_table *)addr;
4723 struct pfr_table *pfrts;
4724 size_t totlen;
4725 int n;
4726
4727 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4728 error = ENODEV;
4729 break;
4730 }
4731 PF_RULES_RLOCK();
4732 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4733 if (n < 0) {
4734 PF_RULES_RUNLOCK();
4735 error = EINVAL;
4736 break;
4737 }
4738 io->pfrio_size = min(io->pfrio_size, n);
4739
4740 totlen = io->pfrio_size * sizeof(struct pfr_table);
4741
4742 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4743 M_TEMP, M_NOWAIT | M_ZERO);
4744 if (pfrts == NULL) {
4745 error = ENOMEM;
4746 PF_RULES_RUNLOCK();
4747 break;
4748 }
4749 error = pfr_get_tables(&io->pfrio_table, pfrts,
4750 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4751 PF_RULES_RUNLOCK();
4752 if (error == 0)
4753 error = copyout(pfrts, io->pfrio_buffer, totlen);
4754 free(pfrts, M_TEMP);
4755 break;
4756 }
4757
4758 case DIOCRGETTSTATS: {
4759 struct pfioc_table *io = (struct pfioc_table *)addr;
4760 struct pfr_tstats *pfrtstats;
4761 size_t totlen;
4762 int n;
4763
4764 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4765 error = ENODEV;
4766 break;
4767 }
4768 PF_TABLE_STATS_LOCK();
4769 PF_RULES_RLOCK();
4770 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4771 if (n < 0) {
4772 PF_RULES_RUNLOCK();
4773 PF_TABLE_STATS_UNLOCK();
4774 error = EINVAL;
4775 break;
4776 }
4777 io->pfrio_size = min(io->pfrio_size, n);
4778
4779 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4780 pfrtstats = mallocarray(io->pfrio_size,
4781 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4782 if (pfrtstats == NULL) {
4783 error = ENOMEM;
4784 PF_RULES_RUNLOCK();
4785 PF_TABLE_STATS_UNLOCK();
4786 break;
4787 }
4788 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4789 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4790 PF_RULES_RUNLOCK();
4791 PF_TABLE_STATS_UNLOCK();
4792 if (error == 0)
4793 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4794 free(pfrtstats, M_TEMP);
4795 break;
4796 }
4797
4798 case DIOCRCLRTSTATS: {
4799 struct pfioc_table *io = (struct pfioc_table *)addr;
4800 struct pfr_table *pfrts;
4801 size_t totlen;
4802
4803 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4804 error = ENODEV;
4805 break;
4806 }
4807
4808 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4809 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4810 /* We used to count tables and use the minimum required
4811 * size, so we didn't fail on overly large requests.
4812 * Keep doing so. */
4813 io->pfrio_size = pf_ioctl_maxcount;
4814 break;
4815 }
4816
4817 totlen = io->pfrio_size * sizeof(struct pfr_table);
4818 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4819 M_TEMP, M_WAITOK);
4820 error = copyin(io->pfrio_buffer, pfrts, totlen);
4821 if (error) {
4822 free(pfrts, M_TEMP);
4823 break;
4824 }
4825
4826 PF_TABLE_STATS_LOCK();
4827 PF_RULES_RLOCK();
4828 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4829 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4830 PF_RULES_RUNLOCK();
4831 PF_TABLE_STATS_UNLOCK();
4832 free(pfrts, M_TEMP);
4833 break;
4834 }
4835
4836 case DIOCRSETTFLAGS: {
4837 struct pfioc_table *io = (struct pfioc_table *)addr;
4838 struct pfr_table *pfrts;
4839 size_t totlen;
4840 int n;
4841
4842 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4843 error = ENODEV;
4844 break;
4845 }
4846
4847 PF_RULES_RLOCK();
4848 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4849 if (n < 0) {
4850 PF_RULES_RUNLOCK();
4851 error = EINVAL;
4852 break;
4853 }
4854
4855 io->pfrio_size = min(io->pfrio_size, n);
4856 PF_RULES_RUNLOCK();
4857
4858 totlen = io->pfrio_size * sizeof(struct pfr_table);
4859 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4860 M_TEMP, M_WAITOK);
4861 error = copyin(io->pfrio_buffer, pfrts, totlen);
4862 if (error) {
4863 free(pfrts, M_TEMP);
4864 break;
4865 }
4866 PF_RULES_WLOCK();
4867 error = pfr_set_tflags(pfrts, io->pfrio_size,
4868 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4869 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4870 PF_RULES_WUNLOCK();
4871 free(pfrts, M_TEMP);
4872 break;
4873 }
4874
4875 case DIOCRCLRADDRS: {
4876 struct pfioc_table *io = (struct pfioc_table *)addr;
4877
4878 if (io->pfrio_esize != 0) {
4879 error = ENODEV;
4880 break;
4881 }
4882 PF_RULES_WLOCK();
4883 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4884 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4885 PF_RULES_WUNLOCK();
4886 break;
4887 }
4888
4889 case DIOCRADDADDRS: {
4890 struct pfioc_table *io = (struct pfioc_table *)addr;
4891 struct pfr_addr *pfras;
4892 size_t totlen;
4893
4894 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4895 error = ENODEV;
4896 break;
4897 }
4898 if (io->pfrio_size < 0 ||
4899 io->pfrio_size > pf_ioctl_maxcount ||
4900 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4901 error = EINVAL;
4902 break;
4903 }
4904 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4905 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4906 M_TEMP, M_WAITOK);
4907 error = copyin(io->pfrio_buffer, pfras, totlen);
4908 if (error) {
4909 free(pfras, M_TEMP);
4910 break;
4911 }
4912 PF_RULES_WLOCK();
4913 error = pfr_add_addrs(&io->pfrio_table, pfras,
4914 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4915 PFR_FLAG_USERIOCTL);
4916 PF_RULES_WUNLOCK();
4917 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4918 error = copyout(pfras, io->pfrio_buffer, totlen);
4919 free(pfras, M_TEMP);
4920 break;
4921 }
4922
4923 case DIOCRDELADDRS: {
4924 struct pfioc_table *io = (struct pfioc_table *)addr;
4925 struct pfr_addr *pfras;
4926 size_t totlen;
4927
4928 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4929 error = ENODEV;
4930 break;
4931 }
4932 if (io->pfrio_size < 0 ||
4933 io->pfrio_size > pf_ioctl_maxcount ||
4934 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4935 error = EINVAL;
4936 break;
4937 }
4938 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4939 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4940 M_TEMP, M_WAITOK);
4941 error = copyin(io->pfrio_buffer, pfras, totlen);
4942 if (error) {
4943 free(pfras, M_TEMP);
4944 break;
4945 }
4946 PF_RULES_WLOCK();
4947 error = pfr_del_addrs(&io->pfrio_table, pfras,
4948 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4949 PFR_FLAG_USERIOCTL);
4950 PF_RULES_WUNLOCK();
4951 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4952 error = copyout(pfras, io->pfrio_buffer, totlen);
4953 free(pfras, M_TEMP);
4954 break;
4955 }
4956
4957 case DIOCRSETADDRS: {
4958 struct pfioc_table *io = (struct pfioc_table *)addr;
4959 struct pfr_addr *pfras;
4960 size_t totlen, count;
4961
4962 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4963 error = ENODEV;
4964 break;
4965 }
4966 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4967 error = EINVAL;
4968 break;
4969 }
4970 count = max(io->pfrio_size, io->pfrio_size2);
4971 if (count > pf_ioctl_maxcount ||
4972 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4973 error = EINVAL;
4974 break;
4975 }
4976 totlen = count * sizeof(struct pfr_addr);
4977 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4978 M_WAITOK);
4979 error = copyin(io->pfrio_buffer, pfras, totlen);
4980 if (error) {
4981 free(pfras, M_TEMP);
4982 break;
4983 }
4984 PF_RULES_WLOCK();
4985 error = pfr_set_addrs(&io->pfrio_table, pfras,
4986 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4987 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4988 PFR_FLAG_USERIOCTL, 0);
4989 PF_RULES_WUNLOCK();
4990 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4991 error = copyout(pfras, io->pfrio_buffer, totlen);
4992 free(pfras, M_TEMP);
4993 break;
4994 }
4995
4996 case DIOCRGETADDRS: {
4997 struct pfioc_table *io = (struct pfioc_table *)addr;
4998 struct pfr_addr *pfras;
4999 size_t totlen;
5000
5001 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5002 error = ENODEV;
5003 break;
5004 }
5005 if (io->pfrio_size < 0 ||
5006 io->pfrio_size > pf_ioctl_maxcount ||
5007 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5008 error = EINVAL;
5009 break;
5010 }
5011 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5012 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5013 M_TEMP, M_WAITOK | M_ZERO);
5014 PF_RULES_RLOCK();
5015 error = pfr_get_addrs(&io->pfrio_table, pfras,
5016 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5017 PF_RULES_RUNLOCK();
5018 if (error == 0)
5019 error = copyout(pfras, io->pfrio_buffer, totlen);
5020 free(pfras, M_TEMP);
5021 break;
5022 }
5023
5024 case DIOCRGETASTATS: {
5025 struct pfioc_table *io = (struct pfioc_table *)addr;
5026 struct pfr_astats *pfrastats;
5027 size_t totlen;
5028
5029 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
5030 error = ENODEV;
5031 break;
5032 }
5033 if (io->pfrio_size < 0 ||
5034 io->pfrio_size > pf_ioctl_maxcount ||
5035 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5036 error = EINVAL;
5037 break;
5038 }
5039 totlen = io->pfrio_size * sizeof(struct pfr_astats);
5040 pfrastats = mallocarray(io->pfrio_size,
5041 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
5042 PF_RULES_RLOCK();
5043 error = pfr_get_astats(&io->pfrio_table, pfrastats,
5044 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5045 PF_RULES_RUNLOCK();
5046 if (error == 0)
5047 error = copyout(pfrastats, io->pfrio_buffer, totlen);
5048 free(pfrastats, M_TEMP);
5049 break;
5050 }
5051
5052 case DIOCRCLRASTATS: {
5053 struct pfioc_table *io = (struct pfioc_table *)addr;
5054 struct pfr_addr *pfras;
5055 size_t totlen;
5056
5057 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5058 error = ENODEV;
5059 break;
5060 }
5061 if (io->pfrio_size < 0 ||
5062 io->pfrio_size > pf_ioctl_maxcount ||
5063 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5064 error = EINVAL;
5065 break;
5066 }
5067 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5068 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5069 M_TEMP, M_WAITOK);
5070 error = copyin(io->pfrio_buffer, pfras, totlen);
5071 if (error) {
5072 free(pfras, M_TEMP);
5073 break;
5074 }
5075 PF_RULES_WLOCK();
5076 error = pfr_clr_astats(&io->pfrio_table, pfras,
5077 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5078 PFR_FLAG_USERIOCTL);
5079 PF_RULES_WUNLOCK();
5080 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5081 error = copyout(pfras, io->pfrio_buffer, totlen);
5082 free(pfras, M_TEMP);
5083 break;
5084 }
5085
5086 case DIOCRTSTADDRS: {
5087 struct pfioc_table *io = (struct pfioc_table *)addr;
5088 struct pfr_addr *pfras;
5089 size_t totlen;
5090
5091 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5092 error = ENODEV;
5093 break;
5094 }
5095 if (io->pfrio_size < 0 ||
5096 io->pfrio_size > pf_ioctl_maxcount ||
5097 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5098 error = EINVAL;
5099 break;
5100 }
5101 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5102 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5103 M_TEMP, M_WAITOK);
5104 error = copyin(io->pfrio_buffer, pfras, totlen);
5105 if (error) {
5106 free(pfras, M_TEMP);
5107 break;
5108 }
5109 PF_RULES_RLOCK();
5110 error = pfr_tst_addrs(&io->pfrio_table, pfras,
5111 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5112 PFR_FLAG_USERIOCTL);
5113 PF_RULES_RUNLOCK();
5114 if (error == 0)
5115 error = copyout(pfras, io->pfrio_buffer, totlen);
5116 free(pfras, M_TEMP);
5117 break;
5118 }
5119
5120 case DIOCRINADEFINE: {
5121 struct pfioc_table *io = (struct pfioc_table *)addr;
5122 struct pfr_addr *pfras;
5123 size_t totlen;
5124
5125 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5126 error = ENODEV;
5127 break;
5128 }
5129 if (io->pfrio_size < 0 ||
5130 io->pfrio_size > pf_ioctl_maxcount ||
5131 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5132 error = EINVAL;
5133 break;
5134 }
5135 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5136 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5137 M_TEMP, M_WAITOK);
5138 error = copyin(io->pfrio_buffer, pfras, totlen);
5139 if (error) {
5140 free(pfras, M_TEMP);
5141 break;
5142 }
5143 PF_RULES_WLOCK();
5144 error = pfr_ina_define(&io->pfrio_table, pfras,
5145 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5146 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5147 PF_RULES_WUNLOCK();
5148 free(pfras, M_TEMP);
5149 break;
5150 }
5151
5152 case DIOCOSFPADD: {
5153 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5154 PF_RULES_WLOCK();
5155 error = pf_osfp_add(io);
5156 PF_RULES_WUNLOCK();
5157 break;
5158 }
5159
5160 case DIOCOSFPGET: {
5161 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5162 PF_RULES_RLOCK();
5163 error = pf_osfp_get(io);
5164 PF_RULES_RUNLOCK();
5165 break;
5166 }
5167
5168 case DIOCXBEGIN: {
5169 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5170 struct pfioc_trans_e *ioes, *ioe;
5171 size_t totlen;
5172 int i;
5173
5174 if (io->esize != sizeof(*ioe)) {
5175 error = ENODEV;
5176 break;
5177 }
5178 if (io->size < 0 ||
5179 io->size > pf_ioctl_maxcount ||
5180 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5181 error = EINVAL;
5182 break;
5183 }
5184 totlen = sizeof(struct pfioc_trans_e) * io->size;
5185 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5186 M_TEMP, M_WAITOK);
5187 error = copyin(io->array, ioes, totlen);
5188 if (error) {
5189 free(ioes, M_TEMP);
5190 break;
5191 }
5192 PF_RULES_WLOCK();
5193 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5194 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5195 switch (ioe->rs_num) {
5196 case PF_RULESET_ETH:
5197 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5198 PF_RULES_WUNLOCK();
5199 free(ioes, M_TEMP);
5200 goto fail;
5201 }
5202 break;
5203 #ifdef ALTQ
5204 case PF_RULESET_ALTQ:
5205 if (ioe->anchor[0]) {
5206 PF_RULES_WUNLOCK();
5207 free(ioes, M_TEMP);
5208 error = EINVAL;
5209 goto fail;
5210 }
5211 if ((error = pf_begin_altq(&ioe->ticket))) {
5212 PF_RULES_WUNLOCK();
5213 free(ioes, M_TEMP);
5214 goto fail;
5215 }
5216 break;
5217 #endif /* ALTQ */
5218 case PF_RULESET_TABLE:
5219 {
5220 struct pfr_table table;
5221
5222 bzero(&table, sizeof(table));
5223 strlcpy(table.pfrt_anchor, ioe->anchor,
5224 sizeof(table.pfrt_anchor));
5225 if ((error = pfr_ina_begin(&table,
5226 &ioe->ticket, NULL, 0))) {
5227 PF_RULES_WUNLOCK();
5228 free(ioes, M_TEMP);
5229 goto fail;
5230 }
5231 break;
5232 }
5233 default:
5234 if ((error = pf_begin_rules(&ioe->ticket,
5235 ioe->rs_num, ioe->anchor))) {
5236 PF_RULES_WUNLOCK();
5237 free(ioes, M_TEMP);
5238 goto fail;
5239 }
5240 break;
5241 }
5242 }
5243 PF_RULES_WUNLOCK();
5244 error = copyout(ioes, io->array, totlen);
5245 free(ioes, M_TEMP);
5246 break;
5247 }
5248
5249 case DIOCXROLLBACK: {
5250 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5251 struct pfioc_trans_e *ioe, *ioes;
5252 size_t totlen;
5253 int i;
5254
5255 if (io->esize != sizeof(*ioe)) {
5256 error = ENODEV;
5257 break;
5258 }
5259 if (io->size < 0 ||
5260 io->size > pf_ioctl_maxcount ||
5261 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5262 error = EINVAL;
5263 break;
5264 }
5265 totlen = sizeof(struct pfioc_trans_e) * io->size;
5266 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5267 M_TEMP, M_WAITOK);
5268 error = copyin(io->array, ioes, totlen);
5269 if (error) {
5270 free(ioes, M_TEMP);
5271 break;
5272 }
5273 PF_RULES_WLOCK();
5274 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5275 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5276 switch (ioe->rs_num) {
5277 case PF_RULESET_ETH:
5278 if ((error = pf_rollback_eth(ioe->ticket,
5279 ioe->anchor))) {
5280 PF_RULES_WUNLOCK();
5281 free(ioes, M_TEMP);
5282 goto fail; /* really bad */
5283 }
5284 break;
5285 #ifdef ALTQ
5286 case PF_RULESET_ALTQ:
5287 if (ioe->anchor[0]) {
5288 PF_RULES_WUNLOCK();
5289 free(ioes, M_TEMP);
5290 error = EINVAL;
5291 goto fail;
5292 }
5293 if ((error = pf_rollback_altq(ioe->ticket))) {
5294 PF_RULES_WUNLOCK();
5295 free(ioes, M_TEMP);
5296 goto fail; /* really bad */
5297 }
5298 break;
5299 #endif /* ALTQ */
5300 case PF_RULESET_TABLE:
5301 {
5302 struct pfr_table table;
5303
5304 bzero(&table, sizeof(table));
5305 strlcpy(table.pfrt_anchor, ioe->anchor,
5306 sizeof(table.pfrt_anchor));
5307 if ((error = pfr_ina_rollback(&table,
5308 ioe->ticket, NULL, 0))) {
5309 PF_RULES_WUNLOCK();
5310 free(ioes, M_TEMP);
5311 goto fail; /* really bad */
5312 }
5313 break;
5314 }
5315 default:
5316 if ((error = pf_rollback_rules(ioe->ticket,
5317 ioe->rs_num, ioe->anchor))) {
5318 PF_RULES_WUNLOCK();
5319 free(ioes, M_TEMP);
5320 goto fail; /* really bad */
5321 }
5322 break;
5323 }
5324 }
5325 PF_RULES_WUNLOCK();
5326 free(ioes, M_TEMP);
5327 break;
5328 }
5329
5330 case DIOCXCOMMIT: {
5331 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5332 struct pfioc_trans_e *ioe, *ioes;
5333 struct pf_kruleset *rs;
5334 struct pf_keth_ruleset *ers;
5335 size_t totlen;
5336 int i;
5337
5338 if (io->esize != sizeof(*ioe)) {
5339 error = ENODEV;
5340 break;
5341 }
5342
5343 if (io->size < 0 ||
5344 io->size > pf_ioctl_maxcount ||
5345 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5346 error = EINVAL;
5347 break;
5348 }
5349
5350 totlen = sizeof(struct pfioc_trans_e) * io->size;
5351 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5352 M_TEMP, M_WAITOK);
5353 error = copyin(io->array, ioes, totlen);
5354 if (error) {
5355 free(ioes, M_TEMP);
5356 break;
5357 }
5358 PF_RULES_WLOCK();
5359 /* First makes sure everything will succeed. */
5360 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5361 ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5362 switch (ioe->rs_num) {
5363 case PF_RULESET_ETH:
5364 ers = pf_find_keth_ruleset(ioe->anchor);
5365 if (ers == NULL || ioe->ticket == 0 ||
5366 ioe->ticket != ers->inactive.ticket) {
5367 PF_RULES_WUNLOCK();
5368 free(ioes, M_TEMP);
5369 error = EINVAL;
5370 goto fail;
5371 }
5372 break;
5373 #ifdef ALTQ
5374 case PF_RULESET_ALTQ:
5375 if (ioe->anchor[0]) {
5376 PF_RULES_WUNLOCK();
5377 free(ioes, M_TEMP);
5378 error = EINVAL;
5379 goto fail;
5380 }
5381 if (!V_altqs_inactive_open || ioe->ticket !=
5382 V_ticket_altqs_inactive) {
5383 PF_RULES_WUNLOCK();
5384 free(ioes, M_TEMP);
5385 error = EBUSY;
5386 goto fail;
5387 }
5388 break;
5389 #endif /* ALTQ */
5390 case PF_RULESET_TABLE:
5391 rs = pf_find_kruleset(ioe->anchor);
5392 if (rs == NULL || !rs->topen || ioe->ticket !=
5393 rs->tticket) {
5394 PF_RULES_WUNLOCK();
5395 free(ioes, M_TEMP);
5396 error = EBUSY;
5397 goto fail;
5398 }
5399 break;
5400 default:
5401 if (ioe->rs_num < 0 || ioe->rs_num >=
5402 PF_RULESET_MAX) {
5403 PF_RULES_WUNLOCK();
5404 free(ioes, M_TEMP);
5405 error = EINVAL;
5406 goto fail;
5407 }
5408 rs = pf_find_kruleset(ioe->anchor);
5409 if (rs == NULL ||
5410 !rs->rules[ioe->rs_num].inactive.open ||
5411 rs->rules[ioe->rs_num].inactive.ticket !=
5412 ioe->ticket) {
5413 PF_RULES_WUNLOCK();
5414 free(ioes, M_TEMP);
5415 error = EBUSY;
5416 goto fail;
5417 }
5418 break;
5419 }
5420 }
5421 /* Now do the commit - no errors should happen here. */
5422 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5423 switch (ioe->rs_num) {
5424 case PF_RULESET_ETH:
5425 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5426 PF_RULES_WUNLOCK();
5427 free(ioes, M_TEMP);
5428 goto fail; /* really bad */
5429 }
5430 break;
5431 #ifdef ALTQ
5432 case PF_RULESET_ALTQ:
5433 if ((error = pf_commit_altq(ioe->ticket))) {
5434 PF_RULES_WUNLOCK();
5435 free(ioes, M_TEMP);
5436 goto fail; /* really bad */
5437 }
5438 break;
5439 #endif /* ALTQ */
5440 case PF_RULESET_TABLE:
5441 {
5442 struct pfr_table table;
5443
5444 bzero(&table, sizeof(table));
5445 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
5446 sizeof(table.pfrt_anchor));
5447 if ((error = pfr_ina_commit(&table,
5448 ioe->ticket, NULL, NULL, 0))) {
5449 PF_RULES_WUNLOCK();
5450 free(ioes, M_TEMP);
5451 goto fail; /* really bad */
5452 }
5453 break;
5454 }
5455 default:
5456 if ((error = pf_commit_rules(ioe->ticket,
5457 ioe->rs_num, ioe->anchor))) {
5458 PF_RULES_WUNLOCK();
5459 free(ioes, M_TEMP);
5460 goto fail; /* really bad */
5461 }
5462 break;
5463 }
5464 }
5465 PF_RULES_WUNLOCK();
5466
5467 /* Only hook into EtherNet taffic if we've got rules for it. */
5468 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5469 hook_pf_eth();
5470 else
5471 dehook_pf_eth();
5472
5473 free(ioes, M_TEMP);
5474 break;
5475 }
5476
5477 case DIOCGETSRCNODES: {
5478 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
5479 struct pf_srchash *sh;
5480 struct pf_ksrc_node *n;
5481 struct pf_src_node *p, *pstore;
5482 uint32_t i, nr = 0;
5483
5484 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5485 i++, sh++) {
5486 PF_HASHROW_LOCK(sh);
5487 LIST_FOREACH(n, &sh->nodes, entry)
5488 nr++;
5489 PF_HASHROW_UNLOCK(sh);
5490 }
5491
5492 psn->psn_len = min(psn->psn_len,
5493 sizeof(struct pf_src_node) * nr);
5494
5495 if (psn->psn_len == 0) {
5496 psn->psn_len = sizeof(struct pf_src_node) * nr;
5497 break;
5498 }
5499
5500 nr = 0;
5501
5502 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5503 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5504 i++, sh++) {
5505 PF_HASHROW_LOCK(sh);
5506 LIST_FOREACH(n, &sh->nodes, entry) {
5507
5508 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5509 break;
5510
5511 pf_src_node_copy(n, p);
5512
5513 p++;
5514 nr++;
5515 }
5516 PF_HASHROW_UNLOCK(sh);
5517 }
5518 error = copyout(pstore, psn->psn_src_nodes,
5519 sizeof(struct pf_src_node) * nr);
5520 if (error) {
5521 free(pstore, M_TEMP);
5522 break;
5523 }
5524 psn->psn_len = sizeof(struct pf_src_node) * nr;
5525 free(pstore, M_TEMP);
5526 break;
5527 }
5528
5529 case DIOCCLRSRCNODES: {
5530 pf_kill_srcnodes(NULL);
5531 break;
5532 }
5533
5534 case DIOCKILLSRCNODES:
5535 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5536 break;
5537
5538 #ifdef COMPAT_FREEBSD13
5539 case DIOCKEEPCOUNTERS_FREEBSD13:
5540 #endif
5541 case DIOCKEEPCOUNTERS:
5542 error = pf_keepcounters((struct pfioc_nv *)addr);
5543 break;
5544
5545 case DIOCGETSYNCOOKIES:
5546 error = pf_get_syncookies((struct pfioc_nv *)addr);
5547 break;
5548
5549 case DIOCSETSYNCOOKIES:
5550 error = pf_set_syncookies((struct pfioc_nv *)addr);
5551 break;
5552
5553 case DIOCSETHOSTID: {
5554 u_int32_t *hostid = (u_int32_t *)addr;
5555
5556 PF_RULES_WLOCK();
5557 if (*hostid == 0)
5558 V_pf_status.hostid = arc4random();
5559 else
5560 V_pf_status.hostid = *hostid;
5561 PF_RULES_WUNLOCK();
5562 break;
5563 }
5564
5565 case DIOCOSFPFLUSH:
5566 PF_RULES_WLOCK();
5567 pf_osfp_flush();
5568 PF_RULES_WUNLOCK();
5569 break;
5570
5571 case DIOCIGETIFACES: {
5572 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5573 struct pfi_kif *ifstore;
5574 size_t bufsiz;
5575
5576 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5577 error = ENODEV;
5578 break;
5579 }
5580
5581 if (io->pfiio_size < 0 ||
5582 io->pfiio_size > pf_ioctl_maxcount ||
5583 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5584 error = EINVAL;
5585 break;
5586 }
5587
5588 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5589
5590 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5591 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5592 M_TEMP, M_WAITOK | M_ZERO);
5593
5594 PF_RULES_RLOCK();
5595 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5596 PF_RULES_RUNLOCK();
5597 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5598 free(ifstore, M_TEMP);
5599 break;
5600 }
5601
5602 case DIOCSETIFFLAG: {
5603 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5604
5605 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5606
5607 PF_RULES_WLOCK();
5608 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5609 PF_RULES_WUNLOCK();
5610 break;
5611 }
5612
5613 case DIOCCLRIFFLAG: {
5614 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5615
5616 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5617
5618 PF_RULES_WLOCK();
5619 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5620 PF_RULES_WUNLOCK();
5621 break;
5622 }
5623
5624 case DIOCSETREASS: {
5625 u_int32_t *reass = (u_int32_t *)addr;
5626
5627 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5628 /* Removal of DF flag without reassembly enabled is not a
5629 * valid combination. Disable reassembly in such case. */
5630 if (!(V_pf_status.reass & PF_REASS_ENABLED))
5631 V_pf_status.reass = 0;
5632 break;
5633 }
5634
5635 default:
5636 error = ENODEV;
5637 break;
5638 }
5639 fail:
5640 CURVNET_RESTORE();
5641
5642 #undef ERROUT_IOCTL
5643
5644 return (error);
5645 }
5646
5647 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5648 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5649 {
5650 bzero(sp, sizeof(union pfsync_state_union));
5651
5652 /* copy from state key */
5653 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5654 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5655 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5656 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5657 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5658 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5659 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5660 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5661 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5662 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5663
5664 /* copy from state */
5665 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5666 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5667 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5668 sp->pfs_1301.expire = pf_state_expires(st);
5669 if (sp->pfs_1301.expire <= time_uptime)
5670 sp->pfs_1301.expire = htonl(0);
5671 else
5672 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5673
5674 sp->pfs_1301.direction = st->direction;
5675 sp->pfs_1301.log = st->act.log;
5676 sp->pfs_1301.timeout = st->timeout;
5677
5678 switch (msg_version) {
5679 case PFSYNC_MSG_VERSION_1301:
5680 sp->pfs_1301.state_flags = st->state_flags;
5681 break;
5682 case PFSYNC_MSG_VERSION_1400:
5683 sp->pfs_1400.state_flags = htons(st->state_flags);
5684 sp->pfs_1400.qid = htons(st->act.qid);
5685 sp->pfs_1400.pqid = htons(st->act.pqid);
5686 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5687 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5688 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5689 sp->pfs_1400.min_ttl = st->act.min_ttl;
5690 sp->pfs_1400.set_tos = st->act.set_tos;
5691 sp->pfs_1400.max_mss = htons(st->act.max_mss);
5692 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5693 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5694 sp->pfs_1400.rt = st->act.rt;
5695 if (st->act.rt_kif)
5696 strlcpy(sp->pfs_1400.rt_ifname,
5697 st->act.rt_kif->pfik_name,
5698 sizeof(sp->pfs_1400.rt_ifname));
5699 break;
5700 default:
5701 panic("%s: Unsupported pfsync_msg_version %d",
5702 __func__, msg_version);
5703 }
5704
5705 /*
5706 * XXX Why do we bother pfsyncing source node information if source
5707 * nodes are not synced? Showing users that there is source tracking
5708 * when there is none seems useless.
5709 */
5710 if (st->sns[PF_SN_LIMIT] != NULL)
5711 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5712 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
5713 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5714
5715 sp->pfs_1301.id = st->id;
5716 sp->pfs_1301.creatorid = st->creatorid;
5717 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5718 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5719
5720 if (st->rule == NULL)
5721 sp->pfs_1301.rule = htonl(-1);
5722 else
5723 sp->pfs_1301.rule = htonl(st->rule->nr);
5724 if (st->anchor == NULL)
5725 sp->pfs_1301.anchor = htonl(-1);
5726 else
5727 sp->pfs_1301.anchor = htonl(st->anchor->nr);
5728 if (st->nat_rule == NULL)
5729 sp->pfs_1301.nat_rule = htonl(-1);
5730 else
5731 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
5732
5733 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5734 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5735 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5736 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5737 }
5738
5739 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5740 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5741 {
5742 bzero(sp, sizeof(*sp));
5743
5744 sp->version = PF_STATE_VERSION;
5745
5746 /* copy from state key */
5747 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5748 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5749 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5750 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5751 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5752 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5753 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5754 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5755 sp->proto = st->key[PF_SK_WIRE]->proto;
5756 sp->af = st->key[PF_SK_WIRE]->af;
5757
5758 /* copy from state */
5759 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5760 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5761 sizeof(sp->orig_ifname));
5762 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr));
5763 sp->creation = htonl(time_uptime - (st->creation / 1000));
5764 sp->expire = pf_state_expires(st);
5765 if (sp->expire <= time_uptime)
5766 sp->expire = htonl(0);
5767 else
5768 sp->expire = htonl(sp->expire - time_uptime);
5769
5770 sp->direction = st->direction;
5771 sp->log = st->act.log;
5772 sp->timeout = st->timeout;
5773 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5774 sp->state_flags_compat = st->state_flags;
5775 sp->state_flags = htons(st->state_flags);
5776 if (st->sns[PF_SN_LIMIT] != NULL)
5777 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5778 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL)
5779 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5780 sp->id = st->id;
5781 sp->creatorid = st->creatorid;
5782 pf_state_peer_hton(&st->src, &sp->src);
5783 pf_state_peer_hton(&st->dst, &sp->dst);
5784
5785 if (st->rule == NULL)
5786 sp->rule = htonl(-1);
5787 else
5788 sp->rule = htonl(st->rule->nr);
5789 if (st->anchor == NULL)
5790 sp->anchor = htonl(-1);
5791 else
5792 sp->anchor = htonl(st->anchor->nr);
5793 if (st->nat_rule == NULL)
5794 sp->nat_rule = htonl(-1);
5795 else
5796 sp->nat_rule = htonl(st->nat_rule->nr);
5797
5798 sp->packets[0] = st->packets[0];
5799 sp->packets[1] = st->packets[1];
5800 sp->bytes[0] = st->bytes[0];
5801 sp->bytes[1] = st->bytes[1];
5802
5803 sp->qid = htons(st->act.qid);
5804 sp->pqid = htons(st->act.pqid);
5805 sp->dnpipe = htons(st->act.dnpipe);
5806 sp->dnrpipe = htons(st->act.dnrpipe);
5807 sp->rtableid = htonl(st->act.rtableid);
5808 sp->min_ttl = st->act.min_ttl;
5809 sp->set_tos = st->act.set_tos;
5810 sp->max_mss = htons(st->act.max_mss);
5811 sp->rt = st->act.rt;
5812 if (st->act.rt_kif)
5813 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
5814 sizeof(sp->rt_ifname));
5815 sp->set_prio[0] = st->act.set_prio[0];
5816 sp->set_prio[1] = st->act.set_prio[1];
5817
5818 }
5819
5820 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5821 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5822 {
5823 struct pfr_ktable *kt;
5824
5825 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5826
5827 kt = aw->p.tbl;
5828 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5829 kt = kt->pfrkt_root;
5830 aw->p.tbl = NULL;
5831 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5832 kt->pfrkt_cnt : -1;
5833 }
5834
5835 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5836 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5837 size_t number, char **names)
5838 {
5839 nvlist_t *nvc;
5840
5841 nvc = nvlist_create(0);
5842 if (nvc == NULL)
5843 return (ENOMEM);
5844
5845 for (int i = 0; i < number; i++) {
5846 nvlist_append_number_array(nvc, "counters",
5847 counter_u64_fetch(counters[i]));
5848 nvlist_append_string_array(nvc, "names",
5849 names[i]);
5850 nvlist_append_number_array(nvc, "ids",
5851 i);
5852 }
5853 nvlist_add_nvlist(nvl, name, nvc);
5854 nvlist_destroy(nvc);
5855
5856 return (0);
5857 }
5858
5859 static int
pf_getstatus(struct pfioc_nv * nv)5860 pf_getstatus(struct pfioc_nv *nv)
5861 {
5862 nvlist_t *nvl = NULL, *nvc = NULL;
5863 void *nvlpacked = NULL;
5864 int error;
5865 struct pf_status s;
5866 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5867 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5868 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5869 PF_RULES_RLOCK_TRACKER;
5870
5871 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
5872
5873 PF_RULES_RLOCK();
5874
5875 nvl = nvlist_create(0);
5876 if (nvl == NULL)
5877 ERROUT(ENOMEM);
5878
5879 nvlist_add_bool(nvl, "running", V_pf_status.running);
5880 nvlist_add_number(nvl, "since", V_pf_status.since);
5881 nvlist_add_number(nvl, "debug", V_pf_status.debug);
5882 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5883 nvlist_add_number(nvl, "states", V_pf_status.states);
5884 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5885 nvlist_add_number(nvl, "reass", V_pf_status.reass);
5886 nvlist_add_bool(nvl, "syncookies_active",
5887 V_pf_status.syncookies_active);
5888 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5889
5890 /* counters */
5891 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5892 PFRES_MAX, pf_reasons);
5893 if (error != 0)
5894 ERROUT(error);
5895
5896 /* lcounters */
5897 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5898 KLCNT_MAX, pf_lcounter);
5899 if (error != 0)
5900 ERROUT(error);
5901
5902 /* fcounters */
5903 nvc = nvlist_create(0);
5904 if (nvc == NULL)
5905 ERROUT(ENOMEM);
5906
5907 for (int i = 0; i < FCNT_MAX; i++) {
5908 nvlist_append_number_array(nvc, "counters",
5909 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5910 nvlist_append_string_array(nvc, "names",
5911 pf_fcounter[i]);
5912 nvlist_append_number_array(nvc, "ids",
5913 i);
5914 }
5915 nvlist_add_nvlist(nvl, "fcounters", nvc);
5916 nvlist_destroy(nvc);
5917 nvc = NULL;
5918
5919 /* scounters */
5920 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5921 SCNT_MAX, pf_fcounter);
5922 if (error != 0)
5923 ERROUT(error);
5924
5925 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5926 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5927 PF_MD5_DIGEST_LENGTH);
5928
5929 pfi_update_status(V_pf_status.ifname, &s);
5930
5931 /* pcounters / bcounters */
5932 for (int i = 0; i < 2; i++) {
5933 for (int j = 0; j < 2; j++) {
5934 for (int k = 0; k < 2; k++) {
5935 nvlist_append_number_array(nvl, "pcounters",
5936 s.pcounters[i][j][k]);
5937 }
5938 nvlist_append_number_array(nvl, "bcounters",
5939 s.bcounters[i][j]);
5940 }
5941 }
5942
5943 nvlpacked = nvlist_pack(nvl, &nv->len);
5944 if (nvlpacked == NULL)
5945 ERROUT(ENOMEM);
5946
5947 if (nv->size == 0)
5948 ERROUT(0);
5949 else if (nv->size < nv->len)
5950 ERROUT(ENOSPC);
5951
5952 PF_RULES_RUNLOCK();
5953 error = copyout(nvlpacked, nv->data, nv->len);
5954 goto done;
5955
5956 #undef ERROUT
5957 errout:
5958 PF_RULES_RUNLOCK();
5959 done:
5960 free(nvlpacked, M_NVLIST);
5961 nvlist_destroy(nvc);
5962 nvlist_destroy(nvl);
5963
5964 return (error);
5965 }
5966
5967 /*
5968 * XXX - Check for version mismatch!!!
5969 */
5970 static void
pf_clear_all_states(void)5971 pf_clear_all_states(void)
5972 {
5973 struct epoch_tracker et;
5974 struct pf_kstate *s;
5975 u_int i;
5976
5977 NET_EPOCH_ENTER(et);
5978 for (i = 0; i <= V_pf_hashmask; i++) {
5979 struct pf_idhash *ih = &V_pf_idhash[i];
5980 relock:
5981 PF_HASHROW_LOCK(ih);
5982 LIST_FOREACH(s, &ih->states, entry) {
5983 s->timeout = PFTM_PURGE;
5984 /* Don't send out individual delete messages. */
5985 s->state_flags |= PFSTATE_NOSYNC;
5986 pf_remove_state(s);
5987 goto relock;
5988 }
5989 PF_HASHROW_UNLOCK(ih);
5990 }
5991 NET_EPOCH_EXIT(et);
5992 }
5993
5994 static int
pf_clear_tables(void)5995 pf_clear_tables(void)
5996 {
5997 struct pfioc_table io;
5998 int error;
5999
6000 bzero(&io, sizeof(io));
6001 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
6002
6003 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
6004 io.pfrio_flags);
6005
6006 return (error);
6007 }
6008
6009 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)6010 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
6011 {
6012 struct pf_ksrc_node_list kill;
6013 u_int killed;
6014
6015 LIST_INIT(&kill);
6016 for (int i = 0; i <= V_pf_srchashmask; i++) {
6017 struct pf_srchash *sh = &V_pf_srchash[i];
6018 struct pf_ksrc_node *sn, *tmp;
6019
6020 PF_HASHROW_LOCK(sh);
6021 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
6022 if (psnk == NULL ||
6023 (PF_MATCHA(psnk->psnk_src.neg,
6024 &psnk->psnk_src.addr.v.a.addr,
6025 &psnk->psnk_src.addr.v.a.mask,
6026 &sn->addr, sn->af) &&
6027 PF_MATCHA(psnk->psnk_dst.neg,
6028 &psnk->psnk_dst.addr.v.a.addr,
6029 &psnk->psnk_dst.addr.v.a.mask,
6030 &sn->raddr, sn->af))) {
6031 pf_unlink_src_node(sn);
6032 LIST_INSERT_HEAD(&kill, sn, entry);
6033 sn->expire = 1;
6034 }
6035 PF_HASHROW_UNLOCK(sh);
6036 }
6037
6038 for (int i = 0; i <= V_pf_hashmask; i++) {
6039 struct pf_idhash *ih = &V_pf_idhash[i];
6040 struct pf_kstate *s;
6041
6042 PF_HASHROW_LOCK(ih);
6043 LIST_FOREACH(s, &ih->states, entry) {
6044 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX;
6045 sn_type++) {
6046 if (s->sns[sn_type] &&
6047 s->sns[sn_type]->expire == 1) {
6048 s->sns[sn_type] = NULL;
6049 }
6050 }
6051 }
6052 PF_HASHROW_UNLOCK(ih);
6053 }
6054
6055 killed = pf_free_src_nodes(&kill);
6056
6057 if (psnk != NULL)
6058 psnk->psnk_killed = killed;
6059 }
6060
6061 static int
pf_keepcounters(struct pfioc_nv * nv)6062 pf_keepcounters(struct pfioc_nv *nv)
6063 {
6064 nvlist_t *nvl = NULL;
6065 void *nvlpacked = NULL;
6066 int error = 0;
6067
6068 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6069
6070 if (nv->len > pf_ioctl_maxcount)
6071 ERROUT(ENOMEM);
6072
6073 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6074 error = copyin(nv->data, nvlpacked, nv->len);
6075 if (error)
6076 ERROUT(error);
6077
6078 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6079 if (nvl == NULL)
6080 ERROUT(EBADMSG);
6081
6082 if (! nvlist_exists_bool(nvl, "keep_counters"))
6083 ERROUT(EBADMSG);
6084
6085 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6086
6087 on_error:
6088 nvlist_destroy(nvl);
6089 free(nvlpacked, M_NVLIST);
6090 return (error);
6091 }
6092
6093 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6094 pf_clear_states(const struct pf_kstate_kill *kill)
6095 {
6096 struct pf_state_key_cmp match_key;
6097 struct pf_kstate *s;
6098 struct pfi_kkif *kif;
6099 int idx;
6100 unsigned int killed = 0, dir;
6101
6102 NET_EPOCH_ASSERT();
6103
6104 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6105 struct pf_idhash *ih = &V_pf_idhash[i];
6106
6107 relock_DIOCCLRSTATES:
6108 PF_HASHROW_LOCK(ih);
6109 LIST_FOREACH(s, &ih->states, entry) {
6110 /* For floating states look at the original kif. */
6111 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6112
6113 if (kill->psk_ifname[0] &&
6114 strcmp(kill->psk_ifname,
6115 kif->pfik_name))
6116 continue;
6117
6118 if (kill->psk_kill_match) {
6119 bzero(&match_key, sizeof(match_key));
6120
6121 if (s->direction == PF_OUT) {
6122 dir = PF_IN;
6123 idx = PF_SK_STACK;
6124 } else {
6125 dir = PF_OUT;
6126 idx = PF_SK_WIRE;
6127 }
6128
6129 match_key.af = s->key[idx]->af;
6130 match_key.proto = s->key[idx]->proto;
6131 PF_ACPY(&match_key.addr[0],
6132 &s->key[idx]->addr[1], match_key.af);
6133 match_key.port[0] = s->key[idx]->port[1];
6134 PF_ACPY(&match_key.addr[1],
6135 &s->key[idx]->addr[0], match_key.af);
6136 match_key.port[1] = s->key[idx]->port[0];
6137 }
6138
6139 /*
6140 * Don't send out individual
6141 * delete messages.
6142 */
6143 s->state_flags |= PFSTATE_NOSYNC;
6144 pf_remove_state(s);
6145 killed++;
6146
6147 if (kill->psk_kill_match)
6148 killed += pf_kill_matching_state(&match_key,
6149 dir);
6150
6151 goto relock_DIOCCLRSTATES;
6152 }
6153 PF_HASHROW_UNLOCK(ih);
6154 }
6155
6156 if (V_pfsync_clear_states_ptr != NULL)
6157 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6158
6159 return (killed);
6160 }
6161
6162 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6163 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6164 {
6165 struct pf_kstate *s;
6166
6167 NET_EPOCH_ASSERT();
6168 if (kill->psk_pfcmp.id) {
6169 if (kill->psk_pfcmp.creatorid == 0)
6170 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6171 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6172 kill->psk_pfcmp.creatorid))) {
6173 pf_remove_state(s);
6174 *killed = 1;
6175 }
6176 return;
6177 }
6178
6179 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6180 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6181 }
6182
6183 static int
pf_killstates_nv(struct pfioc_nv * nv)6184 pf_killstates_nv(struct pfioc_nv *nv)
6185 {
6186 struct pf_kstate_kill kill;
6187 struct epoch_tracker et;
6188 nvlist_t *nvl = NULL;
6189 void *nvlpacked = NULL;
6190 int error = 0;
6191 unsigned int killed = 0;
6192
6193 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6194
6195 if (nv->len > pf_ioctl_maxcount)
6196 ERROUT(ENOMEM);
6197
6198 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6199 error = copyin(nv->data, nvlpacked, nv->len);
6200 if (error)
6201 ERROUT(error);
6202
6203 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6204 if (nvl == NULL)
6205 ERROUT(EBADMSG);
6206
6207 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6208 if (error)
6209 ERROUT(error);
6210
6211 NET_EPOCH_ENTER(et);
6212 pf_killstates(&kill, &killed);
6213 NET_EPOCH_EXIT(et);
6214
6215 free(nvlpacked, M_NVLIST);
6216 nvlpacked = NULL;
6217 nvlist_destroy(nvl);
6218 nvl = nvlist_create(0);
6219 if (nvl == NULL)
6220 ERROUT(ENOMEM);
6221
6222 nvlist_add_number(nvl, "killed", killed);
6223
6224 nvlpacked = nvlist_pack(nvl, &nv->len);
6225 if (nvlpacked == NULL)
6226 ERROUT(ENOMEM);
6227
6228 if (nv->size == 0)
6229 ERROUT(0);
6230 else if (nv->size < nv->len)
6231 ERROUT(ENOSPC);
6232
6233 error = copyout(nvlpacked, nv->data, nv->len);
6234
6235 on_error:
6236 nvlist_destroy(nvl);
6237 free(nvlpacked, M_NVLIST);
6238 return (error);
6239 }
6240
6241 static int
pf_clearstates_nv(struct pfioc_nv * nv)6242 pf_clearstates_nv(struct pfioc_nv *nv)
6243 {
6244 struct pf_kstate_kill kill;
6245 struct epoch_tracker et;
6246 nvlist_t *nvl = NULL;
6247 void *nvlpacked = NULL;
6248 int error = 0;
6249 unsigned int killed;
6250
6251 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6252
6253 if (nv->len > pf_ioctl_maxcount)
6254 ERROUT(ENOMEM);
6255
6256 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6257 error = copyin(nv->data, nvlpacked, nv->len);
6258 if (error)
6259 ERROUT(error);
6260
6261 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6262 if (nvl == NULL)
6263 ERROUT(EBADMSG);
6264
6265 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6266 if (error)
6267 ERROUT(error);
6268
6269 NET_EPOCH_ENTER(et);
6270 killed = pf_clear_states(&kill);
6271 NET_EPOCH_EXIT(et);
6272
6273 free(nvlpacked, M_NVLIST);
6274 nvlpacked = NULL;
6275 nvlist_destroy(nvl);
6276 nvl = nvlist_create(0);
6277 if (nvl == NULL)
6278 ERROUT(ENOMEM);
6279
6280 nvlist_add_number(nvl, "killed", killed);
6281
6282 nvlpacked = nvlist_pack(nvl, &nv->len);
6283 if (nvlpacked == NULL)
6284 ERROUT(ENOMEM);
6285
6286 if (nv->size == 0)
6287 ERROUT(0);
6288 else if (nv->size < nv->len)
6289 ERROUT(ENOSPC);
6290
6291 error = copyout(nvlpacked, nv->data, nv->len);
6292
6293 #undef ERROUT
6294 on_error:
6295 nvlist_destroy(nvl);
6296 free(nvlpacked, M_NVLIST);
6297 return (error);
6298 }
6299
6300 static int
pf_getstate(struct pfioc_nv * nv)6301 pf_getstate(struct pfioc_nv *nv)
6302 {
6303 nvlist_t *nvl = NULL, *nvls;
6304 void *nvlpacked = NULL;
6305 struct pf_kstate *s = NULL;
6306 int error = 0;
6307 uint64_t id, creatorid;
6308
6309 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6310
6311 if (nv->len > pf_ioctl_maxcount)
6312 ERROUT(ENOMEM);
6313
6314 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6315 error = copyin(nv->data, nvlpacked, nv->len);
6316 if (error)
6317 ERROUT(error);
6318
6319 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6320 if (nvl == NULL)
6321 ERROUT(EBADMSG);
6322
6323 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6324 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6325
6326 s = pf_find_state_byid(id, creatorid);
6327 if (s == NULL)
6328 ERROUT(ENOENT);
6329
6330 free(nvlpacked, M_NVLIST);
6331 nvlpacked = NULL;
6332 nvlist_destroy(nvl);
6333 nvl = nvlist_create(0);
6334 if (nvl == NULL)
6335 ERROUT(ENOMEM);
6336
6337 nvls = pf_state_to_nvstate(s);
6338 if (nvls == NULL)
6339 ERROUT(ENOMEM);
6340
6341 nvlist_add_nvlist(nvl, "state", nvls);
6342 nvlist_destroy(nvls);
6343
6344 nvlpacked = nvlist_pack(nvl, &nv->len);
6345 if (nvlpacked == NULL)
6346 ERROUT(ENOMEM);
6347
6348 if (nv->size == 0)
6349 ERROUT(0);
6350 else if (nv->size < nv->len)
6351 ERROUT(ENOSPC);
6352
6353 error = copyout(nvlpacked, nv->data, nv->len);
6354
6355 #undef ERROUT
6356 errout:
6357 if (s != NULL)
6358 PF_STATE_UNLOCK(s);
6359 free(nvlpacked, M_NVLIST);
6360 nvlist_destroy(nvl);
6361 return (error);
6362 }
6363
6364 /*
6365 * XXX - Check for version mismatch!!!
6366 */
6367
6368 /*
6369 * Duplicate pfctl -Fa operation to get rid of as much as we can.
6370 */
6371 static int
shutdown_pf(void)6372 shutdown_pf(void)
6373 {
6374 int error = 0;
6375 u_int32_t t[5];
6376 char nn = '\0';
6377 struct pf_kanchor *anchor;
6378 struct pf_keth_anchor *eth_anchor;
6379 int rs_num;
6380
6381 do {
6382 /* Unlink rules of all user defined anchors */
6383 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6384 /* Wildcard based anchors may not have a respective
6385 * explicit anchor rule or they may be left empty
6386 * without rules. It leads to anchor.refcnt=0, and the
6387 * rest of the logic does not expect it. */
6388 if (anchor->refcnt == 0)
6389 anchor->refcnt = 1;
6390 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6391 if ((error = pf_begin_rules(&t[rs_num], rs_num,
6392 anchor->path)) != 0) {
6393 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6394 "anchor.path=%s rs_num=%d\n",
6395 anchor->path, rs_num));
6396 goto error; /* XXX: rollback? */
6397 }
6398 }
6399 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6400 error = pf_commit_rules(t[rs_num], rs_num,
6401 anchor->path);
6402 MPASS(error == 0);
6403 }
6404 }
6405
6406 /* Unlink rules of all user defined ether anchors */
6407 RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6408 &V_pf_keth_anchors) {
6409 /* Wildcard based anchors may not have a respective
6410 * explicit anchor rule or they may be left empty
6411 * without rules. It leads to anchor.refcnt=0, and the
6412 * rest of the logic does not expect it. */
6413 if (eth_anchor->refcnt == 0)
6414 eth_anchor->refcnt = 1;
6415 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6416 != 0) {
6417 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6418 "anchor.path=%s\n", eth_anchor->path));
6419 goto error;
6420 }
6421 error = pf_commit_eth(t[0], eth_anchor->path);
6422 MPASS(error == 0);
6423 }
6424
6425 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6426 != 0) {
6427 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6428 break;
6429 }
6430 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6431 != 0) {
6432 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6433 break; /* XXX: rollback? */
6434 }
6435 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6436 != 0) {
6437 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6438 break; /* XXX: rollback? */
6439 }
6440 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6441 != 0) {
6442 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6443 break; /* XXX: rollback? */
6444 }
6445 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6446 != 0) {
6447 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6448 break; /* XXX: rollback? */
6449 }
6450
6451 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6452 MPASS(error == 0);
6453 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6454 MPASS(error == 0);
6455 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6456 MPASS(error == 0);
6457 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6458 MPASS(error == 0);
6459 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6460 MPASS(error == 0);
6461
6462 if ((error = pf_clear_tables()) != 0)
6463 break;
6464
6465 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6466 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6467 break;
6468 }
6469 error = pf_commit_eth(t[0], &nn);
6470 MPASS(error == 0);
6471
6472 #ifdef ALTQ
6473 if ((error = pf_begin_altq(&t[0])) != 0) {
6474 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6475 break;
6476 }
6477 pf_commit_altq(t[0]);
6478 #endif
6479
6480 pf_clear_all_states();
6481
6482 pf_kill_srcnodes(NULL);
6483
6484 /* status does not use malloced mem so no need to cleanup */
6485 /* fingerprints and interfaces have their own cleanup code */
6486 } while(0);
6487
6488 error:
6489 return (error);
6490 }
6491
6492 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6493 pf_check_return(int chk, struct mbuf **m)
6494 {
6495
6496 switch (chk) {
6497 case PF_PASS:
6498 if (*m == NULL)
6499 return (PFIL_CONSUMED);
6500 else
6501 return (PFIL_PASS);
6502 break;
6503 default:
6504 if (*m != NULL) {
6505 m_freem(*m);
6506 *m = NULL;
6507 }
6508 return (PFIL_DROPPED);
6509 }
6510 }
6511
6512 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6513 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6514 void *ruleset __unused, struct inpcb *inp)
6515 {
6516 int chk;
6517
6518 CURVNET_ASSERT_SET();
6519
6520 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6521
6522 return (pf_check_return(chk, m));
6523 }
6524
6525 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6526 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6527 void *ruleset __unused, struct inpcb *inp)
6528 {
6529 int chk;
6530
6531 CURVNET_ASSERT_SET();
6532
6533 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6534
6535 return (pf_check_return(chk, m));
6536 }
6537
6538 #ifdef INET
6539 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6540 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6541 void *ruleset __unused, struct inpcb *inp)
6542 {
6543 int chk;
6544
6545 CURVNET_ASSERT_SET();
6546
6547 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
6548
6549 return (pf_check_return(chk, m));
6550 }
6551
6552 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6553 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6554 void *ruleset __unused, struct inpcb *inp)
6555 {
6556 int chk;
6557
6558 CURVNET_ASSERT_SET();
6559
6560 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
6561
6562 return (pf_check_return(chk, m));
6563 }
6564 #endif
6565
6566 #ifdef INET6
6567 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6568 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6569 void *ruleset __unused, struct inpcb *inp)
6570 {
6571 int chk;
6572
6573 CURVNET_ASSERT_SET();
6574
6575 /*
6576 * In case of loopback traffic IPv6 uses the real interface in
6577 * order to support scoped addresses. In order to support stateful
6578 * filtering we have change this to lo0 as it is the case in IPv4.
6579 */
6580 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6581 m, inp, NULL);
6582
6583 return (pf_check_return(chk, m));
6584 }
6585
6586 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6587 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6588 void *ruleset __unused, struct inpcb *inp)
6589 {
6590 int chk;
6591
6592 CURVNET_ASSERT_SET();
6593
6594 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL);
6595
6596 return (pf_check_return(chk, m));
6597 }
6598 #endif /* INET6 */
6599
6600 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6601 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6602 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
6603 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
6604
6605 #ifdef INET
6606 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6607 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6608 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
6609 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
6610 #endif
6611 #ifdef INET6
6612 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6613 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6614 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
6615 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
6616 #endif
6617
6618 static void
hook_pf_eth(void)6619 hook_pf_eth(void)
6620 {
6621 struct pfil_hook_args pha = {
6622 .pa_version = PFIL_VERSION,
6623 .pa_modname = "pf",
6624 .pa_type = PFIL_TYPE_ETHERNET,
6625 };
6626 struct pfil_link_args pla = {
6627 .pa_version = PFIL_VERSION,
6628 };
6629 int ret __diagused;
6630
6631 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6632 return;
6633
6634 pha.pa_mbuf_chk = pf_eth_check_in;
6635 pha.pa_flags = PFIL_IN;
6636 pha.pa_rulname = "eth-in";
6637 V_pf_eth_in_hook = pfil_add_hook(&pha);
6638 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6639 pla.pa_head = V_link_pfil_head;
6640 pla.pa_hook = V_pf_eth_in_hook;
6641 ret = pfil_link(&pla);
6642 MPASS(ret == 0);
6643 pha.pa_mbuf_chk = pf_eth_check_out;
6644 pha.pa_flags = PFIL_OUT;
6645 pha.pa_rulname = "eth-out";
6646 V_pf_eth_out_hook = pfil_add_hook(&pha);
6647 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6648 pla.pa_head = V_link_pfil_head;
6649 pla.pa_hook = V_pf_eth_out_hook;
6650 ret = pfil_link(&pla);
6651 MPASS(ret == 0);
6652
6653 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6654 }
6655
6656 static void
hook_pf(void)6657 hook_pf(void)
6658 {
6659 struct pfil_hook_args pha = {
6660 .pa_version = PFIL_VERSION,
6661 .pa_modname = "pf",
6662 };
6663 struct pfil_link_args pla = {
6664 .pa_version = PFIL_VERSION,
6665 };
6666 int ret __diagused;
6667
6668 if (atomic_load_bool(&V_pf_pfil_hooked))
6669 return;
6670
6671 #ifdef INET
6672 pha.pa_type = PFIL_TYPE_IP4;
6673 pha.pa_mbuf_chk = pf_check_in;
6674 pha.pa_flags = PFIL_IN;
6675 pha.pa_rulname = "default-in";
6676 V_pf_ip4_in_hook = pfil_add_hook(&pha);
6677 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6678 pla.pa_head = V_inet_pfil_head;
6679 pla.pa_hook = V_pf_ip4_in_hook;
6680 ret = pfil_link(&pla);
6681 MPASS(ret == 0);
6682 pha.pa_mbuf_chk = pf_check_out;
6683 pha.pa_flags = PFIL_OUT;
6684 pha.pa_rulname = "default-out";
6685 V_pf_ip4_out_hook = pfil_add_hook(&pha);
6686 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6687 pla.pa_head = V_inet_pfil_head;
6688 pla.pa_hook = V_pf_ip4_out_hook;
6689 ret = pfil_link(&pla);
6690 MPASS(ret == 0);
6691 if (V_pf_filter_local) {
6692 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6693 pla.pa_head = V_inet_local_pfil_head;
6694 pla.pa_hook = V_pf_ip4_out_hook;
6695 ret = pfil_link(&pla);
6696 MPASS(ret == 0);
6697 }
6698 #endif
6699 #ifdef INET6
6700 pha.pa_type = PFIL_TYPE_IP6;
6701 pha.pa_mbuf_chk = pf_check6_in;
6702 pha.pa_flags = PFIL_IN;
6703 pha.pa_rulname = "default-in6";
6704 V_pf_ip6_in_hook = pfil_add_hook(&pha);
6705 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6706 pla.pa_head = V_inet6_pfil_head;
6707 pla.pa_hook = V_pf_ip6_in_hook;
6708 ret = pfil_link(&pla);
6709 MPASS(ret == 0);
6710 pha.pa_mbuf_chk = pf_check6_out;
6711 pha.pa_rulname = "default-out6";
6712 pha.pa_flags = PFIL_OUT;
6713 V_pf_ip6_out_hook = pfil_add_hook(&pha);
6714 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6715 pla.pa_head = V_inet6_pfil_head;
6716 pla.pa_hook = V_pf_ip6_out_hook;
6717 ret = pfil_link(&pla);
6718 MPASS(ret == 0);
6719 if (V_pf_filter_local) {
6720 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6721 pla.pa_head = V_inet6_local_pfil_head;
6722 pla.pa_hook = V_pf_ip6_out_hook;
6723 ret = pfil_link(&pla);
6724 MPASS(ret == 0);
6725 }
6726 #endif
6727
6728 atomic_store_bool(&V_pf_pfil_hooked, true);
6729 }
6730
6731 static void
dehook_pf_eth(void)6732 dehook_pf_eth(void)
6733 {
6734
6735 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6736 return;
6737
6738 pfil_remove_hook(V_pf_eth_in_hook);
6739 pfil_remove_hook(V_pf_eth_out_hook);
6740
6741 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6742 }
6743
6744 static void
dehook_pf(void)6745 dehook_pf(void)
6746 {
6747
6748 if (!atomic_load_bool(&V_pf_pfil_hooked))
6749 return;
6750
6751 #ifdef INET
6752 pfil_remove_hook(V_pf_ip4_in_hook);
6753 pfil_remove_hook(V_pf_ip4_out_hook);
6754 #endif
6755 #ifdef INET6
6756 pfil_remove_hook(V_pf_ip6_in_hook);
6757 pfil_remove_hook(V_pf_ip6_out_hook);
6758 #endif
6759
6760 atomic_store_bool(&V_pf_pfil_hooked, false);
6761 }
6762
6763 static void
pf_load_vnet(void)6764 pf_load_vnet(void)
6765 {
6766 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6767 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6768
6769 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6770 sx_init(&V_pf_ioctl_lock, "pf ioctl");
6771
6772 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6773 PF_RULE_TAG_HASH_SIZE_DEFAULT);
6774 #ifdef ALTQ
6775 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6776 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6777 #endif
6778
6779 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6780
6781 pfattach_vnet();
6782 V_pf_vnet_active = 1;
6783 }
6784
6785 static int
pf_load(void)6786 pf_load(void)
6787 {
6788 int error;
6789
6790 sx_init(&pf_end_lock, "pf end thread");
6791
6792 pf_mtag_initialize();
6793
6794 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6795 if (pf_dev == NULL)
6796 return (ENOMEM);
6797
6798 pf_end_threads = 0;
6799 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6800 if (error != 0)
6801 return (error);
6802
6803 pfi_initialize();
6804
6805 return (0);
6806 }
6807
6808 static void
pf_unload_vnet(void)6809 pf_unload_vnet(void)
6810 {
6811 int ret __diagused;
6812
6813 V_pf_vnet_active = 0;
6814 V_pf_status.running = 0;
6815 dehook_pf();
6816 dehook_pf_eth();
6817
6818 PF_RULES_WLOCK();
6819 pf_syncookies_cleanup();
6820 shutdown_pf();
6821 PF_RULES_WUNLOCK();
6822
6823 ret = swi_remove(V_pf_swi_cookie);
6824 MPASS(ret == 0);
6825 ret = intr_event_destroy(V_pf_swi_ie);
6826 MPASS(ret == 0);
6827
6828 pf_unload_vnet_purge();
6829
6830 pf_normalize_cleanup();
6831 PF_RULES_WLOCK();
6832 pfi_cleanup_vnet();
6833 PF_RULES_WUNLOCK();
6834 pfr_cleanup();
6835 pf_osfp_flush();
6836 pf_cleanup();
6837 if (IS_DEFAULT_VNET(curvnet))
6838 pf_mtag_cleanup();
6839
6840 pf_cleanup_tagset(&V_pf_tags);
6841 #ifdef ALTQ
6842 pf_cleanup_tagset(&V_pf_qids);
6843 #endif
6844 uma_zdestroy(V_pf_tag_z);
6845
6846 #ifdef PF_WANT_32_TO_64_COUNTER
6847 PF_RULES_WLOCK();
6848 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6849
6850 MPASS(LIST_EMPTY(&V_pf_allkiflist));
6851 MPASS(V_pf_allkifcount == 0);
6852
6853 LIST_REMOVE(&V_pf_default_rule, allrulelist);
6854 V_pf_allrulecount--;
6855 LIST_REMOVE(V_pf_rulemarker, allrulelist);
6856
6857 MPASS(LIST_EMPTY(&V_pf_allrulelist));
6858 MPASS(V_pf_allrulecount == 0);
6859
6860 PF_RULES_WUNLOCK();
6861
6862 free(V_pf_kifmarker, PFI_MTYPE);
6863 free(V_pf_rulemarker, M_PFRULE);
6864 #endif
6865
6866 /* Free counters last as we updated them during shutdown. */
6867 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6868 for (int i = 0; i < 2; i++) {
6869 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6870 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6871 }
6872 counter_u64_free(V_pf_default_rule.states_cur);
6873 counter_u64_free(V_pf_default_rule.states_tot);
6874 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
6875 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]);
6876 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6877
6878 for (int i = 0; i < PFRES_MAX; i++)
6879 counter_u64_free(V_pf_status.counters[i]);
6880 for (int i = 0; i < KLCNT_MAX; i++)
6881 counter_u64_free(V_pf_status.lcounters[i]);
6882 for (int i = 0; i < FCNT_MAX; i++)
6883 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6884 for (int i = 0; i < SCNT_MAX; i++)
6885 counter_u64_free(V_pf_status.scounters[i]);
6886
6887 rm_destroy(&V_pf_rules_lock);
6888 sx_destroy(&V_pf_ioctl_lock);
6889 }
6890
6891 static void
pf_unload(void)6892 pf_unload(void)
6893 {
6894
6895 sx_xlock(&pf_end_lock);
6896 pf_end_threads = 1;
6897 while (pf_end_threads < 2) {
6898 wakeup_one(pf_purge_thread);
6899 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6900 }
6901 sx_xunlock(&pf_end_lock);
6902
6903 pf_nl_unregister();
6904
6905 if (pf_dev != NULL)
6906 destroy_dev(pf_dev);
6907
6908 pfi_cleanup();
6909
6910 sx_destroy(&pf_end_lock);
6911 }
6912
6913 static void
vnet_pf_init(void * unused __unused)6914 vnet_pf_init(void *unused __unused)
6915 {
6916
6917 pf_load_vnet();
6918 }
6919 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6920 vnet_pf_init, NULL);
6921
6922 static void
vnet_pf_uninit(const void * unused __unused)6923 vnet_pf_uninit(const void *unused __unused)
6924 {
6925
6926 pf_unload_vnet();
6927 }
6928 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6929 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6930 vnet_pf_uninit, NULL);
6931
6932 static int
pf_modevent(module_t mod,int type,void * data)6933 pf_modevent(module_t mod, int type, void *data)
6934 {
6935 int error = 0;
6936
6937 switch(type) {
6938 case MOD_LOAD:
6939 error = pf_load();
6940 pf_nl_register();
6941 break;
6942 case MOD_UNLOAD:
6943 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6944 * the vnet_pf_uninit()s */
6945 break;
6946 default:
6947 error = EINVAL;
6948 break;
6949 }
6950
6951 return (error);
6952 }
6953
6954 static moduledata_t pf_mod = {
6955 "pf",
6956 pf_modevent,
6957 0
6958 };
6959
6960 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6961 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6962 MODULE_DEPEND(pf, crypto, 1, 1, 1);
6963 MODULE_VERSION(pf, PF_MODVER);
6964