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 MALLOC_DEFINE(M_PF, "pf", "pf(4)");
191
192 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
193 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
194 #endif
195
196 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
197 #define V_pf_filter_local VNET(pf_filter_local)
198 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
199 &VNET_NAME(pf_filter_local), false,
200 "Enable filtering for packets delivered to local network stack");
201
202 #ifdef PF_DEFAULT_TO_DROP
203 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
204 #else
205 VNET_DEFINE_STATIC(bool, default_to_drop);
206 #endif
207 #define V_default_to_drop VNET(default_to_drop)
208 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
209 &VNET_NAME(default_to_drop), false,
210 "Make the default rule drop all packets.");
211
212 static void pf_init_tagset(struct pf_tagset *, unsigned int *,
213 unsigned int);
214 static void pf_cleanup_tagset(struct pf_tagset *);
215 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *);
216 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t);
217 static u_int16_t tagname2tag(struct pf_tagset *, const char *);
218 static u_int16_t pf_tagname2tag(const char *);
219 static void tag_unref(struct pf_tagset *, u_int16_t);
220
221 struct cdev *pf_dev;
222
223 /*
224 * XXX - These are new and need to be checked when moveing to a new version
225 */
226 static void pf_clear_all_states(void);
227 static int pf_killstates_row(struct pf_kstate_kill *,
228 struct pf_idhash *);
229 static int pf_killstates_nv(struct pfioc_nv *);
230 static int pf_clearstates_nv(struct pfioc_nv *);
231 static int pf_getstate(struct pfioc_nv *);
232 static int pf_getstatus(struct pfioc_nv *);
233 static int pf_clear_tables(void);
234 static void pf_kill_srcnodes(struct pfioc_src_node_kill *);
235 static int pf_keepcounters(struct pfioc_nv *);
236 static void pf_tbladdr_copyout(struct pf_addr_wrap *);
237
238 /*
239 * Wrapper functions for pfil(9) hooks
240 */
241 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
242 int flags, void *ruleset __unused, struct inpcb *inp);
243 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
244 int flags, void *ruleset __unused, struct inpcb *inp);
245 #ifdef INET
246 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
247 int flags, void *ruleset __unused, struct inpcb *inp);
248 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
249 int flags, void *ruleset __unused, struct inpcb *inp);
250 #endif
251 #ifdef INET6
252 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
253 int flags, void *ruleset __unused, struct inpcb *inp);
254 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
255 int flags, void *ruleset __unused, struct inpcb *inp);
256 #endif
257
258 static void hook_pf_eth(void);
259 static void hook_pf(void);
260 static void dehook_pf_eth(void);
261 static void dehook_pf(void);
262 static int shutdown_pf(void);
263 static int pf_load(void);
264 static void pf_unload(void);
265
266 static struct cdevsw pf_cdevsw = {
267 .d_ioctl = pfioctl,
268 .d_name = PF_NAME,
269 .d_version = D_VERSION,
270 };
271
272 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
273 #define V_pf_pfil_hooked VNET(pf_pfil_hooked)
274 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
275 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked)
276
277 /*
278 * We need a flag that is neither hooked nor running to know when
279 * the VNET is "valid". We primarily need this to control (global)
280 * external event, e.g., eventhandlers.
281 */
282 VNET_DEFINE(int, pf_vnet_active);
283 #define V_pf_vnet_active VNET(pf_vnet_active)
284
285 int pf_end_threads;
286 struct proc *pf_purge_proc;
287
288 VNET_DEFINE(struct rmlock, pf_rules_lock);
289 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
290 #define V_pf_ioctl_lock VNET(pf_ioctl_lock)
291 struct sx pf_end_lock;
292
293 /* pfsync */
294 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
295 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
296 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
297 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
298 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
299 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
300 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
301 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
302
303 /* pflog */
304 pflog_packet_t *pflog_packet_ptr = NULL;
305
306 /*
307 * Copy a user-provided string, returning an error if truncation would occur.
308 * Avoid scanning past "sz" bytes in the source string since there's no
309 * guarantee that it's nul-terminated.
310 */
311 static int
pf_user_strcpy(char * dst,const char * src,size_t sz)312 pf_user_strcpy(char *dst, const char *src, size_t sz)
313 {
314 if (strnlen(src, sz) == sz)
315 return (EINVAL);
316 (void)strlcpy(dst, src, sz);
317 return (0);
318 }
319
320 static void
pfattach_vnet(void)321 pfattach_vnet(void)
322 {
323 u_int32_t *my_timeout = V_pf_default_rule.timeout;
324
325 bzero(&V_pf_status, sizeof(V_pf_status));
326
327 pf_initialize();
328 pfr_initialize();
329 pfi_initialize_vnet();
330 pf_normalize_init();
331 pf_syncookies_init();
332
333 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
334 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
335 V_pf_limits[PF_LIMIT_ANCHORS].limit = PF_ANCHOR_HIWAT;
336 V_pf_limits[PF_LIMIT_ETH_ANCHORS].limit = PF_ANCHOR_HIWAT;
337
338 RB_INIT(&V_pf_anchors);
339 pf_init_kruleset(&pf_main_ruleset);
340
341 pf_init_keth(V_pf_keth);
342
343 /* default rule should never be garbage collected */
344 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
345 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
346 V_pf_default_rule.nr = (uint32_t)-1;
347 V_pf_default_rule.rtableid = -1;
348
349 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
350 for (int i = 0; i < 2; i++) {
351 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
352 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
353 }
354 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
355 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
356 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++)
357 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
358
359 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
360 M_WAITOK | M_ZERO);
361
362 #ifdef PF_WANT_32_TO_64_COUNTER
363 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
364 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
365 PF_RULES_WLOCK();
366 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
367 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
368 V_pf_allrulecount++;
369 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
370 PF_RULES_WUNLOCK();
371 #endif
372
373 /* initialize default timeouts */
374 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
375 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
376 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
377 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
378 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
379 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
380 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
381 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
382 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
383 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
384 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
385 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
386 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
387 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
388 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
389 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
390 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
391 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
392 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
393 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
394 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
395 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
396 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
397 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
398 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
399
400 V_pf_status.debug = PF_DEBUG_URGENT;
401 /*
402 * XXX This is different than in OpenBSD where reassembly is enabled by
403 * defult. In FreeBSD we expect people to still use scrub rules and
404 * switch to the new syntax later. Only when they switch they must
405 * explicitly enable reassemle. We could change the default once the
406 * scrub rule functionality is hopefully removed some day in future.
407 */
408 V_pf_status.reass = 0;
409
410 V_pf_pfil_hooked = false;
411 V_pf_pfil_eth_hooked = false;
412
413 /* XXX do our best to avoid a conflict */
414 V_pf_status.hostid = arc4random();
415
416 for (int i = 0; i < PFRES_MAX; i++)
417 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
418 for (int i = 0; i < KLCNT_MAX; i++)
419 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
420 for (int i = 0; i < FCNT_MAX; i++)
421 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
422 for (int i = 0; i < SCNT_MAX; i++)
423 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
424 for (int i = 0; i < NCNT_MAX; i++)
425 V_pf_status.ncounters[i] = counter_u64_alloc(M_WAITOK);
426
427 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
428 INTR_MPSAFE, &V_pf_swi_cookie) != 0)
429 /* XXXGL: leaked all above. */
430 return;
431 }
432
433 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)434 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
435 u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
436 u_int8_t check_ticket, int which)
437 {
438 struct pf_kruleset *ruleset;
439 struct pf_krule *rule;
440 int rs_num;
441
442 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT);
443
444 ruleset = pf_find_kruleset(anchor);
445 if (ruleset == NULL)
446 return (NULL);
447 rs_num = pf_get_ruleset_number(rule_action);
448 if (rs_num >= PF_RULESET_MAX)
449 return (NULL);
450 if (active) {
451 if (check_ticket && ticket !=
452 ruleset->rules[rs_num].active.ticket)
453 return (NULL);
454 if (r_last)
455 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
456 pf_krulequeue);
457 else
458 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
459 } else {
460 if (check_ticket && ticket !=
461 ruleset->rules[rs_num].inactive.ticket)
462 return (NULL);
463 if (r_last)
464 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
465 pf_krulequeue);
466 else
467 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
468 }
469 if (!r_last) {
470 while ((rule != NULL) && (rule->nr != rule_number))
471 rule = TAILQ_NEXT(rule, entries);
472 }
473 if (rule == NULL)
474 return (NULL);
475
476 switch (which) {
477 case PF_RDR:
478 return (&rule->rdr);
479 case PF_NAT:
480 return (&rule->nat);
481 case PF_RT:
482 return (&rule->route);
483 default:
484 panic("Unknow pool type %d", which);
485 }
486 }
487
488 static void
pf_mv_kpool(struct pf_kpalist * poola,struct pf_kpalist * poolb)489 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
490 {
491 struct pf_kpooladdr *mv_pool_pa;
492
493 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
494 TAILQ_REMOVE(poola, mv_pool_pa, entries);
495 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
496 }
497 }
498
499 static void
pf_empty_kpool(struct pf_kpalist * poola)500 pf_empty_kpool(struct pf_kpalist *poola)
501 {
502 struct pf_kpooladdr *pa;
503
504 while ((pa = TAILQ_FIRST(poola)) != NULL) {
505 switch (pa->addr.type) {
506 case PF_ADDR_DYNIFTL:
507 pfi_dynaddr_remove(pa->addr.p.dyn);
508 break;
509 case PF_ADDR_TABLE:
510 /* XXX: this could be unfinished pooladdr on pabuf */
511 if (pa->addr.p.tbl != NULL)
512 pfr_detach_table(pa->addr.p.tbl);
513 break;
514 }
515 if (pa->kif)
516 pfi_kkif_unref(pa->kif);
517 TAILQ_REMOVE(poola, pa, entries);
518 free(pa, M_PFRULE);
519 }
520 }
521
522 static void
pf_unlink_rule_locked(struct pf_krulequeue * rulequeue,struct pf_krule * rule)523 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
524 {
525
526 PF_RULES_WASSERT();
527 PF_UNLNKDRULES_ASSERT();
528
529 TAILQ_REMOVE(rulequeue, rule, entries);
530
531 rule->rule_ref |= PFRULE_REFS;
532 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
533 }
534
535 static void
pf_unlink_rule(struct pf_krulequeue * rulequeue,struct pf_krule * rule)536 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
537 {
538
539 PF_RULES_WASSERT();
540
541 PF_UNLNKDRULES_LOCK();
542 pf_unlink_rule_locked(rulequeue, rule);
543 PF_UNLNKDRULES_UNLOCK();
544 }
545
546 static void
pf_free_eth_rule(struct pf_keth_rule * rule)547 pf_free_eth_rule(struct pf_keth_rule *rule)
548 {
549 PF_RULES_WASSERT();
550
551 if (rule == NULL)
552 return;
553
554 if (rule->tag)
555 tag_unref(&V_pf_tags, rule->tag);
556 if (rule->match_tag)
557 tag_unref(&V_pf_tags, rule->match_tag);
558 #ifdef ALTQ
559 pf_qid_unref(rule->qid);
560 #endif
561
562 if (rule->bridge_to)
563 pfi_kkif_unref(rule->bridge_to);
564 if (rule->kif)
565 pfi_kkif_unref(rule->kif);
566
567 if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
568 pfr_detach_table(rule->ipsrc.addr.p.tbl);
569 if (rule->ipdst.addr.type == PF_ADDR_TABLE)
570 pfr_detach_table(rule->ipdst.addr.p.tbl);
571
572 counter_u64_free(rule->evaluations);
573 for (int i = 0; i < 2; i++) {
574 counter_u64_free(rule->packets[i]);
575 counter_u64_free(rule->bytes[i]);
576 }
577 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
578 pf_keth_anchor_remove(rule);
579
580 free(rule, M_PFRULE);
581 }
582
583 void
pf_free_rule(struct pf_krule * rule)584 pf_free_rule(struct pf_krule *rule)
585 {
586
587 PF_RULES_WASSERT();
588 PF_CONFIG_ASSERT();
589
590 if (rule->tag)
591 tag_unref(&V_pf_tags, rule->tag);
592 if (rule->match_tag)
593 tag_unref(&V_pf_tags, rule->match_tag);
594 #ifdef ALTQ
595 if (rule->pqid != rule->qid)
596 pf_qid_unref(rule->pqid);
597 pf_qid_unref(rule->qid);
598 #endif
599 switch (rule->src.addr.type) {
600 case PF_ADDR_DYNIFTL:
601 pfi_dynaddr_remove(rule->src.addr.p.dyn);
602 break;
603 case PF_ADDR_TABLE:
604 pfr_detach_table(rule->src.addr.p.tbl);
605 break;
606 }
607 switch (rule->dst.addr.type) {
608 case PF_ADDR_DYNIFTL:
609 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
610 break;
611 case PF_ADDR_TABLE:
612 pfr_detach_table(rule->dst.addr.p.tbl);
613 break;
614 }
615 if (rule->overload_tbl)
616 pfr_detach_table(rule->overload_tbl);
617 if (rule->kif)
618 pfi_kkif_unref(rule->kif);
619 if (rule->rcv_kif)
620 pfi_kkif_unref(rule->rcv_kif);
621 pf_remove_kanchor(rule);
622 pf_empty_kpool(&rule->rdr.list);
623 pf_empty_kpool(&rule->nat.list);
624 pf_empty_kpool(&rule->route.list);
625
626 pf_krule_free(rule);
627 }
628
629 static void
pf_init_tagset(struct pf_tagset * ts,unsigned int * tunable_size,unsigned int default_size)630 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
631 unsigned int default_size)
632 {
633 unsigned int i;
634 unsigned int hashsize;
635
636 if (*tunable_size == 0 || !powerof2(*tunable_size))
637 *tunable_size = default_size;
638
639 hashsize = *tunable_size;
640 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
641 M_WAITOK);
642 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
643 M_WAITOK);
644 ts->mask = hashsize - 1;
645 ts->seed = arc4random();
646 for (i = 0; i < hashsize; i++) {
647 TAILQ_INIT(&ts->namehash[i]);
648 TAILQ_INIT(&ts->taghash[i]);
649 }
650 BIT_FILL(TAGID_MAX, &ts->avail);
651 }
652
653 static void
pf_cleanup_tagset(struct pf_tagset * ts)654 pf_cleanup_tagset(struct pf_tagset *ts)
655 {
656 unsigned int i;
657 unsigned int hashsize;
658 struct pf_tagname *t, *tmp;
659
660 /*
661 * Only need to clean up one of the hashes as each tag is hashed
662 * into each table.
663 */
664 hashsize = ts->mask + 1;
665 for (i = 0; i < hashsize; i++)
666 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
667 uma_zfree(V_pf_tag_z, t);
668
669 free(ts->namehash, M_PFHASH);
670 free(ts->taghash, M_PFHASH);
671 }
672
673 static uint16_t
tagname2hashindex(const struct pf_tagset * ts,const char * tagname)674 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
675 {
676 size_t len;
677
678 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
679 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
680 }
681
682 static uint16_t
tag2hashindex(const struct pf_tagset * ts,uint16_t tag)683 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
684 {
685
686 return (tag & ts->mask);
687 }
688
689 static u_int16_t
tagname2tag(struct pf_tagset * ts,const char * tagname)690 tagname2tag(struct pf_tagset *ts, const char *tagname)
691 {
692 struct pf_tagname *tag;
693 u_int32_t index;
694 u_int16_t new_tagid;
695
696 PF_RULES_WASSERT();
697
698 index = tagname2hashindex(ts, tagname);
699 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
700 if (strcmp(tagname, tag->name) == 0) {
701 tag->ref++;
702 return (tag->tag);
703 }
704
705 /*
706 * new entry
707 *
708 * to avoid fragmentation, we do a linear search from the beginning
709 * and take the first free slot we find.
710 */
711 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
712 /*
713 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
714 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
715 * set. It may also return a bit number greater than TAGID_MAX due
716 * to rounding of the number of bits in the vector up to a multiple
717 * of the vector word size at declaration/allocation time.
718 */
719 if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
720 return (0);
721
722 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */
723 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
724
725 /* allocate and fill new struct pf_tagname */
726 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
727 if (tag == NULL)
728 return (0);
729 strlcpy(tag->name, tagname, sizeof(tag->name));
730 tag->tag = new_tagid;
731 tag->ref = 1;
732
733 /* Insert into namehash */
734 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
735
736 /* Insert into taghash */
737 index = tag2hashindex(ts, new_tagid);
738 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
739
740 return (tag->tag);
741 }
742
743 static void
tag_unref(struct pf_tagset * ts,u_int16_t tag)744 tag_unref(struct pf_tagset *ts, u_int16_t tag)
745 {
746 struct pf_tagname *t;
747 uint16_t index;
748
749 PF_RULES_WASSERT();
750
751 index = tag2hashindex(ts, tag);
752 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
753 if (tag == t->tag) {
754 if (--t->ref == 0) {
755 TAILQ_REMOVE(&ts->taghash[index], t,
756 taghash_entries);
757 index = tagname2hashindex(ts, t->name);
758 TAILQ_REMOVE(&ts->namehash[index], t,
759 namehash_entries);
760 /* Bits are 0-based for BIT_SET() */
761 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
762 uma_zfree(V_pf_tag_z, t);
763 }
764 break;
765 }
766 }
767
768 static uint16_t
pf_tagname2tag(const char * tagname)769 pf_tagname2tag(const char *tagname)
770 {
771 return (tagname2tag(&V_pf_tags, tagname));
772 }
773
774 static int
pf_begin_eth(uint32_t * ticket,const char * anchor)775 pf_begin_eth(uint32_t *ticket, const char *anchor)
776 {
777 struct pf_keth_rule *rule, *tmp;
778 struct pf_keth_ruleset *rs;
779
780 PF_RULES_WASSERT();
781
782 rs = pf_find_or_create_keth_ruleset(anchor);
783 if (rs == NULL)
784 return (EINVAL);
785
786 /* Purge old inactive rules. */
787 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
788 tmp) {
789 TAILQ_REMOVE(rs->inactive.rules, rule,
790 entries);
791 pf_free_eth_rule(rule);
792 }
793
794 *ticket = ++rs->inactive.ticket;
795 rs->inactive.open = 1;
796
797 return (0);
798 }
799
800 static int
pf_rollback_eth(uint32_t ticket,const char * anchor)801 pf_rollback_eth(uint32_t ticket, const char *anchor)
802 {
803 struct pf_keth_rule *rule, *tmp;
804 struct pf_keth_ruleset *rs;
805
806 PF_RULES_WASSERT();
807
808 rs = pf_find_keth_ruleset(anchor);
809 if (rs == NULL)
810 return (EINVAL);
811
812 if (!rs->inactive.open ||
813 ticket != rs->inactive.ticket)
814 return (0);
815
816 /* Purge old inactive rules. */
817 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
818 tmp) {
819 TAILQ_REMOVE(rs->inactive.rules, rule, entries);
820 pf_free_eth_rule(rule);
821 }
822
823 rs->inactive.open = 0;
824
825 pf_remove_if_empty_keth_ruleset(rs);
826
827 return (0);
828 }
829
830 #define PF_SET_SKIP_STEPS(i) \
831 do { \
832 while (head[i] != cur) { \
833 head[i]->skip[i].ptr = cur; \
834 head[i] = TAILQ_NEXT(head[i], entries); \
835 } \
836 } while (0)
837
838 static void
pf_eth_calc_skip_steps(struct pf_keth_ruleq * rules)839 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
840 {
841 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
842 int i;
843
844 cur = TAILQ_FIRST(rules);
845 prev = cur;
846 for (i = 0; i < PFE_SKIP_COUNT; ++i)
847 head[i] = cur;
848 while (cur != NULL) {
849 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
850 PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
851 if (cur->direction != prev->direction)
852 PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
853 if (cur->proto != prev->proto)
854 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
855 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
856 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
857 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
858 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
859 if (cur->ipsrc.neg != prev->ipsrc.neg ||
860 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
861 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
862 if (cur->ipdst.neg != prev->ipdst.neg ||
863 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
864 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
865
866 prev = cur;
867 cur = TAILQ_NEXT(cur, entries);
868 }
869 for (i = 0; i < PFE_SKIP_COUNT; ++i)
870 PF_SET_SKIP_STEPS(i);
871 }
872
873 static int
pf_commit_eth(uint32_t ticket,const char * anchor)874 pf_commit_eth(uint32_t ticket, const char *anchor)
875 {
876 struct pf_keth_ruleq *rules;
877 struct pf_keth_ruleset *rs;
878
879 rs = pf_find_keth_ruleset(anchor);
880 if (rs == NULL) {
881 return (EINVAL);
882 }
883
884 if (!rs->inactive.open ||
885 ticket != rs->inactive.ticket)
886 return (EBUSY);
887
888 PF_RULES_WASSERT();
889
890 pf_eth_calc_skip_steps(rs->inactive.rules);
891
892 rules = rs->active.rules;
893 atomic_store_ptr(&rs->active.rules, rs->inactive.rules);
894 rs->inactive.rules = rules;
895 rs->inactive.ticket = rs->active.ticket;
896
897 return (pf_rollback_eth(rs->inactive.ticket,
898 rs->anchor ? rs->anchor->path : ""));
899 }
900
901 #ifdef ALTQ
902 static uint16_t
pf_qname2qid(const char * qname)903 pf_qname2qid(const char *qname)
904 {
905 return (tagname2tag(&V_pf_qids, qname));
906 }
907
908 static void
pf_qid_unref(uint16_t qid)909 pf_qid_unref(uint16_t qid)
910 {
911 tag_unref(&V_pf_qids, qid);
912 }
913
914 static int
pf_begin_altq(u_int32_t * ticket)915 pf_begin_altq(u_int32_t *ticket)
916 {
917 struct pf_altq *altq, *tmp;
918 int error = 0;
919
920 PF_RULES_WASSERT();
921
922 /* Purge the old altq lists */
923 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
924 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
925 /* detach and destroy the discipline */
926 error = altq_remove(altq);
927 }
928 free(altq, M_PFALTQ);
929 }
930 TAILQ_INIT(V_pf_altq_ifs_inactive);
931 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
932 pf_qid_unref(altq->qid);
933 free(altq, M_PFALTQ);
934 }
935 TAILQ_INIT(V_pf_altqs_inactive);
936 if (error)
937 return (error);
938 *ticket = ++V_ticket_altqs_inactive;
939 V_altqs_inactive_open = 1;
940 return (0);
941 }
942
943 static int
pf_rollback_altq(u_int32_t ticket)944 pf_rollback_altq(u_int32_t ticket)
945 {
946 struct pf_altq *altq, *tmp;
947 int error = 0;
948
949 PF_RULES_WASSERT();
950
951 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
952 return (0);
953 /* Purge the old altq lists */
954 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
955 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
956 /* detach and destroy the discipline */
957 error = altq_remove(altq);
958 }
959 free(altq, M_PFALTQ);
960 }
961 TAILQ_INIT(V_pf_altq_ifs_inactive);
962 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
963 pf_qid_unref(altq->qid);
964 free(altq, M_PFALTQ);
965 }
966 TAILQ_INIT(V_pf_altqs_inactive);
967 V_altqs_inactive_open = 0;
968 return (error);
969 }
970
971 static int
pf_commit_altq(u_int32_t ticket)972 pf_commit_altq(u_int32_t ticket)
973 {
974 struct pf_altqqueue *old_altqs, *old_altq_ifs;
975 struct pf_altq *altq, *tmp;
976 int err, error = 0;
977
978 PF_RULES_WASSERT();
979
980 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
981 return (EBUSY);
982
983 /* swap altqs, keep the old. */
984 old_altqs = V_pf_altqs_active;
985 old_altq_ifs = V_pf_altq_ifs_active;
986 V_pf_altqs_active = V_pf_altqs_inactive;
987 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
988 V_pf_altqs_inactive = old_altqs;
989 V_pf_altq_ifs_inactive = old_altq_ifs;
990 V_ticket_altqs_active = V_ticket_altqs_inactive;
991
992 /* Attach new disciplines */
993 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
994 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
995 /* attach the discipline */
996 error = altq_pfattach(altq);
997 if (error == 0 && V_pf_altq_running)
998 error = pf_enable_altq(altq);
999 if (error != 0)
1000 return (error);
1001 }
1002 }
1003
1004 /* Purge the old altq lists */
1005 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1006 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1007 /* detach and destroy the discipline */
1008 if (V_pf_altq_running)
1009 error = pf_disable_altq(altq);
1010 err = altq_pfdetach(altq);
1011 if (err != 0 && error == 0)
1012 error = err;
1013 err = altq_remove(altq);
1014 if (err != 0 && error == 0)
1015 error = err;
1016 }
1017 free(altq, M_PFALTQ);
1018 }
1019 TAILQ_INIT(V_pf_altq_ifs_inactive);
1020 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1021 pf_qid_unref(altq->qid);
1022 free(altq, M_PFALTQ);
1023 }
1024 TAILQ_INIT(V_pf_altqs_inactive);
1025
1026 V_altqs_inactive_open = 0;
1027 return (error);
1028 }
1029
1030 static int
pf_enable_altq(struct pf_altq * altq)1031 pf_enable_altq(struct pf_altq *altq)
1032 {
1033 struct ifnet *ifp;
1034 struct tb_profile tb;
1035 int error = 0;
1036
1037 if ((ifp = ifunit(altq->ifname)) == NULL)
1038 return (EINVAL);
1039
1040 if (ifp->if_snd.altq_type != ALTQT_NONE)
1041 error = altq_enable(&ifp->if_snd);
1042
1043 /* set tokenbucket regulator */
1044 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1045 tb.rate = altq->ifbandwidth;
1046 tb.depth = altq->tbrsize;
1047 error = tbr_set(&ifp->if_snd, &tb);
1048 }
1049
1050 return (error);
1051 }
1052
1053 static int
pf_disable_altq(struct pf_altq * altq)1054 pf_disable_altq(struct pf_altq *altq)
1055 {
1056 struct ifnet *ifp;
1057 struct tb_profile tb;
1058 int error;
1059
1060 if ((ifp = ifunit(altq->ifname)) == NULL)
1061 return (EINVAL);
1062
1063 /*
1064 * when the discipline is no longer referenced, it was overridden
1065 * by a new one. if so, just return.
1066 */
1067 if (altq->altq_disc != ifp->if_snd.altq_disc)
1068 return (0);
1069
1070 error = altq_disable(&ifp->if_snd);
1071
1072 if (error == 0) {
1073 /* clear tokenbucket regulator */
1074 tb.rate = 0;
1075 error = tbr_set(&ifp->if_snd, &tb);
1076 }
1077
1078 return (error);
1079 }
1080
1081 static int
pf_altq_ifnet_event_add(struct ifnet * ifp,int remove,u_int32_t ticket,struct pf_altq * altq)1082 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1083 struct pf_altq *altq)
1084 {
1085 struct ifnet *ifp1;
1086 int error = 0;
1087
1088 /* Deactivate the interface in question */
1089 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1090 if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1091 (remove && ifp1 == ifp)) {
1092 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1093 } else {
1094 error = altq_add(ifp1, altq);
1095
1096 if (ticket != V_ticket_altqs_inactive)
1097 error = EBUSY;
1098
1099 if (error)
1100 free(altq, M_PFALTQ);
1101 }
1102
1103 return (error);
1104 }
1105
1106 void
pf_altq_ifnet_event(struct ifnet * ifp,int remove)1107 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1108 {
1109 struct pf_altq *a1, *a2, *a3;
1110 u_int32_t ticket;
1111 int error = 0;
1112
1113 /*
1114 * No need to re-evaluate the configuration for events on interfaces
1115 * that do not support ALTQ, as it's not possible for such
1116 * interfaces to be part of the configuration.
1117 */
1118 if (!ALTQ_IS_READY(&ifp->if_snd))
1119 return;
1120
1121 /* Interrupt userland queue modifications */
1122 if (V_altqs_inactive_open)
1123 pf_rollback_altq(V_ticket_altqs_inactive);
1124
1125 /* Start new altq ruleset */
1126 if (pf_begin_altq(&ticket))
1127 return;
1128
1129 /* Copy the current active set */
1130 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1131 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1132 if (a2 == NULL) {
1133 error = ENOMEM;
1134 break;
1135 }
1136 bcopy(a1, a2, sizeof(struct pf_altq));
1137
1138 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1139 if (error)
1140 break;
1141
1142 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1143 }
1144 if (error)
1145 goto out;
1146 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1147 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1148 if (a2 == NULL) {
1149 error = ENOMEM;
1150 break;
1151 }
1152 bcopy(a1, a2, sizeof(struct pf_altq));
1153
1154 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1155 error = EBUSY;
1156 free(a2, M_PFALTQ);
1157 break;
1158 }
1159 a2->altq_disc = NULL;
1160 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1161 if (strncmp(a3->ifname, a2->ifname,
1162 IFNAMSIZ) == 0) {
1163 a2->altq_disc = a3->altq_disc;
1164 break;
1165 }
1166 }
1167 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1168 if (error)
1169 break;
1170
1171 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1172 }
1173
1174 out:
1175 if (error != 0)
1176 pf_rollback_altq(ticket);
1177 else
1178 pf_commit_altq(ticket);
1179 }
1180 #endif /* ALTQ */
1181
1182 static struct pf_krule_global *
pf_rule_tree_alloc(int flags)1183 pf_rule_tree_alloc(int flags)
1184 {
1185 struct pf_krule_global *tree;
1186
1187 tree = malloc(sizeof(struct pf_krule_global), M_PF, flags);
1188 if (tree == NULL)
1189 return (NULL);
1190 RB_INIT(tree);
1191 return (tree);
1192 }
1193
1194 void
pf_rule_tree_free(struct pf_krule_global * tree)1195 pf_rule_tree_free(struct pf_krule_global *tree)
1196 {
1197
1198 free(tree, M_PF);
1199 }
1200
1201 static int
pf_begin_rules(u_int32_t * ticket,int rs_num,const char * anchor)1202 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1203 {
1204 struct pf_krule_global *tree;
1205 struct pf_kruleset *rs;
1206 struct pf_krule *rule;
1207
1208 PF_RULES_WASSERT();
1209
1210 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1211 return (EINVAL);
1212 tree = pf_rule_tree_alloc(M_NOWAIT);
1213 if (tree == NULL)
1214 return (ENOMEM);
1215 rs = pf_find_or_create_kruleset(anchor);
1216 if (rs == NULL) {
1217 pf_rule_tree_free(tree);
1218 return (EINVAL);
1219 }
1220 pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1221 rs->rules[rs_num].inactive.tree = tree;
1222
1223 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1224 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1225 rs->rules[rs_num].inactive.rcount--;
1226 }
1227 *ticket = ++rs->rules[rs_num].inactive.ticket;
1228 rs->rules[rs_num].inactive.open = 1;
1229 return (0);
1230 }
1231
1232 static int
pf_rollback_rules(u_int32_t ticket,int rs_num,char * anchor)1233 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1234 {
1235 struct pf_kruleset *rs;
1236 struct pf_krule *rule;
1237
1238 PF_RULES_WASSERT();
1239
1240 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1241 return (EINVAL);
1242 rs = pf_find_kruleset(anchor);
1243 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1244 rs->rules[rs_num].inactive.ticket != ticket)
1245 return (0);
1246 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1247 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1248 rs->rules[rs_num].inactive.rcount--;
1249 }
1250 rs->rules[rs_num].inactive.open = 0;
1251 return (0);
1252 }
1253
1254 #define PF_MD5_UPD(st, elm) \
1255 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1256
1257 #define PF_MD5_UPD_STR(st, elm) \
1258 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1259
1260 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \
1261 (stor) = htonl((st)->elm); \
1262 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1263 } while (0)
1264
1265 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \
1266 (stor) = htons((st)->elm); \
1267 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1268 } while (0)
1269
1270 static void
pf_hash_rule_addr(MD5_CTX * ctx,struct pf_rule_addr * pfr)1271 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1272 {
1273 PF_MD5_UPD(pfr, addr.type);
1274 switch (pfr->addr.type) {
1275 case PF_ADDR_DYNIFTL:
1276 PF_MD5_UPD(pfr, addr.v.ifname);
1277 PF_MD5_UPD(pfr, addr.iflags);
1278 break;
1279 case PF_ADDR_TABLE:
1280 if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX,
1281 strlen(PF_OPTIMIZER_TABLE_PFX)))
1282 PF_MD5_UPD(pfr, addr.v.tblname);
1283 break;
1284 case PF_ADDR_ADDRMASK:
1285 /* XXX ignore af? */
1286 PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1287 PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1288 break;
1289 }
1290
1291 PF_MD5_UPD(pfr, port[0]);
1292 PF_MD5_UPD(pfr, port[1]);
1293 PF_MD5_UPD(pfr, neg);
1294 PF_MD5_UPD(pfr, port_op);
1295 }
1296
1297 static void
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)1298 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1299 {
1300 u_int16_t x;
1301 u_int32_t y;
1302
1303 pf_hash_rule_addr(ctx, &rule->src);
1304 pf_hash_rule_addr(ctx, &rule->dst);
1305 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1306 PF_MD5_UPD_STR(rule, label[i]);
1307 PF_MD5_UPD_STR(rule, ifname);
1308 PF_MD5_UPD_STR(rule, rcv_ifname);
1309 PF_MD5_UPD_STR(rule, match_tagname);
1310 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1311 PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1312 PF_MD5_UPD_HTONL(rule, prob, y);
1313 PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1314 PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1315 PF_MD5_UPD(rule, uid.op);
1316 PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1317 PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1318 PF_MD5_UPD(rule, gid.op);
1319 PF_MD5_UPD_HTONL(rule, rule_flag, y);
1320 PF_MD5_UPD(rule, action);
1321 PF_MD5_UPD(rule, direction);
1322 PF_MD5_UPD(rule, af);
1323 PF_MD5_UPD(rule, quick);
1324 PF_MD5_UPD(rule, ifnot);
1325 PF_MD5_UPD(rule, rcvifnot);
1326 PF_MD5_UPD(rule, match_tag_not);
1327 PF_MD5_UPD(rule, natpass);
1328 PF_MD5_UPD(rule, keep_state);
1329 PF_MD5_UPD(rule, proto);
1330 PF_MD5_UPD(rule, type);
1331 PF_MD5_UPD(rule, code);
1332 PF_MD5_UPD(rule, flags);
1333 PF_MD5_UPD(rule, flagset);
1334 PF_MD5_UPD(rule, allow_opts);
1335 PF_MD5_UPD(rule, rt);
1336 PF_MD5_UPD(rule, tos);
1337 PF_MD5_UPD(rule, scrub_flags);
1338 PF_MD5_UPD(rule, min_ttl);
1339 PF_MD5_UPD(rule, set_tos);
1340 if (rule->anchor != NULL)
1341 PF_MD5_UPD_STR(rule, anchor->path);
1342 }
1343
1344 static void
pf_hash_rule(struct pf_krule * rule)1345 pf_hash_rule(struct pf_krule *rule)
1346 {
1347 MD5_CTX ctx;
1348
1349 MD5Init(&ctx);
1350 pf_hash_rule_rolling(&ctx, rule);
1351 MD5Final(rule->md5sum, &ctx);
1352 }
1353
1354 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1355 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1356 {
1357
1358 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1359 }
1360
1361 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1362 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1363 {
1364 struct pf_kruleset *rs;
1365 struct pf_krule *rule, *old_rule;
1366 struct pf_krulequeue *old_rules;
1367 struct pf_krule_global *old_tree;
1368 int error;
1369 u_int32_t old_rcount;
1370
1371 PF_RULES_WASSERT();
1372
1373 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1374 return (EINVAL);
1375 rs = pf_find_kruleset(anchor);
1376 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1377 ticket != rs->rules[rs_num].inactive.ticket)
1378 return (EBUSY);
1379
1380 /* Calculate checksum for the main ruleset */
1381 if (rs == &pf_main_ruleset) {
1382 error = pf_setup_pfsync_matching(rs);
1383 if (error != 0)
1384 return (error);
1385 }
1386
1387 /* Swap rules, keep the old. */
1388 old_rules = rs->rules[rs_num].active.ptr;
1389 old_rcount = rs->rules[rs_num].active.rcount;
1390 old_tree = rs->rules[rs_num].active.tree;
1391
1392 rs->rules[rs_num].active.ptr =
1393 rs->rules[rs_num].inactive.ptr;
1394 rs->rules[rs_num].active.tree =
1395 rs->rules[rs_num].inactive.tree;
1396 rs->rules[rs_num].active.rcount =
1397 rs->rules[rs_num].inactive.rcount;
1398
1399 /* Attempt to preserve counter information. */
1400 if (V_pf_status.keep_counters && old_tree != NULL) {
1401 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1402 entries) {
1403 old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1404 if (old_rule == NULL) {
1405 continue;
1406 }
1407 pf_counter_u64_critical_enter();
1408 pf_counter_u64_rollup_protected(&rule->evaluations,
1409 pf_counter_u64_fetch(&old_rule->evaluations));
1410 pf_counter_u64_rollup_protected(&rule->packets[0],
1411 pf_counter_u64_fetch(&old_rule->packets[0]));
1412 pf_counter_u64_rollup_protected(&rule->packets[1],
1413 pf_counter_u64_fetch(&old_rule->packets[1]));
1414 pf_counter_u64_rollup_protected(&rule->bytes[0],
1415 pf_counter_u64_fetch(&old_rule->bytes[0]));
1416 pf_counter_u64_rollup_protected(&rule->bytes[1],
1417 pf_counter_u64_fetch(&old_rule->bytes[1]));
1418 pf_counter_u64_critical_exit();
1419 }
1420 }
1421
1422 rs->rules[rs_num].inactive.ptr = old_rules;
1423 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1424 rs->rules[rs_num].inactive.rcount = old_rcount;
1425
1426 rs->rules[rs_num].active.ticket =
1427 rs->rules[rs_num].inactive.ticket;
1428 pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1429
1430 /* Purge the old rule list. */
1431 PF_UNLNKDRULES_LOCK();
1432 while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1433 pf_unlink_rule_locked(old_rules, rule);
1434 PF_UNLNKDRULES_UNLOCK();
1435 rs->rules[rs_num].inactive.rcount = 0;
1436 rs->rules[rs_num].inactive.open = 0;
1437 pf_remove_if_empty_kruleset(rs);
1438 pf_rule_tree_free(old_tree);
1439
1440 return (0);
1441 }
1442
1443 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1444 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1445 {
1446 MD5_CTX ctx;
1447 struct pf_krule *rule;
1448 int rs_cnt;
1449 u_int8_t digest[PF_MD5_DIGEST_LENGTH];
1450
1451 MD5Init(&ctx);
1452 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1453 /* XXX PF_RULESET_SCRUB as well? */
1454 if (rs_cnt == PF_RULESET_SCRUB)
1455 continue;
1456
1457 if (rs->rules[rs_cnt].inactive.rcount) {
1458 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1459 entries) {
1460 pf_hash_rule_rolling(&ctx, rule);
1461 }
1462 }
1463 }
1464
1465 MD5Final(digest, &ctx);
1466 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1467 return (0);
1468 }
1469
1470 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1471 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1472 {
1473 int error = 0;
1474
1475 switch (addr->type) {
1476 case PF_ADDR_TABLE:
1477 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1478 if (addr->p.tbl == NULL)
1479 error = ENOMEM;
1480 break;
1481 default:
1482 error = EINVAL;
1483 }
1484
1485 return (error);
1486 }
1487
1488 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1489 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1490 sa_family_t af)
1491 {
1492 int error = 0;
1493
1494 switch (addr->type) {
1495 case PF_ADDR_TABLE:
1496 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1497 if (addr->p.tbl == NULL)
1498 error = ENOMEM;
1499 break;
1500 case PF_ADDR_DYNIFTL:
1501 error = pfi_dynaddr_setup(addr, af);
1502 break;
1503 }
1504
1505 return (error);
1506 }
1507
1508 void
pf_addr_copyout(struct pf_addr_wrap * addr)1509 pf_addr_copyout(struct pf_addr_wrap *addr)
1510 {
1511
1512 switch (addr->type) {
1513 case PF_ADDR_DYNIFTL:
1514 pfi_dynaddr_copyout(addr);
1515 break;
1516 case PF_ADDR_TABLE:
1517 pf_tbladdr_copyout(addr);
1518 break;
1519 }
1520 }
1521
1522 static void
pf_src_node_copy(const struct pf_ksrc_node * in,struct pf_src_node * out)1523 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1524 {
1525 int secs = time_uptime;
1526
1527 bzero(out, sizeof(struct pf_src_node));
1528
1529 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1530 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1531
1532 if (in->rule != NULL)
1533 out->rule.nr = in->rule->nr;
1534
1535 for (int i = 0; i < 2; i++) {
1536 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1537 out->packets[i] = counter_u64_fetch(in->packets[i]);
1538 }
1539
1540 out->states = in->states;
1541 out->conn = in->conn;
1542 out->af = in->af;
1543 out->ruletype = in->ruletype;
1544
1545 out->creation = secs - in->creation;
1546 if (out->expire > secs)
1547 out->expire -= secs;
1548 else
1549 out->expire = 0;
1550
1551 /* Adjust the connection rate estimate. */
1552 out->conn_rate.limit = in->conn_rate.limit;
1553 out->conn_rate.seconds = in->conn_rate.seconds;
1554 /* If there's no limit there's no counter_rate. */
1555 if (in->conn_rate.cr != NULL)
1556 out->conn_rate.count = counter_rate_get(in->conn_rate.cr);
1557 }
1558
1559 #ifdef ALTQ
1560 /*
1561 * Handle export of struct pf_kaltq to user binaries that may be using any
1562 * version of struct pf_altq.
1563 */
1564 static int
pf_export_kaltq(struct pf_altq * q,struct pfioc_altq_v1 * pa,size_t ioc_size)1565 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1566 {
1567 u_int32_t version;
1568
1569 if (ioc_size == sizeof(struct pfioc_altq_v0))
1570 version = 0;
1571 else
1572 version = pa->version;
1573
1574 if (version > PFIOC_ALTQ_VERSION)
1575 return (EINVAL);
1576
1577 #define ASSIGN(x) exported_q->x = q->x
1578 #define COPY(x) \
1579 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1580 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1581 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1582
1583 switch (version) {
1584 case 0: {
1585 struct pf_altq_v0 *exported_q =
1586 &((struct pfioc_altq_v0 *)pa)->altq;
1587
1588 COPY(ifname);
1589
1590 ASSIGN(scheduler);
1591 ASSIGN(tbrsize);
1592 exported_q->tbrsize = SATU16(q->tbrsize);
1593 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1594
1595 COPY(qname);
1596 COPY(parent);
1597 ASSIGN(parent_qid);
1598 exported_q->bandwidth = SATU32(q->bandwidth);
1599 ASSIGN(priority);
1600 ASSIGN(local_flags);
1601
1602 ASSIGN(qlimit);
1603 ASSIGN(flags);
1604
1605 if (q->scheduler == ALTQT_HFSC) {
1606 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1607 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1608 SATU32(q->pq_u.hfsc_opts.x)
1609
1610 ASSIGN_OPT_SATU32(rtsc_m1);
1611 ASSIGN_OPT(rtsc_d);
1612 ASSIGN_OPT_SATU32(rtsc_m2);
1613
1614 ASSIGN_OPT_SATU32(lssc_m1);
1615 ASSIGN_OPT(lssc_d);
1616 ASSIGN_OPT_SATU32(lssc_m2);
1617
1618 ASSIGN_OPT_SATU32(ulsc_m1);
1619 ASSIGN_OPT(ulsc_d);
1620 ASSIGN_OPT_SATU32(ulsc_m2);
1621
1622 ASSIGN_OPT(flags);
1623
1624 #undef ASSIGN_OPT
1625 #undef ASSIGN_OPT_SATU32
1626 } else
1627 COPY(pq_u);
1628
1629 ASSIGN(qid);
1630 break;
1631 }
1632 case 1: {
1633 struct pf_altq_v1 *exported_q =
1634 &((struct pfioc_altq_v1 *)pa)->altq;
1635
1636 COPY(ifname);
1637
1638 ASSIGN(scheduler);
1639 ASSIGN(tbrsize);
1640 ASSIGN(ifbandwidth);
1641
1642 COPY(qname);
1643 COPY(parent);
1644 ASSIGN(parent_qid);
1645 ASSIGN(bandwidth);
1646 ASSIGN(priority);
1647 ASSIGN(local_flags);
1648
1649 ASSIGN(qlimit);
1650 ASSIGN(flags);
1651 COPY(pq_u);
1652
1653 ASSIGN(qid);
1654 break;
1655 }
1656 default:
1657 panic("%s: unhandled struct pfioc_altq version", __func__);
1658 break;
1659 }
1660
1661 #undef ASSIGN
1662 #undef COPY
1663 #undef SATU16
1664 #undef SATU32
1665
1666 return (0);
1667 }
1668
1669 /*
1670 * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1671 * that may be using any version of it.
1672 */
1673 static int
pf_import_kaltq(struct pfioc_altq_v1 * pa,struct pf_altq * q,size_t ioc_size)1674 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1675 {
1676 u_int32_t version;
1677
1678 if (ioc_size == sizeof(struct pfioc_altq_v0))
1679 version = 0;
1680 else
1681 version = pa->version;
1682
1683 if (version > PFIOC_ALTQ_VERSION)
1684 return (EINVAL);
1685
1686 #define ASSIGN(x) q->x = imported_q->x
1687 #define COPY(x) \
1688 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1689
1690 switch (version) {
1691 case 0: {
1692 struct pf_altq_v0 *imported_q =
1693 &((struct pfioc_altq_v0 *)pa)->altq;
1694
1695 COPY(ifname);
1696
1697 ASSIGN(scheduler);
1698 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1699 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1700
1701 COPY(qname);
1702 COPY(parent);
1703 ASSIGN(parent_qid);
1704 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1705 ASSIGN(priority);
1706 ASSIGN(local_flags);
1707
1708 ASSIGN(qlimit);
1709 ASSIGN(flags);
1710
1711 if (imported_q->scheduler == ALTQT_HFSC) {
1712 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1713
1714 /*
1715 * The m1 and m2 parameters are being copied from
1716 * 32-bit to 64-bit.
1717 */
1718 ASSIGN_OPT(rtsc_m1);
1719 ASSIGN_OPT(rtsc_d);
1720 ASSIGN_OPT(rtsc_m2);
1721
1722 ASSIGN_OPT(lssc_m1);
1723 ASSIGN_OPT(lssc_d);
1724 ASSIGN_OPT(lssc_m2);
1725
1726 ASSIGN_OPT(ulsc_m1);
1727 ASSIGN_OPT(ulsc_d);
1728 ASSIGN_OPT(ulsc_m2);
1729
1730 ASSIGN_OPT(flags);
1731
1732 #undef ASSIGN_OPT
1733 } else
1734 COPY(pq_u);
1735
1736 ASSIGN(qid);
1737 break;
1738 }
1739 case 1: {
1740 struct pf_altq_v1 *imported_q =
1741 &((struct pfioc_altq_v1 *)pa)->altq;
1742
1743 COPY(ifname);
1744
1745 ASSIGN(scheduler);
1746 ASSIGN(tbrsize);
1747 ASSIGN(ifbandwidth);
1748
1749 COPY(qname);
1750 COPY(parent);
1751 ASSIGN(parent_qid);
1752 ASSIGN(bandwidth);
1753 ASSIGN(priority);
1754 ASSIGN(local_flags);
1755
1756 ASSIGN(qlimit);
1757 ASSIGN(flags);
1758 COPY(pq_u);
1759
1760 ASSIGN(qid);
1761 break;
1762 }
1763 default:
1764 panic("%s: unhandled struct pfioc_altq version", __func__);
1765 break;
1766 }
1767
1768 #undef ASSIGN
1769 #undef COPY
1770
1771 return (0);
1772 }
1773
1774 static struct pf_altq *
pf_altq_get_nth_active(u_int32_t n)1775 pf_altq_get_nth_active(u_int32_t n)
1776 {
1777 struct pf_altq *altq;
1778 u_int32_t nr;
1779
1780 nr = 0;
1781 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1782 if (nr == n)
1783 return (altq);
1784 nr++;
1785 }
1786
1787 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1788 if (nr == n)
1789 return (altq);
1790 nr++;
1791 }
1792
1793 return (NULL);
1794 }
1795 #endif /* ALTQ */
1796
1797 struct pf_krule *
pf_krule_alloc(void)1798 pf_krule_alloc(void)
1799 {
1800 struct pf_krule *rule;
1801
1802 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1803 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF);
1804 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF);
1805 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF);
1806 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1807 M_WAITOK | M_ZERO);
1808 return (rule);
1809 }
1810
1811 void
pf_krule_free(struct pf_krule * rule)1812 pf_krule_free(struct pf_krule *rule)
1813 {
1814 #ifdef PF_WANT_32_TO_64_COUNTER
1815 bool wowned;
1816 #endif
1817
1818 if (rule == NULL)
1819 return;
1820
1821 #ifdef PF_WANT_32_TO_64_COUNTER
1822 if (rule->allrulelinked) {
1823 wowned = PF_RULES_WOWNED();
1824 if (!wowned)
1825 PF_RULES_WLOCK();
1826 LIST_REMOVE(rule, allrulelist);
1827 V_pf_allrulecount--;
1828 if (!wowned)
1829 PF_RULES_WUNLOCK();
1830 }
1831 #endif
1832
1833 pf_counter_u64_deinit(&rule->evaluations);
1834 for (int i = 0; i < 2; i++) {
1835 pf_counter_u64_deinit(&rule->packets[i]);
1836 pf_counter_u64_deinit(&rule->bytes[i]);
1837 }
1838 counter_u64_free(rule->states_cur);
1839 counter_u64_free(rule->states_tot);
1840 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
1841 counter_u64_free(rule->src_nodes[sn_type]);
1842 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1843
1844 mtx_destroy(&rule->nat.mtx);
1845 mtx_destroy(&rule->rdr.mtx);
1846 mtx_destroy(&rule->route.mtx);
1847 free(rule, M_PFRULE);
1848 }
1849
1850 void
pf_krule_clear_counters(struct pf_krule * rule)1851 pf_krule_clear_counters(struct pf_krule *rule)
1852 {
1853 pf_counter_u64_zero(&rule->evaluations);
1854 for (int i = 0; i < 2; i++) {
1855 pf_counter_u64_zero(&rule->packets[i]);
1856 pf_counter_u64_zero(&rule->bytes[i]);
1857 }
1858 counter_u64_zero(rule->states_tot);
1859 }
1860
1861 static void
pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr * kpool,struct pf_pooladdr * pool)1862 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1863 struct pf_pooladdr *pool)
1864 {
1865
1866 bzero(pool, sizeof(*pool));
1867 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1868 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1869 }
1870
1871 static int
pf_pooladdr_to_kpooladdr(const struct pf_pooladdr * pool,struct pf_kpooladdr * kpool)1872 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1873 struct pf_kpooladdr *kpool)
1874 {
1875 int ret;
1876
1877 bzero(kpool, sizeof(*kpool));
1878 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1879 ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1880 sizeof(kpool->ifname));
1881 return (ret);
1882 }
1883
1884 static void
pf_pool_to_kpool(const struct pf_pool * pool,struct pf_kpool * kpool)1885 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1886 {
1887 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1888 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1889
1890 bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1891 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1892
1893 kpool->tblidx = pool->tblidx;
1894 kpool->proxy_port[0] = pool->proxy_port[0];
1895 kpool->proxy_port[1] = pool->proxy_port[1];
1896 kpool->opts = pool->opts;
1897 }
1898
1899 static int
pf_rule_to_krule(const struct pf_rule * rule,struct pf_krule * krule)1900 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1901 {
1902 int ret;
1903
1904 #ifndef INET
1905 if (rule->af == AF_INET) {
1906 return (EAFNOSUPPORT);
1907 }
1908 #endif /* INET */
1909 #ifndef INET6
1910 if (rule->af == AF_INET6) {
1911 return (EAFNOSUPPORT);
1912 }
1913 #endif /* INET6 */
1914
1915 ret = pf_check_rule_addr(&rule->src);
1916 if (ret != 0)
1917 return (ret);
1918 ret = pf_check_rule_addr(&rule->dst);
1919 if (ret != 0)
1920 return (ret);
1921
1922 bcopy(&rule->src, &krule->src, sizeof(rule->src));
1923 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1924
1925 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1926 if (ret != 0)
1927 return (ret);
1928 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1929 if (ret != 0)
1930 return (ret);
1931 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1932 if (ret != 0)
1933 return (ret);
1934 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1935 if (ret != 0)
1936 return (ret);
1937 ret = pf_user_strcpy(krule->tagname, rule->tagname,
1938 sizeof(rule->tagname));
1939 if (ret != 0)
1940 return (ret);
1941 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1942 sizeof(rule->match_tagname));
1943 if (ret != 0)
1944 return (ret);
1945 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1946 sizeof(rule->overload_tblname));
1947 if (ret != 0)
1948 return (ret);
1949
1950 pf_pool_to_kpool(&rule->rpool, &krule->rdr);
1951
1952 /* Don't allow userspace to set evaluations, packets or bytes. */
1953 /* kif, anchor, overload_tbl are not copied over. */
1954
1955 krule->os_fingerprint = rule->os_fingerprint;
1956
1957 krule->rtableid = rule->rtableid;
1958 /* pf_rule->timeout is smaller than pf_krule->timeout */
1959 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1960 krule->max_states = rule->max_states;
1961 krule->max_src_nodes = rule->max_src_nodes;
1962 krule->max_src_states = rule->max_src_states;
1963 krule->max_src_conn = rule->max_src_conn;
1964 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1965 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1966 krule->qid = rule->qid;
1967 krule->pqid = rule->pqid;
1968 krule->nr = rule->nr;
1969 krule->prob = rule->prob;
1970 krule->cuid = rule->cuid;
1971 krule->cpid = rule->cpid;
1972
1973 krule->return_icmp = rule->return_icmp;
1974 krule->return_icmp6 = rule->return_icmp6;
1975 krule->max_mss = rule->max_mss;
1976 krule->tag = rule->tag;
1977 krule->match_tag = rule->match_tag;
1978 krule->scrub_flags = rule->scrub_flags;
1979
1980 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1981 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1982
1983 krule->rule_flag = rule->rule_flag;
1984 krule->action = rule->action;
1985 krule->direction = rule->direction;
1986 krule->log = rule->log;
1987 krule->logif = rule->logif;
1988 krule->quick = rule->quick;
1989 krule->ifnot = rule->ifnot;
1990 krule->match_tag_not = rule->match_tag_not;
1991 krule->natpass = rule->natpass;
1992
1993 krule->keep_state = rule->keep_state;
1994 krule->af = rule->af;
1995 krule->proto = rule->proto;
1996 krule->type = rule->type;
1997 krule->code = rule->code;
1998 krule->flags = rule->flags;
1999 krule->flagset = rule->flagset;
2000 krule->min_ttl = rule->min_ttl;
2001 krule->allow_opts = rule->allow_opts;
2002 krule->rt = rule->rt;
2003 krule->return_ttl = rule->return_ttl;
2004 krule->tos = rule->tos;
2005 krule->set_tos = rule->set_tos;
2006
2007 krule->flush = rule->flush;
2008 krule->prio = rule->prio;
2009 krule->set_prio[0] = rule->set_prio[0];
2010 krule->set_prio[1] = rule->set_prio[1];
2011
2012 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2013
2014 return (0);
2015 }
2016
2017 int
pf_ioctl_getrules(struct pfioc_rule * pr)2018 pf_ioctl_getrules(struct pfioc_rule *pr)
2019 {
2020 struct pf_kruleset *ruleset;
2021 struct pf_krule *tail;
2022 int rs_num;
2023
2024 PF_RULES_WLOCK();
2025 ruleset = pf_find_kruleset(pr->anchor);
2026 if (ruleset == NULL) {
2027 PF_RULES_WUNLOCK();
2028 return (EINVAL);
2029 }
2030 rs_num = pf_get_ruleset_number(pr->rule.action);
2031 if (rs_num >= PF_RULESET_MAX) {
2032 PF_RULES_WUNLOCK();
2033 return (EINVAL);
2034 }
2035 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2036 pf_krulequeue);
2037 if (tail)
2038 pr->nr = tail->nr + 1;
2039 else
2040 pr->nr = 0;
2041 pr->ticket = ruleset->rules[rs_num].active.ticket;
2042 PF_RULES_WUNLOCK();
2043
2044 return (0);
2045 }
2046
2047 static int
pf_rule_checkaf(struct pf_krule * r)2048 pf_rule_checkaf(struct pf_krule *r)
2049 {
2050 switch (r->af) {
2051 case 0:
2052 if (r->rule_flag & PFRULE_AFTO)
2053 return (EPFNOSUPPORT);
2054 break;
2055 case AF_INET:
2056 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET6)
2057 return (EPFNOSUPPORT);
2058 break;
2059 #ifdef INET6
2060 case AF_INET6:
2061 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET)
2062 return (EPFNOSUPPORT);
2063 break;
2064 #endif /* INET6 */
2065 default:
2066 return (EPFNOSUPPORT);
2067 }
2068
2069 if ((r->rule_flag & PFRULE_AFTO) == 0 && r->naf != 0)
2070 return (EPFNOSUPPORT);
2071
2072 return (0);
2073 }
2074
2075 static int
pf_validate_range(uint8_t op,uint16_t port[2])2076 pf_validate_range(uint8_t op, uint16_t port[2])
2077 {
2078 uint16_t a = ntohs(port[0]);
2079 uint16_t b = ntohs(port[1]);
2080
2081 if ((op == PF_OP_RRG && a > b) || /* 34:12, i.e. none */
2082 (op == PF_OP_IRG && a >= b) || /* 34><12, i.e. none */
2083 (op == PF_OP_XRG && a > b)) /* 34<>22, i.e. all */
2084 return 1;
2085 return 0;
2086 }
2087
2088 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)2089 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2090 uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2091 uid_t uid, pid_t pid)
2092 {
2093 struct pf_kruleset *ruleset;
2094 struct pf_krule *tail;
2095 struct pf_kpooladdr *pa;
2096 struct pfi_kkif *kif = NULL, *rcv_kif = NULL;
2097 int rs_num;
2098 int error = 0;
2099
2100 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
2101 #define ERROUT_UNLOCKED(x) ERROUT_FUNCTION(errout_unlocked, x)
2102
2103 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE)
2104 ERROUT_UNLOCKED(EINVAL);
2105
2106 if ((error = pf_rule_checkaf(rule)))
2107 ERROUT_UNLOCKED(error);
2108 if (pf_validate_range(rule->src.port_op, rule->src.port))
2109 ERROUT_UNLOCKED(EINVAL);
2110 if (pf_validate_range(rule->dst.port_op, rule->dst.port))
2111 ERROUT_UNLOCKED(EINVAL);
2112
2113 if (rule->ifname[0])
2114 kif = pf_kkif_create(M_WAITOK);
2115 if (rule->rcv_ifname[0])
2116 rcv_kif = pf_kkif_create(M_WAITOK);
2117 pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2118 for (int i = 0; i < 2; i++) {
2119 pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2120 pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2121 }
2122 rule->states_cur = counter_u64_alloc(M_WAITOK);
2123 rule->states_tot = counter_u64_alloc(M_WAITOK);
2124 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
2125 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
2126 rule->cuid = uid;
2127 rule->cpid = pid;
2128 TAILQ_INIT(&rule->rdr.list);
2129 TAILQ_INIT(&rule->nat.list);
2130 TAILQ_INIT(&rule->route.list);
2131
2132 PF_CONFIG_LOCK();
2133 PF_RULES_WLOCK();
2134 #ifdef PF_WANT_32_TO_64_COUNTER
2135 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2136 MPASS(!rule->allrulelinked);
2137 rule->allrulelinked = true;
2138 V_pf_allrulecount++;
2139 #endif
2140 ruleset = pf_find_kruleset(anchor);
2141 if (ruleset == NULL)
2142 ERROUT(EINVAL);
2143 rs_num = pf_get_ruleset_number(rule->action);
2144 if (rs_num >= PF_RULESET_MAX)
2145 ERROUT(EINVAL);
2146 if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2147 DPFPRINTF(PF_DEBUG_MISC,
2148 "ticket: %d != [%d]%d", ticket, rs_num,
2149 ruleset->rules[rs_num].inactive.ticket);
2150 ERROUT(EBUSY);
2151 }
2152 if (pool_ticket != V_ticket_pabuf) {
2153 DPFPRINTF(PF_DEBUG_MISC,
2154 "pool_ticket: %d != %d", pool_ticket,
2155 V_ticket_pabuf);
2156 ERROUT(EBUSY);
2157 }
2158 /*
2159 * XXXMJG hack: there is no mechanism to ensure they started the
2160 * transaction. Ticket checked above may happen to match by accident,
2161 * even if nobody called DIOCXBEGIN, let alone this process.
2162 * Partially work around it by checking if the RB tree got allocated,
2163 * see pf_begin_rules.
2164 */
2165 if (ruleset->rules[rs_num].inactive.tree == NULL) {
2166 ERROUT(EINVAL);
2167 }
2168
2169 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2170 pf_krulequeue);
2171 if (tail)
2172 rule->nr = tail->nr + 1;
2173 else
2174 rule->nr = 0;
2175 if (rule->ifname[0]) {
2176 rule->kif = pfi_kkif_attach(kif, rule->ifname);
2177 kif = NULL;
2178 pfi_kkif_ref(rule->kif);
2179 } else
2180 rule->kif = NULL;
2181
2182 if (rule->rcv_ifname[0]) {
2183 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname);
2184 rcv_kif = NULL;
2185 pfi_kkif_ref(rule->rcv_kif);
2186 } else
2187 rule->rcv_kif = NULL;
2188
2189 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2190 ERROUT(EBUSY);
2191 #ifdef ALTQ
2192 /* set queue IDs */
2193 if (rule->qname[0] != 0) {
2194 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2195 ERROUT(EBUSY);
2196 else if (rule->pqname[0] != 0) {
2197 if ((rule->pqid =
2198 pf_qname2qid(rule->pqname)) == 0)
2199 ERROUT(EBUSY);
2200 } else
2201 rule->pqid = rule->qid;
2202 }
2203 #endif
2204 if (rule->tagname[0])
2205 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2206 ERROUT(EBUSY);
2207 if (rule->match_tagname[0])
2208 if ((rule->match_tag =
2209 pf_tagname2tag(rule->match_tagname)) == 0)
2210 ERROUT(EBUSY);
2211 if (rule->rt && !rule->direction)
2212 ERROUT(EINVAL);
2213 if (!rule->log)
2214 rule->logif = 0;
2215 if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit,
2216 rule->pktrate.seconds))
2217 ERROUT(ENOMEM);
2218 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2219 ERROUT(ENOMEM);
2220 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2221 ERROUT(ENOMEM);
2222 if (pf_kanchor_setup(rule, ruleset, anchor_call))
2223 ERROUT(EINVAL);
2224 if (rule->scrub_flags & PFSTATE_SETPRIO &&
2225 (rule->set_prio[0] > PF_PRIO_MAX ||
2226 rule->set_prio[1] > PF_PRIO_MAX))
2227 ERROUT(EINVAL);
2228 for (int i = 0; i < 3; i++) {
2229 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
2230 if (pa->addr.type == PF_ADDR_TABLE) {
2231 pa->addr.p.tbl = pfr_attach_table(ruleset,
2232 pa->addr.v.tblname);
2233 if (pa->addr.p.tbl == NULL)
2234 ERROUT(ENOMEM);
2235 }
2236 }
2237
2238 rule->overload_tbl = NULL;
2239 if (rule->overload_tblname[0]) {
2240 if ((rule->overload_tbl = pfr_attach_table(ruleset,
2241 rule->overload_tblname)) == NULL)
2242 ERROUT(EINVAL);
2243 else
2244 rule->overload_tbl->pfrkt_flags |=
2245 PFR_TFLAG_ACTIVE;
2246 }
2247
2248 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list);
2249
2250 /*
2251 * Old version of pfctl provide route redirection pools in single
2252 * common redirection pool rdr. New versions use rdr only for
2253 * rdr-to rules.
2254 */
2255 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) {
2256 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list);
2257 } else {
2258 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list);
2259 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list);
2260 }
2261
2262 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2263 (rule->action == PF_BINAT)) && rule->anchor == NULL &&
2264 TAILQ_FIRST(&rule->rdr.list) == NULL) {
2265 ERROUT(EINVAL);
2266 }
2267
2268 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) {
2269 ERROUT(EINVAL);
2270 }
2271
2272 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR ||
2273 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) {
2274 ERROUT(EINVAL);
2275 }
2276
2277 MPASS(error == 0);
2278
2279 rule->nat.cur = TAILQ_FIRST(&rule->nat.list);
2280 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list);
2281 rule->route.cur = TAILQ_FIRST(&rule->route.list);
2282 rule->route.ipv6_nexthop_af = AF_INET6;
2283 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2284 rule, entries);
2285 ruleset->rules[rs_num].inactive.rcount++;
2286
2287 PF_RULES_WUNLOCK();
2288 pf_hash_rule(rule);
2289 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2290 PF_RULES_WLOCK();
2291 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2292 ruleset->rules[rs_num].inactive.rcount--;
2293 pf_free_rule(rule);
2294 rule = NULL;
2295 ERROUT(EEXIST);
2296 }
2297 PF_CONFIG_UNLOCK();
2298
2299 return (0);
2300
2301 #undef ERROUT
2302 #undef ERROUT_UNLOCKED
2303 errout:
2304 PF_RULES_WUNLOCK();
2305 PF_CONFIG_UNLOCK();
2306 errout_unlocked:
2307 pf_kkif_free(rcv_kif);
2308 pf_kkif_free(kif);
2309 pf_krule_free(rule);
2310 return (error);
2311 }
2312
2313 static bool
pf_label_match(const struct pf_krule * rule,const char * label)2314 pf_label_match(const struct pf_krule *rule, const char *label)
2315 {
2316 int i = 0;
2317
2318 while (*rule->label[i]) {
2319 if (strcmp(rule->label[i], label) == 0)
2320 return (true);
2321 i++;
2322 }
2323
2324 return (false);
2325 }
2326
2327 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)2328 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2329 {
2330 struct pf_kstate *s;
2331 int more = 0;
2332
2333 s = pf_find_state_all(key, dir, &more);
2334 if (s == NULL)
2335 return (0);
2336
2337 if (more) {
2338 PF_STATE_UNLOCK(s);
2339 return (0);
2340 }
2341
2342 pf_remove_state(s);
2343 return (1);
2344 }
2345
2346 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)2347 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2348 {
2349 struct pf_kstate *s;
2350 struct pf_state_key *sk;
2351 struct pf_addr *srcaddr, *dstaddr;
2352 struct pf_state_key_cmp match_key;
2353 int idx, killed = 0;
2354 unsigned int dir;
2355 u_int16_t srcport, dstport;
2356 struct pfi_kkif *kif;
2357
2358 relock_DIOCKILLSTATES:
2359 PF_HASHROW_LOCK(ih);
2360 LIST_FOREACH(s, &ih->states, entry) {
2361 /* For floating states look at the original kif. */
2362 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2363
2364 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2365 if (s->direction == PF_OUT) {
2366 srcaddr = &sk->addr[1];
2367 dstaddr = &sk->addr[0];
2368 srcport = sk->port[1];
2369 dstport = sk->port[0];
2370 } else {
2371 srcaddr = &sk->addr[0];
2372 dstaddr = &sk->addr[1];
2373 srcport = sk->port[0];
2374 dstport = sk->port[1];
2375 }
2376
2377 if (psk->psk_af && sk->af != psk->psk_af)
2378 continue;
2379
2380 if (psk->psk_proto && psk->psk_proto != sk->proto)
2381 continue;
2382
2383 if (! pf_match_addr(psk->psk_src.neg,
2384 &psk->psk_src.addr.v.a.addr,
2385 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2386 continue;
2387
2388 if (! pf_match_addr(psk->psk_dst.neg,
2389 &psk->psk_dst.addr.v.a.addr,
2390 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2391 continue;
2392
2393 if (! pf_match_addr(psk->psk_rt_addr.neg,
2394 &psk->psk_rt_addr.addr.v.a.addr,
2395 &psk->psk_rt_addr.addr.v.a.mask,
2396 &s->act.rt_addr, sk->af))
2397 continue;
2398
2399 if (psk->psk_src.port_op != 0 &&
2400 ! pf_match_port(psk->psk_src.port_op,
2401 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2402 continue;
2403
2404 if (psk->psk_dst.port_op != 0 &&
2405 ! pf_match_port(psk->psk_dst.port_op,
2406 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2407 continue;
2408
2409 if (psk->psk_label[0] &&
2410 ! pf_label_match(s->rule, psk->psk_label))
2411 continue;
2412
2413 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2414 kif->pfik_name))
2415 continue;
2416
2417 if (psk->psk_kill_match) {
2418 /* Create the key to find matching states, with lock
2419 * held. */
2420
2421 bzero(&match_key, sizeof(match_key));
2422
2423 if (s->direction == PF_OUT) {
2424 dir = PF_IN;
2425 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2426 } else {
2427 dir = PF_OUT;
2428 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2429 }
2430
2431 match_key.af = s->key[idx]->af;
2432 match_key.proto = s->key[idx]->proto;
2433 pf_addrcpy(&match_key.addr[0],
2434 &s->key[idx]->addr[1], match_key.af);
2435 match_key.port[0] = s->key[idx]->port[1];
2436 pf_addrcpy(&match_key.addr[1],
2437 &s->key[idx]->addr[0], match_key.af);
2438 match_key.port[1] = s->key[idx]->port[0];
2439 }
2440
2441 pf_remove_state(s);
2442 killed++;
2443
2444 if (psk->psk_kill_match)
2445 killed += pf_kill_matching_state(&match_key, dir);
2446
2447 goto relock_DIOCKILLSTATES;
2448 }
2449 PF_HASHROW_UNLOCK(ih);
2450
2451 return (killed);
2452 }
2453
2454 void
unhandled_af(int af)2455 unhandled_af(int af)
2456 {
2457 panic("unhandled af %d", af);
2458 }
2459
2460 int
pf_start(void)2461 pf_start(void)
2462 {
2463 int error = 0;
2464
2465 sx_xlock(&V_pf_ioctl_lock);
2466 if (V_pf_status.running)
2467 error = EEXIST;
2468 else {
2469 hook_pf();
2470 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2471 hook_pf_eth();
2472 V_pf_status.running = 1;
2473 V_pf_status.since = time_uptime;
2474 new_unrhdr64(&V_pf_stateid, time_second);
2475
2476 DPFPRINTF(PF_DEBUG_MISC, "pf: started");
2477 }
2478 sx_xunlock(&V_pf_ioctl_lock);
2479
2480 return (error);
2481 }
2482
2483 int
pf_stop(void)2484 pf_stop(void)
2485 {
2486 int error = 0;
2487
2488 sx_xlock(&V_pf_ioctl_lock);
2489 if (!V_pf_status.running)
2490 error = ENOENT;
2491 else {
2492 V_pf_status.running = 0;
2493 dehook_pf();
2494 dehook_pf_eth();
2495 V_pf_status.since = time_uptime;
2496 DPFPRINTF(PF_DEBUG_MISC, "pf: stopped");
2497 }
2498 sx_xunlock(&V_pf_ioctl_lock);
2499
2500 return (error);
2501 }
2502
2503 void
pf_ioctl_clear_status(void)2504 pf_ioctl_clear_status(void)
2505 {
2506 PF_RULES_WLOCK();
2507 for (int i = 0; i < PFRES_MAX; i++)
2508 counter_u64_zero(V_pf_status.counters[i]);
2509 for (int i = 0; i < FCNT_MAX; i++)
2510 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2511 for (int i = 0; i < SCNT_MAX; i++)
2512 counter_u64_zero(V_pf_status.scounters[i]);
2513 for (int i = 0; i < NCNT_MAX; i++)
2514 counter_u64_zero(V_pf_status.ncounters[i]);
2515 for (int i = 0; i < KLCNT_MAX; i++)
2516 counter_u64_zero(V_pf_status.lcounters[i]);
2517 V_pf_status.since = time_uptime;
2518 if (*V_pf_status.ifname)
2519 pfi_update_status(V_pf_status.ifname, NULL);
2520 PF_RULES_WUNLOCK();
2521 }
2522
2523 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2524 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2525 {
2526 uint32_t old;
2527
2528 if (timeout < 0 || timeout >= PFTM_MAX ||
2529 seconds < 0)
2530 return (EINVAL);
2531
2532 PF_RULES_WLOCK();
2533 old = V_pf_default_rule.timeout[timeout];
2534 if (timeout == PFTM_INTERVAL && seconds == 0)
2535 seconds = 1;
2536 V_pf_default_rule.timeout[timeout] = seconds;
2537 if (timeout == PFTM_INTERVAL && seconds < old)
2538 wakeup(pf_purge_thread);
2539
2540 if (prev_seconds != NULL)
2541 *prev_seconds = old;
2542
2543 PF_RULES_WUNLOCK();
2544
2545 return (0);
2546 }
2547
2548 int
pf_ioctl_get_timeout(int timeout,int * seconds)2549 pf_ioctl_get_timeout(int timeout, int *seconds)
2550 {
2551 PF_RULES_RLOCK_TRACKER;
2552
2553 if (timeout < 0 || timeout >= PFTM_MAX)
2554 return (EINVAL);
2555
2556 PF_RULES_RLOCK();
2557 *seconds = V_pf_default_rule.timeout[timeout];
2558 PF_RULES_RUNLOCK();
2559
2560 return (0);
2561 }
2562
2563 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2564 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2565 {
2566
2567 PF_RULES_WLOCK();
2568 if (index < 0 || index >= PF_LIMIT_MAX ||
2569 V_pf_limits[index].zone == NULL) {
2570 PF_RULES_WUNLOCK();
2571 return (EINVAL);
2572 }
2573 uma_zone_set_max(V_pf_limits[index].zone,
2574 limit == 0 ? INT_MAX : limit);
2575 if (old_limit != NULL)
2576 *old_limit = V_pf_limits[index].limit;
2577 V_pf_limits[index].limit = limit;
2578 PF_RULES_WUNLOCK();
2579
2580 return (0);
2581 }
2582
2583 int
pf_ioctl_get_limit(int index,unsigned int * limit)2584 pf_ioctl_get_limit(int index, unsigned int *limit)
2585 {
2586 PF_RULES_RLOCK_TRACKER;
2587
2588 if (index < 0 || index >= PF_LIMIT_MAX)
2589 return (EINVAL);
2590
2591 PF_RULES_RLOCK();
2592 *limit = V_pf_limits[index].limit;
2593 PF_RULES_RUNLOCK();
2594
2595 return (0);
2596 }
2597
2598 int
pf_ioctl_begin_addrs(uint32_t * ticket)2599 pf_ioctl_begin_addrs(uint32_t *ticket)
2600 {
2601 PF_RULES_WLOCK();
2602 pf_empty_kpool(&V_pf_pabuf[0]);
2603 pf_empty_kpool(&V_pf_pabuf[1]);
2604 pf_empty_kpool(&V_pf_pabuf[2]);
2605 *ticket = ++V_ticket_pabuf;
2606 PF_RULES_WUNLOCK();
2607
2608 return (0);
2609 }
2610
2611 int
pf_ioctl_add_addr(struct pf_nl_pooladdr * pp)2612 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
2613 {
2614 struct pf_kpooladdr *pa = NULL;
2615 struct pfi_kkif *kif = NULL;
2616 int error;
2617
2618 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2619 pp->which != PF_RT)
2620 return (EINVAL);
2621
2622 switch (pp->af) {
2623 #ifdef INET
2624 case AF_INET:
2625 /* FALLTHROUGH */
2626 #endif /* INET */
2627 #ifdef INET6
2628 case AF_INET6:
2629 /* FALLTHROUGH */
2630 #endif /* INET6 */
2631 case AF_UNSPEC:
2632 break;
2633 default:
2634 return (EAFNOSUPPORT);
2635 }
2636
2637 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2638 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2639 pp->addr.addr.type != PF_ADDR_TABLE)
2640 return (EINVAL);
2641
2642 if (pp->addr.addr.p.dyn != NULL)
2643 return (EINVAL);
2644
2645 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2646 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2647 if (error != 0)
2648 goto out;
2649 if (pa->ifname[0])
2650 kif = pf_kkif_create(M_WAITOK);
2651 PF_RULES_WLOCK();
2652 if (pp->ticket != V_ticket_pabuf) {
2653 PF_RULES_WUNLOCK();
2654 if (pa->ifname[0])
2655 pf_kkif_free(kif);
2656 error = EBUSY;
2657 goto out;
2658 }
2659 if (pa->ifname[0]) {
2660 pa->kif = pfi_kkif_attach(kif, pa->ifname);
2661 kif = NULL;
2662 pfi_kkif_ref(pa->kif);
2663 } else
2664 pa->kif = NULL;
2665 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2666 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2667 if (pa->ifname[0])
2668 pfi_kkif_unref(pa->kif);
2669 PF_RULES_WUNLOCK();
2670 goto out;
2671 }
2672 pa->af = pp->af;
2673 switch (pp->which) {
2674 case PF_NAT:
2675 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries);
2676 break;
2677 case PF_RDR:
2678 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries);
2679 break;
2680 case PF_RT:
2681 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries);
2682 break;
2683 }
2684 PF_RULES_WUNLOCK();
2685
2686 return (0);
2687
2688 out:
2689 free(pa, M_PFRULE);
2690 return (error);
2691 }
2692
2693 int
pf_ioctl_get_addrs(struct pf_nl_pooladdr * pp)2694 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
2695 {
2696 struct pf_kpool *pool;
2697 struct pf_kpooladdr *pa;
2698
2699 PF_RULES_RLOCK_TRACKER;
2700
2701 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2702 pp->which != PF_RT)
2703 return (EINVAL);
2704
2705 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2706 pp->nr = 0;
2707
2708 PF_RULES_RLOCK();
2709 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2710 pp->r_num, 0, 1, 0, pp->which);
2711 if (pool == NULL) {
2712 PF_RULES_RUNLOCK();
2713 return (EBUSY);
2714 }
2715 TAILQ_FOREACH(pa, &pool->list, entries)
2716 pp->nr++;
2717 PF_RULES_RUNLOCK();
2718
2719 return (0);
2720 }
2721
2722 int
pf_ioctl_get_addr(struct pf_nl_pooladdr * pp)2723 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
2724 {
2725 struct pf_kpool *pool;
2726 struct pf_kpooladdr *pa;
2727 u_int32_t nr = 0;
2728
2729 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2730 pp->which != PF_RT)
2731 return (EINVAL);
2732
2733 PF_RULES_RLOCK_TRACKER;
2734
2735 pp->anchor[sizeof(pp->anchor) - 1] = '\0';
2736
2737 PF_RULES_RLOCK();
2738 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2739 pp->r_num, 0, 1, 1, pp->which);
2740 if (pool == NULL) {
2741 PF_RULES_RUNLOCK();
2742 return (EBUSY);
2743 }
2744 pa = TAILQ_FIRST(&pool->list);
2745 while ((pa != NULL) && (nr < pp->nr)) {
2746 pa = TAILQ_NEXT(pa, entries);
2747 nr++;
2748 }
2749 if (pa == NULL) {
2750 PF_RULES_RUNLOCK();
2751 return (EBUSY);
2752 }
2753 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2754 pp->af = pa->af;
2755 pf_addr_copyout(&pp->addr.addr);
2756 PF_RULES_RUNLOCK();
2757
2758 return (0);
2759 }
2760
2761 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)2762 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
2763 {
2764 struct pf_kruleset *ruleset;
2765 struct pf_kanchor *anchor;
2766
2767 PF_RULES_RLOCK_TRACKER;
2768
2769 pr->path[sizeof(pr->path) - 1] = '\0';
2770
2771 PF_RULES_RLOCK();
2772 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2773 PF_RULES_RUNLOCK();
2774 return (ENOENT);
2775 }
2776 pr->nr = 0;
2777 if (ruleset == &pf_main_ruleset) {
2778 /* XXX kludge for pf_main_ruleset */
2779 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2780 if (anchor->parent == NULL)
2781 pr->nr++;
2782 } else {
2783 RB_FOREACH(anchor, pf_kanchor_node,
2784 &ruleset->anchor->children)
2785 pr->nr++;
2786 }
2787 PF_RULES_RUNLOCK();
2788
2789 return (0);
2790 }
2791
2792 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)2793 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
2794 {
2795 struct pf_kruleset *ruleset;
2796 struct pf_kanchor *anchor;
2797 u_int32_t nr = 0;
2798 int error = 0;
2799
2800 PF_RULES_RLOCK_TRACKER;
2801
2802 PF_RULES_RLOCK();
2803 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2804 PF_RULES_RUNLOCK();
2805 return (ENOENT);
2806 }
2807
2808 pr->name[0] = '\0';
2809 if (ruleset == &pf_main_ruleset) {
2810 /* XXX kludge for pf_main_ruleset */
2811 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2812 if (anchor->parent == NULL && nr++ == pr->nr) {
2813 strlcpy(pr->name, anchor->name,
2814 sizeof(pr->name));
2815 break;
2816 }
2817 } else {
2818 RB_FOREACH(anchor, pf_kanchor_node,
2819 &ruleset->anchor->children)
2820 if (nr++ == pr->nr) {
2821 strlcpy(pr->name, anchor->name,
2822 sizeof(pr->name));
2823 break;
2824 }
2825 }
2826 if (!pr->name[0])
2827 error = EBUSY;
2828 PF_RULES_RUNLOCK();
2829
2830 return (error);
2831 }
2832
2833 int
pf_ioctl_natlook(struct pfioc_natlook * pnl)2834 pf_ioctl_natlook(struct pfioc_natlook *pnl)
2835 {
2836 struct pf_state_key *sk;
2837 struct pf_kstate *state;
2838 struct pf_state_key_cmp key;
2839 int m = 0, direction = pnl->direction;
2840 int sidx, didx;
2841
2842 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
2843 sidx = (direction == PF_IN) ? 1 : 0;
2844 didx = (direction == PF_IN) ? 0 : 1;
2845
2846 if (!pnl->proto ||
2847 PF_AZERO(&pnl->saddr, pnl->af) ||
2848 PF_AZERO(&pnl->daddr, pnl->af) ||
2849 ((pnl->proto == IPPROTO_TCP ||
2850 pnl->proto == IPPROTO_UDP) &&
2851 (!pnl->dport || !pnl->sport)))
2852 return (EINVAL);
2853
2854 switch (pnl->direction) {
2855 case PF_IN:
2856 case PF_OUT:
2857 case PF_INOUT:
2858 break;
2859 default:
2860 return (EINVAL);
2861 }
2862
2863 switch (pnl->af) {
2864 #ifdef INET
2865 case AF_INET:
2866 break;
2867 #endif /* INET */
2868 #ifdef INET6
2869 case AF_INET6:
2870 break;
2871 #endif /* INET6 */
2872 default:
2873 return (EAFNOSUPPORT);
2874 }
2875
2876 bzero(&key, sizeof(key));
2877 key.af = pnl->af;
2878 key.proto = pnl->proto;
2879 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af);
2880 key.port[sidx] = pnl->sport;
2881 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af);
2882 key.port[didx] = pnl->dport;
2883
2884 state = pf_find_state_all(&key, direction, &m);
2885 if (state == NULL)
2886 return (ENOENT);
2887
2888 if (m > 1) {
2889 PF_STATE_UNLOCK(state);
2890 return (E2BIG); /* more than one state */
2891 }
2892
2893 sk = state->key[sidx];
2894 pf_addrcpy(&pnl->rsaddr,
2895 &sk->addr[sidx], sk->af);
2896 pnl->rsport = sk->port[sidx];
2897 pf_addrcpy(&pnl->rdaddr,
2898 &sk->addr[didx], sk->af);
2899 pnl->rdport = sk->port[didx];
2900 PF_STATE_UNLOCK(state);
2901
2902 return (0);
2903 }
2904
2905 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2906 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2907 {
2908 int error = 0;
2909 PF_RULES_RLOCK_TRACKER;
2910
2911 #define ERROUT_IOCTL(target, x) \
2912 do { \
2913 error = (x); \
2914 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
2915 goto target; \
2916 } while (0)
2917
2918
2919 /* XXX keep in sync with switch() below */
2920 if (securelevel_gt(td->td_ucred, 2))
2921 switch (cmd) {
2922 case DIOCGETRULES:
2923 case DIOCGETRULENV:
2924 case DIOCGETADDRS:
2925 case DIOCGETADDR:
2926 case DIOCGETSTATE:
2927 case DIOCGETSTATENV:
2928 case DIOCSETSTATUSIF:
2929 case DIOCGETSTATUSNV:
2930 case DIOCCLRSTATUS:
2931 case DIOCNATLOOK:
2932 case DIOCSETDEBUG:
2933 #ifdef COMPAT_FREEBSD14
2934 case DIOCGETSTATES:
2935 case DIOCGETSTATESV2:
2936 #endif
2937 case DIOCGETTIMEOUT:
2938 case DIOCCLRRULECTRS:
2939 case DIOCGETLIMIT:
2940 case DIOCGETALTQSV0:
2941 case DIOCGETALTQSV1:
2942 case DIOCGETALTQV0:
2943 case DIOCGETALTQV1:
2944 case DIOCGETQSTATSV0:
2945 case DIOCGETQSTATSV1:
2946 case DIOCGETRULESETS:
2947 case DIOCGETRULESET:
2948 case DIOCRGETTABLES:
2949 case DIOCRGETTSTATS:
2950 case DIOCRCLRTSTATS:
2951 case DIOCRCLRADDRS:
2952 case DIOCRADDADDRS:
2953 case DIOCRDELADDRS:
2954 case DIOCRSETADDRS:
2955 case DIOCRGETADDRS:
2956 case DIOCRGETASTATS:
2957 case DIOCRCLRASTATS:
2958 case DIOCRTSTADDRS:
2959 case DIOCOSFPGET:
2960 case DIOCGETSRCNODES:
2961 case DIOCCLRSRCNODES:
2962 case DIOCGETSYNCOOKIES:
2963 case DIOCIGETIFACES:
2964 case DIOCGIFSPEEDV0:
2965 case DIOCGIFSPEEDV1:
2966 case DIOCSETIFFLAG:
2967 case DIOCCLRIFFLAG:
2968 case DIOCGETETHRULES:
2969 case DIOCGETETHRULE:
2970 case DIOCGETETHRULESETS:
2971 case DIOCGETETHRULESET:
2972 break;
2973 case DIOCRCLRTABLES:
2974 case DIOCRADDTABLES:
2975 case DIOCRDELTABLES:
2976 case DIOCRSETTFLAGS:
2977 if (((struct pfioc_table *)addr)->pfrio_flags &
2978 PFR_FLAG_DUMMY)
2979 break; /* dummy operation ok */
2980 return (EPERM);
2981 default:
2982 return (EPERM);
2983 }
2984
2985 if (!(flags & FWRITE))
2986 switch (cmd) {
2987 case DIOCGETRULES:
2988 case DIOCGETADDRS:
2989 case DIOCGETADDR:
2990 case DIOCGETSTATE:
2991 case DIOCGETSTATENV:
2992 case DIOCGETSTATUSNV:
2993 #ifdef COMPAT_FREEBSD14
2994 case DIOCGETSTATES:
2995 case DIOCGETSTATESV2:
2996 #endif
2997 case DIOCGETTIMEOUT:
2998 case DIOCGETLIMIT:
2999 case DIOCGETALTQSV0:
3000 case DIOCGETALTQSV1:
3001 case DIOCGETALTQV0:
3002 case DIOCGETALTQV1:
3003 case DIOCGETQSTATSV0:
3004 case DIOCGETQSTATSV1:
3005 case DIOCGETRULESETS:
3006 case DIOCGETRULESET:
3007 case DIOCNATLOOK:
3008 case DIOCRGETTABLES:
3009 case DIOCRGETTSTATS:
3010 case DIOCRGETADDRS:
3011 case DIOCRGETASTATS:
3012 case DIOCRTSTADDRS:
3013 case DIOCOSFPGET:
3014 case DIOCGETSRCNODES:
3015 case DIOCGETSYNCOOKIES:
3016 case DIOCIGETIFACES:
3017 case DIOCGIFSPEEDV1:
3018 case DIOCGIFSPEEDV0:
3019 case DIOCGETRULENV:
3020 case DIOCGETETHRULES:
3021 case DIOCGETETHRULE:
3022 case DIOCGETETHRULESETS:
3023 case DIOCGETETHRULESET:
3024 break;
3025 case DIOCRCLRTABLES:
3026 case DIOCRADDTABLES:
3027 case DIOCRDELTABLES:
3028 case DIOCRCLRTSTATS:
3029 case DIOCRCLRADDRS:
3030 case DIOCRADDADDRS:
3031 case DIOCRDELADDRS:
3032 case DIOCRSETADDRS:
3033 case DIOCRSETTFLAGS:
3034 if (((struct pfioc_table *)addr)->pfrio_flags &
3035 PFR_FLAG_DUMMY) {
3036 flags |= FWRITE; /* need write lock for dummy */
3037 break; /* dummy operation ok */
3038 }
3039 return (EACCES);
3040 default:
3041 return (EACCES);
3042 }
3043
3044 CURVNET_SET(TD_TO_VNET(td));
3045
3046 switch (cmd) {
3047 #ifdef COMPAT_FREEBSD14
3048 case DIOCSTART:
3049 error = pf_start();
3050 break;
3051
3052 case DIOCSTOP:
3053 error = pf_stop();
3054 break;
3055 #endif
3056
3057 case DIOCGETETHRULES: {
3058 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3059 nvlist_t *nvl;
3060 void *packed;
3061 struct pf_keth_rule *tail;
3062 struct pf_keth_ruleset *rs;
3063 u_int32_t ticket, nr;
3064 const char *anchor = "";
3065
3066 nvl = NULL;
3067 packed = NULL;
3068
3069 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
3070
3071 if (nv->len > pf_ioctl_maxcount)
3072 ERROUT(ENOMEM);
3073
3074 /* Copy the request in */
3075 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
3076 error = copyin(nv->data, packed, nv->len);
3077 if (error)
3078 ERROUT(error);
3079
3080 nvl = nvlist_unpack(packed, nv->len, 0);
3081 if (nvl == NULL)
3082 ERROUT(EBADMSG);
3083
3084 if (! nvlist_exists_string(nvl, "anchor"))
3085 ERROUT(EBADMSG);
3086
3087 anchor = nvlist_get_string(nvl, "anchor");
3088
3089 rs = pf_find_keth_ruleset(anchor);
3090
3091 nvlist_destroy(nvl);
3092 nvl = NULL;
3093 free(packed, M_NVLIST);
3094 packed = NULL;
3095
3096 if (rs == NULL)
3097 ERROUT(ENOENT);
3098
3099 /* Reply */
3100 nvl = nvlist_create(0);
3101 if (nvl == NULL)
3102 ERROUT(ENOMEM);
3103
3104 PF_RULES_RLOCK();
3105
3106 ticket = rs->active.ticket;
3107 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
3108 if (tail)
3109 nr = tail->nr + 1;
3110 else
3111 nr = 0;
3112
3113 PF_RULES_RUNLOCK();
3114
3115 nvlist_add_number(nvl, "ticket", ticket);
3116 nvlist_add_number(nvl, "nr", nr);
3117
3118 packed = nvlist_pack(nvl, &nv->len);
3119 if (packed == NULL)
3120 ERROUT(ENOMEM);
3121
3122 if (nv->size == 0)
3123 ERROUT(0);
3124 else if (nv->size < nv->len)
3125 ERROUT(ENOSPC);
3126
3127 error = copyout(packed, nv->data, nv->len);
3128
3129 #undef ERROUT
3130 DIOCGETETHRULES_error:
3131 free(packed, M_NVLIST);
3132 nvlist_destroy(nvl);
3133 break;
3134 }
3135
3136 case DIOCGETETHRULE: {
3137 struct epoch_tracker et;
3138 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3139 nvlist_t *nvl = NULL;
3140 void *nvlpacked = NULL;
3141 struct pf_keth_rule *rule = NULL;
3142 struct pf_keth_ruleset *rs;
3143 u_int32_t ticket, nr;
3144 bool clear = false;
3145 const char *anchor;
3146
3147 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
3148
3149 if (nv->len > pf_ioctl_maxcount)
3150 ERROUT(ENOMEM);
3151
3152 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3153 error = copyin(nv->data, nvlpacked, nv->len);
3154 if (error)
3155 ERROUT(error);
3156
3157 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3158 if (nvl == NULL)
3159 ERROUT(EBADMSG);
3160 if (! nvlist_exists_number(nvl, "ticket"))
3161 ERROUT(EBADMSG);
3162 ticket = nvlist_get_number(nvl, "ticket");
3163 if (! nvlist_exists_string(nvl, "anchor"))
3164 ERROUT(EBADMSG);
3165 anchor = nvlist_get_string(nvl, "anchor");
3166
3167 if (nvlist_exists_bool(nvl, "clear"))
3168 clear = nvlist_get_bool(nvl, "clear");
3169
3170 if (clear && !(flags & FWRITE))
3171 ERROUT(EACCES);
3172
3173 if (! nvlist_exists_number(nvl, "nr"))
3174 ERROUT(EBADMSG);
3175 nr = nvlist_get_number(nvl, "nr");
3176
3177 PF_RULES_RLOCK();
3178 rs = pf_find_keth_ruleset(anchor);
3179 if (rs == NULL) {
3180 PF_RULES_RUNLOCK();
3181 ERROUT(ENOENT);
3182 }
3183 if (ticket != rs->active.ticket) {
3184 PF_RULES_RUNLOCK();
3185 ERROUT(EBUSY);
3186 }
3187
3188 nvlist_destroy(nvl);
3189 nvl = NULL;
3190 free(nvlpacked, M_NVLIST);
3191 nvlpacked = NULL;
3192
3193 rule = TAILQ_FIRST(rs->active.rules);
3194 while ((rule != NULL) && (rule->nr != nr))
3195 rule = TAILQ_NEXT(rule, entries);
3196 if (rule == NULL) {
3197 PF_RULES_RUNLOCK();
3198 ERROUT(ENOENT);
3199 }
3200 /* Make sure rule can't go away. */
3201 NET_EPOCH_ENTER(et);
3202 PF_RULES_RUNLOCK();
3203 nvl = pf_keth_rule_to_nveth_rule(rule);
3204 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
3205 NET_EPOCH_EXIT(et);
3206 ERROUT(EBUSY);
3207 }
3208 NET_EPOCH_EXIT(et);
3209 if (nvl == NULL)
3210 ERROUT(ENOMEM);
3211
3212 nvlpacked = nvlist_pack(nvl, &nv->len);
3213 if (nvlpacked == NULL)
3214 ERROUT(ENOMEM);
3215
3216 if (nv->size == 0)
3217 ERROUT(0);
3218 else if (nv->size < nv->len)
3219 ERROUT(ENOSPC);
3220
3221 error = copyout(nvlpacked, nv->data, nv->len);
3222 if (error == 0 && clear) {
3223 counter_u64_zero(rule->evaluations);
3224 for (int i = 0; i < 2; i++) {
3225 counter_u64_zero(rule->packets[i]);
3226 counter_u64_zero(rule->bytes[i]);
3227 }
3228 }
3229
3230 #undef ERROUT
3231 DIOCGETETHRULE_error:
3232 free(nvlpacked, M_NVLIST);
3233 nvlist_destroy(nvl);
3234 break;
3235 }
3236
3237 case DIOCADDETHRULE: {
3238 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3239 nvlist_t *nvl = NULL;
3240 void *nvlpacked = NULL;
3241 struct pf_keth_rule *rule = NULL, *tail = NULL;
3242 struct pf_keth_ruleset *ruleset = NULL;
3243 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
3244 const char *anchor = "", *anchor_call = "";
3245
3246 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3247
3248 if (nv->len > pf_ioctl_maxcount)
3249 ERROUT(ENOMEM);
3250
3251 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3252 error = copyin(nv->data, nvlpacked, nv->len);
3253 if (error)
3254 ERROUT(error);
3255
3256 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3257 if (nvl == NULL)
3258 ERROUT(EBADMSG);
3259
3260 if (! nvlist_exists_number(nvl, "ticket"))
3261 ERROUT(EBADMSG);
3262
3263 if (nvlist_exists_string(nvl, "anchor"))
3264 anchor = nvlist_get_string(nvl, "anchor");
3265 if (nvlist_exists_string(nvl, "anchor_call"))
3266 anchor_call = nvlist_get_string(nvl, "anchor_call");
3267
3268 ruleset = pf_find_keth_ruleset(anchor);
3269 if (ruleset == NULL)
3270 ERROUT(EINVAL);
3271
3272 if (nvlist_get_number(nvl, "ticket") !=
3273 ruleset->inactive.ticket) {
3274 DPFPRINTF(PF_DEBUG_MISC,
3275 "ticket: %d != %d",
3276 (u_int32_t)nvlist_get_number(nvl, "ticket"),
3277 ruleset->inactive.ticket);
3278 ERROUT(EBUSY);
3279 }
3280
3281 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3282 rule->timestamp = NULL;
3283
3284 error = pf_nveth_rule_to_keth_rule(nvl, rule);
3285 if (error != 0)
3286 ERROUT(error);
3287
3288 if (rule->ifname[0])
3289 kif = pf_kkif_create(M_WAITOK);
3290 if (rule->bridge_to_name[0])
3291 bridge_to_kif = pf_kkif_create(M_WAITOK);
3292 rule->evaluations = counter_u64_alloc(M_WAITOK);
3293 for (int i = 0; i < 2; i++) {
3294 rule->packets[i] = counter_u64_alloc(M_WAITOK);
3295 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3296 }
3297 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3298 M_WAITOK | M_ZERO);
3299
3300 PF_RULES_WLOCK();
3301
3302 if (rule->ifname[0]) {
3303 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3304 pfi_kkif_ref(rule->kif);
3305 } else
3306 rule->kif = NULL;
3307 if (rule->bridge_to_name[0]) {
3308 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3309 rule->bridge_to_name);
3310 pfi_kkif_ref(rule->bridge_to);
3311 } else
3312 rule->bridge_to = NULL;
3313
3314 #ifdef ALTQ
3315 /* set queue IDs */
3316 if (rule->qname[0] != 0) {
3317 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3318 error = EBUSY;
3319 else
3320 rule->qid = rule->qid;
3321 }
3322 #endif
3323 if (rule->tagname[0])
3324 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3325 error = EBUSY;
3326 if (rule->match_tagname[0])
3327 if ((rule->match_tag = pf_tagname2tag(
3328 rule->match_tagname)) == 0)
3329 error = EBUSY;
3330
3331 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3332 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3333 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3334 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3335
3336 if (error) {
3337 pf_free_eth_rule(rule);
3338 PF_RULES_WUNLOCK();
3339 ERROUT(error);
3340 }
3341
3342 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3343 pf_free_eth_rule(rule);
3344 PF_RULES_WUNLOCK();
3345 ERROUT(EINVAL);
3346 }
3347
3348 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3349 if (tail)
3350 rule->nr = tail->nr + 1;
3351 else
3352 rule->nr = 0;
3353
3354 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3355
3356 PF_RULES_WUNLOCK();
3357
3358 #undef ERROUT
3359 DIOCADDETHRULE_error:
3360 nvlist_destroy(nvl);
3361 free(nvlpacked, M_NVLIST);
3362 break;
3363 }
3364
3365 case DIOCGETETHRULESETS: {
3366 struct epoch_tracker et;
3367 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3368 nvlist_t *nvl = NULL;
3369 void *nvlpacked = NULL;
3370 struct pf_keth_ruleset *ruleset;
3371 struct pf_keth_anchor *anchor;
3372 int nr = 0;
3373
3374 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3375
3376 if (nv->len > pf_ioctl_maxcount)
3377 ERROUT(ENOMEM);
3378
3379 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3380 error = copyin(nv->data, nvlpacked, nv->len);
3381 if (error)
3382 ERROUT(error);
3383
3384 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3385 if (nvl == NULL)
3386 ERROUT(EBADMSG);
3387 if (! nvlist_exists_string(nvl, "path"))
3388 ERROUT(EBADMSG);
3389
3390 NET_EPOCH_ENTER(et);
3391
3392 if ((ruleset = pf_find_keth_ruleset(
3393 nvlist_get_string(nvl, "path"))) == NULL) {
3394 NET_EPOCH_EXIT(et);
3395 ERROUT(ENOENT);
3396 }
3397
3398 if (ruleset->anchor == NULL) {
3399 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3400 if (anchor->parent == NULL)
3401 nr++;
3402 } else {
3403 RB_FOREACH(anchor, pf_keth_anchor_node,
3404 &ruleset->anchor->children)
3405 nr++;
3406 }
3407
3408 NET_EPOCH_EXIT(et);
3409
3410 nvlist_destroy(nvl);
3411 nvl = NULL;
3412 free(nvlpacked, M_NVLIST);
3413 nvlpacked = NULL;
3414
3415 nvl = nvlist_create(0);
3416 if (nvl == NULL)
3417 ERROUT(ENOMEM);
3418
3419 nvlist_add_number(nvl, "nr", nr);
3420
3421 nvlpacked = nvlist_pack(nvl, &nv->len);
3422 if (nvlpacked == NULL)
3423 ERROUT(ENOMEM);
3424
3425 if (nv->size == 0)
3426 ERROUT(0);
3427 else if (nv->size < nv->len)
3428 ERROUT(ENOSPC);
3429
3430 error = copyout(nvlpacked, nv->data, nv->len);
3431
3432 #undef ERROUT
3433 DIOCGETETHRULESETS_error:
3434 free(nvlpacked, M_NVLIST);
3435 nvlist_destroy(nvl);
3436 break;
3437 }
3438
3439 case DIOCGETETHRULESET: {
3440 struct epoch_tracker et;
3441 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3442 nvlist_t *nvl = NULL;
3443 void *nvlpacked = NULL;
3444 struct pf_keth_ruleset *ruleset;
3445 struct pf_keth_anchor *anchor;
3446 int nr = 0, req_nr = 0;
3447 bool found = false;
3448
3449 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3450
3451 if (nv->len > pf_ioctl_maxcount)
3452 ERROUT(ENOMEM);
3453
3454 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3455 error = copyin(nv->data, nvlpacked, nv->len);
3456 if (error)
3457 ERROUT(error);
3458
3459 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3460 if (nvl == NULL)
3461 ERROUT(EBADMSG);
3462 if (! nvlist_exists_string(nvl, "path"))
3463 ERROUT(EBADMSG);
3464 if (! nvlist_exists_number(nvl, "nr"))
3465 ERROUT(EBADMSG);
3466
3467 req_nr = nvlist_get_number(nvl, "nr");
3468
3469 NET_EPOCH_ENTER(et);
3470
3471 if ((ruleset = pf_find_keth_ruleset(
3472 nvlist_get_string(nvl, "path"))) == NULL) {
3473 NET_EPOCH_EXIT(et);
3474 ERROUT(ENOENT);
3475 }
3476
3477 nvlist_destroy(nvl);
3478 nvl = NULL;
3479 free(nvlpacked, M_NVLIST);
3480 nvlpacked = NULL;
3481
3482 nvl = nvlist_create(0);
3483 if (nvl == NULL) {
3484 NET_EPOCH_EXIT(et);
3485 ERROUT(ENOMEM);
3486 }
3487
3488 if (ruleset->anchor == NULL) {
3489 RB_FOREACH(anchor, pf_keth_anchor_global,
3490 &V_pf_keth_anchors) {
3491 if (anchor->parent == NULL && nr++ == req_nr) {
3492 found = true;
3493 break;
3494 }
3495 }
3496 } else {
3497 RB_FOREACH(anchor, pf_keth_anchor_node,
3498 &ruleset->anchor->children) {
3499 if (nr++ == req_nr) {
3500 found = true;
3501 break;
3502 }
3503 }
3504 }
3505
3506 NET_EPOCH_EXIT(et);
3507 if (found) {
3508 nvlist_add_number(nvl, "nr", nr);
3509 nvlist_add_string(nvl, "name", anchor->name);
3510 if (ruleset->anchor)
3511 nvlist_add_string(nvl, "path",
3512 ruleset->anchor->path);
3513 else
3514 nvlist_add_string(nvl, "path", "");
3515 } else {
3516 ERROUT(EBUSY);
3517 }
3518
3519 nvlpacked = nvlist_pack(nvl, &nv->len);
3520 if (nvlpacked == NULL)
3521 ERROUT(ENOMEM);
3522
3523 if (nv->size == 0)
3524 ERROUT(0);
3525 else if (nv->size < nv->len)
3526 ERROUT(ENOSPC);
3527
3528 error = copyout(nvlpacked, nv->data, nv->len);
3529
3530 #undef ERROUT
3531 DIOCGETETHRULESET_error:
3532 free(nvlpacked, M_NVLIST);
3533 nvlist_destroy(nvl);
3534 break;
3535 }
3536
3537 case DIOCADDRULENV: {
3538 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3539 nvlist_t *nvl = NULL;
3540 void *nvlpacked = NULL;
3541 struct pf_krule *rule = NULL;
3542 const char *anchor = "", *anchor_call = "";
3543 uint32_t ticket = 0, pool_ticket = 0;
3544
3545 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
3546
3547 if (nv->len > pf_ioctl_maxcount)
3548 ERROUT(ENOMEM);
3549
3550 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3551 error = copyin(nv->data, nvlpacked, nv->len);
3552 if (error)
3553 ERROUT(error);
3554
3555 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3556 if (nvl == NULL)
3557 ERROUT(EBADMSG);
3558
3559 if (! nvlist_exists_number(nvl, "ticket"))
3560 ERROUT(EINVAL);
3561 ticket = nvlist_get_number(nvl, "ticket");
3562
3563 if (! nvlist_exists_number(nvl, "pool_ticket"))
3564 ERROUT(EINVAL);
3565 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3566
3567 if (! nvlist_exists_nvlist(nvl, "rule"))
3568 ERROUT(EINVAL);
3569
3570 rule = pf_krule_alloc();
3571 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3572 rule);
3573 if (error)
3574 ERROUT(error);
3575
3576 if (nvlist_exists_string(nvl, "anchor"))
3577 anchor = nvlist_get_string(nvl, "anchor");
3578 if (nvlist_exists_string(nvl, "anchor_call"))
3579 anchor_call = nvlist_get_string(nvl, "anchor_call");
3580
3581 if ((error = nvlist_error(nvl)))
3582 ERROUT(error);
3583
3584 /* Frees rule on error */
3585 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3586 anchor_call, td->td_ucred->cr_ruid,
3587 td->td_proc ? td->td_proc->p_pid : 0);
3588
3589 nvlist_destroy(nvl);
3590 free(nvlpacked, M_NVLIST);
3591 break;
3592 #undef ERROUT
3593 DIOCADDRULENV_error:
3594 pf_krule_free(rule);
3595 nvlist_destroy(nvl);
3596 free(nvlpacked, M_NVLIST);
3597
3598 break;
3599 }
3600 case DIOCADDRULE: {
3601 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3602 struct pf_krule *rule;
3603
3604 rule = pf_krule_alloc();
3605 error = pf_rule_to_krule(&pr->rule, rule);
3606 if (error != 0) {
3607 pf_krule_free(rule);
3608 goto fail;
3609 }
3610
3611 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
3612
3613 /* Frees rule on error */
3614 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3615 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3616 td->td_proc ? td->td_proc->p_pid : 0);
3617 break;
3618 }
3619
3620 case DIOCGETRULES: {
3621 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3622
3623 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
3624
3625 error = pf_ioctl_getrules(pr);
3626
3627 break;
3628 }
3629
3630 case DIOCGETRULENV: {
3631 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3632 nvlist_t *nvrule = NULL;
3633 nvlist_t *nvl = NULL;
3634 struct pf_kruleset *ruleset;
3635 struct pf_krule *rule;
3636 void *nvlpacked = NULL;
3637 int rs_num, nr;
3638 bool clear_counter = false;
3639
3640 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
3641
3642 if (nv->len > pf_ioctl_maxcount)
3643 ERROUT(ENOMEM);
3644
3645 /* Copy the request in */
3646 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3647 error = copyin(nv->data, nvlpacked, nv->len);
3648 if (error)
3649 ERROUT(error);
3650
3651 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3652 if (nvl == NULL)
3653 ERROUT(EBADMSG);
3654
3655 if (! nvlist_exists_string(nvl, "anchor"))
3656 ERROUT(EBADMSG);
3657 if (! nvlist_exists_number(nvl, "ruleset"))
3658 ERROUT(EBADMSG);
3659 if (! nvlist_exists_number(nvl, "ticket"))
3660 ERROUT(EBADMSG);
3661 if (! nvlist_exists_number(nvl, "nr"))
3662 ERROUT(EBADMSG);
3663
3664 if (nvlist_exists_bool(nvl, "clear_counter"))
3665 clear_counter = nvlist_get_bool(nvl, "clear_counter");
3666
3667 if (clear_counter && !(flags & FWRITE))
3668 ERROUT(EACCES);
3669
3670 nr = nvlist_get_number(nvl, "nr");
3671
3672 PF_RULES_WLOCK();
3673 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3674 if (ruleset == NULL) {
3675 PF_RULES_WUNLOCK();
3676 ERROUT(ENOENT);
3677 }
3678
3679 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3680 if (rs_num >= PF_RULESET_MAX) {
3681 PF_RULES_WUNLOCK();
3682 ERROUT(EINVAL);
3683 }
3684
3685 if (nvlist_get_number(nvl, "ticket") !=
3686 ruleset->rules[rs_num].active.ticket) {
3687 PF_RULES_WUNLOCK();
3688 ERROUT(EBUSY);
3689 }
3690
3691 if ((error = nvlist_error(nvl))) {
3692 PF_RULES_WUNLOCK();
3693 ERROUT(error);
3694 }
3695
3696 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3697 while ((rule != NULL) && (rule->nr != nr))
3698 rule = TAILQ_NEXT(rule, entries);
3699 if (rule == NULL) {
3700 PF_RULES_WUNLOCK();
3701 ERROUT(EBUSY);
3702 }
3703
3704 nvrule = pf_krule_to_nvrule(rule);
3705
3706 nvlist_destroy(nvl);
3707 nvl = nvlist_create(0);
3708 if (nvl == NULL) {
3709 PF_RULES_WUNLOCK();
3710 ERROUT(ENOMEM);
3711 }
3712 nvlist_add_number(nvl, "nr", nr);
3713 nvlist_add_nvlist(nvl, "rule", nvrule);
3714 nvlist_destroy(nvrule);
3715 nvrule = NULL;
3716 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3717 PF_RULES_WUNLOCK();
3718 ERROUT(EBUSY);
3719 }
3720
3721 free(nvlpacked, M_NVLIST);
3722 nvlpacked = nvlist_pack(nvl, &nv->len);
3723 if (nvlpacked == NULL) {
3724 PF_RULES_WUNLOCK();
3725 ERROUT(ENOMEM);
3726 }
3727
3728 if (nv->size == 0) {
3729 PF_RULES_WUNLOCK();
3730 ERROUT(0);
3731 }
3732 else if (nv->size < nv->len) {
3733 PF_RULES_WUNLOCK();
3734 ERROUT(ENOSPC);
3735 }
3736
3737 if (clear_counter)
3738 pf_krule_clear_counters(rule);
3739
3740 PF_RULES_WUNLOCK();
3741
3742 error = copyout(nvlpacked, nv->data, nv->len);
3743
3744 #undef ERROUT
3745 DIOCGETRULENV_error:
3746 free(nvlpacked, M_NVLIST);
3747 nvlist_destroy(nvrule);
3748 nvlist_destroy(nvl);
3749
3750 break;
3751 }
3752
3753 case DIOCCHANGERULE: {
3754 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
3755 struct pf_kruleset *ruleset;
3756 struct pf_krule *oldrule = NULL, *newrule = NULL;
3757 struct pfi_kkif *kif = NULL;
3758 struct pf_kpooladdr *pa;
3759 u_int32_t nr = 0;
3760 int rs_num;
3761
3762 pcr->anchor[sizeof(pcr->anchor) - 1] = '\0';
3763
3764 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3765 pcr->action > PF_CHANGE_GET_TICKET) {
3766 error = EINVAL;
3767 goto fail;
3768 }
3769 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3770 error = EINVAL;
3771 goto fail;
3772 }
3773
3774 if (pcr->action != PF_CHANGE_REMOVE) {
3775 newrule = pf_krule_alloc();
3776 error = pf_rule_to_krule(&pcr->rule, newrule);
3777 if (error != 0) {
3778 pf_krule_free(newrule);
3779 goto fail;
3780 }
3781
3782 if ((error = pf_rule_checkaf(newrule))) {
3783 pf_krule_free(newrule);
3784 goto fail;
3785 }
3786 if (newrule->ifname[0])
3787 kif = pf_kkif_create(M_WAITOK);
3788 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3789 for (int i = 0; i < 2; i++) {
3790 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3791 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3792 }
3793 newrule->states_cur = counter_u64_alloc(M_WAITOK);
3794 newrule->states_tot = counter_u64_alloc(M_WAITOK);
3795 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
3796 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
3797 newrule->cuid = td->td_ucred->cr_ruid;
3798 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3799 TAILQ_INIT(&newrule->nat.list);
3800 TAILQ_INIT(&newrule->rdr.list);
3801 TAILQ_INIT(&newrule->route.list);
3802 }
3803 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3804
3805 PF_CONFIG_LOCK();
3806 PF_RULES_WLOCK();
3807 #ifdef PF_WANT_32_TO_64_COUNTER
3808 if (newrule != NULL) {
3809 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3810 newrule->allrulelinked = true;
3811 V_pf_allrulecount++;
3812 }
3813 #endif
3814
3815 if (!(pcr->action == PF_CHANGE_REMOVE ||
3816 pcr->action == PF_CHANGE_GET_TICKET) &&
3817 pcr->pool_ticket != V_ticket_pabuf)
3818 ERROUT(EBUSY);
3819
3820 ruleset = pf_find_kruleset(pcr->anchor);
3821 if (ruleset == NULL)
3822 ERROUT(EINVAL);
3823
3824 rs_num = pf_get_ruleset_number(pcr->rule.action);
3825 if (rs_num >= PF_RULESET_MAX)
3826 ERROUT(EINVAL);
3827
3828 /*
3829 * XXXMJG: there is no guarantee that the ruleset was
3830 * created by the usual route of calling DIOCXBEGIN.
3831 * As a result it is possible the rule tree will not
3832 * be allocated yet. Hack around it by doing it here.
3833 * Note it is fine to let the tree persist in case of
3834 * error as it will be freed down the road on future
3835 * updates (if need be).
3836 */
3837 if (ruleset->rules[rs_num].active.tree == NULL) {
3838 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3839 if (ruleset->rules[rs_num].active.tree == NULL) {
3840 ERROUT(ENOMEM);
3841 }
3842 }
3843
3844 if (pcr->action == PF_CHANGE_GET_TICKET) {
3845 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3846 ERROUT(0);
3847 } else if (pcr->ticket !=
3848 ruleset->rules[rs_num].active.ticket)
3849 ERROUT(EINVAL);
3850
3851 if (pcr->action != PF_CHANGE_REMOVE) {
3852 if (newrule->ifname[0]) {
3853 newrule->kif = pfi_kkif_attach(kif,
3854 newrule->ifname);
3855 kif = NULL;
3856 pfi_kkif_ref(newrule->kif);
3857 } else
3858 newrule->kif = NULL;
3859
3860 if (newrule->rtableid > 0 &&
3861 newrule->rtableid >= rt_numfibs)
3862 error = EBUSY;
3863
3864 #ifdef ALTQ
3865 /* set queue IDs */
3866 if (newrule->qname[0] != 0) {
3867 if ((newrule->qid =
3868 pf_qname2qid(newrule->qname)) == 0)
3869 error = EBUSY;
3870 else if (newrule->pqname[0] != 0) {
3871 if ((newrule->pqid =
3872 pf_qname2qid(newrule->pqname)) == 0)
3873 error = EBUSY;
3874 } else
3875 newrule->pqid = newrule->qid;
3876 }
3877 #endif /* ALTQ */
3878 if (newrule->tagname[0])
3879 if ((newrule->tag =
3880 pf_tagname2tag(newrule->tagname)) == 0)
3881 error = EBUSY;
3882 if (newrule->match_tagname[0])
3883 if ((newrule->match_tag = pf_tagname2tag(
3884 newrule->match_tagname)) == 0)
3885 error = EBUSY;
3886 if (newrule->rt && !newrule->direction)
3887 error = EINVAL;
3888 if (!newrule->log)
3889 newrule->logif = 0;
3890 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3891 error = ENOMEM;
3892 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3893 error = ENOMEM;
3894 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3895 error = EINVAL;
3896 for (int i = 0; i < 3; i++) {
3897 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
3898 if (pa->addr.type == PF_ADDR_TABLE) {
3899 pa->addr.p.tbl =
3900 pfr_attach_table(ruleset,
3901 pa->addr.v.tblname);
3902 if (pa->addr.p.tbl == NULL)
3903 error = ENOMEM;
3904 }
3905 }
3906
3907 newrule->overload_tbl = NULL;
3908 if (newrule->overload_tblname[0]) {
3909 if ((newrule->overload_tbl = pfr_attach_table(
3910 ruleset, newrule->overload_tblname)) ==
3911 NULL)
3912 error = EINVAL;
3913 else
3914 newrule->overload_tbl->pfrkt_flags |=
3915 PFR_TFLAG_ACTIVE;
3916 }
3917
3918 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
3919 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
3920 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list);
3921 if (((((newrule->action == PF_NAT) ||
3922 (newrule->action == PF_RDR) ||
3923 (newrule->action == PF_BINAT) ||
3924 (newrule->rt > PF_NOPFROUTE)) &&
3925 !newrule->anchor)) &&
3926 (TAILQ_FIRST(&newrule->rdr.list) == NULL))
3927 error = EINVAL;
3928
3929 if (error) {
3930 pf_free_rule(newrule);
3931 PF_RULES_WUNLOCK();
3932 PF_CONFIG_UNLOCK();
3933 goto fail;
3934 }
3935
3936 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
3937 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
3938 }
3939 pf_empty_kpool(&V_pf_pabuf[0]);
3940 pf_empty_kpool(&V_pf_pabuf[1]);
3941 pf_empty_kpool(&V_pf_pabuf[2]);
3942
3943 if (pcr->action == PF_CHANGE_ADD_HEAD)
3944 oldrule = TAILQ_FIRST(
3945 ruleset->rules[rs_num].active.ptr);
3946 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3947 oldrule = TAILQ_LAST(
3948 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3949 else {
3950 oldrule = TAILQ_FIRST(
3951 ruleset->rules[rs_num].active.ptr);
3952 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3953 oldrule = TAILQ_NEXT(oldrule, entries);
3954 if (oldrule == NULL) {
3955 if (newrule != NULL)
3956 pf_free_rule(newrule);
3957 PF_RULES_WUNLOCK();
3958 PF_CONFIG_UNLOCK();
3959 error = EINVAL;
3960 goto fail;
3961 }
3962 }
3963
3964 if (pcr->action == PF_CHANGE_REMOVE) {
3965 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3966 oldrule);
3967 RB_REMOVE(pf_krule_global,
3968 ruleset->rules[rs_num].active.tree, oldrule);
3969 ruleset->rules[rs_num].active.rcount--;
3970 } else {
3971 pf_hash_rule(newrule);
3972 if (RB_INSERT(pf_krule_global,
3973 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3974 pf_free_rule(newrule);
3975 PF_RULES_WUNLOCK();
3976 PF_CONFIG_UNLOCK();
3977 error = EEXIST;
3978 goto fail;
3979 }
3980
3981 if (oldrule == NULL)
3982 TAILQ_INSERT_TAIL(
3983 ruleset->rules[rs_num].active.ptr,
3984 newrule, entries);
3985 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3986 pcr->action == PF_CHANGE_ADD_BEFORE)
3987 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3988 else
3989 TAILQ_INSERT_AFTER(
3990 ruleset->rules[rs_num].active.ptr,
3991 oldrule, newrule, entries);
3992 ruleset->rules[rs_num].active.rcount++;
3993 }
3994
3995 nr = 0;
3996 TAILQ_FOREACH(oldrule,
3997 ruleset->rules[rs_num].active.ptr, entries)
3998 oldrule->nr = nr++;
3999
4000 ruleset->rules[rs_num].active.ticket++;
4001
4002 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
4003 pf_remove_if_empty_kruleset(ruleset);
4004
4005 PF_RULES_WUNLOCK();
4006 PF_CONFIG_UNLOCK();
4007 break;
4008
4009 #undef ERROUT
4010 DIOCCHANGERULE_error:
4011 PF_RULES_WUNLOCK();
4012 PF_CONFIG_UNLOCK();
4013 pf_krule_free(newrule);
4014 pf_kkif_free(kif);
4015 break;
4016 }
4017
4018 case DIOCCLRSTATESNV: {
4019 error = pf_clearstates_nv((struct pfioc_nv *)addr);
4020 break;
4021 }
4022
4023 case DIOCKILLSTATESNV: {
4024 error = pf_killstates_nv((struct pfioc_nv *)addr);
4025 break;
4026 }
4027
4028 case DIOCADDSTATE: {
4029 struct pfioc_state *ps = (struct pfioc_state *)addr;
4030 struct pfsync_state_1301 *sp = &ps->state;
4031
4032 if (sp->timeout >= PFTM_MAX) {
4033 error = EINVAL;
4034 goto fail;
4035 }
4036 if (V_pfsync_state_import_ptr != NULL) {
4037 PF_RULES_RLOCK();
4038 error = V_pfsync_state_import_ptr(
4039 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
4040 PFSYNC_MSG_VERSION_1301);
4041 PF_RULES_RUNLOCK();
4042 } else
4043 error = EOPNOTSUPP;
4044 break;
4045 }
4046
4047 case DIOCGETSTATE: {
4048 struct pfioc_state *ps = (struct pfioc_state *)addr;
4049 struct pf_kstate *s;
4050
4051 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
4052 if (s == NULL) {
4053 error = ENOENT;
4054 goto fail;
4055 }
4056
4057 pfsync_state_export((union pfsync_state_union*)&ps->state,
4058 s, PFSYNC_MSG_VERSION_1301);
4059 PF_STATE_UNLOCK(s);
4060 break;
4061 }
4062
4063 case DIOCGETSTATENV: {
4064 error = pf_getstate((struct pfioc_nv *)addr);
4065 break;
4066 }
4067
4068 #ifdef COMPAT_FREEBSD14
4069 case DIOCGETSTATES: {
4070 struct pfioc_states *ps = (struct pfioc_states *)addr;
4071 struct pf_kstate *s;
4072 struct pfsync_state_1301 *pstore, *p;
4073 int i, nr;
4074 size_t slice_count = 16, count;
4075 void *out;
4076
4077 if (ps->ps_len <= 0) {
4078 nr = uma_zone_get_cur(V_pf_state_z);
4079 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4080 break;
4081 }
4082
4083 out = ps->ps_states;
4084 pstore = mallocarray(slice_count,
4085 sizeof(struct pfsync_state_1301), M_PF, M_WAITOK | M_ZERO);
4086 nr = 0;
4087
4088 for (i = 0; i <= V_pf_hashmask; i++) {
4089 struct pf_idhash *ih = &V_pf_idhash[i];
4090
4091 DIOCGETSTATES_retry:
4092 p = pstore;
4093
4094 if (LIST_EMPTY(&ih->states))
4095 continue;
4096
4097 PF_HASHROW_LOCK(ih);
4098 count = 0;
4099 LIST_FOREACH(s, &ih->states, entry) {
4100 if (s->timeout == PFTM_UNLINKED)
4101 continue;
4102 count++;
4103 }
4104
4105 if (count > slice_count) {
4106 PF_HASHROW_UNLOCK(ih);
4107 free(pstore, M_PF);
4108 slice_count = count * 2;
4109 pstore = mallocarray(slice_count,
4110 sizeof(struct pfsync_state_1301), M_PF,
4111 M_WAITOK | M_ZERO);
4112 goto DIOCGETSTATES_retry;
4113 }
4114
4115 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4116 PF_HASHROW_UNLOCK(ih);
4117 goto DIOCGETSTATES_full;
4118 }
4119
4120 LIST_FOREACH(s, &ih->states, entry) {
4121 if (s->timeout == PFTM_UNLINKED)
4122 continue;
4123
4124 pfsync_state_export((union pfsync_state_union*)p,
4125 s, PFSYNC_MSG_VERSION_1301);
4126 p++;
4127 nr++;
4128 }
4129 PF_HASHROW_UNLOCK(ih);
4130 error = copyout(pstore, out,
4131 sizeof(struct pfsync_state_1301) * count);
4132 if (error) {
4133 free(pstore, M_PF);
4134 goto fail;
4135 }
4136 out = ps->ps_states + nr;
4137 }
4138 DIOCGETSTATES_full:
4139 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4140 free(pstore, M_PF);
4141
4142 break;
4143 }
4144
4145 case DIOCGETSTATESV2: {
4146 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
4147 struct pf_kstate *s;
4148 struct pf_state_export *pstore, *p;
4149 int i, nr;
4150 size_t slice_count = 16, count;
4151 void *out;
4152
4153 if (ps->ps_req_version > PF_STATE_VERSION) {
4154 error = ENOTSUP;
4155 goto fail;
4156 }
4157
4158 if (ps->ps_len <= 0) {
4159 nr = uma_zone_get_cur(V_pf_state_z);
4160 ps->ps_len = sizeof(struct pf_state_export) * nr;
4161 break;
4162 }
4163
4164 out = ps->ps_states;
4165 pstore = mallocarray(slice_count,
4166 sizeof(struct pf_state_export), M_PF, M_WAITOK | M_ZERO);
4167 nr = 0;
4168
4169 for (i = 0; i <= V_pf_hashmask; i++) {
4170 struct pf_idhash *ih = &V_pf_idhash[i];
4171
4172 DIOCGETSTATESV2_retry:
4173 p = pstore;
4174
4175 if (LIST_EMPTY(&ih->states))
4176 continue;
4177
4178 PF_HASHROW_LOCK(ih);
4179 count = 0;
4180 LIST_FOREACH(s, &ih->states, entry) {
4181 if (s->timeout == PFTM_UNLINKED)
4182 continue;
4183 count++;
4184 }
4185
4186 if (count > slice_count) {
4187 PF_HASHROW_UNLOCK(ih);
4188 free(pstore, M_PF);
4189 slice_count = count * 2;
4190 pstore = mallocarray(slice_count,
4191 sizeof(struct pf_state_export), M_PF,
4192 M_WAITOK | M_ZERO);
4193 goto DIOCGETSTATESV2_retry;
4194 }
4195
4196 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4197 PF_HASHROW_UNLOCK(ih);
4198 goto DIOCGETSTATESV2_full;
4199 }
4200
4201 LIST_FOREACH(s, &ih->states, entry) {
4202 if (s->timeout == PFTM_UNLINKED)
4203 continue;
4204
4205 pf_state_export(p, s);
4206 p++;
4207 nr++;
4208 }
4209 PF_HASHROW_UNLOCK(ih);
4210 error = copyout(pstore, out,
4211 sizeof(struct pf_state_export) * count);
4212 if (error) {
4213 free(pstore, M_PF);
4214 goto fail;
4215 }
4216 out = ps->ps_states + nr;
4217 }
4218 DIOCGETSTATESV2_full:
4219 ps->ps_len = nr * sizeof(struct pf_state_export);
4220 free(pstore, M_PF);
4221
4222 break;
4223 }
4224 #endif
4225 case DIOCGETSTATUSNV: {
4226 error = pf_getstatus((struct pfioc_nv *)addr);
4227 break;
4228 }
4229
4230 case DIOCSETSTATUSIF: {
4231 struct pfioc_if *pi = (struct pfioc_if *)addr;
4232
4233 if (pi->ifname[0] == 0) {
4234 bzero(V_pf_status.ifname, IFNAMSIZ);
4235 break;
4236 }
4237 PF_RULES_WLOCK();
4238 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
4239 PF_RULES_WUNLOCK();
4240 break;
4241 }
4242
4243 case DIOCCLRSTATUS: {
4244 pf_ioctl_clear_status();
4245 break;
4246 }
4247
4248 case DIOCNATLOOK: {
4249 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
4250
4251 error = pf_ioctl_natlook(pnl);
4252 break;
4253 }
4254
4255 case DIOCSETTIMEOUT: {
4256 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4257
4258 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4259 &pt->seconds);
4260 break;
4261 }
4262
4263 case DIOCGETTIMEOUT: {
4264 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4265
4266 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4267 break;
4268 }
4269
4270 case DIOCGETLIMIT: {
4271 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4272
4273 error = pf_ioctl_get_limit(pl->index, &pl->limit);
4274 break;
4275 }
4276
4277 case DIOCSETLIMIT: {
4278 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4279 unsigned int old_limit;
4280
4281 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4282 pl->limit = old_limit;
4283 break;
4284 }
4285
4286 case DIOCSETDEBUG: {
4287 u_int32_t *level = (u_int32_t *)addr;
4288
4289 PF_RULES_WLOCK();
4290 V_pf_status.debug = *level;
4291 PF_RULES_WUNLOCK();
4292 break;
4293 }
4294
4295 case DIOCCLRRULECTRS: {
4296 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4297 struct pf_kruleset *ruleset = &pf_main_ruleset;
4298 struct pf_krule *rule;
4299
4300 PF_RULES_WLOCK();
4301 TAILQ_FOREACH(rule,
4302 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4303 pf_counter_u64_zero(&rule->evaluations);
4304 for (int i = 0; i < 2; i++) {
4305 pf_counter_u64_zero(&rule->packets[i]);
4306 pf_counter_u64_zero(&rule->bytes[i]);
4307 }
4308 }
4309 PF_RULES_WUNLOCK();
4310 break;
4311 }
4312
4313 case DIOCGIFSPEEDV0:
4314 case DIOCGIFSPEEDV1: {
4315 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
4316 struct pf_ifspeed_v1 ps;
4317 struct ifnet *ifp;
4318
4319 if (psp->ifname[0] == '\0') {
4320 error = EINVAL;
4321 goto fail;
4322 }
4323
4324 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4325 if (error != 0)
4326 goto fail;
4327 ifp = ifunit(ps.ifname);
4328 if (ifp != NULL) {
4329 psp->baudrate32 =
4330 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4331 if (cmd == DIOCGIFSPEEDV1)
4332 psp->baudrate = ifp->if_baudrate;
4333 } else {
4334 error = EINVAL;
4335 }
4336 break;
4337 }
4338
4339 #ifdef ALTQ
4340 case DIOCSTARTALTQ: {
4341 struct pf_altq *altq;
4342
4343 PF_RULES_WLOCK();
4344 /* enable all altq interfaces on active list */
4345 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4346 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4347 error = pf_enable_altq(altq);
4348 if (error != 0)
4349 break;
4350 }
4351 }
4352 if (error == 0)
4353 V_pf_altq_running = 1;
4354 PF_RULES_WUNLOCK();
4355 DPFPRINTF(PF_DEBUG_MISC, "altq: started");
4356 break;
4357 }
4358
4359 case DIOCSTOPALTQ: {
4360 struct pf_altq *altq;
4361
4362 PF_RULES_WLOCK();
4363 /* disable all altq interfaces on active list */
4364 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4365 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4366 error = pf_disable_altq(altq);
4367 if (error != 0)
4368 break;
4369 }
4370 }
4371 if (error == 0)
4372 V_pf_altq_running = 0;
4373 PF_RULES_WUNLOCK();
4374 DPFPRINTF(PF_DEBUG_MISC, "altq: stopped");
4375 break;
4376 }
4377
4378 case DIOCADDALTQV0:
4379 case DIOCADDALTQV1: {
4380 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4381 struct pf_altq *altq, *a;
4382 struct ifnet *ifp;
4383
4384 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4385 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4386 if (error)
4387 goto fail;
4388 altq->local_flags = 0;
4389
4390 PF_RULES_WLOCK();
4391 if (pa->ticket != V_ticket_altqs_inactive) {
4392 PF_RULES_WUNLOCK();
4393 free(altq, M_PFALTQ);
4394 error = EBUSY;
4395 goto fail;
4396 }
4397
4398 /*
4399 * if this is for a queue, find the discipline and
4400 * copy the necessary fields
4401 */
4402 if (altq->qname[0] != 0) {
4403 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4404 PF_RULES_WUNLOCK();
4405 error = EBUSY;
4406 free(altq, M_PFALTQ);
4407 goto fail;
4408 }
4409 altq->altq_disc = NULL;
4410 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4411 if (strncmp(a->ifname, altq->ifname,
4412 IFNAMSIZ) == 0) {
4413 altq->altq_disc = a->altq_disc;
4414 break;
4415 }
4416 }
4417 }
4418
4419 if ((ifp = ifunit(altq->ifname)) == NULL)
4420 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4421 else
4422 error = altq_add(ifp, altq);
4423
4424 if (error) {
4425 PF_RULES_WUNLOCK();
4426 free(altq, M_PFALTQ);
4427 goto fail;
4428 }
4429
4430 if (altq->qname[0] != 0)
4431 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4432 else
4433 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4434 /* version error check done on import above */
4435 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4436 PF_RULES_WUNLOCK();
4437 break;
4438 }
4439
4440 case DIOCGETALTQSV0:
4441 case DIOCGETALTQSV1: {
4442 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4443 struct pf_altq *altq;
4444
4445 PF_RULES_RLOCK();
4446 pa->nr = 0;
4447 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4448 pa->nr++;
4449 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4450 pa->nr++;
4451 pa->ticket = V_ticket_altqs_active;
4452 PF_RULES_RUNLOCK();
4453 break;
4454 }
4455
4456 case DIOCGETALTQV0:
4457 case DIOCGETALTQV1: {
4458 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4459 struct pf_altq *altq;
4460
4461 PF_RULES_RLOCK();
4462 if (pa->ticket != V_ticket_altqs_active) {
4463 PF_RULES_RUNLOCK();
4464 error = EBUSY;
4465 goto fail;
4466 }
4467 altq = pf_altq_get_nth_active(pa->nr);
4468 if (altq == NULL) {
4469 PF_RULES_RUNLOCK();
4470 error = EBUSY;
4471 goto fail;
4472 }
4473 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4474 PF_RULES_RUNLOCK();
4475 break;
4476 }
4477
4478 case DIOCCHANGEALTQV0:
4479 case DIOCCHANGEALTQV1:
4480 /* CHANGEALTQ not supported yet! */
4481 error = ENODEV;
4482 break;
4483
4484 case DIOCGETQSTATSV0:
4485 case DIOCGETQSTATSV1: {
4486 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
4487 struct pf_altq *altq;
4488 int nbytes;
4489 u_int32_t version;
4490
4491 PF_RULES_RLOCK();
4492 if (pq->ticket != V_ticket_altqs_active) {
4493 PF_RULES_RUNLOCK();
4494 error = EBUSY;
4495 goto fail;
4496 }
4497 nbytes = pq->nbytes;
4498 altq = pf_altq_get_nth_active(pq->nr);
4499 if (altq == NULL) {
4500 PF_RULES_RUNLOCK();
4501 error = EBUSY;
4502 goto fail;
4503 }
4504
4505 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4506 PF_RULES_RUNLOCK();
4507 error = ENXIO;
4508 goto fail;
4509 }
4510 PF_RULES_RUNLOCK();
4511 if (cmd == DIOCGETQSTATSV0)
4512 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
4513 else
4514 version = pq->version;
4515 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4516 if (error == 0) {
4517 pq->scheduler = altq->scheduler;
4518 pq->nbytes = nbytes;
4519 }
4520 break;
4521 }
4522 #endif /* ALTQ */
4523
4524 case DIOCBEGINADDRS: {
4525 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4526
4527 error = pf_ioctl_begin_addrs(&pp->ticket);
4528 break;
4529 }
4530
4531 case DIOCADDADDR: {
4532 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4533 struct pf_nl_pooladdr npp = {};
4534
4535 npp.which = PF_RDR;
4536 memcpy(&npp, pp, sizeof(*pp));
4537 error = pf_ioctl_add_addr(&npp);
4538 break;
4539 }
4540
4541 case DIOCGETADDRS: {
4542 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4543 struct pf_nl_pooladdr npp = {};
4544
4545 npp.which = PF_RDR;
4546 memcpy(&npp, pp, sizeof(*pp));
4547 error = pf_ioctl_get_addrs(&npp);
4548 memcpy(pp, &npp, sizeof(*pp));
4549
4550 break;
4551 }
4552
4553 case DIOCGETADDR: {
4554 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4555 struct pf_nl_pooladdr npp = {};
4556
4557 npp.which = PF_RDR;
4558 memcpy(&npp, pp, sizeof(*pp));
4559 error = pf_ioctl_get_addr(&npp);
4560 memcpy(pp, &npp, sizeof(*pp));
4561
4562 break;
4563 }
4564
4565 case DIOCCHANGEADDR: {
4566 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
4567 struct pf_kpool *pool;
4568 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
4569 struct pf_kruleset *ruleset;
4570 struct pfi_kkif *kif = NULL;
4571
4572 pca->anchor[sizeof(pca->anchor) - 1] = '\0';
4573
4574 if (pca->action < PF_CHANGE_ADD_HEAD ||
4575 pca->action > PF_CHANGE_REMOVE) {
4576 error = EINVAL;
4577 goto fail;
4578 }
4579 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4580 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4581 pca->addr.addr.type != PF_ADDR_TABLE) {
4582 error = EINVAL;
4583 goto fail;
4584 }
4585 if (pca->addr.addr.p.dyn != NULL) {
4586 error = EINVAL;
4587 goto fail;
4588 }
4589
4590 if (pca->action != PF_CHANGE_REMOVE) {
4591 #ifndef INET
4592 if (pca->af == AF_INET) {
4593 error = EAFNOSUPPORT;
4594 goto fail;
4595 }
4596 #endif /* INET */
4597 #ifndef INET6
4598 if (pca->af == AF_INET6) {
4599 error = EAFNOSUPPORT;
4600 goto fail;
4601 }
4602 #endif /* INET6 */
4603 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4604 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4605 if (newpa->ifname[0])
4606 kif = pf_kkif_create(M_WAITOK);
4607 newpa->kif = NULL;
4608 }
4609 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4610 PF_RULES_WLOCK();
4611 ruleset = pf_find_kruleset(pca->anchor);
4612 if (ruleset == NULL)
4613 ERROUT(EBUSY);
4614
4615 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4616 pca->r_num, pca->r_last, 1, 1, PF_RDR);
4617 if (pool == NULL)
4618 ERROUT(EBUSY);
4619
4620 if (pca->action != PF_CHANGE_REMOVE) {
4621 if (newpa->ifname[0]) {
4622 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4623 pfi_kkif_ref(newpa->kif);
4624 kif = NULL;
4625 }
4626
4627 switch (newpa->addr.type) {
4628 case PF_ADDR_DYNIFTL:
4629 error = pfi_dynaddr_setup(&newpa->addr,
4630 pca->af);
4631 break;
4632 case PF_ADDR_TABLE:
4633 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4634 newpa->addr.v.tblname);
4635 if (newpa->addr.p.tbl == NULL)
4636 error = ENOMEM;
4637 break;
4638 }
4639 if (error)
4640 goto DIOCCHANGEADDR_error;
4641 }
4642
4643 switch (pca->action) {
4644 case PF_CHANGE_ADD_HEAD:
4645 oldpa = TAILQ_FIRST(&pool->list);
4646 break;
4647 case PF_CHANGE_ADD_TAIL:
4648 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4649 break;
4650 default:
4651 oldpa = TAILQ_FIRST(&pool->list);
4652 for (int i = 0; oldpa && i < pca->nr; i++)
4653 oldpa = TAILQ_NEXT(oldpa, entries);
4654
4655 if (oldpa == NULL)
4656 ERROUT(EINVAL);
4657 }
4658
4659 if (pca->action == PF_CHANGE_REMOVE) {
4660 TAILQ_REMOVE(&pool->list, oldpa, entries);
4661 switch (oldpa->addr.type) {
4662 case PF_ADDR_DYNIFTL:
4663 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4664 break;
4665 case PF_ADDR_TABLE:
4666 pfr_detach_table(oldpa->addr.p.tbl);
4667 break;
4668 }
4669 if (oldpa->kif)
4670 pfi_kkif_unref(oldpa->kif);
4671 free(oldpa, M_PFRULE);
4672 } else {
4673 if (oldpa == NULL)
4674 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4675 else if (pca->action == PF_CHANGE_ADD_HEAD ||
4676 pca->action == PF_CHANGE_ADD_BEFORE)
4677 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4678 else
4679 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4680 newpa, entries);
4681 }
4682
4683 pool->cur = TAILQ_FIRST(&pool->list);
4684 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4685 PF_RULES_WUNLOCK();
4686 break;
4687
4688 #undef ERROUT
4689 DIOCCHANGEADDR_error:
4690 if (newpa != NULL) {
4691 if (newpa->kif)
4692 pfi_kkif_unref(newpa->kif);
4693 free(newpa, M_PFRULE);
4694 }
4695 PF_RULES_WUNLOCK();
4696 pf_kkif_free(kif);
4697 break;
4698 }
4699
4700 case DIOCGETRULESETS: {
4701 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4702
4703 pr->path[sizeof(pr->path) - 1] = '\0';
4704
4705 error = pf_ioctl_get_rulesets(pr);
4706 break;
4707 }
4708
4709 case DIOCGETRULESET: {
4710 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4711
4712 pr->path[sizeof(pr->path) - 1] = '\0';
4713
4714 error = pf_ioctl_get_ruleset(pr);
4715 break;
4716 }
4717
4718 case DIOCRCLRTABLES: {
4719 struct pfioc_table *io = (struct pfioc_table *)addr;
4720
4721 if (io->pfrio_esize != 0) {
4722 error = ENODEV;
4723 goto fail;
4724 }
4725 PF_RULES_WLOCK();
4726 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4727 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4728 PF_RULES_WUNLOCK();
4729 break;
4730 }
4731
4732 case DIOCRADDTABLES: {
4733 struct pfioc_table *io = (struct pfioc_table *)addr;
4734 struct pfr_table *pfrts;
4735 size_t totlen;
4736
4737 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4738 error = ENODEV;
4739 goto fail;
4740 }
4741
4742 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4743 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4744 error = ENOMEM;
4745 goto fail;
4746 }
4747
4748 totlen = io->pfrio_size * sizeof(struct pfr_table);
4749 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4750 M_PF, M_WAITOK);
4751 error = copyin(io->pfrio_buffer, pfrts, totlen);
4752 if (error) {
4753 free(pfrts, M_PF);
4754 goto fail;
4755 }
4756 PF_RULES_WLOCK();
4757 error = pfr_add_tables(pfrts, io->pfrio_size,
4758 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4759 PF_RULES_WUNLOCK();
4760 free(pfrts, M_PF);
4761 break;
4762 }
4763
4764 case DIOCRDELTABLES: {
4765 struct pfioc_table *io = (struct pfioc_table *)addr;
4766 struct pfr_table *pfrts;
4767 size_t totlen;
4768
4769 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4770 error = ENODEV;
4771 goto fail;
4772 }
4773
4774 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4775 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4776 error = ENOMEM;
4777 goto fail;
4778 }
4779
4780 totlen = io->pfrio_size * sizeof(struct pfr_table);
4781 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4782 M_PF, M_WAITOK);
4783 error = copyin(io->pfrio_buffer, pfrts, totlen);
4784 if (error) {
4785 free(pfrts, M_PF);
4786 goto fail;
4787 }
4788 PF_RULES_WLOCK();
4789 error = pfr_del_tables(pfrts, io->pfrio_size,
4790 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4791 PF_RULES_WUNLOCK();
4792 free(pfrts, M_PF);
4793 break;
4794 }
4795
4796 case DIOCRGETTABLES: {
4797 struct pfioc_table *io = (struct pfioc_table *)addr;
4798 struct pfr_table *pfrts;
4799 size_t totlen;
4800 int n;
4801
4802 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4803 error = ENODEV;
4804 goto fail;
4805 }
4806 PF_RULES_RLOCK();
4807 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4808 if (n < 0) {
4809 PF_RULES_RUNLOCK();
4810 error = EINVAL;
4811 goto fail;
4812 }
4813 io->pfrio_size = min(io->pfrio_size, n);
4814
4815 totlen = io->pfrio_size * sizeof(struct pfr_table);
4816
4817 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4818 M_PF, M_NOWAIT | M_ZERO);
4819 if (pfrts == NULL) {
4820 error = ENOMEM;
4821 PF_RULES_RUNLOCK();
4822 goto fail;
4823 }
4824 error = pfr_get_tables(&io->pfrio_table, pfrts,
4825 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4826 PF_RULES_RUNLOCK();
4827 if (error == 0)
4828 error = copyout(pfrts, io->pfrio_buffer, totlen);
4829 free(pfrts, M_PF);
4830 break;
4831 }
4832
4833 case DIOCRGETTSTATS: {
4834 struct pfioc_table *io = (struct pfioc_table *)addr;
4835 struct pfr_tstats *pfrtstats;
4836 size_t totlen;
4837 int n;
4838
4839 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4840 error = ENODEV;
4841 goto fail;
4842 }
4843 PF_TABLE_STATS_LOCK();
4844 PF_RULES_RLOCK();
4845 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4846 if (n < 0) {
4847 PF_RULES_RUNLOCK();
4848 PF_TABLE_STATS_UNLOCK();
4849 error = EINVAL;
4850 goto fail;
4851 }
4852 io->pfrio_size = min(io->pfrio_size, n);
4853
4854 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4855 pfrtstats = mallocarray(io->pfrio_size,
4856 sizeof(struct pfr_tstats), M_PF, M_NOWAIT | M_ZERO);
4857 if (pfrtstats == NULL) {
4858 error = ENOMEM;
4859 PF_RULES_RUNLOCK();
4860 PF_TABLE_STATS_UNLOCK();
4861 goto fail;
4862 }
4863 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4864 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4865 PF_RULES_RUNLOCK();
4866 PF_TABLE_STATS_UNLOCK();
4867 if (error == 0)
4868 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4869 free(pfrtstats, M_PF);
4870 break;
4871 }
4872
4873 case DIOCRCLRTSTATS: {
4874 struct pfioc_table *io = (struct pfioc_table *)addr;
4875 struct pfr_table *pfrts;
4876 size_t totlen;
4877
4878 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4879 error = ENODEV;
4880 goto fail;
4881 }
4882
4883 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4884 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4885 /* We used to count tables and use the minimum required
4886 * size, so we didn't fail on overly large requests.
4887 * Keep doing so. */
4888 io->pfrio_size = pf_ioctl_maxcount;
4889 goto fail;
4890 }
4891
4892 totlen = io->pfrio_size * sizeof(struct pfr_table);
4893 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4894 M_PF, M_WAITOK);
4895 error = copyin(io->pfrio_buffer, pfrts, totlen);
4896 if (error) {
4897 free(pfrts, M_PF);
4898 goto fail;
4899 }
4900
4901 PF_TABLE_STATS_LOCK();
4902 PF_RULES_RLOCK();
4903 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4904 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4905 PF_RULES_RUNLOCK();
4906 PF_TABLE_STATS_UNLOCK();
4907 free(pfrts, M_PF);
4908 break;
4909 }
4910
4911 case DIOCRSETTFLAGS: {
4912 struct pfioc_table *io = (struct pfioc_table *)addr;
4913 struct pfr_table *pfrts;
4914 size_t totlen;
4915 int n;
4916
4917 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4918 error = ENODEV;
4919 goto fail;
4920 }
4921
4922 PF_RULES_RLOCK();
4923 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4924 if (n < 0) {
4925 PF_RULES_RUNLOCK();
4926 error = EINVAL;
4927 goto fail;
4928 }
4929
4930 io->pfrio_size = min(io->pfrio_size, n);
4931 PF_RULES_RUNLOCK();
4932
4933 totlen = io->pfrio_size * sizeof(struct pfr_table);
4934 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4935 M_PF, M_WAITOK);
4936 error = copyin(io->pfrio_buffer, pfrts, totlen);
4937 if (error) {
4938 free(pfrts, M_PF);
4939 goto fail;
4940 }
4941 PF_RULES_WLOCK();
4942 error = pfr_set_tflags(pfrts, io->pfrio_size,
4943 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4944 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4945 PF_RULES_WUNLOCK();
4946 free(pfrts, M_PF);
4947 break;
4948 }
4949
4950 case DIOCRCLRADDRS: {
4951 struct pfioc_table *io = (struct pfioc_table *)addr;
4952
4953 if (io->pfrio_esize != 0) {
4954 error = ENODEV;
4955 goto fail;
4956 }
4957 PF_RULES_WLOCK();
4958 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4959 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4960 PF_RULES_WUNLOCK();
4961 break;
4962 }
4963
4964 case DIOCRADDADDRS: {
4965 struct pfioc_table *io = (struct pfioc_table *)addr;
4966 struct pfr_addr *pfras;
4967 size_t totlen;
4968
4969 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4970 error = ENODEV;
4971 goto fail;
4972 }
4973 if (io->pfrio_size < 0 ||
4974 io->pfrio_size > pf_ioctl_maxcount ||
4975 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4976 error = EINVAL;
4977 goto fail;
4978 }
4979 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4980 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4981 M_PF, M_WAITOK);
4982 error = copyin(io->pfrio_buffer, pfras, totlen);
4983 if (error) {
4984 free(pfras, M_PF);
4985 goto fail;
4986 }
4987 PF_RULES_WLOCK();
4988 io->pfrio_nadd = 0;
4989 error = pfr_add_addrs(&io->pfrio_table, pfras,
4990 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4991 PFR_FLAG_USERIOCTL);
4992 PF_RULES_WUNLOCK();
4993 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4994 error = copyout(pfras, io->pfrio_buffer, totlen);
4995 free(pfras, M_PF);
4996 break;
4997 }
4998
4999 case DIOCRDELADDRS: {
5000 struct pfioc_table *io = (struct pfioc_table *)addr;
5001 struct pfr_addr *pfras;
5002 size_t totlen;
5003
5004 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5005 error = ENODEV;
5006 goto fail;
5007 }
5008 if (io->pfrio_size < 0 ||
5009 io->pfrio_size > pf_ioctl_maxcount ||
5010 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5011 error = EINVAL;
5012 goto fail;
5013 }
5014 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5015 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5016 M_PF, M_WAITOK);
5017 error = copyin(io->pfrio_buffer, pfras, totlen);
5018 if (error) {
5019 free(pfras, M_PF);
5020 goto fail;
5021 }
5022 PF_RULES_WLOCK();
5023 error = pfr_del_addrs(&io->pfrio_table, pfras,
5024 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
5025 PFR_FLAG_USERIOCTL);
5026 PF_RULES_WUNLOCK();
5027 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5028 error = copyout(pfras, io->pfrio_buffer, totlen);
5029 free(pfras, M_PF);
5030 break;
5031 }
5032
5033 case DIOCRSETADDRS: {
5034 struct pfioc_table *io = (struct pfioc_table *)addr;
5035 struct pfr_addr *pfras;
5036 size_t totlen, count;
5037
5038 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5039 error = ENODEV;
5040 goto fail;
5041 }
5042 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
5043 error = EINVAL;
5044 goto fail;
5045 }
5046 count = max(io->pfrio_size, io->pfrio_size2);
5047 if (count > pf_ioctl_maxcount ||
5048 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
5049 error = EINVAL;
5050 goto fail;
5051 }
5052 totlen = count * sizeof(struct pfr_addr);
5053 pfras = mallocarray(count, sizeof(struct pfr_addr), M_PF,
5054 M_WAITOK);
5055 error = copyin(io->pfrio_buffer, pfras, totlen);
5056 if (error) {
5057 free(pfras, M_PF);
5058 goto fail;
5059 }
5060 PF_RULES_WLOCK();
5061 error = pfr_set_addrs(&io->pfrio_table, pfras,
5062 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
5063 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
5064 PFR_FLAG_USERIOCTL, 0);
5065 PF_RULES_WUNLOCK();
5066 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5067 error = copyout(pfras, io->pfrio_buffer, totlen);
5068 free(pfras, M_PF);
5069 break;
5070 }
5071
5072 case DIOCRGETADDRS: {
5073 struct pfioc_table *io = (struct pfioc_table *)addr;
5074 struct pfr_addr *pfras;
5075 size_t totlen;
5076
5077 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5078 error = ENODEV;
5079 goto fail;
5080 }
5081 if (io->pfrio_size < 0 ||
5082 io->pfrio_size > pf_ioctl_maxcount ||
5083 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5084 error = EINVAL;
5085 goto fail;
5086 }
5087 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5088 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5089 M_PF, M_WAITOK | M_ZERO);
5090 PF_RULES_RLOCK();
5091 error = pfr_get_addrs(&io->pfrio_table, pfras,
5092 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5093 PF_RULES_RUNLOCK();
5094 if (error == 0)
5095 error = copyout(pfras, io->pfrio_buffer, totlen);
5096 free(pfras, M_PF);
5097 break;
5098 }
5099
5100 case DIOCRGETASTATS: {
5101 struct pfioc_table *io = (struct pfioc_table *)addr;
5102 struct pfr_astats *pfrastats;
5103 size_t totlen;
5104
5105 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
5106 error = ENODEV;
5107 goto fail;
5108 }
5109 if (io->pfrio_size < 0 ||
5110 io->pfrio_size > pf_ioctl_maxcount ||
5111 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5112 error = EINVAL;
5113 goto fail;
5114 }
5115 totlen = io->pfrio_size * sizeof(struct pfr_astats);
5116 pfrastats = mallocarray(io->pfrio_size,
5117 sizeof(struct pfr_astats), M_PF, M_WAITOK | M_ZERO);
5118 PF_RULES_RLOCK();
5119 error = pfr_get_astats(&io->pfrio_table, pfrastats,
5120 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5121 PF_RULES_RUNLOCK();
5122 if (error == 0)
5123 error = copyout(pfrastats, io->pfrio_buffer, totlen);
5124 free(pfrastats, M_PF);
5125 break;
5126 }
5127
5128 case DIOCRCLRASTATS: {
5129 struct pfioc_table *io = (struct pfioc_table *)addr;
5130 struct pfr_addr *pfras;
5131 size_t totlen;
5132
5133 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5134 error = ENODEV;
5135 goto fail;
5136 }
5137 if (io->pfrio_size < 0 ||
5138 io->pfrio_size > pf_ioctl_maxcount ||
5139 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5140 error = EINVAL;
5141 goto fail;
5142 }
5143 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5144 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5145 M_PF, M_WAITOK);
5146 error = copyin(io->pfrio_buffer, pfras, totlen);
5147 if (error) {
5148 free(pfras, M_PF);
5149 goto fail;
5150 }
5151 PF_RULES_WLOCK();
5152 error = pfr_clr_astats(&io->pfrio_table, pfras,
5153 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5154 PFR_FLAG_USERIOCTL);
5155 PF_RULES_WUNLOCK();
5156 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5157 error = copyout(pfras, io->pfrio_buffer, totlen);
5158 free(pfras, M_PF);
5159 break;
5160 }
5161
5162 case DIOCRTSTADDRS: {
5163 struct pfioc_table *io = (struct pfioc_table *)addr;
5164 struct pfr_addr *pfras;
5165 size_t totlen;
5166
5167 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5168 error = ENODEV;
5169 goto fail;
5170 }
5171 if (io->pfrio_size < 0 ||
5172 io->pfrio_size > pf_ioctl_maxcount ||
5173 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5174 error = EINVAL;
5175 goto fail;
5176 }
5177 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5178 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5179 M_PF, M_WAITOK);
5180 error = copyin(io->pfrio_buffer, pfras, totlen);
5181 if (error) {
5182 free(pfras, M_PF);
5183 goto fail;
5184 }
5185 PF_RULES_RLOCK();
5186 error = pfr_tst_addrs(&io->pfrio_table, pfras,
5187 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5188 PFR_FLAG_USERIOCTL);
5189 PF_RULES_RUNLOCK();
5190 if (error == 0)
5191 error = copyout(pfras, io->pfrio_buffer, totlen);
5192 free(pfras, M_PF);
5193 break;
5194 }
5195
5196 case DIOCRINADEFINE: {
5197 struct pfioc_table *io = (struct pfioc_table *)addr;
5198 struct pfr_addr *pfras;
5199 size_t totlen;
5200
5201 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5202 error = ENODEV;
5203 goto fail;
5204 }
5205 if (io->pfrio_size < 0 ||
5206 io->pfrio_size > pf_ioctl_maxcount ||
5207 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5208 error = EINVAL;
5209 goto fail;
5210 }
5211 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5212 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5213 M_PF, M_WAITOK);
5214 error = copyin(io->pfrio_buffer, pfras, totlen);
5215 if (error) {
5216 free(pfras, M_PF);
5217 goto fail;
5218 }
5219 PF_RULES_WLOCK();
5220 error = pfr_ina_define(&io->pfrio_table, pfras,
5221 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5222 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5223 PF_RULES_WUNLOCK();
5224 free(pfras, M_PF);
5225 break;
5226 }
5227
5228 case DIOCOSFPADD: {
5229 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5230 PF_RULES_WLOCK();
5231 error = pf_osfp_add(io);
5232 PF_RULES_WUNLOCK();
5233 break;
5234 }
5235
5236 case DIOCOSFPGET: {
5237 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5238 PF_RULES_RLOCK();
5239 error = pf_osfp_get(io);
5240 PF_RULES_RUNLOCK();
5241 break;
5242 }
5243
5244 case DIOCXBEGIN: {
5245 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5246 struct pfioc_trans_e *ioes, *ioe;
5247 size_t totlen;
5248 int i;
5249
5250 if (io->esize != sizeof(*ioe)) {
5251 error = ENODEV;
5252 goto fail;
5253 }
5254 if (io->size < 0 ||
5255 io->size > pf_ioctl_maxcount ||
5256 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5257 error = EINVAL;
5258 goto fail;
5259 }
5260 totlen = sizeof(struct pfioc_trans_e) * io->size;
5261 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5262 M_PF, M_WAITOK);
5263 error = copyin(io->array, ioes, totlen);
5264 if (error) {
5265 free(ioes, M_PF);
5266 goto fail;
5267 }
5268 PF_RULES_WLOCK();
5269 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5270 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5271 switch (ioe->rs_num) {
5272 case PF_RULESET_ETH:
5273 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5274 PF_RULES_WUNLOCK();
5275 free(ioes, M_PF);
5276 goto fail;
5277 }
5278 break;
5279 #ifdef ALTQ
5280 case PF_RULESET_ALTQ:
5281 if (ioe->anchor[0]) {
5282 PF_RULES_WUNLOCK();
5283 free(ioes, M_PF);
5284 error = EINVAL;
5285 goto fail;
5286 }
5287 if ((error = pf_begin_altq(&ioe->ticket))) {
5288 PF_RULES_WUNLOCK();
5289 free(ioes, M_PF);
5290 goto fail;
5291 }
5292 break;
5293 #endif /* ALTQ */
5294 case PF_RULESET_TABLE:
5295 {
5296 struct pfr_table table;
5297
5298 bzero(&table, sizeof(table));
5299 strlcpy(table.pfrt_anchor, ioe->anchor,
5300 sizeof(table.pfrt_anchor));
5301 if ((error = pfr_ina_begin(&table,
5302 &ioe->ticket, NULL, 0))) {
5303 PF_RULES_WUNLOCK();
5304 free(ioes, M_PF);
5305 goto fail;
5306 }
5307 break;
5308 }
5309 default:
5310 if ((error = pf_begin_rules(&ioe->ticket,
5311 ioe->rs_num, ioe->anchor))) {
5312 PF_RULES_WUNLOCK();
5313 free(ioes, M_PF);
5314 goto fail;
5315 }
5316 break;
5317 }
5318 }
5319 PF_RULES_WUNLOCK();
5320 error = copyout(ioes, io->array, totlen);
5321 free(ioes, M_PF);
5322 break;
5323 }
5324
5325 case DIOCXROLLBACK: {
5326 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5327 struct pfioc_trans_e *ioe, *ioes;
5328 size_t totlen;
5329 int i;
5330
5331 if (io->esize != sizeof(*ioe)) {
5332 error = ENODEV;
5333 goto fail;
5334 }
5335 if (io->size < 0 ||
5336 io->size > pf_ioctl_maxcount ||
5337 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5338 error = EINVAL;
5339 goto fail;
5340 }
5341 totlen = sizeof(struct pfioc_trans_e) * io->size;
5342 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5343 M_PF, M_WAITOK);
5344 error = copyin(io->array, ioes, totlen);
5345 if (error) {
5346 free(ioes, M_PF);
5347 goto fail;
5348 }
5349 PF_RULES_WLOCK();
5350 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5351 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5352 switch (ioe->rs_num) {
5353 case PF_RULESET_ETH:
5354 if ((error = pf_rollback_eth(ioe->ticket,
5355 ioe->anchor))) {
5356 PF_RULES_WUNLOCK();
5357 free(ioes, M_PF);
5358 goto fail; /* really bad */
5359 }
5360 break;
5361 #ifdef ALTQ
5362 case PF_RULESET_ALTQ:
5363 if (ioe->anchor[0]) {
5364 PF_RULES_WUNLOCK();
5365 free(ioes, M_PF);
5366 error = EINVAL;
5367 goto fail;
5368 }
5369 if ((error = pf_rollback_altq(ioe->ticket))) {
5370 PF_RULES_WUNLOCK();
5371 free(ioes, M_PF);
5372 goto fail; /* really bad */
5373 }
5374 break;
5375 #endif /* ALTQ */
5376 case PF_RULESET_TABLE:
5377 {
5378 struct pfr_table table;
5379
5380 bzero(&table, sizeof(table));
5381 strlcpy(table.pfrt_anchor, ioe->anchor,
5382 sizeof(table.pfrt_anchor));
5383 if ((error = pfr_ina_rollback(&table,
5384 ioe->ticket, NULL, 0))) {
5385 PF_RULES_WUNLOCK();
5386 free(ioes, M_PF);
5387 goto fail; /* really bad */
5388 }
5389 break;
5390 }
5391 default:
5392 if ((error = pf_rollback_rules(ioe->ticket,
5393 ioe->rs_num, ioe->anchor))) {
5394 PF_RULES_WUNLOCK();
5395 free(ioes, M_PF);
5396 goto fail; /* really bad */
5397 }
5398 break;
5399 }
5400 }
5401 PF_RULES_WUNLOCK();
5402 free(ioes, M_PF);
5403 break;
5404 }
5405
5406 case DIOCXCOMMIT: {
5407 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5408 struct pfioc_trans_e *ioe, *ioes;
5409 struct pf_kruleset *rs;
5410 struct pf_keth_ruleset *ers;
5411 size_t totlen;
5412 int i;
5413
5414 if (io->esize != sizeof(*ioe)) {
5415 error = ENODEV;
5416 goto fail;
5417 }
5418
5419 if (io->size < 0 ||
5420 io->size > pf_ioctl_maxcount ||
5421 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5422 error = EINVAL;
5423 goto fail;
5424 }
5425
5426 totlen = sizeof(struct pfioc_trans_e) * io->size;
5427 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5428 M_PF, M_WAITOK);
5429 error = copyin(io->array, ioes, totlen);
5430 if (error) {
5431 free(ioes, M_PF);
5432 goto fail;
5433 }
5434 PF_RULES_WLOCK();
5435 /* First makes sure everything will succeed. */
5436 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5437 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5438 switch (ioe->rs_num) {
5439 case PF_RULESET_ETH:
5440 ers = pf_find_keth_ruleset(ioe->anchor);
5441 if (ers == NULL || ioe->ticket == 0 ||
5442 ioe->ticket != ers->inactive.ticket) {
5443 PF_RULES_WUNLOCK();
5444 free(ioes, M_PF);
5445 error = EINVAL;
5446 goto fail;
5447 }
5448 break;
5449 #ifdef ALTQ
5450 case PF_RULESET_ALTQ:
5451 if (ioe->anchor[0]) {
5452 PF_RULES_WUNLOCK();
5453 free(ioes, M_PF);
5454 error = EINVAL;
5455 goto fail;
5456 }
5457 if (!V_altqs_inactive_open || ioe->ticket !=
5458 V_ticket_altqs_inactive) {
5459 PF_RULES_WUNLOCK();
5460 free(ioes, M_PF);
5461 error = EBUSY;
5462 goto fail;
5463 }
5464 break;
5465 #endif /* ALTQ */
5466 case PF_RULESET_TABLE:
5467 rs = pf_find_kruleset(ioe->anchor);
5468 if (rs == NULL || !rs->topen || ioe->ticket !=
5469 rs->tticket) {
5470 PF_RULES_WUNLOCK();
5471 free(ioes, M_PF);
5472 error = EBUSY;
5473 goto fail;
5474 }
5475 break;
5476 default:
5477 if (ioe->rs_num < 0 || ioe->rs_num >=
5478 PF_RULESET_MAX) {
5479 PF_RULES_WUNLOCK();
5480 free(ioes, M_PF);
5481 error = EINVAL;
5482 goto fail;
5483 }
5484 rs = pf_find_kruleset(ioe->anchor);
5485 if (rs == NULL ||
5486 !rs->rules[ioe->rs_num].inactive.open ||
5487 rs->rules[ioe->rs_num].inactive.ticket !=
5488 ioe->ticket) {
5489 PF_RULES_WUNLOCK();
5490 free(ioes, M_PF);
5491 error = EBUSY;
5492 goto fail;
5493 }
5494 break;
5495 }
5496 }
5497 /* Now do the commit - no errors should happen here. */
5498 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5499 switch (ioe->rs_num) {
5500 case PF_RULESET_ETH:
5501 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5502 PF_RULES_WUNLOCK();
5503 free(ioes, M_PF);
5504 goto fail; /* really bad */
5505 }
5506 break;
5507 #ifdef ALTQ
5508 case PF_RULESET_ALTQ:
5509 if ((error = pf_commit_altq(ioe->ticket))) {
5510 PF_RULES_WUNLOCK();
5511 free(ioes, M_PF);
5512 goto fail; /* really bad */
5513 }
5514 break;
5515 #endif /* ALTQ */
5516 case PF_RULESET_TABLE:
5517 {
5518 struct pfr_table table;
5519
5520 bzero(&table, sizeof(table));
5521 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
5522 sizeof(table.pfrt_anchor));
5523 if ((error = pfr_ina_commit(&table,
5524 ioe->ticket, NULL, NULL, 0))) {
5525 PF_RULES_WUNLOCK();
5526 free(ioes, M_PF);
5527 goto fail; /* really bad */
5528 }
5529 break;
5530 }
5531 default:
5532 if ((error = pf_commit_rules(ioe->ticket,
5533 ioe->rs_num, ioe->anchor))) {
5534 PF_RULES_WUNLOCK();
5535 free(ioes, M_PF);
5536 goto fail; /* really bad */
5537 }
5538 break;
5539 }
5540 }
5541 PF_RULES_WUNLOCK();
5542
5543 /* Only hook into EtherNet taffic if we've got rules for it. */
5544 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5545 hook_pf_eth();
5546 else
5547 dehook_pf_eth();
5548
5549 free(ioes, M_PF);
5550 break;
5551 }
5552
5553 case DIOCGETSRCNODES: {
5554 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
5555 struct pf_srchash *sh;
5556 struct pf_ksrc_node *n;
5557 struct pf_src_node *p, *pstore;
5558 uint32_t i, nr = 0;
5559
5560 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5561 i++, sh++) {
5562 PF_HASHROW_LOCK(sh);
5563 LIST_FOREACH(n, &sh->nodes, entry)
5564 nr++;
5565 PF_HASHROW_UNLOCK(sh);
5566 }
5567
5568 psn->psn_len = min(psn->psn_len,
5569 sizeof(struct pf_src_node) * nr);
5570
5571 if (psn->psn_len == 0) {
5572 psn->psn_len = sizeof(struct pf_src_node) * nr;
5573 goto fail;
5574 }
5575
5576 nr = 0;
5577
5578 p = pstore = malloc(psn->psn_len, M_PF, M_WAITOK | M_ZERO);
5579 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5580 i++, sh++) {
5581 PF_HASHROW_LOCK(sh);
5582 LIST_FOREACH(n, &sh->nodes, entry) {
5583
5584 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5585 break;
5586
5587 pf_src_node_copy(n, p);
5588
5589 p++;
5590 nr++;
5591 }
5592 PF_HASHROW_UNLOCK(sh);
5593 }
5594 error = copyout(pstore, psn->psn_src_nodes,
5595 sizeof(struct pf_src_node) * nr);
5596 if (error) {
5597 free(pstore, M_PF);
5598 goto fail;
5599 }
5600 psn->psn_len = sizeof(struct pf_src_node) * nr;
5601 free(pstore, M_PF);
5602 break;
5603 }
5604
5605 case DIOCCLRSRCNODES: {
5606 pf_kill_srcnodes(NULL);
5607 break;
5608 }
5609
5610 case DIOCKILLSRCNODES:
5611 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5612 break;
5613
5614 #ifdef COMPAT_FREEBSD13
5615 case DIOCKEEPCOUNTERS_FREEBSD13:
5616 #endif
5617 case DIOCKEEPCOUNTERS:
5618 error = pf_keepcounters((struct pfioc_nv *)addr);
5619 break;
5620
5621 case DIOCGETSYNCOOKIES:
5622 error = pf_get_syncookies((struct pfioc_nv *)addr);
5623 break;
5624
5625 case DIOCSETSYNCOOKIES:
5626 error = pf_set_syncookies((struct pfioc_nv *)addr);
5627 break;
5628
5629 case DIOCSETHOSTID: {
5630 u_int32_t *hostid = (u_int32_t *)addr;
5631
5632 PF_RULES_WLOCK();
5633 if (*hostid == 0)
5634 V_pf_status.hostid = arc4random();
5635 else
5636 V_pf_status.hostid = *hostid;
5637 PF_RULES_WUNLOCK();
5638 break;
5639 }
5640
5641 case DIOCOSFPFLUSH:
5642 PF_RULES_WLOCK();
5643 pf_osfp_flush();
5644 PF_RULES_WUNLOCK();
5645 break;
5646
5647 case DIOCIGETIFACES: {
5648 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5649 struct pfi_kif *ifstore;
5650 size_t bufsiz;
5651
5652 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5653 error = ENODEV;
5654 goto fail;
5655 }
5656
5657 if (io->pfiio_size < 0 ||
5658 io->pfiio_size > pf_ioctl_maxcount ||
5659 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5660 error = EINVAL;
5661 goto fail;
5662 }
5663
5664 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5665
5666 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5667 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5668 M_PF, M_WAITOK | M_ZERO);
5669
5670 PF_RULES_RLOCK();
5671 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5672 PF_RULES_RUNLOCK();
5673 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5674 free(ifstore, M_PF);
5675 break;
5676 }
5677
5678 case DIOCSETIFFLAG: {
5679 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5680
5681 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5682
5683 PF_RULES_WLOCK();
5684 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5685 PF_RULES_WUNLOCK();
5686 break;
5687 }
5688
5689 case DIOCCLRIFFLAG: {
5690 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5691
5692 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5693
5694 PF_RULES_WLOCK();
5695 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5696 PF_RULES_WUNLOCK();
5697 break;
5698 }
5699
5700 case DIOCSETREASS: {
5701 u_int32_t *reass = (u_int32_t *)addr;
5702
5703 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5704 /* Removal of DF flag without reassembly enabled is not a
5705 * valid combination. Disable reassembly in such case. */
5706 if (!(V_pf_status.reass & PF_REASS_ENABLED))
5707 V_pf_status.reass = 0;
5708 break;
5709 }
5710
5711 default:
5712 error = ENODEV;
5713 break;
5714 }
5715 fail:
5716 CURVNET_RESTORE();
5717
5718 #undef ERROUT_IOCTL
5719
5720 return (error);
5721 }
5722
5723 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5724 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5725 {
5726 bzero(sp, sizeof(union pfsync_state_union));
5727
5728 /* copy from state key */
5729 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5730 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5731 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5732 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5733 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5734 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5735 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5736 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5737 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5738 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5739
5740 /* copy from state */
5741 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5742 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5743 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5744 sp->pfs_1301.expire = pf_state_expires(st);
5745 if (sp->pfs_1301.expire <= time_uptime)
5746 sp->pfs_1301.expire = htonl(0);
5747 else
5748 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5749
5750 sp->pfs_1301.direction = st->direction;
5751 sp->pfs_1301.log = st->act.log;
5752 sp->pfs_1301.timeout = st->timeout;
5753
5754 switch (msg_version) {
5755 case PFSYNC_MSG_VERSION_1301:
5756 sp->pfs_1301.state_flags = st->state_flags;
5757 break;
5758 case PFSYNC_MSG_VERSION_1400:
5759 sp->pfs_1400.state_flags = htons(st->state_flags);
5760 sp->pfs_1400.qid = htons(st->act.qid);
5761 sp->pfs_1400.pqid = htons(st->act.pqid);
5762 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5763 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5764 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5765 sp->pfs_1400.min_ttl = st->act.min_ttl;
5766 sp->pfs_1400.set_tos = st->act.set_tos;
5767 sp->pfs_1400.max_mss = htons(st->act.max_mss);
5768 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5769 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5770 sp->pfs_1400.rt = st->act.rt;
5771 if (st->act.rt_kif)
5772 strlcpy(sp->pfs_1400.rt_ifname,
5773 st->act.rt_kif->pfik_name,
5774 sizeof(sp->pfs_1400.rt_ifname));
5775 break;
5776 default:
5777 panic("%s: Unsupported pfsync_msg_version %d",
5778 __func__, msg_version);
5779 }
5780
5781 /*
5782 * XXX Why do we bother pfsyncing source node information if source
5783 * nodes are not synced? Showing users that there is source tracking
5784 * when there is none seems useless.
5785 */
5786 if (st->sns[PF_SN_LIMIT] != NULL)
5787 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5788 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
5789 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5790
5791 sp->pfs_1301.id = st->id;
5792 sp->pfs_1301.creatorid = st->creatorid;
5793 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5794 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5795
5796 if (st->rule == NULL)
5797 sp->pfs_1301.rule = htonl(-1);
5798 else
5799 sp->pfs_1301.rule = htonl(st->rule->nr);
5800 if (st->anchor == NULL)
5801 sp->pfs_1301.anchor = htonl(-1);
5802 else
5803 sp->pfs_1301.anchor = htonl(st->anchor->nr);
5804 if (st->nat_rule == NULL)
5805 sp->pfs_1301.nat_rule = htonl(-1);
5806 else
5807 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
5808
5809 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5810 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5811 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5812 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5813 }
5814
5815 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5816 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5817 {
5818 bzero(sp, sizeof(*sp));
5819
5820 sp->version = PF_STATE_VERSION;
5821
5822 /* copy from state key */
5823 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5824 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5825 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5826 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5827 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5828 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5829 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5830 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5831 sp->proto = st->key[PF_SK_WIRE]->proto;
5832 sp->af = st->key[PF_SK_WIRE]->af;
5833
5834 /* copy from state */
5835 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5836 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5837 sizeof(sp->orig_ifname));
5838 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr));
5839 sp->creation = htonl(time_uptime - (st->creation / 1000));
5840 sp->expire = pf_state_expires(st);
5841 if (sp->expire <= time_uptime)
5842 sp->expire = htonl(0);
5843 else
5844 sp->expire = htonl(sp->expire - time_uptime);
5845
5846 sp->direction = st->direction;
5847 sp->log = st->act.log;
5848 sp->timeout = st->timeout;
5849 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5850 sp->state_flags_compat = st->state_flags;
5851 sp->state_flags = htons(st->state_flags);
5852 if (st->sns[PF_SN_LIMIT] != NULL)
5853 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5854 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL)
5855 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5856 sp->id = st->id;
5857 sp->creatorid = st->creatorid;
5858 pf_state_peer_hton(&st->src, &sp->src);
5859 pf_state_peer_hton(&st->dst, &sp->dst);
5860
5861 if (st->rule == NULL)
5862 sp->rule = htonl(-1);
5863 else
5864 sp->rule = htonl(st->rule->nr);
5865 if (st->anchor == NULL)
5866 sp->anchor = htonl(-1);
5867 else
5868 sp->anchor = htonl(st->anchor->nr);
5869 if (st->nat_rule == NULL)
5870 sp->nat_rule = htonl(-1);
5871 else
5872 sp->nat_rule = htonl(st->nat_rule->nr);
5873
5874 sp->packets[0] = st->packets[0];
5875 sp->packets[1] = st->packets[1];
5876 sp->bytes[0] = st->bytes[0];
5877 sp->bytes[1] = st->bytes[1];
5878
5879 sp->qid = htons(st->act.qid);
5880 sp->pqid = htons(st->act.pqid);
5881 sp->dnpipe = htons(st->act.dnpipe);
5882 sp->dnrpipe = htons(st->act.dnrpipe);
5883 sp->rtableid = htonl(st->act.rtableid);
5884 sp->min_ttl = st->act.min_ttl;
5885 sp->set_tos = st->act.set_tos;
5886 sp->max_mss = htons(st->act.max_mss);
5887 sp->rt = st->act.rt;
5888 if (st->act.rt_kif)
5889 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
5890 sizeof(sp->rt_ifname));
5891 sp->set_prio[0] = st->act.set_prio[0];
5892 sp->set_prio[1] = st->act.set_prio[1];
5893
5894 }
5895
5896 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5897 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5898 {
5899 struct pfr_ktable *kt;
5900
5901 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5902
5903 kt = aw->p.tbl;
5904 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5905 kt = kt->pfrkt_root;
5906 aw->p.tbl = NULL;
5907 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5908 kt->pfrkt_cnt : -1;
5909 }
5910
5911 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5912 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5913 size_t number, char **names)
5914 {
5915 nvlist_t *nvc;
5916
5917 nvc = nvlist_create(0);
5918 if (nvc == NULL)
5919 return (ENOMEM);
5920
5921 for (int i = 0; i < number; i++) {
5922 nvlist_append_number_array(nvc, "counters",
5923 counter_u64_fetch(counters[i]));
5924 nvlist_append_string_array(nvc, "names",
5925 names[i]);
5926 nvlist_append_number_array(nvc, "ids",
5927 i);
5928 }
5929 nvlist_add_nvlist(nvl, name, nvc);
5930 nvlist_destroy(nvc);
5931
5932 return (0);
5933 }
5934
5935 static int
pf_getstatus(struct pfioc_nv * nv)5936 pf_getstatus(struct pfioc_nv *nv)
5937 {
5938 nvlist_t *nvl = NULL, *nvc = NULL;
5939 void *nvlpacked = NULL;
5940 int error;
5941 struct pf_status s;
5942 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5943 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5944 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5945 time_t since;
5946
5947 PF_RULES_RLOCK_TRACKER;
5948
5949 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
5950
5951 PF_RULES_RLOCK();
5952
5953 nvl = nvlist_create(0);
5954 if (nvl == NULL)
5955 ERROUT(ENOMEM);
5956
5957 since = time_second - (time_uptime - V_pf_status.since);
5958
5959 nvlist_add_bool(nvl, "running", V_pf_status.running);
5960 nvlist_add_number(nvl, "since", since);
5961 nvlist_add_number(nvl, "debug", V_pf_status.debug);
5962 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5963 nvlist_add_number(nvl, "states", V_pf_status.states);
5964 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5965 nvlist_add_number(nvl, "reass", V_pf_status.reass);
5966 nvlist_add_bool(nvl, "syncookies_active",
5967 V_pf_status.syncookies_active);
5968 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5969
5970 /* counters */
5971 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5972 PFRES_MAX, pf_reasons);
5973 if (error != 0)
5974 ERROUT(error);
5975
5976 /* lcounters */
5977 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5978 KLCNT_MAX, pf_lcounter);
5979 if (error != 0)
5980 ERROUT(error);
5981
5982 /* fcounters */
5983 nvc = nvlist_create(0);
5984 if (nvc == NULL)
5985 ERROUT(ENOMEM);
5986
5987 for (int i = 0; i < FCNT_MAX; i++) {
5988 nvlist_append_number_array(nvc, "counters",
5989 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5990 nvlist_append_string_array(nvc, "names",
5991 pf_fcounter[i]);
5992 nvlist_append_number_array(nvc, "ids",
5993 i);
5994 }
5995 nvlist_add_nvlist(nvl, "fcounters", nvc);
5996 nvlist_destroy(nvc);
5997 nvc = NULL;
5998
5999 /* scounters */
6000 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
6001 SCNT_MAX, pf_fcounter);
6002 if (error != 0)
6003 ERROUT(error);
6004
6005 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
6006 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
6007 PF_MD5_DIGEST_LENGTH);
6008
6009 pfi_update_status(V_pf_status.ifname, &s);
6010
6011 /* pcounters / bcounters */
6012 for (int i = 0; i < 2; i++) {
6013 for (int j = 0; j < 2; j++) {
6014 for (int k = 0; k < 2; k++) {
6015 nvlist_append_number_array(nvl, "pcounters",
6016 s.pcounters[i][j][k]);
6017 }
6018 nvlist_append_number_array(nvl, "bcounters",
6019 s.bcounters[i][j]);
6020 }
6021 }
6022
6023 nvlpacked = nvlist_pack(nvl, &nv->len);
6024 if (nvlpacked == NULL)
6025 ERROUT(ENOMEM);
6026
6027 if (nv->size == 0)
6028 ERROUT(0);
6029 else if (nv->size < nv->len)
6030 ERROUT(ENOSPC);
6031
6032 PF_RULES_RUNLOCK();
6033 error = copyout(nvlpacked, nv->data, nv->len);
6034 goto done;
6035
6036 #undef ERROUT
6037 errout:
6038 PF_RULES_RUNLOCK();
6039 done:
6040 free(nvlpacked, M_NVLIST);
6041 nvlist_destroy(nvc);
6042 nvlist_destroy(nvl);
6043
6044 return (error);
6045 }
6046
6047 /*
6048 * XXX - Check for version mismatch!!!
6049 */
6050 static void
pf_clear_all_states(void)6051 pf_clear_all_states(void)
6052 {
6053 struct epoch_tracker et;
6054 struct pf_kstate *s;
6055 u_int i;
6056
6057 NET_EPOCH_ENTER(et);
6058 for (i = 0; i <= V_pf_hashmask; i++) {
6059 struct pf_idhash *ih = &V_pf_idhash[i];
6060 relock:
6061 PF_HASHROW_LOCK(ih);
6062 LIST_FOREACH(s, &ih->states, entry) {
6063 s->timeout = PFTM_PURGE;
6064 /* Don't send out individual delete messages. */
6065 s->state_flags |= PFSTATE_NOSYNC;
6066 pf_remove_state(s);
6067 goto relock;
6068 }
6069 PF_HASHROW_UNLOCK(ih);
6070 }
6071 NET_EPOCH_EXIT(et);
6072 }
6073
6074 static int
pf_clear_tables(void)6075 pf_clear_tables(void)
6076 {
6077 struct pfioc_table io;
6078 int error;
6079
6080 bzero(&io, sizeof(io));
6081 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
6082
6083 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
6084 io.pfrio_flags);
6085
6086 return (error);
6087 }
6088
6089 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)6090 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
6091 {
6092 struct pf_ksrc_node_list kill;
6093 u_int killed;
6094
6095 LIST_INIT(&kill);
6096 for (int i = 0; i <= V_pf_srchashmask; i++) {
6097 struct pf_srchash *sh = &V_pf_srchash[i];
6098 struct pf_ksrc_node *sn, *tmp;
6099
6100 PF_HASHROW_LOCK(sh);
6101 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
6102 if (psnk == NULL ||
6103 (pf_match_addr(psnk->psnk_src.neg,
6104 &psnk->psnk_src.addr.v.a.addr,
6105 &psnk->psnk_src.addr.v.a.mask,
6106 &sn->addr, sn->af) &&
6107 pf_match_addr(psnk->psnk_dst.neg,
6108 &psnk->psnk_dst.addr.v.a.addr,
6109 &psnk->psnk_dst.addr.v.a.mask,
6110 &sn->raddr, sn->af))) {
6111 pf_unlink_src_node(sn);
6112 LIST_INSERT_HEAD(&kill, sn, entry);
6113 sn->expire = 1;
6114 }
6115 PF_HASHROW_UNLOCK(sh);
6116 }
6117
6118 for (int i = 0; i <= V_pf_hashmask; i++) {
6119 struct pf_idhash *ih = &V_pf_idhash[i];
6120 struct pf_kstate *s;
6121
6122 PF_HASHROW_LOCK(ih);
6123 LIST_FOREACH(s, &ih->states, entry) {
6124 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX;
6125 sn_type++) {
6126 if (s->sns[sn_type] &&
6127 s->sns[sn_type]->expire == 1) {
6128 s->sns[sn_type] = NULL;
6129 }
6130 }
6131 }
6132 PF_HASHROW_UNLOCK(ih);
6133 }
6134
6135 killed = pf_free_src_nodes(&kill);
6136
6137 if (psnk != NULL)
6138 psnk->psnk_killed = killed;
6139 }
6140
6141 static int
pf_keepcounters(struct pfioc_nv * nv)6142 pf_keepcounters(struct pfioc_nv *nv)
6143 {
6144 nvlist_t *nvl = NULL;
6145 void *nvlpacked = NULL;
6146 int error = 0;
6147
6148 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6149
6150 if (nv->len > pf_ioctl_maxcount)
6151 ERROUT(ENOMEM);
6152
6153 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6154 error = copyin(nv->data, nvlpacked, nv->len);
6155 if (error)
6156 ERROUT(error);
6157
6158 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6159 if (nvl == NULL)
6160 ERROUT(EBADMSG);
6161
6162 if (! nvlist_exists_bool(nvl, "keep_counters"))
6163 ERROUT(EBADMSG);
6164
6165 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6166
6167 on_error:
6168 nvlist_destroy(nvl);
6169 free(nvlpacked, M_NVLIST);
6170 return (error);
6171 }
6172
6173 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6174 pf_clear_states(const struct pf_kstate_kill *kill)
6175 {
6176 struct pf_state_key_cmp match_key;
6177 struct pf_kstate *s;
6178 struct pfi_kkif *kif;
6179 int idx;
6180 unsigned int killed = 0, dir;
6181
6182 NET_EPOCH_ASSERT();
6183
6184 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6185 struct pf_idhash *ih = &V_pf_idhash[i];
6186
6187 relock_DIOCCLRSTATES:
6188 PF_HASHROW_LOCK(ih);
6189 LIST_FOREACH(s, &ih->states, entry) {
6190 /* For floating states look at the original kif. */
6191 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6192
6193 if (kill->psk_ifname[0] &&
6194 strcmp(kill->psk_ifname,
6195 kif->pfik_name))
6196 continue;
6197
6198 if (kill->psk_kill_match) {
6199 bzero(&match_key, sizeof(match_key));
6200
6201 if (s->direction == PF_OUT) {
6202 dir = PF_IN;
6203 idx = PF_SK_STACK;
6204 } else {
6205 dir = PF_OUT;
6206 idx = PF_SK_WIRE;
6207 }
6208
6209 match_key.af = s->key[idx]->af;
6210 match_key.proto = s->key[idx]->proto;
6211 pf_addrcpy(&match_key.addr[0],
6212 &s->key[idx]->addr[1], match_key.af);
6213 match_key.port[0] = s->key[idx]->port[1];
6214 pf_addrcpy(&match_key.addr[1],
6215 &s->key[idx]->addr[0], match_key.af);
6216 match_key.port[1] = s->key[idx]->port[0];
6217 }
6218
6219 /*
6220 * Don't send out individual
6221 * delete messages.
6222 */
6223 s->state_flags |= PFSTATE_NOSYNC;
6224 pf_remove_state(s);
6225 killed++;
6226
6227 if (kill->psk_kill_match)
6228 killed += pf_kill_matching_state(&match_key,
6229 dir);
6230
6231 goto relock_DIOCCLRSTATES;
6232 }
6233 PF_HASHROW_UNLOCK(ih);
6234 }
6235
6236 if (V_pfsync_clear_states_ptr != NULL)
6237 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6238
6239 return (killed);
6240 }
6241
6242 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6243 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6244 {
6245 struct pf_kstate *s;
6246
6247 NET_EPOCH_ASSERT();
6248 if (kill->psk_pfcmp.id) {
6249 if (kill->psk_pfcmp.creatorid == 0)
6250 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6251 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6252 kill->psk_pfcmp.creatorid))) {
6253 pf_remove_state(s);
6254 *killed = 1;
6255 }
6256 return;
6257 }
6258
6259 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6260 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6261 }
6262
6263 static int
pf_killstates_nv(struct pfioc_nv * nv)6264 pf_killstates_nv(struct pfioc_nv *nv)
6265 {
6266 struct pf_kstate_kill kill;
6267 struct epoch_tracker et;
6268 nvlist_t *nvl = NULL;
6269 void *nvlpacked = NULL;
6270 int error = 0;
6271 unsigned int killed = 0;
6272
6273 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6274
6275 if (nv->len > pf_ioctl_maxcount)
6276 ERROUT(ENOMEM);
6277
6278 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6279 error = copyin(nv->data, nvlpacked, nv->len);
6280 if (error)
6281 ERROUT(error);
6282
6283 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6284 if (nvl == NULL)
6285 ERROUT(EBADMSG);
6286
6287 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6288 if (error)
6289 ERROUT(error);
6290
6291 NET_EPOCH_ENTER(et);
6292 pf_killstates(&kill, &killed);
6293 NET_EPOCH_EXIT(et);
6294
6295 free(nvlpacked, M_NVLIST);
6296 nvlpacked = NULL;
6297 nvlist_destroy(nvl);
6298 nvl = nvlist_create(0);
6299 if (nvl == NULL)
6300 ERROUT(ENOMEM);
6301
6302 nvlist_add_number(nvl, "killed", killed);
6303
6304 nvlpacked = nvlist_pack(nvl, &nv->len);
6305 if (nvlpacked == NULL)
6306 ERROUT(ENOMEM);
6307
6308 if (nv->size == 0)
6309 ERROUT(0);
6310 else if (nv->size < nv->len)
6311 ERROUT(ENOSPC);
6312
6313 error = copyout(nvlpacked, nv->data, nv->len);
6314
6315 on_error:
6316 nvlist_destroy(nvl);
6317 free(nvlpacked, M_NVLIST);
6318 return (error);
6319 }
6320
6321 static int
pf_clearstates_nv(struct pfioc_nv * nv)6322 pf_clearstates_nv(struct pfioc_nv *nv)
6323 {
6324 struct pf_kstate_kill kill;
6325 struct epoch_tracker et;
6326 nvlist_t *nvl = NULL;
6327 void *nvlpacked = NULL;
6328 int error = 0;
6329 unsigned int killed;
6330
6331 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6332
6333 if (nv->len > pf_ioctl_maxcount)
6334 ERROUT(ENOMEM);
6335
6336 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6337 error = copyin(nv->data, nvlpacked, nv->len);
6338 if (error)
6339 ERROUT(error);
6340
6341 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6342 if (nvl == NULL)
6343 ERROUT(EBADMSG);
6344
6345 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6346 if (error)
6347 ERROUT(error);
6348
6349 NET_EPOCH_ENTER(et);
6350 killed = pf_clear_states(&kill);
6351 NET_EPOCH_EXIT(et);
6352
6353 free(nvlpacked, M_NVLIST);
6354 nvlpacked = NULL;
6355 nvlist_destroy(nvl);
6356 nvl = nvlist_create(0);
6357 if (nvl == NULL)
6358 ERROUT(ENOMEM);
6359
6360 nvlist_add_number(nvl, "killed", killed);
6361
6362 nvlpacked = nvlist_pack(nvl, &nv->len);
6363 if (nvlpacked == NULL)
6364 ERROUT(ENOMEM);
6365
6366 if (nv->size == 0)
6367 ERROUT(0);
6368 else if (nv->size < nv->len)
6369 ERROUT(ENOSPC);
6370
6371 error = copyout(nvlpacked, nv->data, nv->len);
6372
6373 #undef ERROUT
6374 on_error:
6375 nvlist_destroy(nvl);
6376 free(nvlpacked, M_NVLIST);
6377 return (error);
6378 }
6379
6380 static int
pf_getstate(struct pfioc_nv * nv)6381 pf_getstate(struct pfioc_nv *nv)
6382 {
6383 nvlist_t *nvl = NULL, *nvls;
6384 void *nvlpacked = NULL;
6385 struct pf_kstate *s = NULL;
6386 int error = 0;
6387 uint64_t id, creatorid;
6388
6389 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6390
6391 if (nv->len > pf_ioctl_maxcount)
6392 ERROUT(ENOMEM);
6393
6394 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6395 error = copyin(nv->data, nvlpacked, nv->len);
6396 if (error)
6397 ERROUT(error);
6398
6399 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6400 if (nvl == NULL)
6401 ERROUT(EBADMSG);
6402
6403 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6404 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6405
6406 s = pf_find_state_byid(id, creatorid);
6407 if (s == NULL)
6408 ERROUT(ENOENT);
6409
6410 free(nvlpacked, M_NVLIST);
6411 nvlpacked = NULL;
6412 nvlist_destroy(nvl);
6413 nvl = nvlist_create(0);
6414 if (nvl == NULL)
6415 ERROUT(ENOMEM);
6416
6417 nvls = pf_state_to_nvstate(s);
6418 if (nvls == NULL)
6419 ERROUT(ENOMEM);
6420
6421 nvlist_add_nvlist(nvl, "state", nvls);
6422 nvlist_destroy(nvls);
6423
6424 nvlpacked = nvlist_pack(nvl, &nv->len);
6425 if (nvlpacked == NULL)
6426 ERROUT(ENOMEM);
6427
6428 if (nv->size == 0)
6429 ERROUT(0);
6430 else if (nv->size < nv->len)
6431 ERROUT(ENOSPC);
6432
6433 error = copyout(nvlpacked, nv->data, nv->len);
6434
6435 #undef ERROUT
6436 errout:
6437 if (s != NULL)
6438 PF_STATE_UNLOCK(s);
6439 free(nvlpacked, M_NVLIST);
6440 nvlist_destroy(nvl);
6441 return (error);
6442 }
6443
6444 /*
6445 * XXX - Check for version mismatch!!!
6446 */
6447
6448 /*
6449 * Duplicate pfctl -Fa operation to get rid of as much as we can.
6450 */
6451 static int
shutdown_pf(void)6452 shutdown_pf(void)
6453 {
6454 int error = 0;
6455 u_int32_t t[5];
6456 char nn = '\0';
6457 struct pf_kanchor *anchor, *tmp_anchor;
6458 struct pf_keth_anchor *eth_anchor, *tmp_eth_anchor;
6459 int rs_num;
6460
6461 do {
6462 /* Unlink rules of all user defined anchors */
6463 RB_FOREACH_SAFE(anchor, pf_kanchor_global, &V_pf_anchors,
6464 tmp_anchor) {
6465 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6466 if ((error = pf_begin_rules(&t[rs_num], rs_num,
6467 anchor->path)) != 0) {
6468 DPFPRINTF(PF_DEBUG_MISC, "%s: "
6469 "anchor.path=%s rs_num=%d",
6470 __func__, anchor->path, rs_num);
6471 goto error; /* XXX: rollback? */
6472 }
6473 }
6474 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6475 error = pf_commit_rules(t[rs_num], rs_num,
6476 anchor->path);
6477 MPASS(error == 0);
6478 }
6479 }
6480
6481 /* Unlink rules of all user defined ether anchors */
6482 RB_FOREACH_SAFE(eth_anchor, pf_keth_anchor_global,
6483 &V_pf_keth_anchors, tmp_eth_anchor) {
6484 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6485 != 0) {
6486 DPFPRINTF(PF_DEBUG_MISC, "%s: eth "
6487 "anchor.path=%s", __func__,
6488 eth_anchor->path);
6489 goto error;
6490 }
6491 error = pf_commit_eth(t[0], eth_anchor->path);
6492 MPASS(error == 0);
6493 }
6494
6495 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6496 != 0) {
6497 DPFPRINTF(PF_DEBUG_MISC, "%s: SCRUB", __func__);
6498 break;
6499 }
6500 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6501 != 0) {
6502 DPFPRINTF(PF_DEBUG_MISC, "%s: FILTER", __func__);
6503 break; /* XXX: rollback? */
6504 }
6505 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6506 != 0) {
6507 DPFPRINTF(PF_DEBUG_MISC, "%s: NAT", __func__);
6508 break; /* XXX: rollback? */
6509 }
6510 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6511 != 0) {
6512 DPFPRINTF(PF_DEBUG_MISC, "%s: BINAT", __func__);
6513 break; /* XXX: rollback? */
6514 }
6515 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6516 != 0) {
6517 DPFPRINTF(PF_DEBUG_MISC, "%s: RDR", __func__);
6518 break; /* XXX: rollback? */
6519 }
6520
6521 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6522 MPASS(error == 0);
6523 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6524 MPASS(error == 0);
6525 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6526 MPASS(error == 0);
6527 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6528 MPASS(error == 0);
6529 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6530 MPASS(error == 0);
6531
6532 if ((error = pf_clear_tables()) != 0)
6533 break;
6534
6535 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6536 DPFPRINTF(PF_DEBUG_MISC, "%s: eth", __func__);
6537 break;
6538 }
6539 error = pf_commit_eth(t[0], &nn);
6540 MPASS(error == 0);
6541
6542 #ifdef ALTQ
6543 if ((error = pf_begin_altq(&t[0])) != 0) {
6544 DPFPRINTF(PF_DEBUG_MISC, "%s: ALTQ", __func__);
6545 break;
6546 }
6547 pf_commit_altq(t[0]);
6548 #endif
6549
6550 pf_clear_all_states();
6551
6552 pf_kill_srcnodes(NULL);
6553
6554 /* status does not use malloced mem so no need to cleanup */
6555 /* fingerprints and interfaces have their own cleanup code */
6556 } while(0);
6557
6558 error:
6559 return (error);
6560 }
6561
6562 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6563 pf_check_return(int chk, struct mbuf **m)
6564 {
6565
6566 switch (chk) {
6567 case PF_PASS:
6568 if (*m == NULL)
6569 return (PFIL_CONSUMED);
6570 else
6571 return (PFIL_PASS);
6572 break;
6573 default:
6574 if (*m != NULL) {
6575 m_freem(*m);
6576 *m = NULL;
6577 }
6578 return (PFIL_DROPPED);
6579 }
6580 }
6581
6582 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6583 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6584 void *ruleset __unused, struct inpcb *inp)
6585 {
6586 int chk;
6587
6588 CURVNET_ASSERT_SET();
6589
6590 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6591
6592 return (pf_check_return(chk, m));
6593 }
6594
6595 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6596 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6597 void *ruleset __unused, struct inpcb *inp)
6598 {
6599 int chk;
6600
6601 CURVNET_ASSERT_SET();
6602
6603 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6604
6605 return (pf_check_return(chk, m));
6606 }
6607
6608 #ifdef INET
6609 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6610 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6611 void *ruleset __unused, struct inpcb *inp)
6612 {
6613 int chk;
6614
6615 CURVNET_ASSERT_SET();
6616
6617 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
6618
6619 return (pf_check_return(chk, m));
6620 }
6621
6622 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6623 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6624 void *ruleset __unused, struct inpcb *inp)
6625 {
6626 int chk;
6627
6628 CURVNET_ASSERT_SET();
6629
6630 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
6631
6632 return (pf_check_return(chk, m));
6633 }
6634 #endif
6635
6636 #ifdef INET6
6637 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6638 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6639 void *ruleset __unused, struct inpcb *inp)
6640 {
6641 int chk;
6642
6643 CURVNET_ASSERT_SET();
6644
6645 /*
6646 * In case of loopback traffic IPv6 uses the real interface in
6647 * order to support scoped addresses. In order to support stateful
6648 * filtering we have change this to lo0 as it is the case in IPv4.
6649 */
6650 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6651 m, inp, NULL);
6652
6653 return (pf_check_return(chk, m));
6654 }
6655
6656 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6657 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6658 void *ruleset __unused, struct inpcb *inp)
6659 {
6660 int chk;
6661
6662 CURVNET_ASSERT_SET();
6663
6664 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL);
6665
6666 return (pf_check_return(chk, m));
6667 }
6668 #endif /* INET6 */
6669
6670 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6671 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6672 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
6673 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
6674
6675 #ifdef INET
6676 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6677 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6678 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
6679 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
6680 #endif
6681 #ifdef INET6
6682 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6683 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6684 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
6685 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
6686 #endif
6687
6688 static void
hook_pf_eth(void)6689 hook_pf_eth(void)
6690 {
6691 struct pfil_hook_args pha = {
6692 .pa_version = PFIL_VERSION,
6693 .pa_modname = "pf",
6694 .pa_type = PFIL_TYPE_ETHERNET,
6695 };
6696 struct pfil_link_args pla = {
6697 .pa_version = PFIL_VERSION,
6698 };
6699 int ret __diagused;
6700
6701 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6702 return;
6703
6704 pha.pa_mbuf_chk = pf_eth_check_in;
6705 pha.pa_flags = PFIL_IN;
6706 pha.pa_rulname = "eth-in";
6707 V_pf_eth_in_hook = pfil_add_hook(&pha);
6708 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6709 pla.pa_head = V_link_pfil_head;
6710 pla.pa_hook = V_pf_eth_in_hook;
6711 ret = pfil_link(&pla);
6712 MPASS(ret == 0);
6713 pha.pa_mbuf_chk = pf_eth_check_out;
6714 pha.pa_flags = PFIL_OUT;
6715 pha.pa_rulname = "eth-out";
6716 V_pf_eth_out_hook = pfil_add_hook(&pha);
6717 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6718 pla.pa_head = V_link_pfil_head;
6719 pla.pa_hook = V_pf_eth_out_hook;
6720 ret = pfil_link(&pla);
6721 MPASS(ret == 0);
6722
6723 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6724 }
6725
6726 static void
hook_pf(void)6727 hook_pf(void)
6728 {
6729 struct pfil_hook_args pha = {
6730 .pa_version = PFIL_VERSION,
6731 .pa_modname = "pf",
6732 };
6733 struct pfil_link_args pla = {
6734 .pa_version = PFIL_VERSION,
6735 };
6736 int ret __diagused;
6737
6738 if (atomic_load_bool(&V_pf_pfil_hooked))
6739 return;
6740
6741 #ifdef INET
6742 pha.pa_type = PFIL_TYPE_IP4;
6743 pha.pa_mbuf_chk = pf_check_in;
6744 pha.pa_flags = PFIL_IN;
6745 pha.pa_rulname = "default-in";
6746 V_pf_ip4_in_hook = pfil_add_hook(&pha);
6747 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6748 pla.pa_head = V_inet_pfil_head;
6749 pla.pa_hook = V_pf_ip4_in_hook;
6750 ret = pfil_link(&pla);
6751 MPASS(ret == 0);
6752 pha.pa_mbuf_chk = pf_check_out;
6753 pha.pa_flags = PFIL_OUT;
6754 pha.pa_rulname = "default-out";
6755 V_pf_ip4_out_hook = pfil_add_hook(&pha);
6756 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6757 pla.pa_head = V_inet_pfil_head;
6758 pla.pa_hook = V_pf_ip4_out_hook;
6759 ret = pfil_link(&pla);
6760 MPASS(ret == 0);
6761 if (V_pf_filter_local) {
6762 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6763 pla.pa_head = V_inet_local_pfil_head;
6764 pla.pa_hook = V_pf_ip4_out_hook;
6765 ret = pfil_link(&pla);
6766 MPASS(ret == 0);
6767 }
6768 #endif
6769 #ifdef INET6
6770 pha.pa_type = PFIL_TYPE_IP6;
6771 pha.pa_mbuf_chk = pf_check6_in;
6772 pha.pa_flags = PFIL_IN;
6773 pha.pa_rulname = "default-in6";
6774 V_pf_ip6_in_hook = pfil_add_hook(&pha);
6775 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6776 pla.pa_head = V_inet6_pfil_head;
6777 pla.pa_hook = V_pf_ip6_in_hook;
6778 ret = pfil_link(&pla);
6779 MPASS(ret == 0);
6780 pha.pa_mbuf_chk = pf_check6_out;
6781 pha.pa_rulname = "default-out6";
6782 pha.pa_flags = PFIL_OUT;
6783 V_pf_ip6_out_hook = pfil_add_hook(&pha);
6784 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6785 pla.pa_head = V_inet6_pfil_head;
6786 pla.pa_hook = V_pf_ip6_out_hook;
6787 ret = pfil_link(&pla);
6788 MPASS(ret == 0);
6789 if (V_pf_filter_local) {
6790 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6791 pla.pa_head = V_inet6_local_pfil_head;
6792 pla.pa_hook = V_pf_ip6_out_hook;
6793 ret = pfil_link(&pla);
6794 MPASS(ret == 0);
6795 }
6796 #endif
6797
6798 atomic_store_bool(&V_pf_pfil_hooked, true);
6799 }
6800
6801 static void
dehook_pf_eth(void)6802 dehook_pf_eth(void)
6803 {
6804
6805 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6806 return;
6807
6808 pfil_remove_hook(V_pf_eth_in_hook);
6809 pfil_remove_hook(V_pf_eth_out_hook);
6810
6811 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6812 }
6813
6814 static void
dehook_pf(void)6815 dehook_pf(void)
6816 {
6817
6818 if (!atomic_load_bool(&V_pf_pfil_hooked))
6819 return;
6820
6821 #ifdef INET
6822 pfil_remove_hook(V_pf_ip4_in_hook);
6823 pfil_remove_hook(V_pf_ip4_out_hook);
6824 #endif
6825 #ifdef INET6
6826 pfil_remove_hook(V_pf_ip6_in_hook);
6827 pfil_remove_hook(V_pf_ip6_out_hook);
6828 #endif
6829
6830 atomic_store_bool(&V_pf_pfil_hooked, false);
6831 }
6832
6833 static void
pf_load_vnet(void)6834 pf_load_vnet(void)
6835 {
6836 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6837 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6838
6839 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6840 sx_init(&V_pf_ioctl_lock, "pf ioctl");
6841
6842 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6843 PF_RULE_TAG_HASH_SIZE_DEFAULT);
6844 #ifdef ALTQ
6845 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6846 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6847 #endif
6848
6849 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6850
6851 pfattach_vnet();
6852 V_pf_vnet_active = 1;
6853 }
6854
6855 static int
pf_load(void)6856 pf_load(void)
6857 {
6858 int error;
6859
6860 sx_init(&pf_end_lock, "pf end thread");
6861
6862 pf_mtag_initialize();
6863
6864 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6865 if (pf_dev == NULL)
6866 return (ENOMEM);
6867
6868 pf_end_threads = 0;
6869 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6870 if (error != 0)
6871 return (error);
6872
6873 pfi_initialize();
6874
6875 return (0);
6876 }
6877
6878 static void
pf_unload_vnet(void)6879 pf_unload_vnet(void)
6880 {
6881 int ret __diagused;
6882
6883 V_pf_vnet_active = 0;
6884 V_pf_status.running = 0;
6885 dehook_pf();
6886 dehook_pf_eth();
6887
6888 PF_RULES_WLOCK();
6889 pf_syncookies_cleanup();
6890 shutdown_pf();
6891 PF_RULES_WUNLOCK();
6892
6893 ret = swi_remove(V_pf_swi_cookie);
6894 MPASS(ret == 0);
6895 ret = intr_event_destroy(V_pf_swi_ie);
6896 MPASS(ret == 0);
6897
6898 pf_unload_vnet_purge();
6899
6900 pf_normalize_cleanup();
6901 PF_RULES_WLOCK();
6902 pfi_cleanup_vnet();
6903 PF_RULES_WUNLOCK();
6904 pfr_cleanup();
6905 pf_osfp_flush();
6906 pf_cleanup();
6907 if (IS_DEFAULT_VNET(curvnet))
6908 pf_mtag_cleanup();
6909
6910 pf_cleanup_tagset(&V_pf_tags);
6911 #ifdef ALTQ
6912 pf_cleanup_tagset(&V_pf_qids);
6913 #endif
6914 uma_zdestroy(V_pf_tag_z);
6915
6916 #ifdef PF_WANT_32_TO_64_COUNTER
6917 PF_RULES_WLOCK();
6918 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6919
6920 MPASS(LIST_EMPTY(&V_pf_allkiflist));
6921 MPASS(V_pf_allkifcount == 0);
6922
6923 LIST_REMOVE(&V_pf_default_rule, allrulelist);
6924 V_pf_allrulecount--;
6925 LIST_REMOVE(V_pf_rulemarker, allrulelist);
6926
6927 MPASS(LIST_EMPTY(&V_pf_allrulelist));
6928 MPASS(V_pf_allrulecount == 0);
6929
6930 PF_RULES_WUNLOCK();
6931
6932 free(V_pf_kifmarker, PFI_MTYPE);
6933 free(V_pf_rulemarker, M_PFRULE);
6934 #endif
6935
6936 /* Free counters last as we updated them during shutdown. */
6937 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6938 for (int i = 0; i < 2; i++) {
6939 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6940 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6941 }
6942 counter_u64_free(V_pf_default_rule.states_cur);
6943 counter_u64_free(V_pf_default_rule.states_tot);
6944 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
6945 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]);
6946 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6947
6948 for (int i = 0; i < PFRES_MAX; i++)
6949 counter_u64_free(V_pf_status.counters[i]);
6950 for (int i = 0; i < KLCNT_MAX; i++)
6951 counter_u64_free(V_pf_status.lcounters[i]);
6952 for (int i = 0; i < FCNT_MAX; i++)
6953 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6954 for (int i = 0; i < SCNT_MAX; i++)
6955 counter_u64_free(V_pf_status.scounters[i]);
6956 for (int i = 0; i < NCNT_MAX; i++)
6957 counter_u64_free(V_pf_status.ncounters[i]);
6958
6959 rm_destroy(&V_pf_rules_lock);
6960 sx_destroy(&V_pf_ioctl_lock);
6961 }
6962
6963 static void
pf_unload(void)6964 pf_unload(void)
6965 {
6966
6967 sx_xlock(&pf_end_lock);
6968 pf_end_threads = 1;
6969 while (pf_end_threads < 2) {
6970 wakeup_one(pf_purge_thread);
6971 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6972 }
6973 sx_xunlock(&pf_end_lock);
6974
6975 pf_nl_unregister();
6976
6977 if (pf_dev != NULL)
6978 destroy_dev(pf_dev);
6979
6980 pfi_cleanup();
6981
6982 sx_destroy(&pf_end_lock);
6983 }
6984
6985 static void
vnet_pf_init(void * unused __unused)6986 vnet_pf_init(void *unused __unused)
6987 {
6988
6989 pf_load_vnet();
6990 }
6991 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6992 vnet_pf_init, NULL);
6993
6994 static void
vnet_pf_uninit(const void * unused __unused)6995 vnet_pf_uninit(const void *unused __unused)
6996 {
6997
6998 pf_unload_vnet();
6999 }
7000 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
7001 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
7002 vnet_pf_uninit, NULL);
7003
7004 static int
pf_modevent(module_t mod,int type,void * data)7005 pf_modevent(module_t mod, int type, void *data)
7006 {
7007 int error = 0;
7008
7009 switch(type) {
7010 case MOD_LOAD:
7011 error = pf_load();
7012 pf_nl_register();
7013 break;
7014 case MOD_UNLOAD:
7015 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
7016 * the vnet_pf_uninit()s */
7017 break;
7018 default:
7019 error = EINVAL;
7020 break;
7021 }
7022
7023 return (error);
7024 }
7025
7026 static moduledata_t pf_mod = {
7027 "pf",
7028 pf_modevent,
7029 0
7030 };
7031
7032 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
7033 MODULE_DEPEND(pf, netlink, 1, 1, 1);
7034 MODULE_DEPEND(pf, crypto, 1, 1, 1);
7035 MODULE_VERSION(pf, PF_MODVER);
7036