1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2001 Daniel Hartmeier
5 * Copyright (c) 2002,2003 Henning Brauer
6 * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
13 * - Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * - Redistributions in binary form must reproduce the above
16 * copyright notice, this list of conditions and the following
17 * disclaimer in the documentation and/or other materials provided
18 * with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 * POSSIBILITY OF SUCH DAMAGE.
32 *
33 * Effort sponsored in part by the Defense Advanced Research Projects
34 * Agency (DARPA) and Air Force Research Laboratory, Air Force
35 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36 *
37 * $OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38 */
39
40 #include <sys/cdefs.h>
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 #include "opt_bpf.h"
44 #include "opt_pf.h"
45
46 #include <sys/param.h>
47 #include <sys/_bitset.h>
48 #include <sys/bitset.h>
49 #include <sys/bus.h>
50 #include <sys/conf.h>
51 #include <sys/endian.h>
52 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/hash.h>
55 #include <sys/interrupt.h>
56 #include <sys/jail.h>
57 #include <sys/kernel.h>
58 #include <sys/kthread.h>
59 #include <sys/lock.h>
60 #include <sys/mbuf.h>
61 #include <sys/module.h>
62 #include <sys/nv.h>
63 #include <sys/proc.h>
64 #include <sys/sdt.h>
65 #include <sys/smp.h>
66 #include <sys/socket.h>
67 #include <sys/sysctl.h>
68 #include <sys/md5.h>
69 #include <sys/ucred.h>
70
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/if_private.h>
74 #include <net/vnet.h>
75 #include <net/route.h>
76 #include <net/pfil.h>
77 #include <net/pfvar.h>
78 #include <net/if_pfsync.h>
79 #include <net/if_pflog.h>
80
81 #include <netinet/in.h>
82 #include <netinet/ip.h>
83 #include <netinet/ip_var.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet/ip_icmp.h>
86 #include <netpfil/pf/pf_nl.h>
87 #include <netpfil/pf/pf_nv.h>
88
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101
102 static struct pf_kpool *pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 u_int32_t, u_int8_t, u_int8_t, u_int8_t, int);
104
105 static void pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void pf_empty_kpool(struct pf_kpalist *);
107 static int pfioctl(struct cdev *, u_long, caddr_t, int,
108 struct thread *);
109 static int pf_begin_eth(uint32_t *, const char *);
110 static int pf_rollback_eth(uint32_t, const char *);
111 static int pf_commit_eth(uint32_t, const char *);
112 static void pf_free_eth_rule(struct pf_keth_rule *);
113 #ifdef ALTQ
114 static int pf_begin_altq(u_int32_t *);
115 static int pf_rollback_altq(u_int32_t);
116 static int pf_commit_altq(u_int32_t);
117 static int pf_enable_altq(struct pf_altq *);
118 static int pf_disable_altq(struct pf_altq *);
119 static uint16_t pf_qname2qid(const char *);
120 static void pf_qid_unref(uint16_t);
121 #endif /* ALTQ */
122 static int pf_begin_rules(u_int32_t *, int, const char *);
123 static int pf_rollback_rules(u_int32_t, int, char *);
124 static int pf_setup_pfsync_matching(struct pf_kruleset *);
125 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
126 static void pf_hash_rule(struct pf_krule *);
127 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
128 static int pf_commit_rules(u_int32_t, int, char *);
129 static int pf_addr_setup(struct pf_kruleset *,
130 struct pf_addr_wrap *, sa_family_t);
131 static void pf_src_node_copy(const struct pf_ksrc_node *,
132 struct pf_src_node *);
133 #ifdef ALTQ
134 static int pf_export_kaltq(struct pf_altq *,
135 struct pfioc_altq_v1 *, size_t);
136 static int pf_import_kaltq(struct pfioc_altq_v1 *,
137 struct pf_altq *, size_t);
138 #endif /* ALTQ */
139
140 VNET_DEFINE(struct pf_krule, pf_default_rule);
141
142 static __inline int pf_krule_compare(struct pf_krule *,
143 struct pf_krule *);
144
145 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
146
147 #ifdef ALTQ
148 VNET_DEFINE_STATIC(int, pf_altq_running);
149 #define V_pf_altq_running VNET(pf_altq_running)
150 #endif
151
152 #define TAGID_MAX 50000
153 struct pf_tagname {
154 TAILQ_ENTRY(pf_tagname) namehash_entries;
155 TAILQ_ENTRY(pf_tagname) taghash_entries;
156 char name[PF_TAG_NAME_SIZE];
157 uint16_t tag;
158 int ref;
159 };
160
161 struct pf_tagset {
162 TAILQ_HEAD(, pf_tagname) *namehash;
163 TAILQ_HEAD(, pf_tagname) *taghash;
164 unsigned int mask;
165 uint32_t seed;
166 BITSET_DEFINE(, TAGID_MAX) avail;
167 };
168
169 VNET_DEFINE(struct pf_tagset, pf_tags);
170 #define V_pf_tags VNET(pf_tags)
171 static unsigned int pf_rule_tag_hashsize;
172 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128
173 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
174 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
175 "Size of pf(4) rule tag hashtable");
176
177 #ifdef ALTQ
178 VNET_DEFINE(struct pf_tagset, pf_qids);
179 #define V_pf_qids VNET(pf_qids)
180 static unsigned int pf_queue_tag_hashsize;
181 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128
182 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
183 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
184 "Size of pf(4) queue tag hashtable");
185 #endif
186 VNET_DEFINE(uma_zone_t, pf_tag_z);
187 #define V_pf_tag_z VNET(pf_tag_z)
188 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
189 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
190
191 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
192 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
193 #endif
194
195 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
196 #define V_pf_filter_local VNET(pf_filter_local)
197 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
198 &VNET_NAME(pf_filter_local), false,
199 "Enable filtering for packets delivered to local network stack");
200
201 #ifdef PF_DEFAULT_TO_DROP
202 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
203 #else
204 VNET_DEFINE_STATIC(bool, default_to_drop);
205 #endif
206 #define V_default_to_drop VNET(default_to_drop)
207 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
208 &VNET_NAME(default_to_drop), false,
209 "Make the default rule drop all packets.");
210
211 static void pf_init_tagset(struct pf_tagset *, unsigned int *,
212 unsigned int);
213 static void pf_cleanup_tagset(struct pf_tagset *);
214 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *);
215 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t);
216 static u_int16_t tagname2tag(struct pf_tagset *, const char *);
217 static u_int16_t pf_tagname2tag(const char *);
218 static void tag_unref(struct pf_tagset *, u_int16_t);
219
220 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
221
222 struct cdev *pf_dev;
223
224 /*
225 * XXX - These are new and need to be checked when moveing to a new version
226 */
227 static void pf_clear_all_states(void);
228 static int pf_killstates_row(struct pf_kstate_kill *,
229 struct pf_idhash *);
230 static int pf_killstates_nv(struct pfioc_nv *);
231 static int pf_clearstates_nv(struct pfioc_nv *);
232 static int pf_getstate(struct pfioc_nv *);
233 static int pf_getstatus(struct pfioc_nv *);
234 static int pf_clear_tables(void);
235 static void pf_kill_srcnodes(struct pfioc_src_node_kill *);
236 static int pf_keepcounters(struct pfioc_nv *);
237 static void pf_tbladdr_copyout(struct pf_addr_wrap *);
238
239 /*
240 * Wrapper functions for pfil(9) hooks
241 */
242 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
243 int flags, void *ruleset __unused, struct inpcb *inp);
244 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
245 int flags, void *ruleset __unused, struct inpcb *inp);
246 #ifdef INET
247 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
248 int flags, void *ruleset __unused, struct inpcb *inp);
249 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
250 int flags, void *ruleset __unused, struct inpcb *inp);
251 #endif
252 #ifdef INET6
253 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
254 int flags, void *ruleset __unused, struct inpcb *inp);
255 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
256 int flags, void *ruleset __unused, struct inpcb *inp);
257 #endif
258
259 static void hook_pf_eth(void);
260 static void hook_pf(void);
261 static void dehook_pf_eth(void);
262 static void dehook_pf(void);
263 static int shutdown_pf(void);
264 static int pf_load(void);
265 static void pf_unload(void);
266
267 static struct cdevsw pf_cdevsw = {
268 .d_ioctl = pfioctl,
269 .d_name = PF_NAME,
270 .d_version = D_VERSION,
271 };
272
273 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
274 #define V_pf_pfil_hooked VNET(pf_pfil_hooked)
275 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
276 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked)
277
278 /*
279 * We need a flag that is neither hooked nor running to know when
280 * the VNET is "valid". We primarily need this to control (global)
281 * external event, e.g., eventhandlers.
282 */
283 VNET_DEFINE(int, pf_vnet_active);
284 #define V_pf_vnet_active VNET(pf_vnet_active)
285
286 int pf_end_threads;
287 struct proc *pf_purge_proc;
288
289 VNET_DEFINE(struct rmlock, pf_rules_lock);
290 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
291 #define V_pf_ioctl_lock VNET(pf_ioctl_lock)
292 struct sx pf_end_lock;
293
294 /* pfsync */
295 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
296 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
297 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
298 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
299 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
300 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
301 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
302 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
303
304 /* pflog */
305 pflog_packet_t *pflog_packet_ptr = NULL;
306
307 /*
308 * Copy a user-provided string, returning an error if truncation would occur.
309 * Avoid scanning past "sz" bytes in the source string since there's no
310 * guarantee that it's nul-terminated.
311 */
312 static int
pf_user_strcpy(char * dst,const char * src,size_t sz)313 pf_user_strcpy(char *dst, const char *src, size_t sz)
314 {
315 if (strnlen(src, sz) == sz)
316 return (EINVAL);
317 (void)strlcpy(dst, src, sz);
318 return (0);
319 }
320
321 static void
pfattach_vnet(void)322 pfattach_vnet(void)
323 {
324 u_int32_t *my_timeout = V_pf_default_rule.timeout;
325
326 bzero(&V_pf_status, sizeof(V_pf_status));
327
328 pf_initialize();
329 pfr_initialize();
330 pfi_initialize_vnet();
331 pf_normalize_init();
332 pf_syncookies_init();
333
334 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
335 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
336
337 RB_INIT(&V_pf_anchors);
338 pf_init_kruleset(&pf_main_ruleset);
339
340 pf_init_keth(V_pf_keth);
341
342 /* default rule should never be garbage collected */
343 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
344 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
345 V_pf_default_rule.nr = (uint32_t)-1;
346 V_pf_default_rule.rtableid = -1;
347
348 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
349 for (int i = 0; i < 2; i++) {
350 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
351 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
352 }
353 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
354 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
355 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++)
356 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
357
358 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
359 M_WAITOK | M_ZERO);
360
361 #ifdef PF_WANT_32_TO_64_COUNTER
362 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
363 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
364 PF_RULES_WLOCK();
365 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
366 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
367 V_pf_allrulecount++;
368 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
369 PF_RULES_WUNLOCK();
370 #endif
371
372 /* initialize default timeouts */
373 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
374 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
375 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
376 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
377 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
378 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
379 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
380 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
381 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
382 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
383 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
384 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
385 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
386 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
387 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
388 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
389 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
390 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
391 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
392 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
393 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
394 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
395 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
396 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
397 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
398
399 V_pf_status.debug = PF_DEBUG_URGENT;
400 /*
401 * XXX This is different than in OpenBSD where reassembly is enabled by
402 * defult. In FreeBSD we expect people to still use scrub rules and
403 * switch to the new syntax later. Only when they switch they must
404 * explicitly enable reassemle. We could change the default once the
405 * scrub rule functionality is hopefully removed some day in future.
406 */
407 V_pf_status.reass = 0;
408
409 V_pf_pfil_hooked = false;
410 V_pf_pfil_eth_hooked = false;
411
412 /* XXX do our best to avoid a conflict */
413 V_pf_status.hostid = arc4random();
414
415 for (int i = 0; i < PFRES_MAX; i++)
416 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
417 for (int i = 0; i < KLCNT_MAX; i++)
418 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
419 for (int i = 0; i < FCNT_MAX; i++)
420 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
421 for (int i = 0; i < SCNT_MAX; i++)
422 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
423
424 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
425 INTR_MPSAFE, &V_pf_swi_cookie) != 0)
426 /* XXXGL: leaked all above. */
427 return;
428 }
429
430 static struct pf_kpool *
pf_get_kpool(const char * anchor,u_int32_t ticket,u_int8_t rule_action,u_int32_t rule_number,u_int8_t r_last,u_int8_t active,u_int8_t check_ticket,int which)431 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
432 u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
433 u_int8_t check_ticket, int which)
434 {
435 struct pf_kruleset *ruleset;
436 struct pf_krule *rule;
437 int rs_num;
438
439 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT);
440
441 ruleset = pf_find_kruleset(anchor);
442 if (ruleset == NULL)
443 return (NULL);
444 rs_num = pf_get_ruleset_number(rule_action);
445 if (rs_num >= PF_RULESET_MAX)
446 return (NULL);
447 if (active) {
448 if (check_ticket && ticket !=
449 ruleset->rules[rs_num].active.ticket)
450 return (NULL);
451 if (r_last)
452 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
453 pf_krulequeue);
454 else
455 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
456 } else {
457 if (check_ticket && ticket !=
458 ruleset->rules[rs_num].inactive.ticket)
459 return (NULL);
460 if (r_last)
461 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
462 pf_krulequeue);
463 else
464 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
465 }
466 if (!r_last) {
467 while ((rule != NULL) && (rule->nr != rule_number))
468 rule = TAILQ_NEXT(rule, entries);
469 }
470 if (rule == NULL)
471 return (NULL);
472
473 switch (which) {
474 case PF_RDR:
475 return (&rule->rdr);
476 case PF_NAT:
477 return (&rule->nat);
478 case PF_RT:
479 return (&rule->route);
480 default:
481 panic("Unknow pool type %d", which);
482 }
483 }
484
485 static void
pf_mv_kpool(struct pf_kpalist * poola,struct pf_kpalist * poolb)486 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
487 {
488 struct pf_kpooladdr *mv_pool_pa;
489
490 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
491 TAILQ_REMOVE(poola, mv_pool_pa, entries);
492 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
493 }
494 }
495
496 static void
pf_empty_kpool(struct pf_kpalist * poola)497 pf_empty_kpool(struct pf_kpalist *poola)
498 {
499 struct pf_kpooladdr *pa;
500
501 while ((pa = TAILQ_FIRST(poola)) != NULL) {
502 switch (pa->addr.type) {
503 case PF_ADDR_DYNIFTL:
504 pfi_dynaddr_remove(pa->addr.p.dyn);
505 break;
506 case PF_ADDR_TABLE:
507 /* XXX: this could be unfinished pooladdr on pabuf */
508 if (pa->addr.p.tbl != NULL)
509 pfr_detach_table(pa->addr.p.tbl);
510 break;
511 }
512 if (pa->kif)
513 pfi_kkif_unref(pa->kif);
514 TAILQ_REMOVE(poola, pa, entries);
515 free(pa, M_PFRULE);
516 }
517 }
518
519 static void
pf_unlink_rule_locked(struct pf_krulequeue * rulequeue,struct pf_krule * rule)520 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
521 {
522
523 PF_RULES_WASSERT();
524 PF_UNLNKDRULES_ASSERT();
525
526 TAILQ_REMOVE(rulequeue, rule, entries);
527
528 rule->rule_ref |= PFRULE_REFS;
529 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
530 }
531
532 static void
pf_unlink_rule(struct pf_krulequeue * rulequeue,struct pf_krule * rule)533 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
534 {
535
536 PF_RULES_WASSERT();
537
538 PF_UNLNKDRULES_LOCK();
539 pf_unlink_rule_locked(rulequeue, rule);
540 PF_UNLNKDRULES_UNLOCK();
541 }
542
543 static void
pf_free_eth_rule(struct pf_keth_rule * rule)544 pf_free_eth_rule(struct pf_keth_rule *rule)
545 {
546 PF_RULES_WASSERT();
547
548 if (rule == NULL)
549 return;
550
551 if (rule->tag)
552 tag_unref(&V_pf_tags, rule->tag);
553 if (rule->match_tag)
554 tag_unref(&V_pf_tags, rule->match_tag);
555 #ifdef ALTQ
556 pf_qid_unref(rule->qid);
557 #endif
558
559 if (rule->bridge_to)
560 pfi_kkif_unref(rule->bridge_to);
561 if (rule->kif)
562 pfi_kkif_unref(rule->kif);
563
564 if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
565 pfr_detach_table(rule->ipsrc.addr.p.tbl);
566 if (rule->ipdst.addr.type == PF_ADDR_TABLE)
567 pfr_detach_table(rule->ipdst.addr.p.tbl);
568
569 counter_u64_free(rule->evaluations);
570 for (int i = 0; i < 2; i++) {
571 counter_u64_free(rule->packets[i]);
572 counter_u64_free(rule->bytes[i]);
573 }
574 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
575 pf_keth_anchor_remove(rule);
576
577 free(rule, M_PFRULE);
578 }
579
580 void
pf_free_rule(struct pf_krule * rule)581 pf_free_rule(struct pf_krule *rule)
582 {
583
584 PF_RULES_WASSERT();
585 PF_CONFIG_ASSERT();
586
587 if (rule->tag)
588 tag_unref(&V_pf_tags, rule->tag);
589 if (rule->match_tag)
590 tag_unref(&V_pf_tags, rule->match_tag);
591 #ifdef ALTQ
592 if (rule->pqid != rule->qid)
593 pf_qid_unref(rule->pqid);
594 pf_qid_unref(rule->qid);
595 #endif
596 switch (rule->src.addr.type) {
597 case PF_ADDR_DYNIFTL:
598 pfi_dynaddr_remove(rule->src.addr.p.dyn);
599 break;
600 case PF_ADDR_TABLE:
601 pfr_detach_table(rule->src.addr.p.tbl);
602 break;
603 }
604 switch (rule->dst.addr.type) {
605 case PF_ADDR_DYNIFTL:
606 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
607 break;
608 case PF_ADDR_TABLE:
609 pfr_detach_table(rule->dst.addr.p.tbl);
610 break;
611 }
612 if (rule->overload_tbl)
613 pfr_detach_table(rule->overload_tbl);
614 if (rule->kif)
615 pfi_kkif_unref(rule->kif);
616 if (rule->rcv_kif)
617 pfi_kkif_unref(rule->rcv_kif);
618 pf_remove_kanchor(rule);
619 pf_empty_kpool(&rule->rdr.list);
620 pf_empty_kpool(&rule->nat.list);
621 pf_empty_kpool(&rule->route.list);
622
623 pf_krule_free(rule);
624 }
625
626 static void
pf_init_tagset(struct pf_tagset * ts,unsigned int * tunable_size,unsigned int default_size)627 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
628 unsigned int default_size)
629 {
630 unsigned int i;
631 unsigned int hashsize;
632
633 if (*tunable_size == 0 || !powerof2(*tunable_size))
634 *tunable_size = default_size;
635
636 hashsize = *tunable_size;
637 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
638 M_WAITOK);
639 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
640 M_WAITOK);
641 ts->mask = hashsize - 1;
642 ts->seed = arc4random();
643 for (i = 0; i < hashsize; i++) {
644 TAILQ_INIT(&ts->namehash[i]);
645 TAILQ_INIT(&ts->taghash[i]);
646 }
647 BIT_FILL(TAGID_MAX, &ts->avail);
648 }
649
650 static void
pf_cleanup_tagset(struct pf_tagset * ts)651 pf_cleanup_tagset(struct pf_tagset *ts)
652 {
653 unsigned int i;
654 unsigned int hashsize;
655 struct pf_tagname *t, *tmp;
656
657 /*
658 * Only need to clean up one of the hashes as each tag is hashed
659 * into each table.
660 */
661 hashsize = ts->mask + 1;
662 for (i = 0; i < hashsize; i++)
663 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
664 uma_zfree(V_pf_tag_z, t);
665
666 free(ts->namehash, M_PFHASH);
667 free(ts->taghash, M_PFHASH);
668 }
669
670 static uint16_t
tagname2hashindex(const struct pf_tagset * ts,const char * tagname)671 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
672 {
673 size_t len;
674
675 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
676 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
677 }
678
679 static uint16_t
tag2hashindex(const struct pf_tagset * ts,uint16_t tag)680 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
681 {
682
683 return (tag & ts->mask);
684 }
685
686 static u_int16_t
tagname2tag(struct pf_tagset * ts,const char * tagname)687 tagname2tag(struct pf_tagset *ts, const char *tagname)
688 {
689 struct pf_tagname *tag;
690 u_int32_t index;
691 u_int16_t new_tagid;
692
693 PF_RULES_WASSERT();
694
695 index = tagname2hashindex(ts, tagname);
696 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
697 if (strcmp(tagname, tag->name) == 0) {
698 tag->ref++;
699 return (tag->tag);
700 }
701
702 /*
703 * new entry
704 *
705 * to avoid fragmentation, we do a linear search from the beginning
706 * and take the first free slot we find.
707 */
708 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
709 /*
710 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
711 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
712 * set. It may also return a bit number greater than TAGID_MAX due
713 * to rounding of the number of bits in the vector up to a multiple
714 * of the vector word size at declaration/allocation time.
715 */
716 if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
717 return (0);
718
719 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */
720 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
721
722 /* allocate and fill new struct pf_tagname */
723 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
724 if (tag == NULL)
725 return (0);
726 strlcpy(tag->name, tagname, sizeof(tag->name));
727 tag->tag = new_tagid;
728 tag->ref = 1;
729
730 /* Insert into namehash */
731 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
732
733 /* Insert into taghash */
734 index = tag2hashindex(ts, new_tagid);
735 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
736
737 return (tag->tag);
738 }
739
740 static void
tag_unref(struct pf_tagset * ts,u_int16_t tag)741 tag_unref(struct pf_tagset *ts, u_int16_t tag)
742 {
743 struct pf_tagname *t;
744 uint16_t index;
745
746 PF_RULES_WASSERT();
747
748 index = tag2hashindex(ts, tag);
749 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
750 if (tag == t->tag) {
751 if (--t->ref == 0) {
752 TAILQ_REMOVE(&ts->taghash[index], t,
753 taghash_entries);
754 index = tagname2hashindex(ts, t->name);
755 TAILQ_REMOVE(&ts->namehash[index], t,
756 namehash_entries);
757 /* Bits are 0-based for BIT_SET() */
758 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
759 uma_zfree(V_pf_tag_z, t);
760 }
761 break;
762 }
763 }
764
765 static uint16_t
pf_tagname2tag(const char * tagname)766 pf_tagname2tag(const char *tagname)
767 {
768 return (tagname2tag(&V_pf_tags, tagname));
769 }
770
771 static int
pf_begin_eth(uint32_t * ticket,const char * anchor)772 pf_begin_eth(uint32_t *ticket, const char *anchor)
773 {
774 struct pf_keth_rule *rule, *tmp;
775 struct pf_keth_ruleset *rs;
776
777 PF_RULES_WASSERT();
778
779 rs = pf_find_or_create_keth_ruleset(anchor);
780 if (rs == NULL)
781 return (EINVAL);
782
783 /* Purge old inactive rules. */
784 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
785 tmp) {
786 TAILQ_REMOVE(rs->inactive.rules, rule,
787 entries);
788 pf_free_eth_rule(rule);
789 }
790
791 *ticket = ++rs->inactive.ticket;
792 rs->inactive.open = 1;
793
794 return (0);
795 }
796
797 static int
pf_rollback_eth(uint32_t ticket,const char * anchor)798 pf_rollback_eth(uint32_t ticket, const char *anchor)
799 {
800 struct pf_keth_rule *rule, *tmp;
801 struct pf_keth_ruleset *rs;
802
803 PF_RULES_WASSERT();
804
805 rs = pf_find_keth_ruleset(anchor);
806 if (rs == NULL)
807 return (EINVAL);
808
809 if (!rs->inactive.open ||
810 ticket != rs->inactive.ticket)
811 return (0);
812
813 /* Purge old inactive rules. */
814 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
815 tmp) {
816 TAILQ_REMOVE(rs->inactive.rules, rule, entries);
817 pf_free_eth_rule(rule);
818 }
819
820 rs->inactive.open = 0;
821
822 pf_remove_if_empty_keth_ruleset(rs);
823
824 return (0);
825 }
826
827 #define PF_SET_SKIP_STEPS(i) \
828 do { \
829 while (head[i] != cur) { \
830 head[i]->skip[i].ptr = cur; \
831 head[i] = TAILQ_NEXT(head[i], entries); \
832 } \
833 } while (0)
834
835 static void
pf_eth_calc_skip_steps(struct pf_keth_ruleq * rules)836 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
837 {
838 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
839 int i;
840
841 cur = TAILQ_FIRST(rules);
842 prev = cur;
843 for (i = 0; i < PFE_SKIP_COUNT; ++i)
844 head[i] = cur;
845 while (cur != NULL) {
846 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
847 PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
848 if (cur->direction != prev->direction)
849 PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
850 if (cur->proto != prev->proto)
851 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
852 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
853 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
854 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
855 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
856 if (cur->ipsrc.neg != prev->ipsrc.neg ||
857 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
858 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
859 if (cur->ipdst.neg != prev->ipdst.neg ||
860 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
861 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
862
863 prev = cur;
864 cur = TAILQ_NEXT(cur, entries);
865 }
866 for (i = 0; i < PFE_SKIP_COUNT; ++i)
867 PF_SET_SKIP_STEPS(i);
868 }
869
870 static int
pf_commit_eth(uint32_t ticket,const char * anchor)871 pf_commit_eth(uint32_t ticket, const char *anchor)
872 {
873 struct pf_keth_ruleq *rules;
874 struct pf_keth_ruleset *rs;
875
876 rs = pf_find_keth_ruleset(anchor);
877 if (rs == NULL) {
878 return (EINVAL);
879 }
880
881 if (!rs->inactive.open ||
882 ticket != rs->inactive.ticket)
883 return (EBUSY);
884
885 PF_RULES_WASSERT();
886
887 pf_eth_calc_skip_steps(rs->inactive.rules);
888
889 rules = rs->active.rules;
890 atomic_store_ptr(&rs->active.rules, rs->inactive.rules);
891 rs->inactive.rules = rules;
892 rs->inactive.ticket = rs->active.ticket;
893
894 return (pf_rollback_eth(rs->inactive.ticket,
895 rs->anchor ? rs->anchor->path : ""));
896 }
897
898 #ifdef ALTQ
899 static uint16_t
pf_qname2qid(const char * qname)900 pf_qname2qid(const char *qname)
901 {
902 return (tagname2tag(&V_pf_qids, qname));
903 }
904
905 static void
pf_qid_unref(uint16_t qid)906 pf_qid_unref(uint16_t qid)
907 {
908 tag_unref(&V_pf_qids, qid);
909 }
910
911 static int
pf_begin_altq(u_int32_t * ticket)912 pf_begin_altq(u_int32_t *ticket)
913 {
914 struct pf_altq *altq, *tmp;
915 int error = 0;
916
917 PF_RULES_WASSERT();
918
919 /* Purge the old altq lists */
920 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
921 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
922 /* detach and destroy the discipline */
923 error = altq_remove(altq);
924 }
925 free(altq, M_PFALTQ);
926 }
927 TAILQ_INIT(V_pf_altq_ifs_inactive);
928 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
929 pf_qid_unref(altq->qid);
930 free(altq, M_PFALTQ);
931 }
932 TAILQ_INIT(V_pf_altqs_inactive);
933 if (error)
934 return (error);
935 *ticket = ++V_ticket_altqs_inactive;
936 V_altqs_inactive_open = 1;
937 return (0);
938 }
939
940 static int
pf_rollback_altq(u_int32_t ticket)941 pf_rollback_altq(u_int32_t ticket)
942 {
943 struct pf_altq *altq, *tmp;
944 int error = 0;
945
946 PF_RULES_WASSERT();
947
948 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
949 return (0);
950 /* Purge the old altq lists */
951 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
952 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
953 /* detach and destroy the discipline */
954 error = altq_remove(altq);
955 }
956 free(altq, M_PFALTQ);
957 }
958 TAILQ_INIT(V_pf_altq_ifs_inactive);
959 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
960 pf_qid_unref(altq->qid);
961 free(altq, M_PFALTQ);
962 }
963 TAILQ_INIT(V_pf_altqs_inactive);
964 V_altqs_inactive_open = 0;
965 return (error);
966 }
967
968 static int
pf_commit_altq(u_int32_t ticket)969 pf_commit_altq(u_int32_t ticket)
970 {
971 struct pf_altqqueue *old_altqs, *old_altq_ifs;
972 struct pf_altq *altq, *tmp;
973 int err, error = 0;
974
975 PF_RULES_WASSERT();
976
977 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
978 return (EBUSY);
979
980 /* swap altqs, keep the old. */
981 old_altqs = V_pf_altqs_active;
982 old_altq_ifs = V_pf_altq_ifs_active;
983 V_pf_altqs_active = V_pf_altqs_inactive;
984 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
985 V_pf_altqs_inactive = old_altqs;
986 V_pf_altq_ifs_inactive = old_altq_ifs;
987 V_ticket_altqs_active = V_ticket_altqs_inactive;
988
989 /* Attach new disciplines */
990 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
991 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
992 /* attach the discipline */
993 error = altq_pfattach(altq);
994 if (error == 0 && V_pf_altq_running)
995 error = pf_enable_altq(altq);
996 if (error != 0)
997 return (error);
998 }
999 }
1000
1001 /* Purge the old altq lists */
1002 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1003 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1004 /* detach and destroy the discipline */
1005 if (V_pf_altq_running)
1006 error = pf_disable_altq(altq);
1007 err = altq_pfdetach(altq);
1008 if (err != 0 && error == 0)
1009 error = err;
1010 err = altq_remove(altq);
1011 if (err != 0 && error == 0)
1012 error = err;
1013 }
1014 free(altq, M_PFALTQ);
1015 }
1016 TAILQ_INIT(V_pf_altq_ifs_inactive);
1017 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1018 pf_qid_unref(altq->qid);
1019 free(altq, M_PFALTQ);
1020 }
1021 TAILQ_INIT(V_pf_altqs_inactive);
1022
1023 V_altqs_inactive_open = 0;
1024 return (error);
1025 }
1026
1027 static int
pf_enable_altq(struct pf_altq * altq)1028 pf_enable_altq(struct pf_altq *altq)
1029 {
1030 struct ifnet *ifp;
1031 struct tb_profile tb;
1032 int error = 0;
1033
1034 if ((ifp = ifunit(altq->ifname)) == NULL)
1035 return (EINVAL);
1036
1037 if (ifp->if_snd.altq_type != ALTQT_NONE)
1038 error = altq_enable(&ifp->if_snd);
1039
1040 /* set tokenbucket regulator */
1041 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1042 tb.rate = altq->ifbandwidth;
1043 tb.depth = altq->tbrsize;
1044 error = tbr_set(&ifp->if_snd, &tb);
1045 }
1046
1047 return (error);
1048 }
1049
1050 static int
pf_disable_altq(struct pf_altq * altq)1051 pf_disable_altq(struct pf_altq *altq)
1052 {
1053 struct ifnet *ifp;
1054 struct tb_profile tb;
1055 int error;
1056
1057 if ((ifp = ifunit(altq->ifname)) == NULL)
1058 return (EINVAL);
1059
1060 /*
1061 * when the discipline is no longer referenced, it was overridden
1062 * by a new one. if so, just return.
1063 */
1064 if (altq->altq_disc != ifp->if_snd.altq_disc)
1065 return (0);
1066
1067 error = altq_disable(&ifp->if_snd);
1068
1069 if (error == 0) {
1070 /* clear tokenbucket regulator */
1071 tb.rate = 0;
1072 error = tbr_set(&ifp->if_snd, &tb);
1073 }
1074
1075 return (error);
1076 }
1077
1078 static int
pf_altq_ifnet_event_add(struct ifnet * ifp,int remove,u_int32_t ticket,struct pf_altq * altq)1079 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1080 struct pf_altq *altq)
1081 {
1082 struct ifnet *ifp1;
1083 int error = 0;
1084
1085 /* Deactivate the interface in question */
1086 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1087 if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1088 (remove && ifp1 == ifp)) {
1089 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1090 } else {
1091 error = altq_add(ifp1, altq);
1092
1093 if (ticket != V_ticket_altqs_inactive)
1094 error = EBUSY;
1095
1096 if (error)
1097 free(altq, M_PFALTQ);
1098 }
1099
1100 return (error);
1101 }
1102
1103 void
pf_altq_ifnet_event(struct ifnet * ifp,int remove)1104 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1105 {
1106 struct pf_altq *a1, *a2, *a3;
1107 u_int32_t ticket;
1108 int error = 0;
1109
1110 /*
1111 * No need to re-evaluate the configuration for events on interfaces
1112 * that do not support ALTQ, as it's not possible for such
1113 * interfaces to be part of the configuration.
1114 */
1115 if (!ALTQ_IS_READY(&ifp->if_snd))
1116 return;
1117
1118 /* Interrupt userland queue modifications */
1119 if (V_altqs_inactive_open)
1120 pf_rollback_altq(V_ticket_altqs_inactive);
1121
1122 /* Start new altq ruleset */
1123 if (pf_begin_altq(&ticket))
1124 return;
1125
1126 /* Copy the current active set */
1127 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1128 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1129 if (a2 == NULL) {
1130 error = ENOMEM;
1131 break;
1132 }
1133 bcopy(a1, a2, sizeof(struct pf_altq));
1134
1135 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1136 if (error)
1137 break;
1138
1139 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1140 }
1141 if (error)
1142 goto out;
1143 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1144 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1145 if (a2 == NULL) {
1146 error = ENOMEM;
1147 break;
1148 }
1149 bcopy(a1, a2, sizeof(struct pf_altq));
1150
1151 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1152 error = EBUSY;
1153 free(a2, M_PFALTQ);
1154 break;
1155 }
1156 a2->altq_disc = NULL;
1157 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1158 if (strncmp(a3->ifname, a2->ifname,
1159 IFNAMSIZ) == 0) {
1160 a2->altq_disc = a3->altq_disc;
1161 break;
1162 }
1163 }
1164 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1165 if (error)
1166 break;
1167
1168 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1169 }
1170
1171 out:
1172 if (error != 0)
1173 pf_rollback_altq(ticket);
1174 else
1175 pf_commit_altq(ticket);
1176 }
1177 #endif /* ALTQ */
1178
1179 static struct pf_krule_global *
pf_rule_tree_alloc(int flags)1180 pf_rule_tree_alloc(int flags)
1181 {
1182 struct pf_krule_global *tree;
1183
1184 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1185 if (tree == NULL)
1186 return (NULL);
1187 RB_INIT(tree);
1188 return (tree);
1189 }
1190
1191 static void
pf_rule_tree_free(struct pf_krule_global * tree)1192 pf_rule_tree_free(struct pf_krule_global *tree)
1193 {
1194
1195 free(tree, M_TEMP);
1196 }
1197
1198 static int
pf_begin_rules(u_int32_t * ticket,int rs_num,const char * anchor)1199 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1200 {
1201 struct pf_krule_global *tree;
1202 struct pf_kruleset *rs;
1203 struct pf_krule *rule;
1204
1205 PF_RULES_WASSERT();
1206
1207 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1208 return (EINVAL);
1209 tree = pf_rule_tree_alloc(M_NOWAIT);
1210 if (tree == NULL)
1211 return (ENOMEM);
1212 rs = pf_find_or_create_kruleset(anchor);
1213 if (rs == NULL) {
1214 free(tree, M_TEMP);
1215 return (EINVAL);
1216 }
1217 pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1218 rs->rules[rs_num].inactive.tree = tree;
1219
1220 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1221 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1222 rs->rules[rs_num].inactive.rcount--;
1223 }
1224 *ticket = ++rs->rules[rs_num].inactive.ticket;
1225 rs->rules[rs_num].inactive.open = 1;
1226 return (0);
1227 }
1228
1229 static int
pf_rollback_rules(u_int32_t ticket,int rs_num,char * anchor)1230 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1231 {
1232 struct pf_kruleset *rs;
1233 struct pf_krule *rule;
1234
1235 PF_RULES_WASSERT();
1236
1237 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1238 return (EINVAL);
1239 rs = pf_find_kruleset(anchor);
1240 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1241 rs->rules[rs_num].inactive.ticket != ticket)
1242 return (0);
1243 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1244 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1245 rs->rules[rs_num].inactive.rcount--;
1246 }
1247 rs->rules[rs_num].inactive.open = 0;
1248 return (0);
1249 }
1250
1251 #define PF_MD5_UPD(st, elm) \
1252 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1253
1254 #define PF_MD5_UPD_STR(st, elm) \
1255 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1256
1257 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \
1258 (stor) = htonl((st)->elm); \
1259 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1260 } while (0)
1261
1262 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \
1263 (stor) = htons((st)->elm); \
1264 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1265 } while (0)
1266
1267 static void
pf_hash_rule_addr(MD5_CTX * ctx,struct pf_rule_addr * pfr)1268 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1269 {
1270 PF_MD5_UPD(pfr, addr.type);
1271 switch (pfr->addr.type) {
1272 case PF_ADDR_DYNIFTL:
1273 PF_MD5_UPD(pfr, addr.v.ifname);
1274 PF_MD5_UPD(pfr, addr.iflags);
1275 break;
1276 case PF_ADDR_TABLE:
1277 if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX,
1278 strlen(PF_OPTIMIZER_TABLE_PFX)))
1279 PF_MD5_UPD(pfr, addr.v.tblname);
1280 break;
1281 case PF_ADDR_ADDRMASK:
1282 /* XXX ignore af? */
1283 PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1284 PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1285 break;
1286 }
1287
1288 PF_MD5_UPD(pfr, port[0]);
1289 PF_MD5_UPD(pfr, port[1]);
1290 PF_MD5_UPD(pfr, neg);
1291 PF_MD5_UPD(pfr, port_op);
1292 }
1293
1294 static void
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)1295 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1296 {
1297 u_int16_t x;
1298 u_int32_t y;
1299
1300 pf_hash_rule_addr(ctx, &rule->src);
1301 pf_hash_rule_addr(ctx, &rule->dst);
1302 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1303 PF_MD5_UPD_STR(rule, label[i]);
1304 PF_MD5_UPD_STR(rule, ifname);
1305 PF_MD5_UPD_STR(rule, rcv_ifname);
1306 PF_MD5_UPD_STR(rule, match_tagname);
1307 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1308 PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1309 PF_MD5_UPD_HTONL(rule, prob, y);
1310 PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1311 PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1312 PF_MD5_UPD(rule, uid.op);
1313 PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1314 PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1315 PF_MD5_UPD(rule, gid.op);
1316 PF_MD5_UPD_HTONL(rule, rule_flag, y);
1317 PF_MD5_UPD(rule, action);
1318 PF_MD5_UPD(rule, direction);
1319 PF_MD5_UPD(rule, af);
1320 PF_MD5_UPD(rule, quick);
1321 PF_MD5_UPD(rule, ifnot);
1322 PF_MD5_UPD(rule, rcvifnot);
1323 PF_MD5_UPD(rule, match_tag_not);
1324 PF_MD5_UPD(rule, natpass);
1325 PF_MD5_UPD(rule, keep_state);
1326 PF_MD5_UPD(rule, proto);
1327 PF_MD5_UPD(rule, type);
1328 PF_MD5_UPD(rule, code);
1329 PF_MD5_UPD(rule, flags);
1330 PF_MD5_UPD(rule, flagset);
1331 PF_MD5_UPD(rule, allow_opts);
1332 PF_MD5_UPD(rule, rt);
1333 PF_MD5_UPD(rule, tos);
1334 PF_MD5_UPD(rule, scrub_flags);
1335 PF_MD5_UPD(rule, min_ttl);
1336 PF_MD5_UPD(rule, set_tos);
1337 if (rule->anchor != NULL)
1338 PF_MD5_UPD_STR(rule, anchor->path);
1339 }
1340
1341 static void
pf_hash_rule(struct pf_krule * rule)1342 pf_hash_rule(struct pf_krule *rule)
1343 {
1344 MD5_CTX ctx;
1345
1346 MD5Init(&ctx);
1347 pf_hash_rule_rolling(&ctx, rule);
1348 MD5Final(rule->md5sum, &ctx);
1349 }
1350
1351 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1352 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1353 {
1354
1355 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1356 }
1357
1358 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1359 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1360 {
1361 struct pf_kruleset *rs;
1362 struct pf_krule *rule, **old_array, *old_rule;
1363 struct pf_krulequeue *old_rules;
1364 struct pf_krule_global *old_tree;
1365 int error;
1366 u_int32_t old_rcount;
1367
1368 PF_RULES_WASSERT();
1369
1370 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1371 return (EINVAL);
1372 rs = pf_find_kruleset(anchor);
1373 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1374 ticket != rs->rules[rs_num].inactive.ticket)
1375 return (EBUSY);
1376
1377 /* Calculate checksum for the main ruleset */
1378 if (rs == &pf_main_ruleset) {
1379 error = pf_setup_pfsync_matching(rs);
1380 if (error != 0)
1381 return (error);
1382 }
1383
1384 /* Swap rules, keep the old. */
1385 old_rules = rs->rules[rs_num].active.ptr;
1386 old_rcount = rs->rules[rs_num].active.rcount;
1387 old_array = rs->rules[rs_num].active.ptr_array;
1388 old_tree = rs->rules[rs_num].active.tree;
1389
1390 rs->rules[rs_num].active.ptr =
1391 rs->rules[rs_num].inactive.ptr;
1392 rs->rules[rs_num].active.ptr_array =
1393 rs->rules[rs_num].inactive.ptr_array;
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.ptr_array = old_array;
1424 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1425 rs->rules[rs_num].inactive.rcount = old_rcount;
1426
1427 rs->rules[rs_num].active.ticket =
1428 rs->rules[rs_num].inactive.ticket;
1429 pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1430
1431 /* Purge the old rule list. */
1432 PF_UNLNKDRULES_LOCK();
1433 while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1434 pf_unlink_rule_locked(old_rules, rule);
1435 PF_UNLNKDRULES_UNLOCK();
1436 if (rs->rules[rs_num].inactive.ptr_array)
1437 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1438 rs->rules[rs_num].inactive.ptr_array = NULL;
1439 rs->rules[rs_num].inactive.rcount = 0;
1440 rs->rules[rs_num].inactive.open = 0;
1441 pf_remove_if_empty_kruleset(rs);
1442 free(old_tree, M_TEMP);
1443
1444 return (0);
1445 }
1446
1447 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1448 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1449 {
1450 MD5_CTX ctx;
1451 struct pf_krule *rule;
1452 int rs_cnt;
1453 u_int8_t digest[PF_MD5_DIGEST_LENGTH];
1454
1455 MD5Init(&ctx);
1456 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1457 /* XXX PF_RULESET_SCRUB as well? */
1458 if (rs_cnt == PF_RULESET_SCRUB)
1459 continue;
1460
1461 if (rs->rules[rs_cnt].inactive.ptr_array)
1462 free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1463 rs->rules[rs_cnt].inactive.ptr_array = NULL;
1464
1465 if (rs->rules[rs_cnt].inactive.rcount) {
1466 rs->rules[rs_cnt].inactive.ptr_array =
1467 mallocarray(rs->rules[rs_cnt].inactive.rcount,
1468 sizeof(struct pf_rule **),
1469 M_TEMP, M_NOWAIT);
1470
1471 if (!rs->rules[rs_cnt].inactive.ptr_array)
1472 return (ENOMEM);
1473 }
1474
1475 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1476 entries) {
1477 pf_hash_rule_rolling(&ctx, rule);
1478 (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1479 }
1480 }
1481
1482 MD5Final(digest, &ctx);
1483 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1484 return (0);
1485 }
1486
1487 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1488 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1489 {
1490 int error = 0;
1491
1492 switch (addr->type) {
1493 case PF_ADDR_TABLE:
1494 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1495 if (addr->p.tbl == NULL)
1496 error = ENOMEM;
1497 break;
1498 default:
1499 error = EINVAL;
1500 }
1501
1502 return (error);
1503 }
1504
1505 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1506 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1507 sa_family_t af)
1508 {
1509 int error = 0;
1510
1511 switch (addr->type) {
1512 case PF_ADDR_TABLE:
1513 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1514 if (addr->p.tbl == NULL)
1515 error = ENOMEM;
1516 break;
1517 case PF_ADDR_DYNIFTL:
1518 error = pfi_dynaddr_setup(addr, af);
1519 break;
1520 }
1521
1522 return (error);
1523 }
1524
1525 void
pf_addr_copyout(struct pf_addr_wrap * addr)1526 pf_addr_copyout(struct pf_addr_wrap *addr)
1527 {
1528
1529 switch (addr->type) {
1530 case PF_ADDR_DYNIFTL:
1531 pfi_dynaddr_copyout(addr);
1532 break;
1533 case PF_ADDR_TABLE:
1534 pf_tbladdr_copyout(addr);
1535 break;
1536 }
1537 }
1538
1539 static void
pf_src_node_copy(const struct pf_ksrc_node * in,struct pf_src_node * out)1540 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1541 {
1542 int secs = time_uptime;
1543
1544 bzero(out, sizeof(struct pf_src_node));
1545
1546 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1547 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1548
1549 if (in->rule != NULL)
1550 out->rule.nr = in->rule->nr;
1551
1552 for (int i = 0; i < 2; i++) {
1553 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1554 out->packets[i] = counter_u64_fetch(in->packets[i]);
1555 }
1556
1557 out->states = in->states;
1558 out->conn = in->conn;
1559 out->af = in->af;
1560 out->ruletype = in->ruletype;
1561
1562 out->creation = secs - in->creation;
1563 if (out->expire > secs)
1564 out->expire -= secs;
1565 else
1566 out->expire = 0;
1567
1568 /* Adjust the connection rate estimate. */
1569 out->conn_rate.limit = in->conn_rate.limit;
1570 out->conn_rate.seconds = in->conn_rate.seconds;
1571 /* If there's no limit there's no counter_rate. */
1572 if (in->conn_rate.cr != NULL)
1573 out->conn_rate.count = counter_rate_get(in->conn_rate.cr);
1574 }
1575
1576 #ifdef ALTQ
1577 /*
1578 * Handle export of struct pf_kaltq to user binaries that may be using any
1579 * version of struct pf_altq.
1580 */
1581 static int
pf_export_kaltq(struct pf_altq * q,struct pfioc_altq_v1 * pa,size_t ioc_size)1582 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1583 {
1584 u_int32_t version;
1585
1586 if (ioc_size == sizeof(struct pfioc_altq_v0))
1587 version = 0;
1588 else
1589 version = pa->version;
1590
1591 if (version > PFIOC_ALTQ_VERSION)
1592 return (EINVAL);
1593
1594 #define ASSIGN(x) exported_q->x = q->x
1595 #define COPY(x) \
1596 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1597 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1598 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1599
1600 switch (version) {
1601 case 0: {
1602 struct pf_altq_v0 *exported_q =
1603 &((struct pfioc_altq_v0 *)pa)->altq;
1604
1605 COPY(ifname);
1606
1607 ASSIGN(scheduler);
1608 ASSIGN(tbrsize);
1609 exported_q->tbrsize = SATU16(q->tbrsize);
1610 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1611
1612 COPY(qname);
1613 COPY(parent);
1614 ASSIGN(parent_qid);
1615 exported_q->bandwidth = SATU32(q->bandwidth);
1616 ASSIGN(priority);
1617 ASSIGN(local_flags);
1618
1619 ASSIGN(qlimit);
1620 ASSIGN(flags);
1621
1622 if (q->scheduler == ALTQT_HFSC) {
1623 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1624 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1625 SATU32(q->pq_u.hfsc_opts.x)
1626
1627 ASSIGN_OPT_SATU32(rtsc_m1);
1628 ASSIGN_OPT(rtsc_d);
1629 ASSIGN_OPT_SATU32(rtsc_m2);
1630
1631 ASSIGN_OPT_SATU32(lssc_m1);
1632 ASSIGN_OPT(lssc_d);
1633 ASSIGN_OPT_SATU32(lssc_m2);
1634
1635 ASSIGN_OPT_SATU32(ulsc_m1);
1636 ASSIGN_OPT(ulsc_d);
1637 ASSIGN_OPT_SATU32(ulsc_m2);
1638
1639 ASSIGN_OPT(flags);
1640
1641 #undef ASSIGN_OPT
1642 #undef ASSIGN_OPT_SATU32
1643 } else
1644 COPY(pq_u);
1645
1646 ASSIGN(qid);
1647 break;
1648 }
1649 case 1: {
1650 struct pf_altq_v1 *exported_q =
1651 &((struct pfioc_altq_v1 *)pa)->altq;
1652
1653 COPY(ifname);
1654
1655 ASSIGN(scheduler);
1656 ASSIGN(tbrsize);
1657 ASSIGN(ifbandwidth);
1658
1659 COPY(qname);
1660 COPY(parent);
1661 ASSIGN(parent_qid);
1662 ASSIGN(bandwidth);
1663 ASSIGN(priority);
1664 ASSIGN(local_flags);
1665
1666 ASSIGN(qlimit);
1667 ASSIGN(flags);
1668 COPY(pq_u);
1669
1670 ASSIGN(qid);
1671 break;
1672 }
1673 default:
1674 panic("%s: unhandled struct pfioc_altq version", __func__);
1675 break;
1676 }
1677
1678 #undef ASSIGN
1679 #undef COPY
1680 #undef SATU16
1681 #undef SATU32
1682
1683 return (0);
1684 }
1685
1686 /*
1687 * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1688 * that may be using any version of it.
1689 */
1690 static int
pf_import_kaltq(struct pfioc_altq_v1 * pa,struct pf_altq * q,size_t ioc_size)1691 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1692 {
1693 u_int32_t version;
1694
1695 if (ioc_size == sizeof(struct pfioc_altq_v0))
1696 version = 0;
1697 else
1698 version = pa->version;
1699
1700 if (version > PFIOC_ALTQ_VERSION)
1701 return (EINVAL);
1702
1703 #define ASSIGN(x) q->x = imported_q->x
1704 #define COPY(x) \
1705 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1706
1707 switch (version) {
1708 case 0: {
1709 struct pf_altq_v0 *imported_q =
1710 &((struct pfioc_altq_v0 *)pa)->altq;
1711
1712 COPY(ifname);
1713
1714 ASSIGN(scheduler);
1715 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1716 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1717
1718 COPY(qname);
1719 COPY(parent);
1720 ASSIGN(parent_qid);
1721 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1722 ASSIGN(priority);
1723 ASSIGN(local_flags);
1724
1725 ASSIGN(qlimit);
1726 ASSIGN(flags);
1727
1728 if (imported_q->scheduler == ALTQT_HFSC) {
1729 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1730
1731 /*
1732 * The m1 and m2 parameters are being copied from
1733 * 32-bit to 64-bit.
1734 */
1735 ASSIGN_OPT(rtsc_m1);
1736 ASSIGN_OPT(rtsc_d);
1737 ASSIGN_OPT(rtsc_m2);
1738
1739 ASSIGN_OPT(lssc_m1);
1740 ASSIGN_OPT(lssc_d);
1741 ASSIGN_OPT(lssc_m2);
1742
1743 ASSIGN_OPT(ulsc_m1);
1744 ASSIGN_OPT(ulsc_d);
1745 ASSIGN_OPT(ulsc_m2);
1746
1747 ASSIGN_OPT(flags);
1748
1749 #undef ASSIGN_OPT
1750 } else
1751 COPY(pq_u);
1752
1753 ASSIGN(qid);
1754 break;
1755 }
1756 case 1: {
1757 struct pf_altq_v1 *imported_q =
1758 &((struct pfioc_altq_v1 *)pa)->altq;
1759
1760 COPY(ifname);
1761
1762 ASSIGN(scheduler);
1763 ASSIGN(tbrsize);
1764 ASSIGN(ifbandwidth);
1765
1766 COPY(qname);
1767 COPY(parent);
1768 ASSIGN(parent_qid);
1769 ASSIGN(bandwidth);
1770 ASSIGN(priority);
1771 ASSIGN(local_flags);
1772
1773 ASSIGN(qlimit);
1774 ASSIGN(flags);
1775 COPY(pq_u);
1776
1777 ASSIGN(qid);
1778 break;
1779 }
1780 default:
1781 panic("%s: unhandled struct pfioc_altq version", __func__);
1782 break;
1783 }
1784
1785 #undef ASSIGN
1786 #undef COPY
1787
1788 return (0);
1789 }
1790
1791 static struct pf_altq *
pf_altq_get_nth_active(u_int32_t n)1792 pf_altq_get_nth_active(u_int32_t n)
1793 {
1794 struct pf_altq *altq;
1795 u_int32_t nr;
1796
1797 nr = 0;
1798 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1799 if (nr == n)
1800 return (altq);
1801 nr++;
1802 }
1803
1804 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1805 if (nr == n)
1806 return (altq);
1807 nr++;
1808 }
1809
1810 return (NULL);
1811 }
1812 #endif /* ALTQ */
1813
1814 struct pf_krule *
pf_krule_alloc(void)1815 pf_krule_alloc(void)
1816 {
1817 struct pf_krule *rule;
1818
1819 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1820 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF);
1821 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF);
1822 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF);
1823 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1824 M_WAITOK | M_ZERO);
1825 return (rule);
1826 }
1827
1828 void
pf_krule_free(struct pf_krule * rule)1829 pf_krule_free(struct pf_krule *rule)
1830 {
1831 #ifdef PF_WANT_32_TO_64_COUNTER
1832 bool wowned;
1833 #endif
1834
1835 if (rule == NULL)
1836 return;
1837
1838 #ifdef PF_WANT_32_TO_64_COUNTER
1839 if (rule->allrulelinked) {
1840 wowned = PF_RULES_WOWNED();
1841 if (!wowned)
1842 PF_RULES_WLOCK();
1843 LIST_REMOVE(rule, allrulelist);
1844 V_pf_allrulecount--;
1845 if (!wowned)
1846 PF_RULES_WUNLOCK();
1847 }
1848 #endif
1849
1850 pf_counter_u64_deinit(&rule->evaluations);
1851 for (int i = 0; i < 2; i++) {
1852 pf_counter_u64_deinit(&rule->packets[i]);
1853 pf_counter_u64_deinit(&rule->bytes[i]);
1854 }
1855 counter_u64_free(rule->states_cur);
1856 counter_u64_free(rule->states_tot);
1857 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
1858 counter_u64_free(rule->src_nodes[sn_type]);
1859 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1860
1861 mtx_destroy(&rule->nat.mtx);
1862 mtx_destroy(&rule->rdr.mtx);
1863 mtx_destroy(&rule->route.mtx);
1864 free(rule, M_PFRULE);
1865 }
1866
1867 void
pf_krule_clear_counters(struct pf_krule * rule)1868 pf_krule_clear_counters(struct pf_krule *rule)
1869 {
1870 pf_counter_u64_zero(&rule->evaluations);
1871 for (int i = 0; i < 2; i++) {
1872 pf_counter_u64_zero(&rule->packets[i]);
1873 pf_counter_u64_zero(&rule->bytes[i]);
1874 }
1875 counter_u64_zero(rule->states_tot);
1876 }
1877
1878 static void
pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr * kpool,struct pf_pooladdr * pool)1879 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1880 struct pf_pooladdr *pool)
1881 {
1882
1883 bzero(pool, sizeof(*pool));
1884 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1885 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1886 }
1887
1888 static int
pf_pooladdr_to_kpooladdr(const struct pf_pooladdr * pool,struct pf_kpooladdr * kpool)1889 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1890 struct pf_kpooladdr *kpool)
1891 {
1892 int ret;
1893
1894 bzero(kpool, sizeof(*kpool));
1895 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1896 ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1897 sizeof(kpool->ifname));
1898 return (ret);
1899 }
1900
1901 static void
pf_pool_to_kpool(const struct pf_pool * pool,struct pf_kpool * kpool)1902 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1903 {
1904 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1905 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1906
1907 bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1908 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1909
1910 kpool->tblidx = pool->tblidx;
1911 kpool->proxy_port[0] = pool->proxy_port[0];
1912 kpool->proxy_port[1] = pool->proxy_port[1];
1913 kpool->opts = pool->opts;
1914 }
1915
1916 static int
pf_rule_to_krule(const struct pf_rule * rule,struct pf_krule * krule)1917 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1918 {
1919 int ret;
1920
1921 #ifndef INET
1922 if (rule->af == AF_INET) {
1923 return (EAFNOSUPPORT);
1924 }
1925 #endif /* INET */
1926 #ifndef INET6
1927 if (rule->af == AF_INET6) {
1928 return (EAFNOSUPPORT);
1929 }
1930 #endif /* INET6 */
1931
1932 ret = pf_check_rule_addr(&rule->src);
1933 if (ret != 0)
1934 return (ret);
1935 ret = pf_check_rule_addr(&rule->dst);
1936 if (ret != 0)
1937 return (ret);
1938
1939 bcopy(&rule->src, &krule->src, sizeof(rule->src));
1940 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1941
1942 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1943 if (ret != 0)
1944 return (ret);
1945 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1946 if (ret != 0)
1947 return (ret);
1948 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1949 if (ret != 0)
1950 return (ret);
1951 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1952 if (ret != 0)
1953 return (ret);
1954 ret = pf_user_strcpy(krule->tagname, rule->tagname,
1955 sizeof(rule->tagname));
1956 if (ret != 0)
1957 return (ret);
1958 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1959 sizeof(rule->match_tagname));
1960 if (ret != 0)
1961 return (ret);
1962 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1963 sizeof(rule->overload_tblname));
1964 if (ret != 0)
1965 return (ret);
1966
1967 pf_pool_to_kpool(&rule->rpool, &krule->rdr);
1968
1969 /* Don't allow userspace to set evaluations, packets or bytes. */
1970 /* kif, anchor, overload_tbl are not copied over. */
1971
1972 krule->os_fingerprint = rule->os_fingerprint;
1973
1974 krule->rtableid = rule->rtableid;
1975 /* pf_rule->timeout is smaller than pf_krule->timeout */
1976 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1977 krule->max_states = rule->max_states;
1978 krule->max_src_nodes = rule->max_src_nodes;
1979 krule->max_src_states = rule->max_src_states;
1980 krule->max_src_conn = rule->max_src_conn;
1981 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1982 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1983 krule->qid = rule->qid;
1984 krule->pqid = rule->pqid;
1985 krule->nr = rule->nr;
1986 krule->prob = rule->prob;
1987 krule->cuid = rule->cuid;
1988 krule->cpid = rule->cpid;
1989
1990 krule->return_icmp = rule->return_icmp;
1991 krule->return_icmp6 = rule->return_icmp6;
1992 krule->max_mss = rule->max_mss;
1993 krule->tag = rule->tag;
1994 krule->match_tag = rule->match_tag;
1995 krule->scrub_flags = rule->scrub_flags;
1996
1997 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1998 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1999
2000 krule->rule_flag = rule->rule_flag;
2001 krule->action = rule->action;
2002 krule->direction = rule->direction;
2003 krule->log = rule->log;
2004 krule->logif = rule->logif;
2005 krule->quick = rule->quick;
2006 krule->ifnot = rule->ifnot;
2007 krule->match_tag_not = rule->match_tag_not;
2008 krule->natpass = rule->natpass;
2009
2010 krule->keep_state = rule->keep_state;
2011 krule->af = rule->af;
2012 krule->proto = rule->proto;
2013 krule->type = rule->type;
2014 krule->code = rule->code;
2015 krule->flags = rule->flags;
2016 krule->flagset = rule->flagset;
2017 krule->min_ttl = rule->min_ttl;
2018 krule->allow_opts = rule->allow_opts;
2019 krule->rt = rule->rt;
2020 krule->return_ttl = rule->return_ttl;
2021 krule->tos = rule->tos;
2022 krule->set_tos = rule->set_tos;
2023
2024 krule->flush = rule->flush;
2025 krule->prio = rule->prio;
2026 krule->set_prio[0] = rule->set_prio[0];
2027 krule->set_prio[1] = rule->set_prio[1];
2028
2029 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2030
2031 return (0);
2032 }
2033
2034 int
pf_ioctl_getrules(struct pfioc_rule * pr)2035 pf_ioctl_getrules(struct pfioc_rule *pr)
2036 {
2037 struct pf_kruleset *ruleset;
2038 struct pf_krule *tail;
2039 int rs_num;
2040
2041 PF_RULES_WLOCK();
2042 ruleset = pf_find_kruleset(pr->anchor);
2043 if (ruleset == NULL) {
2044 PF_RULES_WUNLOCK();
2045 return (EINVAL);
2046 }
2047 rs_num = pf_get_ruleset_number(pr->rule.action);
2048 if (rs_num >= PF_RULESET_MAX) {
2049 PF_RULES_WUNLOCK();
2050 return (EINVAL);
2051 }
2052 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2053 pf_krulequeue);
2054 if (tail)
2055 pr->nr = tail->nr + 1;
2056 else
2057 pr->nr = 0;
2058 pr->ticket = ruleset->rules[rs_num].active.ticket;
2059 PF_RULES_WUNLOCK();
2060
2061 return (0);
2062 }
2063
2064 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)2065 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2066 uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2067 uid_t uid, pid_t pid)
2068 {
2069 struct pf_kruleset *ruleset;
2070 struct pf_krule *tail;
2071 struct pf_kpooladdr *pa;
2072 struct pfi_kkif *kif = NULL, *rcv_kif = NULL;
2073 int rs_num;
2074 int error = 0;
2075
2076 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2077 error = EINVAL;
2078 goto errout_unlocked;
2079 }
2080
2081 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
2082
2083 if (rule->ifname[0])
2084 kif = pf_kkif_create(M_WAITOK);
2085 if (rule->rcv_ifname[0])
2086 rcv_kif = pf_kkif_create(M_WAITOK);
2087 pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2088 for (int i = 0; i < 2; i++) {
2089 pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2090 pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2091 }
2092 rule->states_cur = counter_u64_alloc(M_WAITOK);
2093 rule->states_tot = counter_u64_alloc(M_WAITOK);
2094 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
2095 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
2096 rule->cuid = uid;
2097 rule->cpid = pid;
2098 TAILQ_INIT(&rule->rdr.list);
2099 TAILQ_INIT(&rule->nat.list);
2100 TAILQ_INIT(&rule->route.list);
2101
2102 PF_CONFIG_LOCK();
2103 PF_RULES_WLOCK();
2104 #ifdef PF_WANT_32_TO_64_COUNTER
2105 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2106 MPASS(!rule->allrulelinked);
2107 rule->allrulelinked = true;
2108 V_pf_allrulecount++;
2109 #endif
2110 ruleset = pf_find_kruleset(anchor);
2111 if (ruleset == NULL)
2112 ERROUT(EINVAL);
2113 rs_num = pf_get_ruleset_number(rule->action);
2114 if (rs_num >= PF_RULESET_MAX)
2115 ERROUT(EINVAL);
2116 if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2117 DPFPRINTF(PF_DEBUG_MISC,
2118 ("ticket: %d != [%d]%d\n", ticket, rs_num,
2119 ruleset->rules[rs_num].inactive.ticket));
2120 ERROUT(EBUSY);
2121 }
2122 if (pool_ticket != V_ticket_pabuf) {
2123 DPFPRINTF(PF_DEBUG_MISC,
2124 ("pool_ticket: %d != %d\n", pool_ticket,
2125 V_ticket_pabuf));
2126 ERROUT(EBUSY);
2127 }
2128 /*
2129 * XXXMJG hack: there is no mechanism to ensure they started the
2130 * transaction. Ticket checked above may happen to match by accident,
2131 * even if nobody called DIOCXBEGIN, let alone this process.
2132 * Partially work around it by checking if the RB tree got allocated,
2133 * see pf_begin_rules.
2134 */
2135 if (ruleset->rules[rs_num].inactive.tree == NULL) {
2136 ERROUT(EINVAL);
2137 }
2138
2139 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2140 pf_krulequeue);
2141 if (tail)
2142 rule->nr = tail->nr + 1;
2143 else
2144 rule->nr = 0;
2145 if (rule->ifname[0]) {
2146 rule->kif = pfi_kkif_attach(kif, rule->ifname);
2147 kif = NULL;
2148 pfi_kkif_ref(rule->kif);
2149 } else
2150 rule->kif = NULL;
2151
2152 if (rule->rcv_ifname[0]) {
2153 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname);
2154 rcv_kif = NULL;
2155 pfi_kkif_ref(rule->rcv_kif);
2156 } else
2157 rule->rcv_kif = NULL;
2158
2159 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2160 ERROUT(EBUSY);
2161 #ifdef ALTQ
2162 /* set queue IDs */
2163 if (rule->qname[0] != 0) {
2164 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2165 ERROUT(EBUSY);
2166 else if (rule->pqname[0] != 0) {
2167 if ((rule->pqid =
2168 pf_qname2qid(rule->pqname)) == 0)
2169 ERROUT(EBUSY);
2170 } else
2171 rule->pqid = rule->qid;
2172 }
2173 #endif
2174 if (rule->tagname[0])
2175 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2176 ERROUT(EBUSY);
2177 if (rule->match_tagname[0])
2178 if ((rule->match_tag =
2179 pf_tagname2tag(rule->match_tagname)) == 0)
2180 ERROUT(EBUSY);
2181 if (rule->rt && !rule->direction)
2182 ERROUT(EINVAL);
2183 if (!rule->log)
2184 rule->logif = 0;
2185 if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit,
2186 rule->pktrate.seconds))
2187 ERROUT(ENOMEM);
2188 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2189 ERROUT(ENOMEM);
2190 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2191 ERROUT(ENOMEM);
2192 if (pf_kanchor_setup(rule, ruleset, anchor_call))
2193 ERROUT(EINVAL);
2194 if (rule->scrub_flags & PFSTATE_SETPRIO &&
2195 (rule->set_prio[0] > PF_PRIO_MAX ||
2196 rule->set_prio[1] > PF_PRIO_MAX))
2197 ERROUT(EINVAL);
2198 for (int i = 0; i < 3; i++) {
2199 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
2200 if (pa->addr.type == PF_ADDR_TABLE) {
2201 pa->addr.p.tbl = pfr_attach_table(ruleset,
2202 pa->addr.v.tblname);
2203 if (pa->addr.p.tbl == NULL)
2204 ERROUT(ENOMEM);
2205 }
2206 }
2207
2208 rule->overload_tbl = NULL;
2209 if (rule->overload_tblname[0]) {
2210 if ((rule->overload_tbl = pfr_attach_table(ruleset,
2211 rule->overload_tblname)) == NULL)
2212 ERROUT(EINVAL);
2213 else
2214 rule->overload_tbl->pfrkt_flags |=
2215 PFR_TFLAG_ACTIVE;
2216 }
2217
2218 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list);
2219
2220 /*
2221 * Old version of pfctl provide route redirection pools in single
2222 * common redirection pool rdr. New versions use rdr only for
2223 * rdr-to rules.
2224 */
2225 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) {
2226 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list);
2227 } else {
2228 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list);
2229 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list);
2230 }
2231
2232 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2233 (rule->action == PF_BINAT)) && rule->anchor == NULL &&
2234 TAILQ_FIRST(&rule->rdr.list) == NULL) {
2235 ERROUT(EINVAL);
2236 }
2237
2238 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) {
2239 ERROUT(EINVAL);
2240 }
2241
2242 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR ||
2243 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) {
2244 ERROUT(EINVAL);
2245 }
2246
2247 MPASS(error == 0);
2248
2249 rule->nat.cur = TAILQ_FIRST(&rule->nat.list);
2250 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list);
2251 rule->route.cur = TAILQ_FIRST(&rule->route.list);
2252 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2253 rule, entries);
2254 ruleset->rules[rs_num].inactive.rcount++;
2255
2256 PF_RULES_WUNLOCK();
2257 pf_hash_rule(rule);
2258 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2259 PF_RULES_WLOCK();
2260 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2261 ruleset->rules[rs_num].inactive.rcount--;
2262 pf_free_rule(rule);
2263 rule = NULL;
2264 ERROUT(EEXIST);
2265 }
2266 PF_CONFIG_UNLOCK();
2267
2268 return (0);
2269
2270 #undef ERROUT
2271 errout:
2272 PF_RULES_WUNLOCK();
2273 PF_CONFIG_UNLOCK();
2274 errout_unlocked:
2275 pf_kkif_free(rcv_kif);
2276 pf_kkif_free(kif);
2277 pf_krule_free(rule);
2278 return (error);
2279 }
2280
2281 static bool
pf_label_match(const struct pf_krule * rule,const char * label)2282 pf_label_match(const struct pf_krule *rule, const char *label)
2283 {
2284 int i = 0;
2285
2286 while (*rule->label[i]) {
2287 if (strcmp(rule->label[i], label) == 0)
2288 return (true);
2289 i++;
2290 }
2291
2292 return (false);
2293 }
2294
2295 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)2296 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2297 {
2298 struct pf_kstate *s;
2299 int more = 0;
2300
2301 s = pf_find_state_all(key, dir, &more);
2302 if (s == NULL)
2303 return (0);
2304
2305 if (more) {
2306 PF_STATE_UNLOCK(s);
2307 return (0);
2308 }
2309
2310 pf_remove_state(s);
2311 return (1);
2312 }
2313
2314 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)2315 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2316 {
2317 struct pf_kstate *s;
2318 struct pf_state_key *sk;
2319 struct pf_addr *srcaddr, *dstaddr;
2320 struct pf_state_key_cmp match_key;
2321 int idx, killed = 0;
2322 unsigned int dir;
2323 u_int16_t srcport, dstport;
2324 struct pfi_kkif *kif;
2325
2326 relock_DIOCKILLSTATES:
2327 PF_HASHROW_LOCK(ih);
2328 LIST_FOREACH(s, &ih->states, entry) {
2329 /* For floating states look at the original kif. */
2330 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2331
2332 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2333 if (s->direction == PF_OUT) {
2334 srcaddr = &sk->addr[1];
2335 dstaddr = &sk->addr[0];
2336 srcport = sk->port[1];
2337 dstport = sk->port[0];
2338 } else {
2339 srcaddr = &sk->addr[0];
2340 dstaddr = &sk->addr[1];
2341 srcport = sk->port[0];
2342 dstport = sk->port[1];
2343 }
2344
2345 if (psk->psk_af && sk->af != psk->psk_af)
2346 continue;
2347
2348 if (psk->psk_proto && psk->psk_proto != sk->proto)
2349 continue;
2350
2351 if (! pf_match_addr(psk->psk_src.neg,
2352 &psk->psk_src.addr.v.a.addr,
2353 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2354 continue;
2355
2356 if (! pf_match_addr(psk->psk_dst.neg,
2357 &psk->psk_dst.addr.v.a.addr,
2358 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2359 continue;
2360
2361 if (! pf_match_addr(psk->psk_rt_addr.neg,
2362 &psk->psk_rt_addr.addr.v.a.addr,
2363 &psk->psk_rt_addr.addr.v.a.mask,
2364 &s->act.rt_addr, sk->af))
2365 continue;
2366
2367 if (psk->psk_src.port_op != 0 &&
2368 ! pf_match_port(psk->psk_src.port_op,
2369 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2370 continue;
2371
2372 if (psk->psk_dst.port_op != 0 &&
2373 ! pf_match_port(psk->psk_dst.port_op,
2374 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2375 continue;
2376
2377 if (psk->psk_label[0] &&
2378 ! pf_label_match(s->rule, psk->psk_label))
2379 continue;
2380
2381 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2382 kif->pfik_name))
2383 continue;
2384
2385 if (psk->psk_kill_match) {
2386 /* Create the key to find matching states, with lock
2387 * held. */
2388
2389 bzero(&match_key, sizeof(match_key));
2390
2391 if (s->direction == PF_OUT) {
2392 dir = PF_IN;
2393 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2394 } else {
2395 dir = PF_OUT;
2396 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2397 }
2398
2399 match_key.af = s->key[idx]->af;
2400 match_key.proto = s->key[idx]->proto;
2401 pf_addrcpy(&match_key.addr[0],
2402 &s->key[idx]->addr[1], match_key.af);
2403 match_key.port[0] = s->key[idx]->port[1];
2404 pf_addrcpy(&match_key.addr[1],
2405 &s->key[idx]->addr[0], match_key.af);
2406 match_key.port[1] = s->key[idx]->port[0];
2407 }
2408
2409 pf_remove_state(s);
2410 killed++;
2411
2412 if (psk->psk_kill_match)
2413 killed += pf_kill_matching_state(&match_key, dir);
2414
2415 goto relock_DIOCKILLSTATES;
2416 }
2417 PF_HASHROW_UNLOCK(ih);
2418
2419 return (killed);
2420 }
2421
2422 void
unhandled_af(int af)2423 unhandled_af(int af)
2424 {
2425 panic("unhandled af %d", af);
2426 }
2427
2428 int
pf_start(void)2429 pf_start(void)
2430 {
2431 int error = 0;
2432
2433 sx_xlock(&V_pf_ioctl_lock);
2434 if (V_pf_status.running)
2435 error = EEXIST;
2436 else {
2437 hook_pf();
2438 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2439 hook_pf_eth();
2440 V_pf_status.running = 1;
2441 V_pf_status.since = time_uptime;
2442 new_unrhdr64(&V_pf_stateid, time_second);
2443
2444 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2445 }
2446 sx_xunlock(&V_pf_ioctl_lock);
2447
2448 return (error);
2449 }
2450
2451 int
pf_stop(void)2452 pf_stop(void)
2453 {
2454 int error = 0;
2455
2456 sx_xlock(&V_pf_ioctl_lock);
2457 if (!V_pf_status.running)
2458 error = ENOENT;
2459 else {
2460 V_pf_status.running = 0;
2461 dehook_pf();
2462 dehook_pf_eth();
2463 V_pf_status.since = time_uptime;
2464 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2465 }
2466 sx_xunlock(&V_pf_ioctl_lock);
2467
2468 return (error);
2469 }
2470
2471 void
pf_ioctl_clear_status(void)2472 pf_ioctl_clear_status(void)
2473 {
2474 PF_RULES_WLOCK();
2475 for (int i = 0; i < PFRES_MAX; i++)
2476 counter_u64_zero(V_pf_status.counters[i]);
2477 for (int i = 0; i < FCNT_MAX; i++)
2478 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2479 for (int i = 0; i < SCNT_MAX; i++)
2480 counter_u64_zero(V_pf_status.scounters[i]);
2481 for (int i = 0; i < KLCNT_MAX; i++)
2482 counter_u64_zero(V_pf_status.lcounters[i]);
2483 V_pf_status.since = time_uptime;
2484 if (*V_pf_status.ifname)
2485 pfi_update_status(V_pf_status.ifname, NULL);
2486 PF_RULES_WUNLOCK();
2487 }
2488
2489 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2490 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2491 {
2492 uint32_t old;
2493
2494 if (timeout < 0 || timeout >= PFTM_MAX ||
2495 seconds < 0)
2496 return (EINVAL);
2497
2498 PF_RULES_WLOCK();
2499 old = V_pf_default_rule.timeout[timeout];
2500 if (timeout == PFTM_INTERVAL && seconds == 0)
2501 seconds = 1;
2502 V_pf_default_rule.timeout[timeout] = seconds;
2503 if (timeout == PFTM_INTERVAL && seconds < old)
2504 wakeup(pf_purge_thread);
2505
2506 if (prev_seconds != NULL)
2507 *prev_seconds = old;
2508
2509 PF_RULES_WUNLOCK();
2510
2511 return (0);
2512 }
2513
2514 int
pf_ioctl_get_timeout(int timeout,int * seconds)2515 pf_ioctl_get_timeout(int timeout, int *seconds)
2516 {
2517 PF_RULES_RLOCK_TRACKER;
2518
2519 if (timeout < 0 || timeout >= PFTM_MAX)
2520 return (EINVAL);
2521
2522 PF_RULES_RLOCK();
2523 *seconds = V_pf_default_rule.timeout[timeout];
2524 PF_RULES_RUNLOCK();
2525
2526 return (0);
2527 }
2528
2529 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2530 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2531 {
2532
2533 PF_RULES_WLOCK();
2534 if (index < 0 || index >= PF_LIMIT_MAX ||
2535 V_pf_limits[index].zone == NULL) {
2536 PF_RULES_WUNLOCK();
2537 return (EINVAL);
2538 }
2539 uma_zone_set_max(V_pf_limits[index].zone,
2540 limit == 0 ? INT_MAX : limit);
2541 if (old_limit != NULL)
2542 *old_limit = V_pf_limits[index].limit;
2543 V_pf_limits[index].limit = limit;
2544 PF_RULES_WUNLOCK();
2545
2546 return (0);
2547 }
2548
2549 int
pf_ioctl_get_limit(int index,unsigned int * limit)2550 pf_ioctl_get_limit(int index, unsigned int *limit)
2551 {
2552 PF_RULES_RLOCK_TRACKER;
2553
2554 if (index < 0 || index >= PF_LIMIT_MAX)
2555 return (EINVAL);
2556
2557 PF_RULES_RLOCK();
2558 *limit = V_pf_limits[index].limit;
2559 PF_RULES_RUNLOCK();
2560
2561 return (0);
2562 }
2563
2564 int
pf_ioctl_begin_addrs(uint32_t * ticket)2565 pf_ioctl_begin_addrs(uint32_t *ticket)
2566 {
2567 PF_RULES_WLOCK();
2568 pf_empty_kpool(&V_pf_pabuf[0]);
2569 pf_empty_kpool(&V_pf_pabuf[1]);
2570 pf_empty_kpool(&V_pf_pabuf[2]);
2571 *ticket = ++V_ticket_pabuf;
2572 PF_RULES_WUNLOCK();
2573
2574 return (0);
2575 }
2576
2577 int
pf_ioctl_add_addr(struct pf_nl_pooladdr * pp)2578 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
2579 {
2580 struct pf_kpooladdr *pa = NULL;
2581 struct pfi_kkif *kif = NULL;
2582 int error;
2583
2584 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2585 pp->which != PF_RT)
2586 return (EINVAL);
2587
2588 switch (pp->af) {
2589 #ifdef INET
2590 case AF_INET:
2591 /* FALLTHROUGH */
2592 #endif /* INET */
2593 #ifdef INET6
2594 case AF_INET6:
2595 /* FALLTHROUGH */
2596 #endif /* INET6 */
2597 case AF_UNSPEC:
2598 break;
2599 default:
2600 return (EAFNOSUPPORT);
2601 }
2602
2603 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2604 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2605 pp->addr.addr.type != PF_ADDR_TABLE)
2606 return (EINVAL);
2607
2608 if (pp->addr.addr.p.dyn != NULL)
2609 return (EINVAL);
2610
2611 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2612 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2613 if (error != 0)
2614 goto out;
2615 if (pa->ifname[0])
2616 kif = pf_kkif_create(M_WAITOK);
2617 PF_RULES_WLOCK();
2618 if (pp->ticket != V_ticket_pabuf) {
2619 PF_RULES_WUNLOCK();
2620 if (pa->ifname[0])
2621 pf_kkif_free(kif);
2622 error = EBUSY;
2623 goto out;
2624 }
2625 if (pa->ifname[0]) {
2626 pa->kif = pfi_kkif_attach(kif, pa->ifname);
2627 kif = NULL;
2628 pfi_kkif_ref(pa->kif);
2629 } else
2630 pa->kif = NULL;
2631 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2632 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2633 if (pa->ifname[0])
2634 pfi_kkif_unref(pa->kif);
2635 PF_RULES_WUNLOCK();
2636 goto out;
2637 }
2638 switch (pp->which) {
2639 case PF_NAT:
2640 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries);
2641 break;
2642 case PF_RDR:
2643 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries);
2644 break;
2645 case PF_RT:
2646 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries);
2647 break;
2648 }
2649 PF_RULES_WUNLOCK();
2650
2651 return (0);
2652
2653 out:
2654 free(pa, M_PFRULE);
2655 return (error);
2656 }
2657
2658 int
pf_ioctl_get_addrs(struct pf_nl_pooladdr * pp)2659 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
2660 {
2661 struct pf_kpool *pool;
2662 struct pf_kpooladdr *pa;
2663
2664 PF_RULES_RLOCK_TRACKER;
2665
2666 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2667 pp->which != PF_RT)
2668 return (EINVAL);
2669
2670 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2671 pp->nr = 0;
2672
2673 PF_RULES_RLOCK();
2674 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2675 pp->r_num, 0, 1, 0, pp->which);
2676 if (pool == NULL) {
2677 PF_RULES_RUNLOCK();
2678 return (EBUSY);
2679 }
2680 TAILQ_FOREACH(pa, &pool->list, entries)
2681 pp->nr++;
2682 PF_RULES_RUNLOCK();
2683
2684 return (0);
2685 }
2686
2687 int
pf_ioctl_get_addr(struct pf_nl_pooladdr * pp)2688 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
2689 {
2690 struct pf_kpool *pool;
2691 struct pf_kpooladdr *pa;
2692 u_int32_t nr = 0;
2693
2694 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2695 pp->which != PF_RT)
2696 return (EINVAL);
2697
2698 PF_RULES_RLOCK_TRACKER;
2699
2700 pp->anchor[sizeof(pp->anchor) - 1] = '\0';
2701
2702 PF_RULES_RLOCK();
2703 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2704 pp->r_num, 0, 1, 1, pp->which);
2705 if (pool == NULL) {
2706 PF_RULES_RUNLOCK();
2707 return (EBUSY);
2708 }
2709 pa = TAILQ_FIRST(&pool->list);
2710 while ((pa != NULL) && (nr < pp->nr)) {
2711 pa = TAILQ_NEXT(pa, entries);
2712 nr++;
2713 }
2714 if (pa == NULL) {
2715 PF_RULES_RUNLOCK();
2716 return (EBUSY);
2717 }
2718 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2719 pf_addr_copyout(&pp->addr.addr);
2720 PF_RULES_RUNLOCK();
2721
2722 return (0);
2723 }
2724
2725 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)2726 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
2727 {
2728 struct pf_kruleset *ruleset;
2729 struct pf_kanchor *anchor;
2730
2731 PF_RULES_RLOCK_TRACKER;
2732
2733 pr->path[sizeof(pr->path) - 1] = '\0';
2734
2735 PF_RULES_RLOCK();
2736 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2737 PF_RULES_RUNLOCK();
2738 return (ENOENT);
2739 }
2740 pr->nr = 0;
2741 if (ruleset == &pf_main_ruleset) {
2742 /* XXX kludge for pf_main_ruleset */
2743 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2744 if (anchor->parent == NULL)
2745 pr->nr++;
2746 } else {
2747 RB_FOREACH(anchor, pf_kanchor_node,
2748 &ruleset->anchor->children)
2749 pr->nr++;
2750 }
2751 PF_RULES_RUNLOCK();
2752
2753 return (0);
2754 }
2755
2756 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)2757 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
2758 {
2759 struct pf_kruleset *ruleset;
2760 struct pf_kanchor *anchor;
2761 u_int32_t nr = 0;
2762 int error = 0;
2763
2764 PF_RULES_RLOCK_TRACKER;
2765
2766 PF_RULES_RLOCK();
2767 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2768 PF_RULES_RUNLOCK();
2769 return (ENOENT);
2770 }
2771
2772 pr->name[0] = '\0';
2773 if (ruleset == &pf_main_ruleset) {
2774 /* XXX kludge for pf_main_ruleset */
2775 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2776 if (anchor->parent == NULL && nr++ == pr->nr) {
2777 strlcpy(pr->name, anchor->name,
2778 sizeof(pr->name));
2779 break;
2780 }
2781 } else {
2782 RB_FOREACH(anchor, pf_kanchor_node,
2783 &ruleset->anchor->children)
2784 if (nr++ == pr->nr) {
2785 strlcpy(pr->name, anchor->name,
2786 sizeof(pr->name));
2787 break;
2788 }
2789 }
2790 if (!pr->name[0])
2791 error = EBUSY;
2792 PF_RULES_RUNLOCK();
2793
2794 return (error);
2795 }
2796
2797 int
pf_ioctl_natlook(struct pfioc_natlook * pnl)2798 pf_ioctl_natlook(struct pfioc_natlook *pnl)
2799 {
2800 struct pf_state_key *sk;
2801 struct pf_kstate *state;
2802 struct pf_state_key_cmp key;
2803 int m = 0, direction = pnl->direction;
2804 int sidx, didx;
2805
2806 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
2807 sidx = (direction == PF_IN) ? 1 : 0;
2808 didx = (direction == PF_IN) ? 0 : 1;
2809
2810 if (!pnl->proto ||
2811 PF_AZERO(&pnl->saddr, pnl->af) ||
2812 PF_AZERO(&pnl->daddr, pnl->af) ||
2813 ((pnl->proto == IPPROTO_TCP ||
2814 pnl->proto == IPPROTO_UDP) &&
2815 (!pnl->dport || !pnl->sport)))
2816 return (EINVAL);
2817
2818 switch (pnl->direction) {
2819 case PF_IN:
2820 case PF_OUT:
2821 case PF_INOUT:
2822 break;
2823 default:
2824 return (EINVAL);
2825 }
2826
2827 switch (pnl->af) {
2828 #ifdef INET
2829 case AF_INET:
2830 break;
2831 #endif /* INET */
2832 #ifdef INET6
2833 case AF_INET6:
2834 break;
2835 #endif /* INET6 */
2836 default:
2837 return (EAFNOSUPPORT);
2838 }
2839
2840 bzero(&key, sizeof(key));
2841 key.af = pnl->af;
2842 key.proto = pnl->proto;
2843 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af);
2844 key.port[sidx] = pnl->sport;
2845 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af);
2846 key.port[didx] = pnl->dport;
2847
2848 state = pf_find_state_all(&key, direction, &m);
2849 if (state == NULL)
2850 return (ENOENT);
2851
2852 if (m > 1) {
2853 PF_STATE_UNLOCK(state);
2854 return (E2BIG); /* more than one state */
2855 }
2856
2857 sk = state->key[sidx];
2858 pf_addrcpy(&pnl->rsaddr,
2859 &sk->addr[sidx], sk->af);
2860 pnl->rsport = sk->port[sidx];
2861 pf_addrcpy(&pnl->rdaddr,
2862 &sk->addr[didx], sk->af);
2863 pnl->rdport = sk->port[didx];
2864 PF_STATE_UNLOCK(state);
2865
2866 return (0);
2867 }
2868
2869 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2870 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2871 {
2872 int error = 0;
2873 PF_RULES_RLOCK_TRACKER;
2874
2875 #define ERROUT_IOCTL(target, x) \
2876 do { \
2877 error = (x); \
2878 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
2879 goto target; \
2880 } while (0)
2881
2882
2883 /* XXX keep in sync with switch() below */
2884 if (securelevel_gt(td->td_ucred, 2))
2885 switch (cmd) {
2886 case DIOCGETRULES:
2887 case DIOCGETRULENV:
2888 case DIOCGETADDRS:
2889 case DIOCGETADDR:
2890 case DIOCGETSTATE:
2891 case DIOCGETSTATENV:
2892 case DIOCSETSTATUSIF:
2893 case DIOCGETSTATUSNV:
2894 case DIOCCLRSTATUS:
2895 case DIOCNATLOOK:
2896 case DIOCSETDEBUG:
2897 #ifdef COMPAT_FREEBSD14
2898 case DIOCGETSTATES:
2899 case DIOCGETSTATESV2:
2900 #endif
2901 case DIOCGETTIMEOUT:
2902 case DIOCCLRRULECTRS:
2903 case DIOCGETLIMIT:
2904 case DIOCGETALTQSV0:
2905 case DIOCGETALTQSV1:
2906 case DIOCGETALTQV0:
2907 case DIOCGETALTQV1:
2908 case DIOCGETQSTATSV0:
2909 case DIOCGETQSTATSV1:
2910 case DIOCGETRULESETS:
2911 case DIOCGETRULESET:
2912 case DIOCRGETTABLES:
2913 case DIOCRGETTSTATS:
2914 case DIOCRCLRTSTATS:
2915 case DIOCRCLRADDRS:
2916 case DIOCRADDADDRS:
2917 case DIOCRDELADDRS:
2918 case DIOCRSETADDRS:
2919 case DIOCRGETADDRS:
2920 case DIOCRGETASTATS:
2921 case DIOCRCLRASTATS:
2922 case DIOCRTSTADDRS:
2923 case DIOCOSFPGET:
2924 case DIOCGETSRCNODES:
2925 case DIOCCLRSRCNODES:
2926 case DIOCGETSYNCOOKIES:
2927 case DIOCIGETIFACES:
2928 case DIOCGIFSPEEDV0:
2929 case DIOCGIFSPEEDV1:
2930 case DIOCSETIFFLAG:
2931 case DIOCCLRIFFLAG:
2932 case DIOCGETETHRULES:
2933 case DIOCGETETHRULE:
2934 case DIOCGETETHRULESETS:
2935 case DIOCGETETHRULESET:
2936 break;
2937 case DIOCRCLRTABLES:
2938 case DIOCRADDTABLES:
2939 case DIOCRDELTABLES:
2940 case DIOCRSETTFLAGS:
2941 if (((struct pfioc_table *)addr)->pfrio_flags &
2942 PFR_FLAG_DUMMY)
2943 break; /* dummy operation ok */
2944 return (EPERM);
2945 default:
2946 return (EPERM);
2947 }
2948
2949 if (!(flags & FWRITE))
2950 switch (cmd) {
2951 case DIOCGETRULES:
2952 case DIOCGETADDRS:
2953 case DIOCGETADDR:
2954 case DIOCGETSTATE:
2955 case DIOCGETSTATENV:
2956 case DIOCGETSTATUSNV:
2957 #ifdef COMPAT_FREEBSD14
2958 case DIOCGETSTATES:
2959 case DIOCGETSTATESV2:
2960 #endif
2961 case DIOCGETTIMEOUT:
2962 case DIOCGETLIMIT:
2963 case DIOCGETALTQSV0:
2964 case DIOCGETALTQSV1:
2965 case DIOCGETALTQV0:
2966 case DIOCGETALTQV1:
2967 case DIOCGETQSTATSV0:
2968 case DIOCGETQSTATSV1:
2969 case DIOCGETRULESETS:
2970 case DIOCGETRULESET:
2971 case DIOCNATLOOK:
2972 case DIOCRGETTABLES:
2973 case DIOCRGETTSTATS:
2974 case DIOCRGETADDRS:
2975 case DIOCRGETASTATS:
2976 case DIOCRTSTADDRS:
2977 case DIOCOSFPGET:
2978 case DIOCGETSRCNODES:
2979 case DIOCGETSYNCOOKIES:
2980 case DIOCIGETIFACES:
2981 case DIOCGIFSPEEDV1:
2982 case DIOCGIFSPEEDV0:
2983 case DIOCGETRULENV:
2984 case DIOCGETETHRULES:
2985 case DIOCGETETHRULE:
2986 case DIOCGETETHRULESETS:
2987 case DIOCGETETHRULESET:
2988 break;
2989 case DIOCRCLRTABLES:
2990 case DIOCRADDTABLES:
2991 case DIOCRDELTABLES:
2992 case DIOCRCLRTSTATS:
2993 case DIOCRCLRADDRS:
2994 case DIOCRADDADDRS:
2995 case DIOCRDELADDRS:
2996 case DIOCRSETADDRS:
2997 case DIOCRSETTFLAGS:
2998 if (((struct pfioc_table *)addr)->pfrio_flags &
2999 PFR_FLAG_DUMMY) {
3000 flags |= FWRITE; /* need write lock for dummy */
3001 break; /* dummy operation ok */
3002 }
3003 return (EACCES);
3004 default:
3005 return (EACCES);
3006 }
3007
3008 CURVNET_SET(TD_TO_VNET(td));
3009
3010 switch (cmd) {
3011 #ifdef COMPAT_FREEBSD14
3012 case DIOCSTART:
3013 error = pf_start();
3014 break;
3015
3016 case DIOCSTOP:
3017 error = pf_stop();
3018 break;
3019 #endif
3020
3021 case DIOCGETETHRULES: {
3022 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3023 nvlist_t *nvl;
3024 void *packed;
3025 struct pf_keth_rule *tail;
3026 struct pf_keth_ruleset *rs;
3027 u_int32_t ticket, nr;
3028 const char *anchor = "";
3029
3030 nvl = NULL;
3031 packed = NULL;
3032
3033 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
3034
3035 if (nv->len > pf_ioctl_maxcount)
3036 ERROUT(ENOMEM);
3037
3038 /* Copy the request in */
3039 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
3040 error = copyin(nv->data, packed, nv->len);
3041 if (error)
3042 ERROUT(error);
3043
3044 nvl = nvlist_unpack(packed, nv->len, 0);
3045 if (nvl == NULL)
3046 ERROUT(EBADMSG);
3047
3048 if (! nvlist_exists_string(nvl, "anchor"))
3049 ERROUT(EBADMSG);
3050
3051 anchor = nvlist_get_string(nvl, "anchor");
3052
3053 rs = pf_find_keth_ruleset(anchor);
3054
3055 nvlist_destroy(nvl);
3056 nvl = NULL;
3057 free(packed, M_NVLIST);
3058 packed = NULL;
3059
3060 if (rs == NULL)
3061 ERROUT(ENOENT);
3062
3063 /* Reply */
3064 nvl = nvlist_create(0);
3065 if (nvl == NULL)
3066 ERROUT(ENOMEM);
3067
3068 PF_RULES_RLOCK();
3069
3070 ticket = rs->active.ticket;
3071 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
3072 if (tail)
3073 nr = tail->nr + 1;
3074 else
3075 nr = 0;
3076
3077 PF_RULES_RUNLOCK();
3078
3079 nvlist_add_number(nvl, "ticket", ticket);
3080 nvlist_add_number(nvl, "nr", nr);
3081
3082 packed = nvlist_pack(nvl, &nv->len);
3083 if (packed == NULL)
3084 ERROUT(ENOMEM);
3085
3086 if (nv->size == 0)
3087 ERROUT(0);
3088 else if (nv->size < nv->len)
3089 ERROUT(ENOSPC);
3090
3091 error = copyout(packed, nv->data, nv->len);
3092
3093 #undef ERROUT
3094 DIOCGETETHRULES_error:
3095 free(packed, M_NVLIST);
3096 nvlist_destroy(nvl);
3097 break;
3098 }
3099
3100 case DIOCGETETHRULE: {
3101 struct epoch_tracker et;
3102 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3103 nvlist_t *nvl = NULL;
3104 void *nvlpacked = NULL;
3105 struct pf_keth_rule *rule = NULL;
3106 struct pf_keth_ruleset *rs;
3107 u_int32_t ticket, nr;
3108 bool clear = false;
3109 const char *anchor;
3110
3111 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
3112
3113 if (nv->len > pf_ioctl_maxcount)
3114 ERROUT(ENOMEM);
3115
3116 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3117 error = copyin(nv->data, nvlpacked, nv->len);
3118 if (error)
3119 ERROUT(error);
3120
3121 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3122 if (nvl == NULL)
3123 ERROUT(EBADMSG);
3124 if (! nvlist_exists_number(nvl, "ticket"))
3125 ERROUT(EBADMSG);
3126 ticket = nvlist_get_number(nvl, "ticket");
3127 if (! nvlist_exists_string(nvl, "anchor"))
3128 ERROUT(EBADMSG);
3129 anchor = nvlist_get_string(nvl, "anchor");
3130
3131 if (nvlist_exists_bool(nvl, "clear"))
3132 clear = nvlist_get_bool(nvl, "clear");
3133
3134 if (clear && !(flags & FWRITE))
3135 ERROUT(EACCES);
3136
3137 if (! nvlist_exists_number(nvl, "nr"))
3138 ERROUT(EBADMSG);
3139 nr = nvlist_get_number(nvl, "nr");
3140
3141 PF_RULES_RLOCK();
3142 rs = pf_find_keth_ruleset(anchor);
3143 if (rs == NULL) {
3144 PF_RULES_RUNLOCK();
3145 ERROUT(ENOENT);
3146 }
3147 if (ticket != rs->active.ticket) {
3148 PF_RULES_RUNLOCK();
3149 ERROUT(EBUSY);
3150 }
3151
3152 nvlist_destroy(nvl);
3153 nvl = NULL;
3154 free(nvlpacked, M_NVLIST);
3155 nvlpacked = NULL;
3156
3157 rule = TAILQ_FIRST(rs->active.rules);
3158 while ((rule != NULL) && (rule->nr != nr))
3159 rule = TAILQ_NEXT(rule, entries);
3160 if (rule == NULL) {
3161 PF_RULES_RUNLOCK();
3162 ERROUT(ENOENT);
3163 }
3164 /* Make sure rule can't go away. */
3165 NET_EPOCH_ENTER(et);
3166 PF_RULES_RUNLOCK();
3167 nvl = pf_keth_rule_to_nveth_rule(rule);
3168 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
3169 NET_EPOCH_EXIT(et);
3170 ERROUT(EBUSY);
3171 }
3172 NET_EPOCH_EXIT(et);
3173 if (nvl == NULL)
3174 ERROUT(ENOMEM);
3175
3176 nvlpacked = nvlist_pack(nvl, &nv->len);
3177 if (nvlpacked == NULL)
3178 ERROUT(ENOMEM);
3179
3180 if (nv->size == 0)
3181 ERROUT(0);
3182 else if (nv->size < nv->len)
3183 ERROUT(ENOSPC);
3184
3185 error = copyout(nvlpacked, nv->data, nv->len);
3186 if (error == 0 && clear) {
3187 counter_u64_zero(rule->evaluations);
3188 for (int i = 0; i < 2; i++) {
3189 counter_u64_zero(rule->packets[i]);
3190 counter_u64_zero(rule->bytes[i]);
3191 }
3192 }
3193
3194 #undef ERROUT
3195 DIOCGETETHRULE_error:
3196 free(nvlpacked, M_NVLIST);
3197 nvlist_destroy(nvl);
3198 break;
3199 }
3200
3201 case DIOCADDETHRULE: {
3202 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3203 nvlist_t *nvl = NULL;
3204 void *nvlpacked = NULL;
3205 struct pf_keth_rule *rule = NULL, *tail = NULL;
3206 struct pf_keth_ruleset *ruleset = NULL;
3207 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
3208 const char *anchor = "", *anchor_call = "";
3209
3210 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3211
3212 if (nv->len > pf_ioctl_maxcount)
3213 ERROUT(ENOMEM);
3214
3215 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3216 error = copyin(nv->data, nvlpacked, nv->len);
3217 if (error)
3218 ERROUT(error);
3219
3220 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3221 if (nvl == NULL)
3222 ERROUT(EBADMSG);
3223
3224 if (! nvlist_exists_number(nvl, "ticket"))
3225 ERROUT(EBADMSG);
3226
3227 if (nvlist_exists_string(nvl, "anchor"))
3228 anchor = nvlist_get_string(nvl, "anchor");
3229 if (nvlist_exists_string(nvl, "anchor_call"))
3230 anchor_call = nvlist_get_string(nvl, "anchor_call");
3231
3232 ruleset = pf_find_keth_ruleset(anchor);
3233 if (ruleset == NULL)
3234 ERROUT(EINVAL);
3235
3236 if (nvlist_get_number(nvl, "ticket") !=
3237 ruleset->inactive.ticket) {
3238 DPFPRINTF(PF_DEBUG_MISC,
3239 ("ticket: %d != %d\n",
3240 (u_int32_t)nvlist_get_number(nvl, "ticket"),
3241 ruleset->inactive.ticket));
3242 ERROUT(EBUSY);
3243 }
3244
3245 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3246 rule->timestamp = NULL;
3247
3248 error = pf_nveth_rule_to_keth_rule(nvl, rule);
3249 if (error != 0)
3250 ERROUT(error);
3251
3252 if (rule->ifname[0])
3253 kif = pf_kkif_create(M_WAITOK);
3254 if (rule->bridge_to_name[0])
3255 bridge_to_kif = pf_kkif_create(M_WAITOK);
3256 rule->evaluations = counter_u64_alloc(M_WAITOK);
3257 for (int i = 0; i < 2; i++) {
3258 rule->packets[i] = counter_u64_alloc(M_WAITOK);
3259 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3260 }
3261 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3262 M_WAITOK | M_ZERO);
3263
3264 PF_RULES_WLOCK();
3265
3266 if (rule->ifname[0]) {
3267 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3268 pfi_kkif_ref(rule->kif);
3269 } else
3270 rule->kif = NULL;
3271 if (rule->bridge_to_name[0]) {
3272 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3273 rule->bridge_to_name);
3274 pfi_kkif_ref(rule->bridge_to);
3275 } else
3276 rule->bridge_to = NULL;
3277
3278 #ifdef ALTQ
3279 /* set queue IDs */
3280 if (rule->qname[0] != 0) {
3281 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3282 error = EBUSY;
3283 else
3284 rule->qid = rule->qid;
3285 }
3286 #endif
3287 if (rule->tagname[0])
3288 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3289 error = EBUSY;
3290 if (rule->match_tagname[0])
3291 if ((rule->match_tag = pf_tagname2tag(
3292 rule->match_tagname)) == 0)
3293 error = EBUSY;
3294
3295 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3296 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3297 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3298 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3299
3300 if (error) {
3301 pf_free_eth_rule(rule);
3302 PF_RULES_WUNLOCK();
3303 ERROUT(error);
3304 }
3305
3306 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3307 pf_free_eth_rule(rule);
3308 PF_RULES_WUNLOCK();
3309 ERROUT(EINVAL);
3310 }
3311
3312 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3313 if (tail)
3314 rule->nr = tail->nr + 1;
3315 else
3316 rule->nr = 0;
3317
3318 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3319
3320 PF_RULES_WUNLOCK();
3321
3322 #undef ERROUT
3323 DIOCADDETHRULE_error:
3324 nvlist_destroy(nvl);
3325 free(nvlpacked, M_NVLIST);
3326 break;
3327 }
3328
3329 case DIOCGETETHRULESETS: {
3330 struct epoch_tracker et;
3331 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3332 nvlist_t *nvl = NULL;
3333 void *nvlpacked = NULL;
3334 struct pf_keth_ruleset *ruleset;
3335 struct pf_keth_anchor *anchor;
3336 int nr = 0;
3337
3338 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3339
3340 if (nv->len > pf_ioctl_maxcount)
3341 ERROUT(ENOMEM);
3342
3343 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3344 error = copyin(nv->data, nvlpacked, nv->len);
3345 if (error)
3346 ERROUT(error);
3347
3348 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3349 if (nvl == NULL)
3350 ERROUT(EBADMSG);
3351 if (! nvlist_exists_string(nvl, "path"))
3352 ERROUT(EBADMSG);
3353
3354 NET_EPOCH_ENTER(et);
3355
3356 if ((ruleset = pf_find_keth_ruleset(
3357 nvlist_get_string(nvl, "path"))) == NULL) {
3358 NET_EPOCH_EXIT(et);
3359 ERROUT(ENOENT);
3360 }
3361
3362 if (ruleset->anchor == NULL) {
3363 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3364 if (anchor->parent == NULL)
3365 nr++;
3366 } else {
3367 RB_FOREACH(anchor, pf_keth_anchor_node,
3368 &ruleset->anchor->children)
3369 nr++;
3370 }
3371
3372 NET_EPOCH_EXIT(et);
3373
3374 nvlist_destroy(nvl);
3375 nvl = NULL;
3376 free(nvlpacked, M_NVLIST);
3377 nvlpacked = NULL;
3378
3379 nvl = nvlist_create(0);
3380 if (nvl == NULL)
3381 ERROUT(ENOMEM);
3382
3383 nvlist_add_number(nvl, "nr", nr);
3384
3385 nvlpacked = nvlist_pack(nvl, &nv->len);
3386 if (nvlpacked == NULL)
3387 ERROUT(ENOMEM);
3388
3389 if (nv->size == 0)
3390 ERROUT(0);
3391 else if (nv->size < nv->len)
3392 ERROUT(ENOSPC);
3393
3394 error = copyout(nvlpacked, nv->data, nv->len);
3395
3396 #undef ERROUT
3397 DIOCGETETHRULESETS_error:
3398 free(nvlpacked, M_NVLIST);
3399 nvlist_destroy(nvl);
3400 break;
3401 }
3402
3403 case DIOCGETETHRULESET: {
3404 struct epoch_tracker et;
3405 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3406 nvlist_t *nvl = NULL;
3407 void *nvlpacked = NULL;
3408 struct pf_keth_ruleset *ruleset;
3409 struct pf_keth_anchor *anchor;
3410 int nr = 0, req_nr = 0;
3411 bool found = false;
3412
3413 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3414
3415 if (nv->len > pf_ioctl_maxcount)
3416 ERROUT(ENOMEM);
3417
3418 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3419 error = copyin(nv->data, nvlpacked, nv->len);
3420 if (error)
3421 ERROUT(error);
3422
3423 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3424 if (nvl == NULL)
3425 ERROUT(EBADMSG);
3426 if (! nvlist_exists_string(nvl, "path"))
3427 ERROUT(EBADMSG);
3428 if (! nvlist_exists_number(nvl, "nr"))
3429 ERROUT(EBADMSG);
3430
3431 req_nr = nvlist_get_number(nvl, "nr");
3432
3433 NET_EPOCH_ENTER(et);
3434
3435 if ((ruleset = pf_find_keth_ruleset(
3436 nvlist_get_string(nvl, "path"))) == NULL) {
3437 NET_EPOCH_EXIT(et);
3438 ERROUT(ENOENT);
3439 }
3440
3441 nvlist_destroy(nvl);
3442 nvl = NULL;
3443 free(nvlpacked, M_NVLIST);
3444 nvlpacked = NULL;
3445
3446 nvl = nvlist_create(0);
3447 if (nvl == NULL) {
3448 NET_EPOCH_EXIT(et);
3449 ERROUT(ENOMEM);
3450 }
3451
3452 if (ruleset->anchor == NULL) {
3453 RB_FOREACH(anchor, pf_keth_anchor_global,
3454 &V_pf_keth_anchors) {
3455 if (anchor->parent == NULL && nr++ == req_nr) {
3456 found = true;
3457 break;
3458 }
3459 }
3460 } else {
3461 RB_FOREACH(anchor, pf_keth_anchor_node,
3462 &ruleset->anchor->children) {
3463 if (nr++ == req_nr) {
3464 found = true;
3465 break;
3466 }
3467 }
3468 }
3469
3470 NET_EPOCH_EXIT(et);
3471 if (found) {
3472 nvlist_add_number(nvl, "nr", nr);
3473 nvlist_add_string(nvl, "name", anchor->name);
3474 if (ruleset->anchor)
3475 nvlist_add_string(nvl, "path",
3476 ruleset->anchor->path);
3477 else
3478 nvlist_add_string(nvl, "path", "");
3479 } else {
3480 ERROUT(EBUSY);
3481 }
3482
3483 nvlpacked = nvlist_pack(nvl, &nv->len);
3484 if (nvlpacked == NULL)
3485 ERROUT(ENOMEM);
3486
3487 if (nv->size == 0)
3488 ERROUT(0);
3489 else if (nv->size < nv->len)
3490 ERROUT(ENOSPC);
3491
3492 error = copyout(nvlpacked, nv->data, nv->len);
3493
3494 #undef ERROUT
3495 DIOCGETETHRULESET_error:
3496 free(nvlpacked, M_NVLIST);
3497 nvlist_destroy(nvl);
3498 break;
3499 }
3500
3501 case DIOCADDRULENV: {
3502 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3503 nvlist_t *nvl = NULL;
3504 void *nvlpacked = NULL;
3505 struct pf_krule *rule = NULL;
3506 const char *anchor = "", *anchor_call = "";
3507 uint32_t ticket = 0, pool_ticket = 0;
3508
3509 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
3510
3511 if (nv->len > pf_ioctl_maxcount)
3512 ERROUT(ENOMEM);
3513
3514 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3515 error = copyin(nv->data, nvlpacked, nv->len);
3516 if (error)
3517 ERROUT(error);
3518
3519 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3520 if (nvl == NULL)
3521 ERROUT(EBADMSG);
3522
3523 if (! nvlist_exists_number(nvl, "ticket"))
3524 ERROUT(EINVAL);
3525 ticket = nvlist_get_number(nvl, "ticket");
3526
3527 if (! nvlist_exists_number(nvl, "pool_ticket"))
3528 ERROUT(EINVAL);
3529 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3530
3531 if (! nvlist_exists_nvlist(nvl, "rule"))
3532 ERROUT(EINVAL);
3533
3534 rule = pf_krule_alloc();
3535 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3536 rule);
3537 if (error)
3538 ERROUT(error);
3539
3540 if (nvlist_exists_string(nvl, "anchor"))
3541 anchor = nvlist_get_string(nvl, "anchor");
3542 if (nvlist_exists_string(nvl, "anchor_call"))
3543 anchor_call = nvlist_get_string(nvl, "anchor_call");
3544
3545 if ((error = nvlist_error(nvl)))
3546 ERROUT(error);
3547
3548 /* Frees rule on error */
3549 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3550 anchor_call, td->td_ucred->cr_ruid,
3551 td->td_proc ? td->td_proc->p_pid : 0);
3552
3553 nvlist_destroy(nvl);
3554 free(nvlpacked, M_NVLIST);
3555 break;
3556 #undef ERROUT
3557 DIOCADDRULENV_error:
3558 pf_krule_free(rule);
3559 nvlist_destroy(nvl);
3560 free(nvlpacked, M_NVLIST);
3561
3562 break;
3563 }
3564 case DIOCADDRULE: {
3565 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3566 struct pf_krule *rule;
3567
3568 rule = pf_krule_alloc();
3569 error = pf_rule_to_krule(&pr->rule, rule);
3570 if (error != 0) {
3571 pf_krule_free(rule);
3572 break;
3573 }
3574
3575 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
3576
3577 /* Frees rule on error */
3578 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3579 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3580 td->td_proc ? td->td_proc->p_pid : 0);
3581 break;
3582 }
3583
3584 case DIOCGETRULES: {
3585 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3586
3587 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
3588
3589 error = pf_ioctl_getrules(pr);
3590
3591 break;
3592 }
3593
3594 case DIOCGETRULENV: {
3595 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3596 nvlist_t *nvrule = NULL;
3597 nvlist_t *nvl = NULL;
3598 struct pf_kruleset *ruleset;
3599 struct pf_krule *rule;
3600 void *nvlpacked = NULL;
3601 int rs_num, nr;
3602 bool clear_counter = false;
3603
3604 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
3605
3606 if (nv->len > pf_ioctl_maxcount)
3607 ERROUT(ENOMEM);
3608
3609 /* Copy the request in */
3610 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3611 error = copyin(nv->data, nvlpacked, nv->len);
3612 if (error)
3613 ERROUT(error);
3614
3615 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3616 if (nvl == NULL)
3617 ERROUT(EBADMSG);
3618
3619 if (! nvlist_exists_string(nvl, "anchor"))
3620 ERROUT(EBADMSG);
3621 if (! nvlist_exists_number(nvl, "ruleset"))
3622 ERROUT(EBADMSG);
3623 if (! nvlist_exists_number(nvl, "ticket"))
3624 ERROUT(EBADMSG);
3625 if (! nvlist_exists_number(nvl, "nr"))
3626 ERROUT(EBADMSG);
3627
3628 if (nvlist_exists_bool(nvl, "clear_counter"))
3629 clear_counter = nvlist_get_bool(nvl, "clear_counter");
3630
3631 if (clear_counter && !(flags & FWRITE))
3632 ERROUT(EACCES);
3633
3634 nr = nvlist_get_number(nvl, "nr");
3635
3636 PF_RULES_WLOCK();
3637 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3638 if (ruleset == NULL) {
3639 PF_RULES_WUNLOCK();
3640 ERROUT(ENOENT);
3641 }
3642
3643 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3644 if (rs_num >= PF_RULESET_MAX) {
3645 PF_RULES_WUNLOCK();
3646 ERROUT(EINVAL);
3647 }
3648
3649 if (nvlist_get_number(nvl, "ticket") !=
3650 ruleset->rules[rs_num].active.ticket) {
3651 PF_RULES_WUNLOCK();
3652 ERROUT(EBUSY);
3653 }
3654
3655 if ((error = nvlist_error(nvl))) {
3656 PF_RULES_WUNLOCK();
3657 ERROUT(error);
3658 }
3659
3660 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3661 while ((rule != NULL) && (rule->nr != nr))
3662 rule = TAILQ_NEXT(rule, entries);
3663 if (rule == NULL) {
3664 PF_RULES_WUNLOCK();
3665 ERROUT(EBUSY);
3666 }
3667
3668 nvrule = pf_krule_to_nvrule(rule);
3669
3670 nvlist_destroy(nvl);
3671 nvl = nvlist_create(0);
3672 if (nvl == NULL) {
3673 PF_RULES_WUNLOCK();
3674 ERROUT(ENOMEM);
3675 }
3676 nvlist_add_number(nvl, "nr", nr);
3677 nvlist_add_nvlist(nvl, "rule", nvrule);
3678 nvlist_destroy(nvrule);
3679 nvrule = NULL;
3680 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3681 PF_RULES_WUNLOCK();
3682 ERROUT(EBUSY);
3683 }
3684
3685 free(nvlpacked, M_NVLIST);
3686 nvlpacked = nvlist_pack(nvl, &nv->len);
3687 if (nvlpacked == NULL) {
3688 PF_RULES_WUNLOCK();
3689 ERROUT(ENOMEM);
3690 }
3691
3692 if (nv->size == 0) {
3693 PF_RULES_WUNLOCK();
3694 ERROUT(0);
3695 }
3696 else if (nv->size < nv->len) {
3697 PF_RULES_WUNLOCK();
3698 ERROUT(ENOSPC);
3699 }
3700
3701 if (clear_counter)
3702 pf_krule_clear_counters(rule);
3703
3704 PF_RULES_WUNLOCK();
3705
3706 error = copyout(nvlpacked, nv->data, nv->len);
3707
3708 #undef ERROUT
3709 DIOCGETRULENV_error:
3710 free(nvlpacked, M_NVLIST);
3711 nvlist_destroy(nvrule);
3712 nvlist_destroy(nvl);
3713
3714 break;
3715 }
3716
3717 case DIOCCHANGERULE: {
3718 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
3719 struct pf_kruleset *ruleset;
3720 struct pf_krule *oldrule = NULL, *newrule = NULL;
3721 struct pfi_kkif *kif = NULL;
3722 struct pf_kpooladdr *pa;
3723 u_int32_t nr = 0;
3724 int rs_num;
3725
3726 pcr->anchor[sizeof(pcr->anchor) - 1] = '\0';
3727
3728 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3729 pcr->action > PF_CHANGE_GET_TICKET) {
3730 error = EINVAL;
3731 break;
3732 }
3733 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3734 error = EINVAL;
3735 break;
3736 }
3737
3738 if (pcr->action != PF_CHANGE_REMOVE) {
3739 newrule = pf_krule_alloc();
3740 error = pf_rule_to_krule(&pcr->rule, newrule);
3741 if (error != 0) {
3742 pf_krule_free(newrule);
3743 break;
3744 }
3745
3746 if (newrule->ifname[0])
3747 kif = pf_kkif_create(M_WAITOK);
3748 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3749 for (int i = 0; i < 2; i++) {
3750 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3751 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3752 }
3753 newrule->states_cur = counter_u64_alloc(M_WAITOK);
3754 newrule->states_tot = counter_u64_alloc(M_WAITOK);
3755 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
3756 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
3757 newrule->cuid = td->td_ucred->cr_ruid;
3758 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3759 TAILQ_INIT(&newrule->nat.list);
3760 TAILQ_INIT(&newrule->rdr.list);
3761 TAILQ_INIT(&newrule->route.list);
3762 }
3763 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3764
3765 PF_CONFIG_LOCK();
3766 PF_RULES_WLOCK();
3767 #ifdef PF_WANT_32_TO_64_COUNTER
3768 if (newrule != NULL) {
3769 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3770 newrule->allrulelinked = true;
3771 V_pf_allrulecount++;
3772 }
3773 #endif
3774
3775 if (!(pcr->action == PF_CHANGE_REMOVE ||
3776 pcr->action == PF_CHANGE_GET_TICKET) &&
3777 pcr->pool_ticket != V_ticket_pabuf)
3778 ERROUT(EBUSY);
3779
3780 ruleset = pf_find_kruleset(pcr->anchor);
3781 if (ruleset == NULL)
3782 ERROUT(EINVAL);
3783
3784 rs_num = pf_get_ruleset_number(pcr->rule.action);
3785 if (rs_num >= PF_RULESET_MAX)
3786 ERROUT(EINVAL);
3787
3788 /*
3789 * XXXMJG: there is no guarantee that the ruleset was
3790 * created by the usual route of calling DIOCXBEGIN.
3791 * As a result it is possible the rule tree will not
3792 * be allocated yet. Hack around it by doing it here.
3793 * Note it is fine to let the tree persist in case of
3794 * error as it will be freed down the road on future
3795 * updates (if need be).
3796 */
3797 if (ruleset->rules[rs_num].active.tree == NULL) {
3798 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3799 if (ruleset->rules[rs_num].active.tree == NULL) {
3800 ERROUT(ENOMEM);
3801 }
3802 }
3803
3804 if (pcr->action == PF_CHANGE_GET_TICKET) {
3805 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3806 ERROUT(0);
3807 } else if (pcr->ticket !=
3808 ruleset->rules[rs_num].active.ticket)
3809 ERROUT(EINVAL);
3810
3811 if (pcr->action != PF_CHANGE_REMOVE) {
3812 if (newrule->ifname[0]) {
3813 newrule->kif = pfi_kkif_attach(kif,
3814 newrule->ifname);
3815 kif = NULL;
3816 pfi_kkif_ref(newrule->kif);
3817 } else
3818 newrule->kif = NULL;
3819
3820 if (newrule->rtableid > 0 &&
3821 newrule->rtableid >= rt_numfibs)
3822 error = EBUSY;
3823
3824 #ifdef ALTQ
3825 /* set queue IDs */
3826 if (newrule->qname[0] != 0) {
3827 if ((newrule->qid =
3828 pf_qname2qid(newrule->qname)) == 0)
3829 error = EBUSY;
3830 else if (newrule->pqname[0] != 0) {
3831 if ((newrule->pqid =
3832 pf_qname2qid(newrule->pqname)) == 0)
3833 error = EBUSY;
3834 } else
3835 newrule->pqid = newrule->qid;
3836 }
3837 #endif /* ALTQ */
3838 if (newrule->tagname[0])
3839 if ((newrule->tag =
3840 pf_tagname2tag(newrule->tagname)) == 0)
3841 error = EBUSY;
3842 if (newrule->match_tagname[0])
3843 if ((newrule->match_tag = pf_tagname2tag(
3844 newrule->match_tagname)) == 0)
3845 error = EBUSY;
3846 if (newrule->rt && !newrule->direction)
3847 error = EINVAL;
3848 if (!newrule->log)
3849 newrule->logif = 0;
3850 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3851 error = ENOMEM;
3852 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3853 error = ENOMEM;
3854 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3855 error = EINVAL;
3856 for (int i = 0; i < 3; i++) {
3857 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
3858 if (pa->addr.type == PF_ADDR_TABLE) {
3859 pa->addr.p.tbl =
3860 pfr_attach_table(ruleset,
3861 pa->addr.v.tblname);
3862 if (pa->addr.p.tbl == NULL)
3863 error = ENOMEM;
3864 }
3865 }
3866
3867 newrule->overload_tbl = NULL;
3868 if (newrule->overload_tblname[0]) {
3869 if ((newrule->overload_tbl = pfr_attach_table(
3870 ruleset, newrule->overload_tblname)) ==
3871 NULL)
3872 error = EINVAL;
3873 else
3874 newrule->overload_tbl->pfrkt_flags |=
3875 PFR_TFLAG_ACTIVE;
3876 }
3877
3878 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
3879 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
3880 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list);
3881 if (((((newrule->action == PF_NAT) ||
3882 (newrule->action == PF_RDR) ||
3883 (newrule->action == PF_BINAT) ||
3884 (newrule->rt > PF_NOPFROUTE)) &&
3885 !newrule->anchor)) &&
3886 (TAILQ_FIRST(&newrule->rdr.list) == NULL))
3887 error = EINVAL;
3888
3889 if (error) {
3890 pf_free_rule(newrule);
3891 PF_RULES_WUNLOCK();
3892 PF_CONFIG_UNLOCK();
3893 break;
3894 }
3895
3896 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
3897 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
3898 }
3899 pf_empty_kpool(&V_pf_pabuf[0]);
3900 pf_empty_kpool(&V_pf_pabuf[1]);
3901 pf_empty_kpool(&V_pf_pabuf[2]);
3902
3903 if (pcr->action == PF_CHANGE_ADD_HEAD)
3904 oldrule = TAILQ_FIRST(
3905 ruleset->rules[rs_num].active.ptr);
3906 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3907 oldrule = TAILQ_LAST(
3908 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3909 else {
3910 oldrule = TAILQ_FIRST(
3911 ruleset->rules[rs_num].active.ptr);
3912 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3913 oldrule = TAILQ_NEXT(oldrule, entries);
3914 if (oldrule == NULL) {
3915 if (newrule != NULL)
3916 pf_free_rule(newrule);
3917 PF_RULES_WUNLOCK();
3918 PF_CONFIG_UNLOCK();
3919 error = EINVAL;
3920 break;
3921 }
3922 }
3923
3924 if (pcr->action == PF_CHANGE_REMOVE) {
3925 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3926 oldrule);
3927 RB_REMOVE(pf_krule_global,
3928 ruleset->rules[rs_num].active.tree, oldrule);
3929 ruleset->rules[rs_num].active.rcount--;
3930 } else {
3931 pf_hash_rule(newrule);
3932 if (RB_INSERT(pf_krule_global,
3933 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3934 pf_free_rule(newrule);
3935 PF_RULES_WUNLOCK();
3936 PF_CONFIG_UNLOCK();
3937 error = EEXIST;
3938 break;
3939 }
3940
3941 if (oldrule == NULL)
3942 TAILQ_INSERT_TAIL(
3943 ruleset->rules[rs_num].active.ptr,
3944 newrule, entries);
3945 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3946 pcr->action == PF_CHANGE_ADD_BEFORE)
3947 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3948 else
3949 TAILQ_INSERT_AFTER(
3950 ruleset->rules[rs_num].active.ptr,
3951 oldrule, newrule, entries);
3952 ruleset->rules[rs_num].active.rcount++;
3953 }
3954
3955 nr = 0;
3956 TAILQ_FOREACH(oldrule,
3957 ruleset->rules[rs_num].active.ptr, entries)
3958 oldrule->nr = nr++;
3959
3960 ruleset->rules[rs_num].active.ticket++;
3961
3962 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3963 pf_remove_if_empty_kruleset(ruleset);
3964
3965 PF_RULES_WUNLOCK();
3966 PF_CONFIG_UNLOCK();
3967 break;
3968
3969 #undef ERROUT
3970 DIOCCHANGERULE_error:
3971 PF_RULES_WUNLOCK();
3972 PF_CONFIG_UNLOCK();
3973 pf_krule_free(newrule);
3974 pf_kkif_free(kif);
3975 break;
3976 }
3977
3978 case DIOCCLRSTATESNV: {
3979 error = pf_clearstates_nv((struct pfioc_nv *)addr);
3980 break;
3981 }
3982
3983 case DIOCKILLSTATESNV: {
3984 error = pf_killstates_nv((struct pfioc_nv *)addr);
3985 break;
3986 }
3987
3988 case DIOCADDSTATE: {
3989 struct pfioc_state *ps = (struct pfioc_state *)addr;
3990 struct pfsync_state_1301 *sp = &ps->state;
3991
3992 if (sp->timeout >= PFTM_MAX) {
3993 error = EINVAL;
3994 break;
3995 }
3996 if (V_pfsync_state_import_ptr != NULL) {
3997 PF_RULES_RLOCK();
3998 error = V_pfsync_state_import_ptr(
3999 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
4000 PFSYNC_MSG_VERSION_1301);
4001 PF_RULES_RUNLOCK();
4002 } else
4003 error = EOPNOTSUPP;
4004 break;
4005 }
4006
4007 case DIOCGETSTATE: {
4008 struct pfioc_state *ps = (struct pfioc_state *)addr;
4009 struct pf_kstate *s;
4010
4011 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
4012 if (s == NULL) {
4013 error = ENOENT;
4014 break;
4015 }
4016
4017 pfsync_state_export((union pfsync_state_union*)&ps->state,
4018 s, PFSYNC_MSG_VERSION_1301);
4019 PF_STATE_UNLOCK(s);
4020 break;
4021 }
4022
4023 case DIOCGETSTATENV: {
4024 error = pf_getstate((struct pfioc_nv *)addr);
4025 break;
4026 }
4027
4028 #ifdef COMPAT_FREEBSD14
4029 case DIOCGETSTATES: {
4030 struct pfioc_states *ps = (struct pfioc_states *)addr;
4031 struct pf_kstate *s;
4032 struct pfsync_state_1301 *pstore, *p;
4033 int i, nr;
4034 size_t slice_count = 16, count;
4035 void *out;
4036
4037 if (ps->ps_len <= 0) {
4038 nr = uma_zone_get_cur(V_pf_state_z);
4039 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4040 break;
4041 }
4042
4043 out = ps->ps_states;
4044 pstore = mallocarray(slice_count,
4045 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
4046 nr = 0;
4047
4048 for (i = 0; i <= V_pf_hashmask; i++) {
4049 struct pf_idhash *ih = &V_pf_idhash[i];
4050
4051 DIOCGETSTATES_retry:
4052 p = pstore;
4053
4054 if (LIST_EMPTY(&ih->states))
4055 continue;
4056
4057 PF_HASHROW_LOCK(ih);
4058 count = 0;
4059 LIST_FOREACH(s, &ih->states, entry) {
4060 if (s->timeout == PFTM_UNLINKED)
4061 continue;
4062 count++;
4063 }
4064
4065 if (count > slice_count) {
4066 PF_HASHROW_UNLOCK(ih);
4067 free(pstore, M_TEMP);
4068 slice_count = count * 2;
4069 pstore = mallocarray(slice_count,
4070 sizeof(struct pfsync_state_1301), M_TEMP,
4071 M_WAITOK | M_ZERO);
4072 goto DIOCGETSTATES_retry;
4073 }
4074
4075 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4076 PF_HASHROW_UNLOCK(ih);
4077 goto DIOCGETSTATES_full;
4078 }
4079
4080 LIST_FOREACH(s, &ih->states, entry) {
4081 if (s->timeout == PFTM_UNLINKED)
4082 continue;
4083
4084 pfsync_state_export((union pfsync_state_union*)p,
4085 s, PFSYNC_MSG_VERSION_1301);
4086 p++;
4087 nr++;
4088 }
4089 PF_HASHROW_UNLOCK(ih);
4090 error = copyout(pstore, out,
4091 sizeof(struct pfsync_state_1301) * count);
4092 if (error)
4093 break;
4094 out = ps->ps_states + nr;
4095 }
4096 DIOCGETSTATES_full:
4097 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4098 free(pstore, M_TEMP);
4099
4100 break;
4101 }
4102
4103 case DIOCGETSTATESV2: {
4104 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
4105 struct pf_kstate *s;
4106 struct pf_state_export *pstore, *p;
4107 int i, nr;
4108 size_t slice_count = 16, count;
4109 void *out;
4110
4111 if (ps->ps_req_version > PF_STATE_VERSION) {
4112 error = ENOTSUP;
4113 break;
4114 }
4115
4116 if (ps->ps_len <= 0) {
4117 nr = uma_zone_get_cur(V_pf_state_z);
4118 ps->ps_len = sizeof(struct pf_state_export) * nr;
4119 break;
4120 }
4121
4122 out = ps->ps_states;
4123 pstore = mallocarray(slice_count,
4124 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
4125 nr = 0;
4126
4127 for (i = 0; i <= V_pf_hashmask; i++) {
4128 struct pf_idhash *ih = &V_pf_idhash[i];
4129
4130 DIOCGETSTATESV2_retry:
4131 p = pstore;
4132
4133 if (LIST_EMPTY(&ih->states))
4134 continue;
4135
4136 PF_HASHROW_LOCK(ih);
4137 count = 0;
4138 LIST_FOREACH(s, &ih->states, entry) {
4139 if (s->timeout == PFTM_UNLINKED)
4140 continue;
4141 count++;
4142 }
4143
4144 if (count > slice_count) {
4145 PF_HASHROW_UNLOCK(ih);
4146 free(pstore, M_TEMP);
4147 slice_count = count * 2;
4148 pstore = mallocarray(slice_count,
4149 sizeof(struct pf_state_export), M_TEMP,
4150 M_WAITOK | M_ZERO);
4151 goto DIOCGETSTATESV2_retry;
4152 }
4153
4154 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4155 PF_HASHROW_UNLOCK(ih);
4156 goto DIOCGETSTATESV2_full;
4157 }
4158
4159 LIST_FOREACH(s, &ih->states, entry) {
4160 if (s->timeout == PFTM_UNLINKED)
4161 continue;
4162
4163 pf_state_export(p, s);
4164 p++;
4165 nr++;
4166 }
4167 PF_HASHROW_UNLOCK(ih);
4168 error = copyout(pstore, out,
4169 sizeof(struct pf_state_export) * count);
4170 if (error)
4171 break;
4172 out = ps->ps_states + nr;
4173 }
4174 DIOCGETSTATESV2_full:
4175 ps->ps_len = nr * sizeof(struct pf_state_export);
4176 free(pstore, M_TEMP);
4177
4178 break;
4179 }
4180 #endif
4181 case DIOCGETSTATUSNV: {
4182 error = pf_getstatus((struct pfioc_nv *)addr);
4183 break;
4184 }
4185
4186 case DIOCSETSTATUSIF: {
4187 struct pfioc_if *pi = (struct pfioc_if *)addr;
4188
4189 if (pi->ifname[0] == 0) {
4190 bzero(V_pf_status.ifname, IFNAMSIZ);
4191 break;
4192 }
4193 PF_RULES_WLOCK();
4194 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
4195 PF_RULES_WUNLOCK();
4196 break;
4197 }
4198
4199 case DIOCCLRSTATUS: {
4200 pf_ioctl_clear_status();
4201 break;
4202 }
4203
4204 case DIOCNATLOOK: {
4205 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
4206
4207 error = pf_ioctl_natlook(pnl);
4208 break;
4209 }
4210
4211 case DIOCSETTIMEOUT: {
4212 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4213
4214 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4215 &pt->seconds);
4216 break;
4217 }
4218
4219 case DIOCGETTIMEOUT: {
4220 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4221
4222 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4223 break;
4224 }
4225
4226 case DIOCGETLIMIT: {
4227 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4228
4229 error = pf_ioctl_get_limit(pl->index, &pl->limit);
4230 break;
4231 }
4232
4233 case DIOCSETLIMIT: {
4234 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4235 unsigned int old_limit;
4236
4237 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4238 pl->limit = old_limit;
4239 break;
4240 }
4241
4242 case DIOCSETDEBUG: {
4243 u_int32_t *level = (u_int32_t *)addr;
4244
4245 PF_RULES_WLOCK();
4246 V_pf_status.debug = *level;
4247 PF_RULES_WUNLOCK();
4248 break;
4249 }
4250
4251 case DIOCCLRRULECTRS: {
4252 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4253 struct pf_kruleset *ruleset = &pf_main_ruleset;
4254 struct pf_krule *rule;
4255
4256 PF_RULES_WLOCK();
4257 TAILQ_FOREACH(rule,
4258 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4259 pf_counter_u64_zero(&rule->evaluations);
4260 for (int i = 0; i < 2; i++) {
4261 pf_counter_u64_zero(&rule->packets[i]);
4262 pf_counter_u64_zero(&rule->bytes[i]);
4263 }
4264 }
4265 PF_RULES_WUNLOCK();
4266 break;
4267 }
4268
4269 case DIOCGIFSPEEDV0:
4270 case DIOCGIFSPEEDV1: {
4271 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
4272 struct pf_ifspeed_v1 ps;
4273 struct ifnet *ifp;
4274
4275 if (psp->ifname[0] == '\0') {
4276 error = EINVAL;
4277 break;
4278 }
4279
4280 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4281 if (error != 0)
4282 break;
4283 ifp = ifunit(ps.ifname);
4284 if (ifp != NULL) {
4285 psp->baudrate32 =
4286 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4287 if (cmd == DIOCGIFSPEEDV1)
4288 psp->baudrate = ifp->if_baudrate;
4289 } else {
4290 error = EINVAL;
4291 }
4292 break;
4293 }
4294
4295 #ifdef ALTQ
4296 case DIOCSTARTALTQ: {
4297 struct pf_altq *altq;
4298
4299 PF_RULES_WLOCK();
4300 /* enable all altq interfaces on active list */
4301 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4302 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4303 error = pf_enable_altq(altq);
4304 if (error != 0)
4305 break;
4306 }
4307 }
4308 if (error == 0)
4309 V_pf_altq_running = 1;
4310 PF_RULES_WUNLOCK();
4311 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4312 break;
4313 }
4314
4315 case DIOCSTOPALTQ: {
4316 struct pf_altq *altq;
4317
4318 PF_RULES_WLOCK();
4319 /* disable all altq interfaces on active list */
4320 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4321 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4322 error = pf_disable_altq(altq);
4323 if (error != 0)
4324 break;
4325 }
4326 }
4327 if (error == 0)
4328 V_pf_altq_running = 0;
4329 PF_RULES_WUNLOCK();
4330 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4331 break;
4332 }
4333
4334 case DIOCADDALTQV0:
4335 case DIOCADDALTQV1: {
4336 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4337 struct pf_altq *altq, *a;
4338 struct ifnet *ifp;
4339
4340 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4341 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4342 if (error)
4343 break;
4344 altq->local_flags = 0;
4345
4346 PF_RULES_WLOCK();
4347 if (pa->ticket != V_ticket_altqs_inactive) {
4348 PF_RULES_WUNLOCK();
4349 free(altq, M_PFALTQ);
4350 error = EBUSY;
4351 break;
4352 }
4353
4354 /*
4355 * if this is for a queue, find the discipline and
4356 * copy the necessary fields
4357 */
4358 if (altq->qname[0] != 0) {
4359 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4360 PF_RULES_WUNLOCK();
4361 error = EBUSY;
4362 free(altq, M_PFALTQ);
4363 break;
4364 }
4365 altq->altq_disc = NULL;
4366 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4367 if (strncmp(a->ifname, altq->ifname,
4368 IFNAMSIZ) == 0) {
4369 altq->altq_disc = a->altq_disc;
4370 break;
4371 }
4372 }
4373 }
4374
4375 if ((ifp = ifunit(altq->ifname)) == NULL)
4376 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4377 else
4378 error = altq_add(ifp, altq);
4379
4380 if (error) {
4381 PF_RULES_WUNLOCK();
4382 free(altq, M_PFALTQ);
4383 break;
4384 }
4385
4386 if (altq->qname[0] != 0)
4387 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4388 else
4389 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4390 /* version error check done on import above */
4391 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4392 PF_RULES_WUNLOCK();
4393 break;
4394 }
4395
4396 case DIOCGETALTQSV0:
4397 case DIOCGETALTQSV1: {
4398 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4399 struct pf_altq *altq;
4400
4401 PF_RULES_RLOCK();
4402 pa->nr = 0;
4403 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4404 pa->nr++;
4405 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4406 pa->nr++;
4407 pa->ticket = V_ticket_altqs_active;
4408 PF_RULES_RUNLOCK();
4409 break;
4410 }
4411
4412 case DIOCGETALTQV0:
4413 case DIOCGETALTQV1: {
4414 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4415 struct pf_altq *altq;
4416
4417 PF_RULES_RLOCK();
4418 if (pa->ticket != V_ticket_altqs_active) {
4419 PF_RULES_RUNLOCK();
4420 error = EBUSY;
4421 break;
4422 }
4423 altq = pf_altq_get_nth_active(pa->nr);
4424 if (altq == NULL) {
4425 PF_RULES_RUNLOCK();
4426 error = EBUSY;
4427 break;
4428 }
4429 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4430 PF_RULES_RUNLOCK();
4431 break;
4432 }
4433
4434 case DIOCCHANGEALTQV0:
4435 case DIOCCHANGEALTQV1:
4436 /* CHANGEALTQ not supported yet! */
4437 error = ENODEV;
4438 break;
4439
4440 case DIOCGETQSTATSV0:
4441 case DIOCGETQSTATSV1: {
4442 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
4443 struct pf_altq *altq;
4444 int nbytes;
4445 u_int32_t version;
4446
4447 PF_RULES_RLOCK();
4448 if (pq->ticket != V_ticket_altqs_active) {
4449 PF_RULES_RUNLOCK();
4450 error = EBUSY;
4451 break;
4452 }
4453 nbytes = pq->nbytes;
4454 altq = pf_altq_get_nth_active(pq->nr);
4455 if (altq == NULL) {
4456 PF_RULES_RUNLOCK();
4457 error = EBUSY;
4458 break;
4459 }
4460
4461 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4462 PF_RULES_RUNLOCK();
4463 error = ENXIO;
4464 break;
4465 }
4466 PF_RULES_RUNLOCK();
4467 if (cmd == DIOCGETQSTATSV0)
4468 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
4469 else
4470 version = pq->version;
4471 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4472 if (error == 0) {
4473 pq->scheduler = altq->scheduler;
4474 pq->nbytes = nbytes;
4475 }
4476 break;
4477 }
4478 #endif /* ALTQ */
4479
4480 case DIOCBEGINADDRS: {
4481 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4482
4483 error = pf_ioctl_begin_addrs(&pp->ticket);
4484 break;
4485 }
4486
4487 case DIOCADDADDR: {
4488 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4489 struct pf_nl_pooladdr npp = {};
4490
4491 npp.which = PF_RDR;
4492 memcpy(&npp, pp, sizeof(*pp));
4493 error = pf_ioctl_add_addr(&npp);
4494 break;
4495 }
4496
4497 case DIOCGETADDRS: {
4498 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4499 struct pf_nl_pooladdr npp = {};
4500
4501 npp.which = PF_RDR;
4502 memcpy(&npp, pp, sizeof(*pp));
4503 error = pf_ioctl_get_addrs(&npp);
4504 memcpy(pp, &npp, sizeof(*pp));
4505
4506 break;
4507 }
4508
4509 case DIOCGETADDR: {
4510 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4511 struct pf_nl_pooladdr npp = {};
4512
4513 npp.which = PF_RDR;
4514 memcpy(&npp, pp, sizeof(*pp));
4515 error = pf_ioctl_get_addr(&npp);
4516 memcpy(pp, &npp, sizeof(*pp));
4517
4518 break;
4519 }
4520
4521 case DIOCCHANGEADDR: {
4522 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
4523 struct pf_kpool *pool;
4524 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
4525 struct pf_kruleset *ruleset;
4526 struct pfi_kkif *kif = NULL;
4527
4528 pca->anchor[sizeof(pca->anchor) - 1] = '\0';
4529
4530 if (pca->action < PF_CHANGE_ADD_HEAD ||
4531 pca->action > PF_CHANGE_REMOVE) {
4532 error = EINVAL;
4533 break;
4534 }
4535 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4536 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4537 pca->addr.addr.type != PF_ADDR_TABLE) {
4538 error = EINVAL;
4539 break;
4540 }
4541 if (pca->addr.addr.p.dyn != NULL) {
4542 error = EINVAL;
4543 break;
4544 }
4545
4546 if (pca->action != PF_CHANGE_REMOVE) {
4547 #ifndef INET
4548 if (pca->af == AF_INET) {
4549 error = EAFNOSUPPORT;
4550 break;
4551 }
4552 #endif /* INET */
4553 #ifndef INET6
4554 if (pca->af == AF_INET6) {
4555 error = EAFNOSUPPORT;
4556 break;
4557 }
4558 #endif /* INET6 */
4559 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4560 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4561 if (newpa->ifname[0])
4562 kif = pf_kkif_create(M_WAITOK);
4563 newpa->kif = NULL;
4564 }
4565 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4566 PF_RULES_WLOCK();
4567 ruleset = pf_find_kruleset(pca->anchor);
4568 if (ruleset == NULL)
4569 ERROUT(EBUSY);
4570
4571 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4572 pca->r_num, pca->r_last, 1, 1, PF_RDR);
4573 if (pool == NULL)
4574 ERROUT(EBUSY);
4575
4576 if (pca->action != PF_CHANGE_REMOVE) {
4577 if (newpa->ifname[0]) {
4578 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4579 pfi_kkif_ref(newpa->kif);
4580 kif = NULL;
4581 }
4582
4583 switch (newpa->addr.type) {
4584 case PF_ADDR_DYNIFTL:
4585 error = pfi_dynaddr_setup(&newpa->addr,
4586 pca->af);
4587 break;
4588 case PF_ADDR_TABLE:
4589 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4590 newpa->addr.v.tblname);
4591 if (newpa->addr.p.tbl == NULL)
4592 error = ENOMEM;
4593 break;
4594 }
4595 if (error)
4596 goto DIOCCHANGEADDR_error;
4597 }
4598
4599 switch (pca->action) {
4600 case PF_CHANGE_ADD_HEAD:
4601 oldpa = TAILQ_FIRST(&pool->list);
4602 break;
4603 case PF_CHANGE_ADD_TAIL:
4604 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4605 break;
4606 default:
4607 oldpa = TAILQ_FIRST(&pool->list);
4608 for (int i = 0; oldpa && i < pca->nr; i++)
4609 oldpa = TAILQ_NEXT(oldpa, entries);
4610
4611 if (oldpa == NULL)
4612 ERROUT(EINVAL);
4613 }
4614
4615 if (pca->action == PF_CHANGE_REMOVE) {
4616 TAILQ_REMOVE(&pool->list, oldpa, entries);
4617 switch (oldpa->addr.type) {
4618 case PF_ADDR_DYNIFTL:
4619 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4620 break;
4621 case PF_ADDR_TABLE:
4622 pfr_detach_table(oldpa->addr.p.tbl);
4623 break;
4624 }
4625 if (oldpa->kif)
4626 pfi_kkif_unref(oldpa->kif);
4627 free(oldpa, M_PFRULE);
4628 } else {
4629 if (oldpa == NULL)
4630 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4631 else if (pca->action == PF_CHANGE_ADD_HEAD ||
4632 pca->action == PF_CHANGE_ADD_BEFORE)
4633 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4634 else
4635 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4636 newpa, entries);
4637 }
4638
4639 pool->cur = TAILQ_FIRST(&pool->list);
4640 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4641 PF_RULES_WUNLOCK();
4642 break;
4643
4644 #undef ERROUT
4645 DIOCCHANGEADDR_error:
4646 if (newpa != NULL) {
4647 if (newpa->kif)
4648 pfi_kkif_unref(newpa->kif);
4649 free(newpa, M_PFRULE);
4650 }
4651 PF_RULES_WUNLOCK();
4652 pf_kkif_free(kif);
4653 break;
4654 }
4655
4656 case DIOCGETRULESETS: {
4657 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4658
4659 pr->path[sizeof(pr->path) - 1] = '\0';
4660
4661 error = pf_ioctl_get_rulesets(pr);
4662 break;
4663 }
4664
4665 case DIOCGETRULESET: {
4666 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4667
4668 pr->path[sizeof(pr->path) - 1] = '\0';
4669
4670 error = pf_ioctl_get_ruleset(pr);
4671 break;
4672 }
4673
4674 case DIOCRCLRTABLES: {
4675 struct pfioc_table *io = (struct pfioc_table *)addr;
4676
4677 if (io->pfrio_esize != 0) {
4678 error = ENODEV;
4679 break;
4680 }
4681 PF_RULES_WLOCK();
4682 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4683 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4684 PF_RULES_WUNLOCK();
4685 break;
4686 }
4687
4688 case DIOCRADDTABLES: {
4689 struct pfioc_table *io = (struct pfioc_table *)addr;
4690 struct pfr_table *pfrts;
4691 size_t totlen;
4692
4693 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4694 error = ENODEV;
4695 break;
4696 }
4697
4698 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4699 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4700 error = ENOMEM;
4701 break;
4702 }
4703
4704 totlen = io->pfrio_size * sizeof(struct pfr_table);
4705 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4706 M_TEMP, M_WAITOK);
4707 error = copyin(io->pfrio_buffer, pfrts, totlen);
4708 if (error) {
4709 free(pfrts, M_TEMP);
4710 break;
4711 }
4712 PF_RULES_WLOCK();
4713 error = pfr_add_tables(pfrts, io->pfrio_size,
4714 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4715 PF_RULES_WUNLOCK();
4716 free(pfrts, M_TEMP);
4717 break;
4718 }
4719
4720 case DIOCRDELTABLES: {
4721 struct pfioc_table *io = (struct pfioc_table *)addr;
4722 struct pfr_table *pfrts;
4723 size_t totlen;
4724
4725 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4726 error = ENODEV;
4727 break;
4728 }
4729
4730 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4731 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4732 error = ENOMEM;
4733 break;
4734 }
4735
4736 totlen = io->pfrio_size * sizeof(struct pfr_table);
4737 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4738 M_TEMP, M_WAITOK);
4739 error = copyin(io->pfrio_buffer, pfrts, totlen);
4740 if (error) {
4741 free(pfrts, M_TEMP);
4742 break;
4743 }
4744 PF_RULES_WLOCK();
4745 error = pfr_del_tables(pfrts, io->pfrio_size,
4746 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4747 PF_RULES_WUNLOCK();
4748 free(pfrts, M_TEMP);
4749 break;
4750 }
4751
4752 case DIOCRGETTABLES: {
4753 struct pfioc_table *io = (struct pfioc_table *)addr;
4754 struct pfr_table *pfrts;
4755 size_t totlen;
4756 int n;
4757
4758 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4759 error = ENODEV;
4760 break;
4761 }
4762 PF_RULES_RLOCK();
4763 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4764 if (n < 0) {
4765 PF_RULES_RUNLOCK();
4766 error = EINVAL;
4767 break;
4768 }
4769 io->pfrio_size = min(io->pfrio_size, n);
4770
4771 totlen = io->pfrio_size * sizeof(struct pfr_table);
4772
4773 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4774 M_TEMP, M_NOWAIT | M_ZERO);
4775 if (pfrts == NULL) {
4776 error = ENOMEM;
4777 PF_RULES_RUNLOCK();
4778 break;
4779 }
4780 error = pfr_get_tables(&io->pfrio_table, pfrts,
4781 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4782 PF_RULES_RUNLOCK();
4783 if (error == 0)
4784 error = copyout(pfrts, io->pfrio_buffer, totlen);
4785 free(pfrts, M_TEMP);
4786 break;
4787 }
4788
4789 case DIOCRGETTSTATS: {
4790 struct pfioc_table *io = (struct pfioc_table *)addr;
4791 struct pfr_tstats *pfrtstats;
4792 size_t totlen;
4793 int n;
4794
4795 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4796 error = ENODEV;
4797 break;
4798 }
4799 PF_TABLE_STATS_LOCK();
4800 PF_RULES_RLOCK();
4801 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4802 if (n < 0) {
4803 PF_RULES_RUNLOCK();
4804 PF_TABLE_STATS_UNLOCK();
4805 error = EINVAL;
4806 break;
4807 }
4808 io->pfrio_size = min(io->pfrio_size, n);
4809
4810 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4811 pfrtstats = mallocarray(io->pfrio_size,
4812 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4813 if (pfrtstats == NULL) {
4814 error = ENOMEM;
4815 PF_RULES_RUNLOCK();
4816 PF_TABLE_STATS_UNLOCK();
4817 break;
4818 }
4819 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4820 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4821 PF_RULES_RUNLOCK();
4822 PF_TABLE_STATS_UNLOCK();
4823 if (error == 0)
4824 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4825 free(pfrtstats, M_TEMP);
4826 break;
4827 }
4828
4829 case DIOCRCLRTSTATS: {
4830 struct pfioc_table *io = (struct pfioc_table *)addr;
4831 struct pfr_table *pfrts;
4832 size_t totlen;
4833
4834 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4835 error = ENODEV;
4836 break;
4837 }
4838
4839 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4840 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4841 /* We used to count tables and use the minimum required
4842 * size, so we didn't fail on overly large requests.
4843 * Keep doing so. */
4844 io->pfrio_size = pf_ioctl_maxcount;
4845 break;
4846 }
4847
4848 totlen = io->pfrio_size * sizeof(struct pfr_table);
4849 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4850 M_TEMP, M_WAITOK);
4851 error = copyin(io->pfrio_buffer, pfrts, totlen);
4852 if (error) {
4853 free(pfrts, M_TEMP);
4854 break;
4855 }
4856
4857 PF_TABLE_STATS_LOCK();
4858 PF_RULES_RLOCK();
4859 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4860 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4861 PF_RULES_RUNLOCK();
4862 PF_TABLE_STATS_UNLOCK();
4863 free(pfrts, M_TEMP);
4864 break;
4865 }
4866
4867 case DIOCRSETTFLAGS: {
4868 struct pfioc_table *io = (struct pfioc_table *)addr;
4869 struct pfr_table *pfrts;
4870 size_t totlen;
4871 int n;
4872
4873 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4874 error = ENODEV;
4875 break;
4876 }
4877
4878 PF_RULES_RLOCK();
4879 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4880 if (n < 0) {
4881 PF_RULES_RUNLOCK();
4882 error = EINVAL;
4883 break;
4884 }
4885
4886 io->pfrio_size = min(io->pfrio_size, n);
4887 PF_RULES_RUNLOCK();
4888
4889 totlen = io->pfrio_size * sizeof(struct pfr_table);
4890 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4891 M_TEMP, M_WAITOK);
4892 error = copyin(io->pfrio_buffer, pfrts, totlen);
4893 if (error) {
4894 free(pfrts, M_TEMP);
4895 break;
4896 }
4897 PF_RULES_WLOCK();
4898 error = pfr_set_tflags(pfrts, io->pfrio_size,
4899 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4900 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4901 PF_RULES_WUNLOCK();
4902 free(pfrts, M_TEMP);
4903 break;
4904 }
4905
4906 case DIOCRCLRADDRS: {
4907 struct pfioc_table *io = (struct pfioc_table *)addr;
4908
4909 if (io->pfrio_esize != 0) {
4910 error = ENODEV;
4911 break;
4912 }
4913 PF_RULES_WLOCK();
4914 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4915 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4916 PF_RULES_WUNLOCK();
4917 break;
4918 }
4919
4920 case DIOCRADDADDRS: {
4921 struct pfioc_table *io = (struct pfioc_table *)addr;
4922 struct pfr_addr *pfras;
4923 size_t totlen;
4924
4925 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4926 error = ENODEV;
4927 break;
4928 }
4929 if (io->pfrio_size < 0 ||
4930 io->pfrio_size > pf_ioctl_maxcount ||
4931 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4932 error = EINVAL;
4933 break;
4934 }
4935 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4936 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4937 M_TEMP, M_WAITOK);
4938 error = copyin(io->pfrio_buffer, pfras, totlen);
4939 if (error) {
4940 free(pfras, M_TEMP);
4941 break;
4942 }
4943 PF_RULES_WLOCK();
4944 error = pfr_add_addrs(&io->pfrio_table, pfras,
4945 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4946 PFR_FLAG_USERIOCTL);
4947 PF_RULES_WUNLOCK();
4948 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4949 error = copyout(pfras, io->pfrio_buffer, totlen);
4950 free(pfras, M_TEMP);
4951 break;
4952 }
4953
4954 case DIOCRDELADDRS: {
4955 struct pfioc_table *io = (struct pfioc_table *)addr;
4956 struct pfr_addr *pfras;
4957 size_t totlen;
4958
4959 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4960 error = ENODEV;
4961 break;
4962 }
4963 if (io->pfrio_size < 0 ||
4964 io->pfrio_size > pf_ioctl_maxcount ||
4965 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4966 error = EINVAL;
4967 break;
4968 }
4969 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4970 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4971 M_TEMP, M_WAITOK);
4972 error = copyin(io->pfrio_buffer, pfras, totlen);
4973 if (error) {
4974 free(pfras, M_TEMP);
4975 break;
4976 }
4977 PF_RULES_WLOCK();
4978 error = pfr_del_addrs(&io->pfrio_table, pfras,
4979 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4980 PFR_FLAG_USERIOCTL);
4981 PF_RULES_WUNLOCK();
4982 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4983 error = copyout(pfras, io->pfrio_buffer, totlen);
4984 free(pfras, M_TEMP);
4985 break;
4986 }
4987
4988 case DIOCRSETADDRS: {
4989 struct pfioc_table *io = (struct pfioc_table *)addr;
4990 struct pfr_addr *pfras;
4991 size_t totlen, count;
4992
4993 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4994 error = ENODEV;
4995 break;
4996 }
4997 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4998 error = EINVAL;
4999 break;
5000 }
5001 count = max(io->pfrio_size, io->pfrio_size2);
5002 if (count > pf_ioctl_maxcount ||
5003 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
5004 error = EINVAL;
5005 break;
5006 }
5007 totlen = count * sizeof(struct pfr_addr);
5008 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
5009 M_WAITOK);
5010 error = copyin(io->pfrio_buffer, pfras, totlen);
5011 if (error) {
5012 free(pfras, M_TEMP);
5013 break;
5014 }
5015 PF_RULES_WLOCK();
5016 error = pfr_set_addrs(&io->pfrio_table, pfras,
5017 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
5018 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
5019 PFR_FLAG_USERIOCTL, 0);
5020 PF_RULES_WUNLOCK();
5021 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5022 error = copyout(pfras, io->pfrio_buffer, totlen);
5023 free(pfras, M_TEMP);
5024 break;
5025 }
5026
5027 case DIOCRGETADDRS: {
5028 struct pfioc_table *io = (struct pfioc_table *)addr;
5029 struct pfr_addr *pfras;
5030 size_t totlen;
5031
5032 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5033 error = ENODEV;
5034 break;
5035 }
5036 if (io->pfrio_size < 0 ||
5037 io->pfrio_size > pf_ioctl_maxcount ||
5038 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5039 error = EINVAL;
5040 break;
5041 }
5042 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5043 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5044 M_TEMP, M_WAITOK | M_ZERO);
5045 PF_RULES_RLOCK();
5046 error = pfr_get_addrs(&io->pfrio_table, pfras,
5047 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5048 PF_RULES_RUNLOCK();
5049 if (error == 0)
5050 error = copyout(pfras, io->pfrio_buffer, totlen);
5051 free(pfras, M_TEMP);
5052 break;
5053 }
5054
5055 case DIOCRGETASTATS: {
5056 struct pfioc_table *io = (struct pfioc_table *)addr;
5057 struct pfr_astats *pfrastats;
5058 size_t totlen;
5059
5060 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
5061 error = ENODEV;
5062 break;
5063 }
5064 if (io->pfrio_size < 0 ||
5065 io->pfrio_size > pf_ioctl_maxcount ||
5066 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5067 error = EINVAL;
5068 break;
5069 }
5070 totlen = io->pfrio_size * sizeof(struct pfr_astats);
5071 pfrastats = mallocarray(io->pfrio_size,
5072 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
5073 PF_RULES_RLOCK();
5074 error = pfr_get_astats(&io->pfrio_table, pfrastats,
5075 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5076 PF_RULES_RUNLOCK();
5077 if (error == 0)
5078 error = copyout(pfrastats, io->pfrio_buffer, totlen);
5079 free(pfrastats, M_TEMP);
5080 break;
5081 }
5082
5083 case DIOCRCLRASTATS: {
5084 struct pfioc_table *io = (struct pfioc_table *)addr;
5085 struct pfr_addr *pfras;
5086 size_t totlen;
5087
5088 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5089 error = ENODEV;
5090 break;
5091 }
5092 if (io->pfrio_size < 0 ||
5093 io->pfrio_size > pf_ioctl_maxcount ||
5094 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5095 error = EINVAL;
5096 break;
5097 }
5098 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5099 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5100 M_TEMP, M_WAITOK);
5101 error = copyin(io->pfrio_buffer, pfras, totlen);
5102 if (error) {
5103 free(pfras, M_TEMP);
5104 break;
5105 }
5106 PF_RULES_WLOCK();
5107 error = pfr_clr_astats(&io->pfrio_table, pfras,
5108 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5109 PFR_FLAG_USERIOCTL);
5110 PF_RULES_WUNLOCK();
5111 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5112 error = copyout(pfras, io->pfrio_buffer, totlen);
5113 free(pfras, M_TEMP);
5114 break;
5115 }
5116
5117 case DIOCRTSTADDRS: {
5118 struct pfioc_table *io = (struct pfioc_table *)addr;
5119 struct pfr_addr *pfras;
5120 size_t totlen;
5121
5122 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5123 error = ENODEV;
5124 break;
5125 }
5126 if (io->pfrio_size < 0 ||
5127 io->pfrio_size > pf_ioctl_maxcount ||
5128 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5129 error = EINVAL;
5130 break;
5131 }
5132 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5133 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5134 M_TEMP, M_WAITOK);
5135 error = copyin(io->pfrio_buffer, pfras, totlen);
5136 if (error) {
5137 free(pfras, M_TEMP);
5138 break;
5139 }
5140 PF_RULES_RLOCK();
5141 error = pfr_tst_addrs(&io->pfrio_table, pfras,
5142 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5143 PFR_FLAG_USERIOCTL);
5144 PF_RULES_RUNLOCK();
5145 if (error == 0)
5146 error = copyout(pfras, io->pfrio_buffer, totlen);
5147 free(pfras, M_TEMP);
5148 break;
5149 }
5150
5151 case DIOCRINADEFINE: {
5152 struct pfioc_table *io = (struct pfioc_table *)addr;
5153 struct pfr_addr *pfras;
5154 size_t totlen;
5155
5156 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5157 error = ENODEV;
5158 break;
5159 }
5160 if (io->pfrio_size < 0 ||
5161 io->pfrio_size > pf_ioctl_maxcount ||
5162 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5163 error = EINVAL;
5164 break;
5165 }
5166 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5167 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5168 M_TEMP, M_WAITOK);
5169 error = copyin(io->pfrio_buffer, pfras, totlen);
5170 if (error) {
5171 free(pfras, M_TEMP);
5172 break;
5173 }
5174 PF_RULES_WLOCK();
5175 error = pfr_ina_define(&io->pfrio_table, pfras,
5176 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5177 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5178 PF_RULES_WUNLOCK();
5179 free(pfras, M_TEMP);
5180 break;
5181 }
5182
5183 case DIOCOSFPADD: {
5184 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5185 PF_RULES_WLOCK();
5186 error = pf_osfp_add(io);
5187 PF_RULES_WUNLOCK();
5188 break;
5189 }
5190
5191 case DIOCOSFPGET: {
5192 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5193 PF_RULES_RLOCK();
5194 error = pf_osfp_get(io);
5195 PF_RULES_RUNLOCK();
5196 break;
5197 }
5198
5199 case DIOCXBEGIN: {
5200 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5201 struct pfioc_trans_e *ioes, *ioe;
5202 size_t totlen;
5203 int i;
5204
5205 if (io->esize != sizeof(*ioe)) {
5206 error = ENODEV;
5207 break;
5208 }
5209 if (io->size < 0 ||
5210 io->size > pf_ioctl_maxcount ||
5211 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5212 error = EINVAL;
5213 break;
5214 }
5215 totlen = sizeof(struct pfioc_trans_e) * io->size;
5216 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5217 M_TEMP, M_WAITOK);
5218 error = copyin(io->array, ioes, totlen);
5219 if (error) {
5220 free(ioes, M_TEMP);
5221 break;
5222 }
5223 PF_RULES_WLOCK();
5224 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5225 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5226 switch (ioe->rs_num) {
5227 case PF_RULESET_ETH:
5228 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5229 PF_RULES_WUNLOCK();
5230 free(ioes, M_TEMP);
5231 goto fail;
5232 }
5233 break;
5234 #ifdef ALTQ
5235 case PF_RULESET_ALTQ:
5236 if (ioe->anchor[0]) {
5237 PF_RULES_WUNLOCK();
5238 free(ioes, M_TEMP);
5239 error = EINVAL;
5240 goto fail;
5241 }
5242 if ((error = pf_begin_altq(&ioe->ticket))) {
5243 PF_RULES_WUNLOCK();
5244 free(ioes, M_TEMP);
5245 goto fail;
5246 }
5247 break;
5248 #endif /* ALTQ */
5249 case PF_RULESET_TABLE:
5250 {
5251 struct pfr_table table;
5252
5253 bzero(&table, sizeof(table));
5254 strlcpy(table.pfrt_anchor, ioe->anchor,
5255 sizeof(table.pfrt_anchor));
5256 if ((error = pfr_ina_begin(&table,
5257 &ioe->ticket, NULL, 0))) {
5258 PF_RULES_WUNLOCK();
5259 free(ioes, M_TEMP);
5260 goto fail;
5261 }
5262 break;
5263 }
5264 default:
5265 if ((error = pf_begin_rules(&ioe->ticket,
5266 ioe->rs_num, ioe->anchor))) {
5267 PF_RULES_WUNLOCK();
5268 free(ioes, M_TEMP);
5269 goto fail;
5270 }
5271 break;
5272 }
5273 }
5274 PF_RULES_WUNLOCK();
5275 error = copyout(ioes, io->array, totlen);
5276 free(ioes, M_TEMP);
5277 break;
5278 }
5279
5280 case DIOCXROLLBACK: {
5281 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5282 struct pfioc_trans_e *ioe, *ioes;
5283 size_t totlen;
5284 int i;
5285
5286 if (io->esize != sizeof(*ioe)) {
5287 error = ENODEV;
5288 break;
5289 }
5290 if (io->size < 0 ||
5291 io->size > pf_ioctl_maxcount ||
5292 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5293 error = EINVAL;
5294 break;
5295 }
5296 totlen = sizeof(struct pfioc_trans_e) * io->size;
5297 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5298 M_TEMP, M_WAITOK);
5299 error = copyin(io->array, ioes, totlen);
5300 if (error) {
5301 free(ioes, M_TEMP);
5302 break;
5303 }
5304 PF_RULES_WLOCK();
5305 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5306 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5307 switch (ioe->rs_num) {
5308 case PF_RULESET_ETH:
5309 if ((error = pf_rollback_eth(ioe->ticket,
5310 ioe->anchor))) {
5311 PF_RULES_WUNLOCK();
5312 free(ioes, M_TEMP);
5313 goto fail; /* really bad */
5314 }
5315 break;
5316 #ifdef ALTQ
5317 case PF_RULESET_ALTQ:
5318 if (ioe->anchor[0]) {
5319 PF_RULES_WUNLOCK();
5320 free(ioes, M_TEMP);
5321 error = EINVAL;
5322 goto fail;
5323 }
5324 if ((error = pf_rollback_altq(ioe->ticket))) {
5325 PF_RULES_WUNLOCK();
5326 free(ioes, M_TEMP);
5327 goto fail; /* really bad */
5328 }
5329 break;
5330 #endif /* ALTQ */
5331 case PF_RULESET_TABLE:
5332 {
5333 struct pfr_table table;
5334
5335 bzero(&table, sizeof(table));
5336 strlcpy(table.pfrt_anchor, ioe->anchor,
5337 sizeof(table.pfrt_anchor));
5338 if ((error = pfr_ina_rollback(&table,
5339 ioe->ticket, NULL, 0))) {
5340 PF_RULES_WUNLOCK();
5341 free(ioes, M_TEMP);
5342 goto fail; /* really bad */
5343 }
5344 break;
5345 }
5346 default:
5347 if ((error = pf_rollback_rules(ioe->ticket,
5348 ioe->rs_num, ioe->anchor))) {
5349 PF_RULES_WUNLOCK();
5350 free(ioes, M_TEMP);
5351 goto fail; /* really bad */
5352 }
5353 break;
5354 }
5355 }
5356 PF_RULES_WUNLOCK();
5357 free(ioes, M_TEMP);
5358 break;
5359 }
5360
5361 case DIOCXCOMMIT: {
5362 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5363 struct pfioc_trans_e *ioe, *ioes;
5364 struct pf_kruleset *rs;
5365 struct pf_keth_ruleset *ers;
5366 size_t totlen;
5367 int i;
5368
5369 if (io->esize != sizeof(*ioe)) {
5370 error = ENODEV;
5371 break;
5372 }
5373
5374 if (io->size < 0 ||
5375 io->size > pf_ioctl_maxcount ||
5376 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5377 error = EINVAL;
5378 break;
5379 }
5380
5381 totlen = sizeof(struct pfioc_trans_e) * io->size;
5382 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5383 M_TEMP, M_WAITOK);
5384 error = copyin(io->array, ioes, totlen);
5385 if (error) {
5386 free(ioes, M_TEMP);
5387 break;
5388 }
5389 PF_RULES_WLOCK();
5390 /* First makes sure everything will succeed. */
5391 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5392 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5393 switch (ioe->rs_num) {
5394 case PF_RULESET_ETH:
5395 ers = pf_find_keth_ruleset(ioe->anchor);
5396 if (ers == NULL || ioe->ticket == 0 ||
5397 ioe->ticket != ers->inactive.ticket) {
5398 PF_RULES_WUNLOCK();
5399 free(ioes, M_TEMP);
5400 error = EINVAL;
5401 goto fail;
5402 }
5403 break;
5404 #ifdef ALTQ
5405 case PF_RULESET_ALTQ:
5406 if (ioe->anchor[0]) {
5407 PF_RULES_WUNLOCK();
5408 free(ioes, M_TEMP);
5409 error = EINVAL;
5410 goto fail;
5411 }
5412 if (!V_altqs_inactive_open || ioe->ticket !=
5413 V_ticket_altqs_inactive) {
5414 PF_RULES_WUNLOCK();
5415 free(ioes, M_TEMP);
5416 error = EBUSY;
5417 goto fail;
5418 }
5419 break;
5420 #endif /* ALTQ */
5421 case PF_RULESET_TABLE:
5422 rs = pf_find_kruleset(ioe->anchor);
5423 if (rs == NULL || !rs->topen || ioe->ticket !=
5424 rs->tticket) {
5425 PF_RULES_WUNLOCK();
5426 free(ioes, M_TEMP);
5427 error = EBUSY;
5428 goto fail;
5429 }
5430 break;
5431 default:
5432 if (ioe->rs_num < 0 || ioe->rs_num >=
5433 PF_RULESET_MAX) {
5434 PF_RULES_WUNLOCK();
5435 free(ioes, M_TEMP);
5436 error = EINVAL;
5437 goto fail;
5438 }
5439 rs = pf_find_kruleset(ioe->anchor);
5440 if (rs == NULL ||
5441 !rs->rules[ioe->rs_num].inactive.open ||
5442 rs->rules[ioe->rs_num].inactive.ticket !=
5443 ioe->ticket) {
5444 PF_RULES_WUNLOCK();
5445 free(ioes, M_TEMP);
5446 error = EBUSY;
5447 goto fail;
5448 }
5449 break;
5450 }
5451 }
5452 /* Now do the commit - no errors should happen here. */
5453 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5454 switch (ioe->rs_num) {
5455 case PF_RULESET_ETH:
5456 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5457 PF_RULES_WUNLOCK();
5458 free(ioes, M_TEMP);
5459 goto fail; /* really bad */
5460 }
5461 break;
5462 #ifdef ALTQ
5463 case PF_RULESET_ALTQ:
5464 if ((error = pf_commit_altq(ioe->ticket))) {
5465 PF_RULES_WUNLOCK();
5466 free(ioes, M_TEMP);
5467 goto fail; /* really bad */
5468 }
5469 break;
5470 #endif /* ALTQ */
5471 case PF_RULESET_TABLE:
5472 {
5473 struct pfr_table table;
5474
5475 bzero(&table, sizeof(table));
5476 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
5477 sizeof(table.pfrt_anchor));
5478 if ((error = pfr_ina_commit(&table,
5479 ioe->ticket, NULL, NULL, 0))) {
5480 PF_RULES_WUNLOCK();
5481 free(ioes, M_TEMP);
5482 goto fail; /* really bad */
5483 }
5484 break;
5485 }
5486 default:
5487 if ((error = pf_commit_rules(ioe->ticket,
5488 ioe->rs_num, ioe->anchor))) {
5489 PF_RULES_WUNLOCK();
5490 free(ioes, M_TEMP);
5491 goto fail; /* really bad */
5492 }
5493 break;
5494 }
5495 }
5496 PF_RULES_WUNLOCK();
5497
5498 /* Only hook into EtherNet taffic if we've got rules for it. */
5499 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5500 hook_pf_eth();
5501 else
5502 dehook_pf_eth();
5503
5504 free(ioes, M_TEMP);
5505 break;
5506 }
5507
5508 case DIOCGETSRCNODES: {
5509 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
5510 struct pf_srchash *sh;
5511 struct pf_ksrc_node *n;
5512 struct pf_src_node *p, *pstore;
5513 uint32_t i, nr = 0;
5514
5515 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5516 i++, sh++) {
5517 PF_HASHROW_LOCK(sh);
5518 LIST_FOREACH(n, &sh->nodes, entry)
5519 nr++;
5520 PF_HASHROW_UNLOCK(sh);
5521 }
5522
5523 psn->psn_len = min(psn->psn_len,
5524 sizeof(struct pf_src_node) * nr);
5525
5526 if (psn->psn_len == 0) {
5527 psn->psn_len = sizeof(struct pf_src_node) * nr;
5528 break;
5529 }
5530
5531 nr = 0;
5532
5533 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5534 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5535 i++, sh++) {
5536 PF_HASHROW_LOCK(sh);
5537 LIST_FOREACH(n, &sh->nodes, entry) {
5538
5539 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5540 break;
5541
5542 pf_src_node_copy(n, p);
5543
5544 p++;
5545 nr++;
5546 }
5547 PF_HASHROW_UNLOCK(sh);
5548 }
5549 error = copyout(pstore, psn->psn_src_nodes,
5550 sizeof(struct pf_src_node) * nr);
5551 if (error) {
5552 free(pstore, M_TEMP);
5553 break;
5554 }
5555 psn->psn_len = sizeof(struct pf_src_node) * nr;
5556 free(pstore, M_TEMP);
5557 break;
5558 }
5559
5560 case DIOCCLRSRCNODES: {
5561 pf_kill_srcnodes(NULL);
5562 break;
5563 }
5564
5565 case DIOCKILLSRCNODES:
5566 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5567 break;
5568
5569 #ifdef COMPAT_FREEBSD13
5570 case DIOCKEEPCOUNTERS_FREEBSD13:
5571 #endif
5572 case DIOCKEEPCOUNTERS:
5573 error = pf_keepcounters((struct pfioc_nv *)addr);
5574 break;
5575
5576 case DIOCGETSYNCOOKIES:
5577 error = pf_get_syncookies((struct pfioc_nv *)addr);
5578 break;
5579
5580 case DIOCSETSYNCOOKIES:
5581 error = pf_set_syncookies((struct pfioc_nv *)addr);
5582 break;
5583
5584 case DIOCSETHOSTID: {
5585 u_int32_t *hostid = (u_int32_t *)addr;
5586
5587 PF_RULES_WLOCK();
5588 if (*hostid == 0)
5589 V_pf_status.hostid = arc4random();
5590 else
5591 V_pf_status.hostid = *hostid;
5592 PF_RULES_WUNLOCK();
5593 break;
5594 }
5595
5596 case DIOCOSFPFLUSH:
5597 PF_RULES_WLOCK();
5598 pf_osfp_flush();
5599 PF_RULES_WUNLOCK();
5600 break;
5601
5602 case DIOCIGETIFACES: {
5603 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5604 struct pfi_kif *ifstore;
5605 size_t bufsiz;
5606
5607 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5608 error = ENODEV;
5609 break;
5610 }
5611
5612 if (io->pfiio_size < 0 ||
5613 io->pfiio_size > pf_ioctl_maxcount ||
5614 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5615 error = EINVAL;
5616 break;
5617 }
5618
5619 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5620
5621 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5622 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5623 M_TEMP, M_WAITOK | M_ZERO);
5624
5625 PF_RULES_RLOCK();
5626 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5627 PF_RULES_RUNLOCK();
5628 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5629 free(ifstore, M_TEMP);
5630 break;
5631 }
5632
5633 case DIOCSETIFFLAG: {
5634 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5635
5636 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5637
5638 PF_RULES_WLOCK();
5639 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5640 PF_RULES_WUNLOCK();
5641 break;
5642 }
5643
5644 case DIOCCLRIFFLAG: {
5645 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5646
5647 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5648
5649 PF_RULES_WLOCK();
5650 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5651 PF_RULES_WUNLOCK();
5652 break;
5653 }
5654
5655 case DIOCSETREASS: {
5656 u_int32_t *reass = (u_int32_t *)addr;
5657
5658 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5659 /* Removal of DF flag without reassembly enabled is not a
5660 * valid combination. Disable reassembly in such case. */
5661 if (!(V_pf_status.reass & PF_REASS_ENABLED))
5662 V_pf_status.reass = 0;
5663 break;
5664 }
5665
5666 default:
5667 error = ENODEV;
5668 break;
5669 }
5670 fail:
5671 CURVNET_RESTORE();
5672
5673 #undef ERROUT_IOCTL
5674
5675 return (error);
5676 }
5677
5678 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5679 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5680 {
5681 bzero(sp, sizeof(union pfsync_state_union));
5682
5683 /* copy from state key */
5684 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5685 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5686 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5687 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5688 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5689 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5690 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5691 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5692 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5693 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5694
5695 /* copy from state */
5696 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5697 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5698 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5699 sp->pfs_1301.expire = pf_state_expires(st);
5700 if (sp->pfs_1301.expire <= time_uptime)
5701 sp->pfs_1301.expire = htonl(0);
5702 else
5703 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5704
5705 sp->pfs_1301.direction = st->direction;
5706 sp->pfs_1301.log = st->act.log;
5707 sp->pfs_1301.timeout = st->timeout;
5708
5709 switch (msg_version) {
5710 case PFSYNC_MSG_VERSION_1301:
5711 sp->pfs_1301.state_flags = st->state_flags;
5712 break;
5713 case PFSYNC_MSG_VERSION_1400:
5714 sp->pfs_1400.state_flags = htons(st->state_flags);
5715 sp->pfs_1400.qid = htons(st->act.qid);
5716 sp->pfs_1400.pqid = htons(st->act.pqid);
5717 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5718 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5719 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5720 sp->pfs_1400.min_ttl = st->act.min_ttl;
5721 sp->pfs_1400.set_tos = st->act.set_tos;
5722 sp->pfs_1400.max_mss = htons(st->act.max_mss);
5723 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5724 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5725 sp->pfs_1400.rt = st->act.rt;
5726 if (st->act.rt_kif)
5727 strlcpy(sp->pfs_1400.rt_ifname,
5728 st->act.rt_kif->pfik_name,
5729 sizeof(sp->pfs_1400.rt_ifname));
5730 break;
5731 default:
5732 panic("%s: Unsupported pfsync_msg_version %d",
5733 __func__, msg_version);
5734 }
5735
5736 /*
5737 * XXX Why do we bother pfsyncing source node information if source
5738 * nodes are not synced? Showing users that there is source tracking
5739 * when there is none seems useless.
5740 */
5741 if (st->sns[PF_SN_LIMIT] != NULL)
5742 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5743 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
5744 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5745
5746 sp->pfs_1301.id = st->id;
5747 sp->pfs_1301.creatorid = st->creatorid;
5748 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5749 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5750
5751 if (st->rule == NULL)
5752 sp->pfs_1301.rule = htonl(-1);
5753 else
5754 sp->pfs_1301.rule = htonl(st->rule->nr);
5755 if (st->anchor == NULL)
5756 sp->pfs_1301.anchor = htonl(-1);
5757 else
5758 sp->pfs_1301.anchor = htonl(st->anchor->nr);
5759 if (st->nat_rule == NULL)
5760 sp->pfs_1301.nat_rule = htonl(-1);
5761 else
5762 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
5763
5764 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5765 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5766 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5767 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5768 }
5769
5770 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5771 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5772 {
5773 bzero(sp, sizeof(*sp));
5774
5775 sp->version = PF_STATE_VERSION;
5776
5777 /* copy from state key */
5778 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5779 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5780 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5781 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5782 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5783 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5784 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5785 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5786 sp->proto = st->key[PF_SK_WIRE]->proto;
5787 sp->af = st->key[PF_SK_WIRE]->af;
5788
5789 /* copy from state */
5790 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5791 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5792 sizeof(sp->orig_ifname));
5793 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr));
5794 sp->creation = htonl(time_uptime - (st->creation / 1000));
5795 sp->expire = pf_state_expires(st);
5796 if (sp->expire <= time_uptime)
5797 sp->expire = htonl(0);
5798 else
5799 sp->expire = htonl(sp->expire - time_uptime);
5800
5801 sp->direction = st->direction;
5802 sp->log = st->act.log;
5803 sp->timeout = st->timeout;
5804 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5805 sp->state_flags_compat = st->state_flags;
5806 sp->state_flags = htons(st->state_flags);
5807 if (st->sns[PF_SN_LIMIT] != NULL)
5808 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5809 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL)
5810 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5811 sp->id = st->id;
5812 sp->creatorid = st->creatorid;
5813 pf_state_peer_hton(&st->src, &sp->src);
5814 pf_state_peer_hton(&st->dst, &sp->dst);
5815
5816 if (st->rule == NULL)
5817 sp->rule = htonl(-1);
5818 else
5819 sp->rule = htonl(st->rule->nr);
5820 if (st->anchor == NULL)
5821 sp->anchor = htonl(-1);
5822 else
5823 sp->anchor = htonl(st->anchor->nr);
5824 if (st->nat_rule == NULL)
5825 sp->nat_rule = htonl(-1);
5826 else
5827 sp->nat_rule = htonl(st->nat_rule->nr);
5828
5829 sp->packets[0] = st->packets[0];
5830 sp->packets[1] = st->packets[1];
5831 sp->bytes[0] = st->bytes[0];
5832 sp->bytes[1] = st->bytes[1];
5833
5834 sp->qid = htons(st->act.qid);
5835 sp->pqid = htons(st->act.pqid);
5836 sp->dnpipe = htons(st->act.dnpipe);
5837 sp->dnrpipe = htons(st->act.dnrpipe);
5838 sp->rtableid = htonl(st->act.rtableid);
5839 sp->min_ttl = st->act.min_ttl;
5840 sp->set_tos = st->act.set_tos;
5841 sp->max_mss = htons(st->act.max_mss);
5842 sp->rt = st->act.rt;
5843 if (st->act.rt_kif)
5844 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
5845 sizeof(sp->rt_ifname));
5846 sp->set_prio[0] = st->act.set_prio[0];
5847 sp->set_prio[1] = st->act.set_prio[1];
5848
5849 }
5850
5851 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5852 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5853 {
5854 struct pfr_ktable *kt;
5855
5856 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5857
5858 kt = aw->p.tbl;
5859 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5860 kt = kt->pfrkt_root;
5861 aw->p.tbl = NULL;
5862 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5863 kt->pfrkt_cnt : -1;
5864 }
5865
5866 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5867 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5868 size_t number, char **names)
5869 {
5870 nvlist_t *nvc;
5871
5872 nvc = nvlist_create(0);
5873 if (nvc == NULL)
5874 return (ENOMEM);
5875
5876 for (int i = 0; i < number; i++) {
5877 nvlist_append_number_array(nvc, "counters",
5878 counter_u64_fetch(counters[i]));
5879 nvlist_append_string_array(nvc, "names",
5880 names[i]);
5881 nvlist_append_number_array(nvc, "ids",
5882 i);
5883 }
5884 nvlist_add_nvlist(nvl, name, nvc);
5885 nvlist_destroy(nvc);
5886
5887 return (0);
5888 }
5889
5890 static int
pf_getstatus(struct pfioc_nv * nv)5891 pf_getstatus(struct pfioc_nv *nv)
5892 {
5893 nvlist_t *nvl = NULL, *nvc = NULL;
5894 void *nvlpacked = NULL;
5895 int error;
5896 struct pf_status s;
5897 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5898 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5899 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5900 time_t since;
5901
5902 PF_RULES_RLOCK_TRACKER;
5903
5904 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
5905
5906 PF_RULES_RLOCK();
5907
5908 nvl = nvlist_create(0);
5909 if (nvl == NULL)
5910 ERROUT(ENOMEM);
5911
5912 since = time_second - (time_uptime - V_pf_status.since);
5913
5914 nvlist_add_bool(nvl, "running", V_pf_status.running);
5915 nvlist_add_number(nvl, "since", since);
5916 nvlist_add_number(nvl, "debug", V_pf_status.debug);
5917 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5918 nvlist_add_number(nvl, "states", V_pf_status.states);
5919 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5920 nvlist_add_number(nvl, "reass", V_pf_status.reass);
5921 nvlist_add_bool(nvl, "syncookies_active",
5922 V_pf_status.syncookies_active);
5923 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5924
5925 /* counters */
5926 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5927 PFRES_MAX, pf_reasons);
5928 if (error != 0)
5929 ERROUT(error);
5930
5931 /* lcounters */
5932 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5933 KLCNT_MAX, pf_lcounter);
5934 if (error != 0)
5935 ERROUT(error);
5936
5937 /* fcounters */
5938 nvc = nvlist_create(0);
5939 if (nvc == NULL)
5940 ERROUT(ENOMEM);
5941
5942 for (int i = 0; i < FCNT_MAX; i++) {
5943 nvlist_append_number_array(nvc, "counters",
5944 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5945 nvlist_append_string_array(nvc, "names",
5946 pf_fcounter[i]);
5947 nvlist_append_number_array(nvc, "ids",
5948 i);
5949 }
5950 nvlist_add_nvlist(nvl, "fcounters", nvc);
5951 nvlist_destroy(nvc);
5952 nvc = NULL;
5953
5954 /* scounters */
5955 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5956 SCNT_MAX, pf_fcounter);
5957 if (error != 0)
5958 ERROUT(error);
5959
5960 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5961 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5962 PF_MD5_DIGEST_LENGTH);
5963
5964 pfi_update_status(V_pf_status.ifname, &s);
5965
5966 /* pcounters / bcounters */
5967 for (int i = 0; i < 2; i++) {
5968 for (int j = 0; j < 2; j++) {
5969 for (int k = 0; k < 2; k++) {
5970 nvlist_append_number_array(nvl, "pcounters",
5971 s.pcounters[i][j][k]);
5972 }
5973 nvlist_append_number_array(nvl, "bcounters",
5974 s.bcounters[i][j]);
5975 }
5976 }
5977
5978 nvlpacked = nvlist_pack(nvl, &nv->len);
5979 if (nvlpacked == NULL)
5980 ERROUT(ENOMEM);
5981
5982 if (nv->size == 0)
5983 ERROUT(0);
5984 else if (nv->size < nv->len)
5985 ERROUT(ENOSPC);
5986
5987 PF_RULES_RUNLOCK();
5988 error = copyout(nvlpacked, nv->data, nv->len);
5989 goto done;
5990
5991 #undef ERROUT
5992 errout:
5993 PF_RULES_RUNLOCK();
5994 done:
5995 free(nvlpacked, M_NVLIST);
5996 nvlist_destroy(nvc);
5997 nvlist_destroy(nvl);
5998
5999 return (error);
6000 }
6001
6002 /*
6003 * XXX - Check for version mismatch!!!
6004 */
6005 static void
pf_clear_all_states(void)6006 pf_clear_all_states(void)
6007 {
6008 struct epoch_tracker et;
6009 struct pf_kstate *s;
6010 u_int i;
6011
6012 NET_EPOCH_ENTER(et);
6013 for (i = 0; i <= V_pf_hashmask; i++) {
6014 struct pf_idhash *ih = &V_pf_idhash[i];
6015 relock:
6016 PF_HASHROW_LOCK(ih);
6017 LIST_FOREACH(s, &ih->states, entry) {
6018 s->timeout = PFTM_PURGE;
6019 /* Don't send out individual delete messages. */
6020 s->state_flags |= PFSTATE_NOSYNC;
6021 pf_remove_state(s);
6022 goto relock;
6023 }
6024 PF_HASHROW_UNLOCK(ih);
6025 }
6026 NET_EPOCH_EXIT(et);
6027 }
6028
6029 static int
pf_clear_tables(void)6030 pf_clear_tables(void)
6031 {
6032 struct pfioc_table io;
6033 int error;
6034
6035 bzero(&io, sizeof(io));
6036 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
6037
6038 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
6039 io.pfrio_flags);
6040
6041 return (error);
6042 }
6043
6044 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)6045 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
6046 {
6047 struct pf_ksrc_node_list kill;
6048 u_int killed;
6049
6050 LIST_INIT(&kill);
6051 for (int i = 0; i <= V_pf_srchashmask; i++) {
6052 struct pf_srchash *sh = &V_pf_srchash[i];
6053 struct pf_ksrc_node *sn, *tmp;
6054
6055 PF_HASHROW_LOCK(sh);
6056 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
6057 if (psnk == NULL ||
6058 (pf_match_addr(psnk->psnk_src.neg,
6059 &psnk->psnk_src.addr.v.a.addr,
6060 &psnk->psnk_src.addr.v.a.mask,
6061 &sn->addr, sn->af) &&
6062 pf_match_addr(psnk->psnk_dst.neg,
6063 &psnk->psnk_dst.addr.v.a.addr,
6064 &psnk->psnk_dst.addr.v.a.mask,
6065 &sn->raddr, sn->af))) {
6066 pf_unlink_src_node(sn);
6067 LIST_INSERT_HEAD(&kill, sn, entry);
6068 sn->expire = 1;
6069 }
6070 PF_HASHROW_UNLOCK(sh);
6071 }
6072
6073 for (int i = 0; i <= V_pf_hashmask; i++) {
6074 struct pf_idhash *ih = &V_pf_idhash[i];
6075 struct pf_kstate *s;
6076
6077 PF_HASHROW_LOCK(ih);
6078 LIST_FOREACH(s, &ih->states, entry) {
6079 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX;
6080 sn_type++) {
6081 if (s->sns[sn_type] &&
6082 s->sns[sn_type]->expire == 1) {
6083 s->sns[sn_type] = NULL;
6084 }
6085 }
6086 }
6087 PF_HASHROW_UNLOCK(ih);
6088 }
6089
6090 killed = pf_free_src_nodes(&kill);
6091
6092 if (psnk != NULL)
6093 psnk->psnk_killed = killed;
6094 }
6095
6096 static int
pf_keepcounters(struct pfioc_nv * nv)6097 pf_keepcounters(struct pfioc_nv *nv)
6098 {
6099 nvlist_t *nvl = NULL;
6100 void *nvlpacked = NULL;
6101 int error = 0;
6102
6103 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6104
6105 if (nv->len > pf_ioctl_maxcount)
6106 ERROUT(ENOMEM);
6107
6108 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6109 error = copyin(nv->data, nvlpacked, nv->len);
6110 if (error)
6111 ERROUT(error);
6112
6113 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6114 if (nvl == NULL)
6115 ERROUT(EBADMSG);
6116
6117 if (! nvlist_exists_bool(nvl, "keep_counters"))
6118 ERROUT(EBADMSG);
6119
6120 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6121
6122 on_error:
6123 nvlist_destroy(nvl);
6124 free(nvlpacked, M_NVLIST);
6125 return (error);
6126 }
6127
6128 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6129 pf_clear_states(const struct pf_kstate_kill *kill)
6130 {
6131 struct pf_state_key_cmp match_key;
6132 struct pf_kstate *s;
6133 struct pfi_kkif *kif;
6134 int idx;
6135 unsigned int killed = 0, dir;
6136
6137 NET_EPOCH_ASSERT();
6138
6139 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6140 struct pf_idhash *ih = &V_pf_idhash[i];
6141
6142 relock_DIOCCLRSTATES:
6143 PF_HASHROW_LOCK(ih);
6144 LIST_FOREACH(s, &ih->states, entry) {
6145 /* For floating states look at the original kif. */
6146 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6147
6148 if (kill->psk_ifname[0] &&
6149 strcmp(kill->psk_ifname,
6150 kif->pfik_name))
6151 continue;
6152
6153 if (kill->psk_kill_match) {
6154 bzero(&match_key, sizeof(match_key));
6155
6156 if (s->direction == PF_OUT) {
6157 dir = PF_IN;
6158 idx = PF_SK_STACK;
6159 } else {
6160 dir = PF_OUT;
6161 idx = PF_SK_WIRE;
6162 }
6163
6164 match_key.af = s->key[idx]->af;
6165 match_key.proto = s->key[idx]->proto;
6166 pf_addrcpy(&match_key.addr[0],
6167 &s->key[idx]->addr[1], match_key.af);
6168 match_key.port[0] = s->key[idx]->port[1];
6169 pf_addrcpy(&match_key.addr[1],
6170 &s->key[idx]->addr[0], match_key.af);
6171 match_key.port[1] = s->key[idx]->port[0];
6172 }
6173
6174 /*
6175 * Don't send out individual
6176 * delete messages.
6177 */
6178 s->state_flags |= PFSTATE_NOSYNC;
6179 pf_remove_state(s);
6180 killed++;
6181
6182 if (kill->psk_kill_match)
6183 killed += pf_kill_matching_state(&match_key,
6184 dir);
6185
6186 goto relock_DIOCCLRSTATES;
6187 }
6188 PF_HASHROW_UNLOCK(ih);
6189 }
6190
6191 if (V_pfsync_clear_states_ptr != NULL)
6192 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6193
6194 return (killed);
6195 }
6196
6197 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6198 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6199 {
6200 struct pf_kstate *s;
6201
6202 NET_EPOCH_ASSERT();
6203 if (kill->psk_pfcmp.id) {
6204 if (kill->psk_pfcmp.creatorid == 0)
6205 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6206 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6207 kill->psk_pfcmp.creatorid))) {
6208 pf_remove_state(s);
6209 *killed = 1;
6210 }
6211 return;
6212 }
6213
6214 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6215 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6216 }
6217
6218 static int
pf_killstates_nv(struct pfioc_nv * nv)6219 pf_killstates_nv(struct pfioc_nv *nv)
6220 {
6221 struct pf_kstate_kill kill;
6222 struct epoch_tracker et;
6223 nvlist_t *nvl = NULL;
6224 void *nvlpacked = NULL;
6225 int error = 0;
6226 unsigned int killed = 0;
6227
6228 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6229
6230 if (nv->len > pf_ioctl_maxcount)
6231 ERROUT(ENOMEM);
6232
6233 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6234 error = copyin(nv->data, nvlpacked, nv->len);
6235 if (error)
6236 ERROUT(error);
6237
6238 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6239 if (nvl == NULL)
6240 ERROUT(EBADMSG);
6241
6242 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6243 if (error)
6244 ERROUT(error);
6245
6246 NET_EPOCH_ENTER(et);
6247 pf_killstates(&kill, &killed);
6248 NET_EPOCH_EXIT(et);
6249
6250 free(nvlpacked, M_NVLIST);
6251 nvlpacked = NULL;
6252 nvlist_destroy(nvl);
6253 nvl = nvlist_create(0);
6254 if (nvl == NULL)
6255 ERROUT(ENOMEM);
6256
6257 nvlist_add_number(nvl, "killed", killed);
6258
6259 nvlpacked = nvlist_pack(nvl, &nv->len);
6260 if (nvlpacked == NULL)
6261 ERROUT(ENOMEM);
6262
6263 if (nv->size == 0)
6264 ERROUT(0);
6265 else if (nv->size < nv->len)
6266 ERROUT(ENOSPC);
6267
6268 error = copyout(nvlpacked, nv->data, nv->len);
6269
6270 on_error:
6271 nvlist_destroy(nvl);
6272 free(nvlpacked, M_NVLIST);
6273 return (error);
6274 }
6275
6276 static int
pf_clearstates_nv(struct pfioc_nv * nv)6277 pf_clearstates_nv(struct pfioc_nv *nv)
6278 {
6279 struct pf_kstate_kill kill;
6280 struct epoch_tracker et;
6281 nvlist_t *nvl = NULL;
6282 void *nvlpacked = NULL;
6283 int error = 0;
6284 unsigned int killed;
6285
6286 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6287
6288 if (nv->len > pf_ioctl_maxcount)
6289 ERROUT(ENOMEM);
6290
6291 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6292 error = copyin(nv->data, nvlpacked, nv->len);
6293 if (error)
6294 ERROUT(error);
6295
6296 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6297 if (nvl == NULL)
6298 ERROUT(EBADMSG);
6299
6300 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6301 if (error)
6302 ERROUT(error);
6303
6304 NET_EPOCH_ENTER(et);
6305 killed = pf_clear_states(&kill);
6306 NET_EPOCH_EXIT(et);
6307
6308 free(nvlpacked, M_NVLIST);
6309 nvlpacked = NULL;
6310 nvlist_destroy(nvl);
6311 nvl = nvlist_create(0);
6312 if (nvl == NULL)
6313 ERROUT(ENOMEM);
6314
6315 nvlist_add_number(nvl, "killed", killed);
6316
6317 nvlpacked = nvlist_pack(nvl, &nv->len);
6318 if (nvlpacked == NULL)
6319 ERROUT(ENOMEM);
6320
6321 if (nv->size == 0)
6322 ERROUT(0);
6323 else if (nv->size < nv->len)
6324 ERROUT(ENOSPC);
6325
6326 error = copyout(nvlpacked, nv->data, nv->len);
6327
6328 #undef ERROUT
6329 on_error:
6330 nvlist_destroy(nvl);
6331 free(nvlpacked, M_NVLIST);
6332 return (error);
6333 }
6334
6335 static int
pf_getstate(struct pfioc_nv * nv)6336 pf_getstate(struct pfioc_nv *nv)
6337 {
6338 nvlist_t *nvl = NULL, *nvls;
6339 void *nvlpacked = NULL;
6340 struct pf_kstate *s = NULL;
6341 int error = 0;
6342 uint64_t id, creatorid;
6343
6344 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6345
6346 if (nv->len > pf_ioctl_maxcount)
6347 ERROUT(ENOMEM);
6348
6349 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6350 error = copyin(nv->data, nvlpacked, nv->len);
6351 if (error)
6352 ERROUT(error);
6353
6354 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6355 if (nvl == NULL)
6356 ERROUT(EBADMSG);
6357
6358 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6359 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6360
6361 s = pf_find_state_byid(id, creatorid);
6362 if (s == NULL)
6363 ERROUT(ENOENT);
6364
6365 free(nvlpacked, M_NVLIST);
6366 nvlpacked = NULL;
6367 nvlist_destroy(nvl);
6368 nvl = nvlist_create(0);
6369 if (nvl == NULL)
6370 ERROUT(ENOMEM);
6371
6372 nvls = pf_state_to_nvstate(s);
6373 if (nvls == NULL)
6374 ERROUT(ENOMEM);
6375
6376 nvlist_add_nvlist(nvl, "state", nvls);
6377 nvlist_destroy(nvls);
6378
6379 nvlpacked = nvlist_pack(nvl, &nv->len);
6380 if (nvlpacked == NULL)
6381 ERROUT(ENOMEM);
6382
6383 if (nv->size == 0)
6384 ERROUT(0);
6385 else if (nv->size < nv->len)
6386 ERROUT(ENOSPC);
6387
6388 error = copyout(nvlpacked, nv->data, nv->len);
6389
6390 #undef ERROUT
6391 errout:
6392 if (s != NULL)
6393 PF_STATE_UNLOCK(s);
6394 free(nvlpacked, M_NVLIST);
6395 nvlist_destroy(nvl);
6396 return (error);
6397 }
6398
6399 /*
6400 * XXX - Check for version mismatch!!!
6401 */
6402
6403 /*
6404 * Duplicate pfctl -Fa operation to get rid of as much as we can.
6405 */
6406 static int
shutdown_pf(void)6407 shutdown_pf(void)
6408 {
6409 int error = 0;
6410 u_int32_t t[5];
6411 char nn = '\0';
6412 struct pf_kanchor *anchor;
6413 struct pf_keth_anchor *eth_anchor;
6414 int rs_num;
6415
6416 do {
6417 /* Unlink rules of all user defined anchors */
6418 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6419 /* Wildcard based anchors may not have a respective
6420 * explicit anchor rule or they may be left empty
6421 * without rules. It leads to anchor.refcnt=0, and the
6422 * rest of the logic does not expect it. */
6423 if (anchor->refcnt == 0)
6424 anchor->refcnt = 1;
6425 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6426 if ((error = pf_begin_rules(&t[rs_num], rs_num,
6427 anchor->path)) != 0) {
6428 DPFPRINTF(PF_DEBUG_MISC, ("%s: "
6429 "anchor.path=%s rs_num=%d\n",
6430 __func__, anchor->path, rs_num));
6431 goto error; /* XXX: rollback? */
6432 }
6433 }
6434 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6435 error = pf_commit_rules(t[rs_num], rs_num,
6436 anchor->path);
6437 MPASS(error == 0);
6438 }
6439 }
6440
6441 /* Unlink rules of all user defined ether anchors */
6442 RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6443 &V_pf_keth_anchors) {
6444 /* Wildcard based anchors may not have a respective
6445 * explicit anchor rule or they may be left empty
6446 * without rules. It leads to anchor.refcnt=0, and the
6447 * rest of the logic does not expect it. */
6448 if (eth_anchor->refcnt == 0)
6449 eth_anchor->refcnt = 1;
6450 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6451 != 0) {
6452 DPFPRINTF(PF_DEBUG_MISC, ("%s: eth "
6453 "anchor.path=%s\n", __func__,
6454 eth_anchor->path));
6455 goto error;
6456 }
6457 error = pf_commit_eth(t[0], eth_anchor->path);
6458 MPASS(error == 0);
6459 }
6460
6461 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6462 != 0) {
6463 DPFPRINTF(PF_DEBUG_MISC, ("%s: SCRUB\n", __func__));
6464 break;
6465 }
6466 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6467 != 0) {
6468 DPFPRINTF(PF_DEBUG_MISC, ("%s: FILTER\n", __func__));
6469 break; /* XXX: rollback? */
6470 }
6471 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6472 != 0) {
6473 DPFPRINTF(PF_DEBUG_MISC, ("%s: NAT\n", __func__));
6474 break; /* XXX: rollback? */
6475 }
6476 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6477 != 0) {
6478 DPFPRINTF(PF_DEBUG_MISC, ("%s: BINAT\n", __func__));
6479 break; /* XXX: rollback? */
6480 }
6481 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6482 != 0) {
6483 DPFPRINTF(PF_DEBUG_MISC, ("%s: RDR\n", __func__));
6484 break; /* XXX: rollback? */
6485 }
6486
6487 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6488 MPASS(error == 0);
6489 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6490 MPASS(error == 0);
6491 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6492 MPASS(error == 0);
6493 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6494 MPASS(error == 0);
6495 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6496 MPASS(error == 0);
6497
6498 if ((error = pf_clear_tables()) != 0)
6499 break;
6500
6501 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6502 DPFPRINTF(PF_DEBUG_MISC, ("%s: eth\n", __func__));
6503 break;
6504 }
6505 error = pf_commit_eth(t[0], &nn);
6506 MPASS(error == 0);
6507
6508 #ifdef ALTQ
6509 if ((error = pf_begin_altq(&t[0])) != 0) {
6510 DPFPRINTF(PF_DEBUG_MISC, ("%s: ALTQ\n", __func__));
6511 break;
6512 }
6513 pf_commit_altq(t[0]);
6514 #endif
6515
6516 pf_clear_all_states();
6517
6518 pf_kill_srcnodes(NULL);
6519
6520 /* status does not use malloced mem so no need to cleanup */
6521 /* fingerprints and interfaces have their own cleanup code */
6522 } while(0);
6523
6524 error:
6525 return (error);
6526 }
6527
6528 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6529 pf_check_return(int chk, struct mbuf **m)
6530 {
6531
6532 switch (chk) {
6533 case PF_PASS:
6534 if (*m == NULL)
6535 return (PFIL_CONSUMED);
6536 else
6537 return (PFIL_PASS);
6538 break;
6539 default:
6540 if (*m != NULL) {
6541 m_freem(*m);
6542 *m = NULL;
6543 }
6544 return (PFIL_DROPPED);
6545 }
6546 }
6547
6548 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6549 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6550 void *ruleset __unused, struct inpcb *inp)
6551 {
6552 int chk;
6553
6554 CURVNET_ASSERT_SET();
6555
6556 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6557
6558 return (pf_check_return(chk, m));
6559 }
6560
6561 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6562 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6563 void *ruleset __unused, struct inpcb *inp)
6564 {
6565 int chk;
6566
6567 CURVNET_ASSERT_SET();
6568
6569 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6570
6571 return (pf_check_return(chk, m));
6572 }
6573
6574 #ifdef INET
6575 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6576 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6577 void *ruleset __unused, struct inpcb *inp)
6578 {
6579 int chk;
6580
6581 CURVNET_ASSERT_SET();
6582
6583 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
6584
6585 return (pf_check_return(chk, m));
6586 }
6587
6588 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6589 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6590 void *ruleset __unused, struct inpcb *inp)
6591 {
6592 int chk;
6593
6594 CURVNET_ASSERT_SET();
6595
6596 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
6597
6598 return (pf_check_return(chk, m));
6599 }
6600 #endif
6601
6602 #ifdef INET6
6603 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6604 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6605 void *ruleset __unused, struct inpcb *inp)
6606 {
6607 int chk;
6608
6609 CURVNET_ASSERT_SET();
6610
6611 /*
6612 * In case of loopback traffic IPv6 uses the real interface in
6613 * order to support scoped addresses. In order to support stateful
6614 * filtering we have change this to lo0 as it is the case in IPv4.
6615 */
6616 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6617 m, inp, NULL);
6618
6619 return (pf_check_return(chk, m));
6620 }
6621
6622 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6623 pf_check6_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_INET6, PF_OUT, flags, ifp, m, inp, NULL);
6631
6632 return (pf_check_return(chk, m));
6633 }
6634 #endif /* INET6 */
6635
6636 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6637 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6638 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
6639 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
6640
6641 #ifdef INET
6642 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6643 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6644 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
6645 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
6646 #endif
6647 #ifdef INET6
6648 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6649 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6650 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
6651 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
6652 #endif
6653
6654 static void
hook_pf_eth(void)6655 hook_pf_eth(void)
6656 {
6657 struct pfil_hook_args pha = {
6658 .pa_version = PFIL_VERSION,
6659 .pa_modname = "pf",
6660 .pa_type = PFIL_TYPE_ETHERNET,
6661 };
6662 struct pfil_link_args pla = {
6663 .pa_version = PFIL_VERSION,
6664 };
6665 int ret __diagused;
6666
6667 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6668 return;
6669
6670 pha.pa_mbuf_chk = pf_eth_check_in;
6671 pha.pa_flags = PFIL_IN;
6672 pha.pa_rulname = "eth-in";
6673 V_pf_eth_in_hook = pfil_add_hook(&pha);
6674 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6675 pla.pa_head = V_link_pfil_head;
6676 pla.pa_hook = V_pf_eth_in_hook;
6677 ret = pfil_link(&pla);
6678 MPASS(ret == 0);
6679 pha.pa_mbuf_chk = pf_eth_check_out;
6680 pha.pa_flags = PFIL_OUT;
6681 pha.pa_rulname = "eth-out";
6682 V_pf_eth_out_hook = pfil_add_hook(&pha);
6683 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6684 pla.pa_head = V_link_pfil_head;
6685 pla.pa_hook = V_pf_eth_out_hook;
6686 ret = pfil_link(&pla);
6687 MPASS(ret == 0);
6688
6689 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6690 }
6691
6692 static void
hook_pf(void)6693 hook_pf(void)
6694 {
6695 struct pfil_hook_args pha = {
6696 .pa_version = PFIL_VERSION,
6697 .pa_modname = "pf",
6698 };
6699 struct pfil_link_args pla = {
6700 .pa_version = PFIL_VERSION,
6701 };
6702 int ret __diagused;
6703
6704 if (atomic_load_bool(&V_pf_pfil_hooked))
6705 return;
6706
6707 #ifdef INET
6708 pha.pa_type = PFIL_TYPE_IP4;
6709 pha.pa_mbuf_chk = pf_check_in;
6710 pha.pa_flags = PFIL_IN;
6711 pha.pa_rulname = "default-in";
6712 V_pf_ip4_in_hook = pfil_add_hook(&pha);
6713 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6714 pla.pa_head = V_inet_pfil_head;
6715 pla.pa_hook = V_pf_ip4_in_hook;
6716 ret = pfil_link(&pla);
6717 MPASS(ret == 0);
6718 pha.pa_mbuf_chk = pf_check_out;
6719 pha.pa_flags = PFIL_OUT;
6720 pha.pa_rulname = "default-out";
6721 V_pf_ip4_out_hook = pfil_add_hook(&pha);
6722 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6723 pla.pa_head = V_inet_pfil_head;
6724 pla.pa_hook = V_pf_ip4_out_hook;
6725 ret = pfil_link(&pla);
6726 MPASS(ret == 0);
6727 if (V_pf_filter_local) {
6728 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6729 pla.pa_head = V_inet_local_pfil_head;
6730 pla.pa_hook = V_pf_ip4_out_hook;
6731 ret = pfil_link(&pla);
6732 MPASS(ret == 0);
6733 }
6734 #endif
6735 #ifdef INET6
6736 pha.pa_type = PFIL_TYPE_IP6;
6737 pha.pa_mbuf_chk = pf_check6_in;
6738 pha.pa_flags = PFIL_IN;
6739 pha.pa_rulname = "default-in6";
6740 V_pf_ip6_in_hook = pfil_add_hook(&pha);
6741 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6742 pla.pa_head = V_inet6_pfil_head;
6743 pla.pa_hook = V_pf_ip6_in_hook;
6744 ret = pfil_link(&pla);
6745 MPASS(ret == 0);
6746 pha.pa_mbuf_chk = pf_check6_out;
6747 pha.pa_rulname = "default-out6";
6748 pha.pa_flags = PFIL_OUT;
6749 V_pf_ip6_out_hook = pfil_add_hook(&pha);
6750 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6751 pla.pa_head = V_inet6_pfil_head;
6752 pla.pa_hook = V_pf_ip6_out_hook;
6753 ret = pfil_link(&pla);
6754 MPASS(ret == 0);
6755 if (V_pf_filter_local) {
6756 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6757 pla.pa_head = V_inet6_local_pfil_head;
6758 pla.pa_hook = V_pf_ip6_out_hook;
6759 ret = pfil_link(&pla);
6760 MPASS(ret == 0);
6761 }
6762 #endif
6763
6764 atomic_store_bool(&V_pf_pfil_hooked, true);
6765 }
6766
6767 static void
dehook_pf_eth(void)6768 dehook_pf_eth(void)
6769 {
6770
6771 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6772 return;
6773
6774 pfil_remove_hook(V_pf_eth_in_hook);
6775 pfil_remove_hook(V_pf_eth_out_hook);
6776
6777 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6778 }
6779
6780 static void
dehook_pf(void)6781 dehook_pf(void)
6782 {
6783
6784 if (!atomic_load_bool(&V_pf_pfil_hooked))
6785 return;
6786
6787 #ifdef INET
6788 pfil_remove_hook(V_pf_ip4_in_hook);
6789 pfil_remove_hook(V_pf_ip4_out_hook);
6790 #endif
6791 #ifdef INET6
6792 pfil_remove_hook(V_pf_ip6_in_hook);
6793 pfil_remove_hook(V_pf_ip6_out_hook);
6794 #endif
6795
6796 atomic_store_bool(&V_pf_pfil_hooked, false);
6797 }
6798
6799 static void
pf_load_vnet(void)6800 pf_load_vnet(void)
6801 {
6802 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6803 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6804
6805 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6806 sx_init(&V_pf_ioctl_lock, "pf ioctl");
6807
6808 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6809 PF_RULE_TAG_HASH_SIZE_DEFAULT);
6810 #ifdef ALTQ
6811 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6812 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6813 #endif
6814
6815 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6816
6817 pfattach_vnet();
6818 V_pf_vnet_active = 1;
6819 }
6820
6821 static int
pf_load(void)6822 pf_load(void)
6823 {
6824 int error;
6825
6826 sx_init(&pf_end_lock, "pf end thread");
6827
6828 pf_mtag_initialize();
6829
6830 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6831 if (pf_dev == NULL)
6832 return (ENOMEM);
6833
6834 pf_end_threads = 0;
6835 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6836 if (error != 0)
6837 return (error);
6838
6839 pfi_initialize();
6840
6841 return (0);
6842 }
6843
6844 static void
pf_unload_vnet(void)6845 pf_unload_vnet(void)
6846 {
6847 int ret __diagused;
6848
6849 V_pf_vnet_active = 0;
6850 V_pf_status.running = 0;
6851 dehook_pf();
6852 dehook_pf_eth();
6853
6854 PF_RULES_WLOCK();
6855 pf_syncookies_cleanup();
6856 shutdown_pf();
6857 PF_RULES_WUNLOCK();
6858
6859 ret = swi_remove(V_pf_swi_cookie);
6860 MPASS(ret == 0);
6861 ret = intr_event_destroy(V_pf_swi_ie);
6862 MPASS(ret == 0);
6863
6864 pf_unload_vnet_purge();
6865
6866 pf_normalize_cleanup();
6867 PF_RULES_WLOCK();
6868 pfi_cleanup_vnet();
6869 PF_RULES_WUNLOCK();
6870 pfr_cleanup();
6871 pf_osfp_flush();
6872 pf_cleanup();
6873 if (IS_DEFAULT_VNET(curvnet))
6874 pf_mtag_cleanup();
6875
6876 pf_cleanup_tagset(&V_pf_tags);
6877 #ifdef ALTQ
6878 pf_cleanup_tagset(&V_pf_qids);
6879 #endif
6880 uma_zdestroy(V_pf_tag_z);
6881
6882 #ifdef PF_WANT_32_TO_64_COUNTER
6883 PF_RULES_WLOCK();
6884 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6885
6886 MPASS(LIST_EMPTY(&V_pf_allkiflist));
6887 MPASS(V_pf_allkifcount == 0);
6888
6889 LIST_REMOVE(&V_pf_default_rule, allrulelist);
6890 V_pf_allrulecount--;
6891 LIST_REMOVE(V_pf_rulemarker, allrulelist);
6892
6893 MPASS(LIST_EMPTY(&V_pf_allrulelist));
6894 MPASS(V_pf_allrulecount == 0);
6895
6896 PF_RULES_WUNLOCK();
6897
6898 free(V_pf_kifmarker, PFI_MTYPE);
6899 free(V_pf_rulemarker, M_PFRULE);
6900 #endif
6901
6902 /* Free counters last as we updated them during shutdown. */
6903 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6904 for (int i = 0; i < 2; i++) {
6905 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6906 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6907 }
6908 counter_u64_free(V_pf_default_rule.states_cur);
6909 counter_u64_free(V_pf_default_rule.states_tot);
6910 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
6911 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]);
6912 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6913
6914 for (int i = 0; i < PFRES_MAX; i++)
6915 counter_u64_free(V_pf_status.counters[i]);
6916 for (int i = 0; i < KLCNT_MAX; i++)
6917 counter_u64_free(V_pf_status.lcounters[i]);
6918 for (int i = 0; i < FCNT_MAX; i++)
6919 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6920 for (int i = 0; i < SCNT_MAX; i++)
6921 counter_u64_free(V_pf_status.scounters[i]);
6922
6923 rm_destroy(&V_pf_rules_lock);
6924 sx_destroy(&V_pf_ioctl_lock);
6925 }
6926
6927 static void
pf_unload(void)6928 pf_unload(void)
6929 {
6930
6931 sx_xlock(&pf_end_lock);
6932 pf_end_threads = 1;
6933 while (pf_end_threads < 2) {
6934 wakeup_one(pf_purge_thread);
6935 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6936 }
6937 sx_xunlock(&pf_end_lock);
6938
6939 pf_nl_unregister();
6940
6941 if (pf_dev != NULL)
6942 destroy_dev(pf_dev);
6943
6944 pfi_cleanup();
6945
6946 sx_destroy(&pf_end_lock);
6947 }
6948
6949 static void
vnet_pf_init(void * unused __unused)6950 vnet_pf_init(void *unused __unused)
6951 {
6952
6953 pf_load_vnet();
6954 }
6955 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6956 vnet_pf_init, NULL);
6957
6958 static void
vnet_pf_uninit(const void * unused __unused)6959 vnet_pf_uninit(const void *unused __unused)
6960 {
6961
6962 pf_unload_vnet();
6963 }
6964 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6965 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6966 vnet_pf_uninit, NULL);
6967
6968 static int
pf_modevent(module_t mod,int type,void * data)6969 pf_modevent(module_t mod, int type, void *data)
6970 {
6971 int error = 0;
6972
6973 switch(type) {
6974 case MOD_LOAD:
6975 error = pf_load();
6976 pf_nl_register();
6977 break;
6978 case MOD_UNLOAD:
6979 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6980 * the vnet_pf_uninit()s */
6981 break;
6982 default:
6983 error = EINVAL;
6984 break;
6985 }
6986
6987 return (error);
6988 }
6989
6990 static moduledata_t pf_mod = {
6991 "pf",
6992 pf_modevent,
6993 0
6994 };
6995
6996 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6997 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6998 MODULE_DEPEND(pf, crypto, 1, 1, 1);
6999 MODULE_VERSION(pf, PF_MODVER);
7000