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);
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 void pf_rollback_eth_cb(struct epoch_context *);
111 static int pf_rollback_eth(uint32_t, const char *);
112 static int pf_commit_eth(uint32_t, const char *);
113 static void pf_free_eth_rule(struct pf_keth_rule *);
114 #ifdef ALTQ
115 static int pf_begin_altq(u_int32_t *);
116 static int pf_rollback_altq(u_int32_t);
117 static int pf_commit_altq(u_int32_t);
118 static int pf_enable_altq(struct pf_altq *);
119 static int pf_disable_altq(struct pf_altq *);
120 static uint16_t pf_qname2qid(const char *);
121 static void pf_qid_unref(uint16_t);
122 #endif /* ALTQ */
123 static int pf_begin_rules(u_int32_t *, int, const char *);
124 static int pf_rollback_rules(u_int32_t, int, char *);
125 static int pf_setup_pfsync_matching(struct pf_kruleset *);
126 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
127 static void pf_hash_rule(struct pf_krule *);
128 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
129 static int pf_commit_rules(u_int32_t, int, char *);
130 static int pf_addr_setup(struct pf_kruleset *,
131 struct pf_addr_wrap *, sa_family_t);
132 static void pf_src_node_copy(const struct pf_ksrc_node *,
133 struct pf_src_node *);
134 #ifdef ALTQ
135 static int pf_export_kaltq(struct pf_altq *,
136 struct pfioc_altq_v1 *, size_t);
137 static int pf_import_kaltq(struct pfioc_altq_v1 *,
138 struct pf_altq *, size_t);
139 #endif /* ALTQ */
140
141 VNET_DEFINE(struct pf_krule, pf_default_rule);
142
143 static __inline int pf_krule_compare(struct pf_krule *,
144 struct pf_krule *);
145
146 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
147
148 #ifdef ALTQ
149 VNET_DEFINE_STATIC(int, pf_altq_running);
150 #define V_pf_altq_running VNET(pf_altq_running)
151 #endif
152
153 #define TAGID_MAX 50000
154 struct pf_tagname {
155 TAILQ_ENTRY(pf_tagname) namehash_entries;
156 TAILQ_ENTRY(pf_tagname) taghash_entries;
157 char name[PF_TAG_NAME_SIZE];
158 uint16_t tag;
159 int ref;
160 };
161
162 struct pf_tagset {
163 TAILQ_HEAD(, pf_tagname) *namehash;
164 TAILQ_HEAD(, pf_tagname) *taghash;
165 unsigned int mask;
166 uint32_t seed;
167 BITSET_DEFINE(, TAGID_MAX) avail;
168 };
169
170 VNET_DEFINE(struct pf_tagset, pf_tags);
171 #define V_pf_tags VNET(pf_tags)
172 static unsigned int pf_rule_tag_hashsize;
173 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128
174 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
175 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
176 "Size of pf(4) rule tag hashtable");
177
178 #ifdef ALTQ
179 VNET_DEFINE(struct pf_tagset, pf_qids);
180 #define V_pf_qids VNET(pf_qids)
181 static unsigned int pf_queue_tag_hashsize;
182 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128
183 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
184 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
185 "Size of pf(4) queue tag hashtable");
186 #endif
187 VNET_DEFINE(uma_zone_t, pf_tag_z);
188 #define V_pf_tag_z VNET(pf_tag_z)
189 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
190 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
191
192 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
193 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
194 #endif
195
196 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
197 #define V_pf_filter_local VNET(pf_filter_local)
198 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
199 &VNET_NAME(pf_filter_local), false,
200 "Enable filtering for packets delivered to local network stack");
201
202 #ifdef PF_DEFAULT_TO_DROP
203 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
204 #else
205 VNET_DEFINE_STATIC(bool, default_to_drop);
206 #endif
207 #define V_default_to_drop VNET(default_to_drop)
208 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
209 &VNET_NAME(default_to_drop), false,
210 "Make the default rule drop all packets.");
211
212 static void pf_init_tagset(struct pf_tagset *, unsigned int *,
213 unsigned int);
214 static void pf_cleanup_tagset(struct pf_tagset *);
215 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *);
216 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t);
217 static u_int16_t tagname2tag(struct pf_tagset *, const char *);
218 static u_int16_t pf_tagname2tag(const char *);
219 static void tag_unref(struct pf_tagset *, u_int16_t);
220
221 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
222
223 struct cdev *pf_dev;
224
225 /*
226 * XXX - These are new and need to be checked when moveing to a new version
227 */
228 static void pf_clear_all_states(void);
229 static int pf_killstates_row(struct pf_kstate_kill *,
230 struct pf_idhash *);
231 static int pf_killstates_nv(struct pfioc_nv *);
232 static int pf_clearstates_nv(struct pfioc_nv *);
233 static int pf_getstate(struct pfioc_nv *);
234 static int pf_getstatus(struct pfioc_nv *);
235 static int pf_clear_tables(void);
236 static void pf_clear_srcnodes(struct pf_ksrc_node *);
237 static void pf_kill_srcnodes(struct pfioc_src_node_kill *);
238 static int pf_keepcounters(struct pfioc_nv *);
239 static void pf_tbladdr_copyout(struct pf_addr_wrap *);
240
241 /*
242 * Wrapper functions for pfil(9) hooks
243 */
244 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
245 int flags, void *ruleset __unused, struct inpcb *inp);
246 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
247 int flags, void *ruleset __unused, struct inpcb *inp);
248 #ifdef INET
249 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
250 int flags, void *ruleset __unused, struct inpcb *inp);
251 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
252 int flags, void *ruleset __unused, struct inpcb *inp);
253 #endif
254 #ifdef INET6
255 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
256 int flags, void *ruleset __unused, struct inpcb *inp);
257 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
258 int flags, void *ruleset __unused, struct inpcb *inp);
259 #endif
260
261 static void hook_pf_eth(void);
262 static void hook_pf(void);
263 static void dehook_pf_eth(void);
264 static void dehook_pf(void);
265 static int shutdown_pf(void);
266 static int pf_load(void);
267 static void pf_unload(void);
268
269 static struct cdevsw pf_cdevsw = {
270 .d_ioctl = pfioctl,
271 .d_name = PF_NAME,
272 .d_version = D_VERSION,
273 };
274
275 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
276 #define V_pf_pfil_hooked VNET(pf_pfil_hooked)
277 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
278 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked)
279
280 /*
281 * We need a flag that is neither hooked nor running to know when
282 * the VNET is "valid". We primarily need this to control (global)
283 * external event, e.g., eventhandlers.
284 */
285 VNET_DEFINE(int, pf_vnet_active);
286 #define V_pf_vnet_active VNET(pf_vnet_active)
287
288 int pf_end_threads;
289 struct proc *pf_purge_proc;
290
291 VNET_DEFINE(struct rmlock, pf_rules_lock);
292 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
293 #define V_pf_ioctl_lock VNET(pf_ioctl_lock)
294 struct sx pf_end_lock;
295
296 /* pfsync */
297 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
298 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
299 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
300 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
301 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
302 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
303 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
304 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
305
306 /* pflog */
307 pflog_packet_t *pflog_packet_ptr = NULL;
308
309 /*
310 * Copy a user-provided string, returning an error if truncation would occur.
311 * Avoid scanning past "sz" bytes in the source string since there's no
312 * guarantee that it's nul-terminated.
313 */
314 static int
pf_user_strcpy(char * dst,const char * src,size_t sz)315 pf_user_strcpy(char *dst, const char *src, size_t sz)
316 {
317 if (strnlen(src, sz) == sz)
318 return (EINVAL);
319 (void)strlcpy(dst, src, sz);
320 return (0);
321 }
322
323 static void
pfattach_vnet(void)324 pfattach_vnet(void)
325 {
326 u_int32_t *my_timeout = V_pf_default_rule.timeout;
327
328 bzero(&V_pf_status, sizeof(V_pf_status));
329
330 pf_initialize();
331 pfr_initialize();
332 pfi_initialize_vnet();
333 pf_normalize_init();
334 pf_syncookies_init();
335
336 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
337 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
338
339 RB_INIT(&V_pf_anchors);
340 pf_init_kruleset(&pf_main_ruleset);
341
342 pf_init_keth(V_pf_keth);
343
344 /* default rule should never be garbage collected */
345 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
346 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
347 V_pf_default_rule.nr = -1;
348 V_pf_default_rule.rtableid = -1;
349
350 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
351 for (int i = 0; i < 2; i++) {
352 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
353 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
354 }
355 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
356 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
357 V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
358
359 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
360 M_WAITOK | M_ZERO);
361
362 #ifdef PF_WANT_32_TO_64_COUNTER
363 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
364 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
365 PF_RULES_WLOCK();
366 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
367 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
368 V_pf_allrulecount++;
369 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
370 PF_RULES_WUNLOCK();
371 #endif
372
373 /* initialize default timeouts */
374 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
375 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
376 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
377 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
378 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
379 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
380 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
381 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
382 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
383 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
384 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
385 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
386 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
387 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
388 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
389 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
390 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
391 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
392 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
393 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
394 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
395 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
396 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
397 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
398 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
399
400 V_pf_status.debug = PF_DEBUG_URGENT;
401 /*
402 * XXX This is different than in OpenBSD where reassembly is enabled by
403 * defult. In FreeBSD we expect people to still use scrub rules and
404 * switch to the new syntax later. Only when they switch they must
405 * explicitly enable reassemle. We could change the default once the
406 * scrub rule functionality is hopefully removed some day in future.
407 */
408 V_pf_status.reass = 0;
409
410 V_pf_pfil_hooked = false;
411 V_pf_pfil_eth_hooked = false;
412
413 /* XXX do our best to avoid a conflict */
414 V_pf_status.hostid = arc4random();
415
416 for (int i = 0; i < PFRES_MAX; i++)
417 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
418 for (int i = 0; i < KLCNT_MAX; i++)
419 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
420 for (int i = 0; i < FCNT_MAX; i++)
421 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
422 for (int i = 0; i < SCNT_MAX; i++)
423 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
424
425 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
426 INTR_MPSAFE, &V_pf_swi_cookie) != 0)
427 /* XXXGL: leaked all above. */
428 return;
429 }
430
431 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)432 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
433 u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
434 u_int8_t check_ticket)
435 {
436 struct pf_kruleset *ruleset;
437 struct pf_krule *rule;
438 int rs_num;
439
440 ruleset = pf_find_kruleset(anchor);
441 if (ruleset == NULL)
442 return (NULL);
443 rs_num = pf_get_ruleset_number(rule_action);
444 if (rs_num >= PF_RULESET_MAX)
445 return (NULL);
446 if (active) {
447 if (check_ticket && ticket !=
448 ruleset->rules[rs_num].active.ticket)
449 return (NULL);
450 if (r_last)
451 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
452 pf_krulequeue);
453 else
454 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
455 } else {
456 if (check_ticket && ticket !=
457 ruleset->rules[rs_num].inactive.ticket)
458 return (NULL);
459 if (r_last)
460 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
461 pf_krulequeue);
462 else
463 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
464 }
465 if (!r_last) {
466 while ((rule != NULL) && (rule->nr != rule_number))
467 rule = TAILQ_NEXT(rule, entries);
468 }
469 if (rule == NULL)
470 return (NULL);
471
472 return (&rule->rpool);
473 }
474
475 static void
pf_mv_kpool(struct pf_kpalist * poola,struct pf_kpalist * poolb)476 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
477 {
478 struct pf_kpooladdr *mv_pool_pa;
479
480 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
481 TAILQ_REMOVE(poola, mv_pool_pa, entries);
482 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
483 }
484 }
485
486 static void
pf_empty_kpool(struct pf_kpalist * poola)487 pf_empty_kpool(struct pf_kpalist *poola)
488 {
489 struct pf_kpooladdr *pa;
490
491 while ((pa = TAILQ_FIRST(poola)) != NULL) {
492 switch (pa->addr.type) {
493 case PF_ADDR_DYNIFTL:
494 pfi_dynaddr_remove(pa->addr.p.dyn);
495 break;
496 case PF_ADDR_TABLE:
497 /* XXX: this could be unfinished pooladdr on pabuf */
498 if (pa->addr.p.tbl != NULL)
499 pfr_detach_table(pa->addr.p.tbl);
500 break;
501 }
502 if (pa->kif)
503 pfi_kkif_unref(pa->kif);
504 TAILQ_REMOVE(poola, pa, entries);
505 free(pa, M_PFRULE);
506 }
507 }
508
509 static void
pf_unlink_rule_locked(struct pf_krulequeue * rulequeue,struct pf_krule * rule)510 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
511 {
512
513 PF_RULES_WASSERT();
514 PF_UNLNKDRULES_ASSERT();
515
516 TAILQ_REMOVE(rulequeue, rule, entries);
517
518 rule->rule_ref |= PFRULE_REFS;
519 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
520 }
521
522 static void
pf_unlink_rule(struct pf_krulequeue * rulequeue,struct pf_krule * rule)523 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
524 {
525
526 PF_RULES_WASSERT();
527
528 PF_UNLNKDRULES_LOCK();
529 pf_unlink_rule_locked(rulequeue, rule);
530 PF_UNLNKDRULES_UNLOCK();
531 }
532
533 static void
pf_free_eth_rule(struct pf_keth_rule * rule)534 pf_free_eth_rule(struct pf_keth_rule *rule)
535 {
536 PF_RULES_WASSERT();
537
538 if (rule == NULL)
539 return;
540
541 if (rule->tag)
542 tag_unref(&V_pf_tags, rule->tag);
543 if (rule->match_tag)
544 tag_unref(&V_pf_tags, rule->match_tag);
545 #ifdef ALTQ
546 pf_qid_unref(rule->qid);
547 #endif
548
549 if (rule->bridge_to)
550 pfi_kkif_unref(rule->bridge_to);
551 if (rule->kif)
552 pfi_kkif_unref(rule->kif);
553
554 if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
555 pfr_detach_table(rule->ipsrc.addr.p.tbl);
556 if (rule->ipdst.addr.type == PF_ADDR_TABLE)
557 pfr_detach_table(rule->ipdst.addr.p.tbl);
558
559 counter_u64_free(rule->evaluations);
560 for (int i = 0; i < 2; i++) {
561 counter_u64_free(rule->packets[i]);
562 counter_u64_free(rule->bytes[i]);
563 }
564 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
565 pf_keth_anchor_remove(rule);
566
567 free(rule, M_PFRULE);
568 }
569
570 void
pf_free_rule(struct pf_krule * rule)571 pf_free_rule(struct pf_krule *rule)
572 {
573
574 PF_RULES_WASSERT();
575 PF_CONFIG_ASSERT();
576
577 if (rule->tag)
578 tag_unref(&V_pf_tags, rule->tag);
579 if (rule->match_tag)
580 tag_unref(&V_pf_tags, rule->match_tag);
581 #ifdef ALTQ
582 if (rule->pqid != rule->qid)
583 pf_qid_unref(rule->pqid);
584 pf_qid_unref(rule->qid);
585 #endif
586 switch (rule->src.addr.type) {
587 case PF_ADDR_DYNIFTL:
588 pfi_dynaddr_remove(rule->src.addr.p.dyn);
589 break;
590 case PF_ADDR_TABLE:
591 pfr_detach_table(rule->src.addr.p.tbl);
592 break;
593 }
594 switch (rule->dst.addr.type) {
595 case PF_ADDR_DYNIFTL:
596 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
597 break;
598 case PF_ADDR_TABLE:
599 pfr_detach_table(rule->dst.addr.p.tbl);
600 break;
601 }
602 if (rule->overload_tbl)
603 pfr_detach_table(rule->overload_tbl);
604 if (rule->kif)
605 pfi_kkif_unref(rule->kif);
606 if (rule->rcv_kif)
607 pfi_kkif_unref(rule->rcv_kif);
608 pf_kanchor_remove(rule);
609 pf_empty_kpool(&rule->rpool.list);
610
611 pf_krule_free(rule);
612 }
613
614 static void
pf_init_tagset(struct pf_tagset * ts,unsigned int * tunable_size,unsigned int default_size)615 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
616 unsigned int default_size)
617 {
618 unsigned int i;
619 unsigned int hashsize;
620
621 if (*tunable_size == 0 || !powerof2(*tunable_size))
622 *tunable_size = default_size;
623
624 hashsize = *tunable_size;
625 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
626 M_WAITOK);
627 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
628 M_WAITOK);
629 ts->mask = hashsize - 1;
630 ts->seed = arc4random();
631 for (i = 0; i < hashsize; i++) {
632 TAILQ_INIT(&ts->namehash[i]);
633 TAILQ_INIT(&ts->taghash[i]);
634 }
635 BIT_FILL(TAGID_MAX, &ts->avail);
636 }
637
638 static void
pf_cleanup_tagset(struct pf_tagset * ts)639 pf_cleanup_tagset(struct pf_tagset *ts)
640 {
641 unsigned int i;
642 unsigned int hashsize;
643 struct pf_tagname *t, *tmp;
644
645 /*
646 * Only need to clean up one of the hashes as each tag is hashed
647 * into each table.
648 */
649 hashsize = ts->mask + 1;
650 for (i = 0; i < hashsize; i++)
651 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
652 uma_zfree(V_pf_tag_z, t);
653
654 free(ts->namehash, M_PFHASH);
655 free(ts->taghash, M_PFHASH);
656 }
657
658 static uint16_t
tagname2hashindex(const struct pf_tagset * ts,const char * tagname)659 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
660 {
661 size_t len;
662
663 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
664 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
665 }
666
667 static uint16_t
tag2hashindex(const struct pf_tagset * ts,uint16_t tag)668 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
669 {
670
671 return (tag & ts->mask);
672 }
673
674 static u_int16_t
tagname2tag(struct pf_tagset * ts,const char * tagname)675 tagname2tag(struct pf_tagset *ts, const char *tagname)
676 {
677 struct pf_tagname *tag;
678 u_int32_t index;
679 u_int16_t new_tagid;
680
681 PF_RULES_WASSERT();
682
683 index = tagname2hashindex(ts, tagname);
684 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
685 if (strcmp(tagname, tag->name) == 0) {
686 tag->ref++;
687 return (tag->tag);
688 }
689
690 /*
691 * new entry
692 *
693 * to avoid fragmentation, we do a linear search from the beginning
694 * and take the first free slot we find.
695 */
696 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
697 /*
698 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
699 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
700 * set. It may also return a bit number greater than TAGID_MAX due
701 * to rounding of the number of bits in the vector up to a multiple
702 * of the vector word size at declaration/allocation time.
703 */
704 if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
705 return (0);
706
707 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */
708 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
709
710 /* allocate and fill new struct pf_tagname */
711 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
712 if (tag == NULL)
713 return (0);
714 strlcpy(tag->name, tagname, sizeof(tag->name));
715 tag->tag = new_tagid;
716 tag->ref = 1;
717
718 /* Insert into namehash */
719 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
720
721 /* Insert into taghash */
722 index = tag2hashindex(ts, new_tagid);
723 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
724
725 return (tag->tag);
726 }
727
728 static void
tag_unref(struct pf_tagset * ts,u_int16_t tag)729 tag_unref(struct pf_tagset *ts, u_int16_t tag)
730 {
731 struct pf_tagname *t;
732 uint16_t index;
733
734 PF_RULES_WASSERT();
735
736 index = tag2hashindex(ts, tag);
737 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
738 if (tag == t->tag) {
739 if (--t->ref == 0) {
740 TAILQ_REMOVE(&ts->taghash[index], t,
741 taghash_entries);
742 index = tagname2hashindex(ts, t->name);
743 TAILQ_REMOVE(&ts->namehash[index], t,
744 namehash_entries);
745 /* Bits are 0-based for BIT_SET() */
746 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
747 uma_zfree(V_pf_tag_z, t);
748 }
749 break;
750 }
751 }
752
753 static uint16_t
pf_tagname2tag(const char * tagname)754 pf_tagname2tag(const char *tagname)
755 {
756 return (tagname2tag(&V_pf_tags, tagname));
757 }
758
759 static int
pf_begin_eth(uint32_t * ticket,const char * anchor)760 pf_begin_eth(uint32_t *ticket, const char *anchor)
761 {
762 struct pf_keth_rule *rule, *tmp;
763 struct pf_keth_ruleset *rs;
764
765 PF_RULES_WASSERT();
766
767 rs = pf_find_or_create_keth_ruleset(anchor);
768 if (rs == NULL)
769 return (EINVAL);
770
771 /* Purge old inactive rules. */
772 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
773 tmp) {
774 TAILQ_REMOVE(rs->inactive.rules, rule,
775 entries);
776 pf_free_eth_rule(rule);
777 }
778
779 *ticket = ++rs->inactive.ticket;
780 rs->inactive.open = 1;
781
782 return (0);
783 }
784
785 static void
pf_rollback_eth_cb(struct epoch_context * ctx)786 pf_rollback_eth_cb(struct epoch_context *ctx)
787 {
788 struct pf_keth_ruleset *rs;
789
790 rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
791
792 CURVNET_SET(rs->vnet);
793
794 PF_RULES_WLOCK();
795 pf_rollback_eth(rs->inactive.ticket,
796 rs->anchor ? rs->anchor->path : "");
797 PF_RULES_WUNLOCK();
798
799 CURVNET_RESTORE();
800 }
801
802 static int
pf_rollback_eth(uint32_t ticket,const char * anchor)803 pf_rollback_eth(uint32_t ticket, const char *anchor)
804 {
805 struct pf_keth_rule *rule, *tmp;
806 struct pf_keth_ruleset *rs;
807
808 PF_RULES_WASSERT();
809
810 rs = pf_find_keth_ruleset(anchor);
811 if (rs == NULL)
812 return (EINVAL);
813
814 if (!rs->inactive.open ||
815 ticket != rs->inactive.ticket)
816 return (0);
817
818 /* Purge old inactive rules. */
819 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
820 tmp) {
821 TAILQ_REMOVE(rs->inactive.rules, rule, entries);
822 pf_free_eth_rule(rule);
823 }
824
825 rs->inactive.open = 0;
826
827 pf_remove_if_empty_keth_ruleset(rs);
828
829 return (0);
830 }
831
832 #define PF_SET_SKIP_STEPS(i) \
833 do { \
834 while (head[i] != cur) { \
835 head[i]->skip[i].ptr = cur; \
836 head[i] = TAILQ_NEXT(head[i], entries); \
837 } \
838 } while (0)
839
840 static void
pf_eth_calc_skip_steps(struct pf_keth_ruleq * rules)841 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
842 {
843 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
844 int i;
845
846 cur = TAILQ_FIRST(rules);
847 prev = cur;
848 for (i = 0; i < PFE_SKIP_COUNT; ++i)
849 head[i] = cur;
850 while (cur != NULL) {
851 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
852 PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
853 if (cur->direction != prev->direction)
854 PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
855 if (cur->proto != prev->proto)
856 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
857 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
858 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
859 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
860 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
861 if (cur->ipsrc.neg != prev->ipsrc.neg ||
862 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
863 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
864 if (cur->ipdst.neg != prev->ipdst.neg ||
865 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
866 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
867
868 prev = cur;
869 cur = TAILQ_NEXT(cur, entries);
870 }
871 for (i = 0; i < PFE_SKIP_COUNT; ++i)
872 PF_SET_SKIP_STEPS(i);
873 }
874
875 static int
pf_commit_eth(uint32_t ticket,const char * anchor)876 pf_commit_eth(uint32_t ticket, const char *anchor)
877 {
878 struct pf_keth_ruleq *rules;
879 struct pf_keth_ruleset *rs;
880
881 rs = pf_find_keth_ruleset(anchor);
882 if (rs == NULL) {
883 return (EINVAL);
884 }
885
886 if (!rs->inactive.open ||
887 ticket != rs->inactive.ticket)
888 return (EBUSY);
889
890 PF_RULES_WASSERT();
891
892 pf_eth_calc_skip_steps(rs->inactive.rules);
893
894 rules = rs->active.rules;
895 ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
896 rs->inactive.rules = rules;
897 rs->inactive.ticket = rs->active.ticket;
898
899 /* Clean up inactive rules (i.e. previously active rules), only when
900 * we're sure they're no longer used. */
901 NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
902
903 return (0);
904 }
905
906 #ifdef ALTQ
907 static uint16_t
pf_qname2qid(const char * qname)908 pf_qname2qid(const char *qname)
909 {
910 return (tagname2tag(&V_pf_qids, qname));
911 }
912
913 static void
pf_qid_unref(uint16_t qid)914 pf_qid_unref(uint16_t qid)
915 {
916 tag_unref(&V_pf_qids, qid);
917 }
918
919 static int
pf_begin_altq(u_int32_t * ticket)920 pf_begin_altq(u_int32_t *ticket)
921 {
922 struct pf_altq *altq, *tmp;
923 int error = 0;
924
925 PF_RULES_WASSERT();
926
927 /* Purge the old altq lists */
928 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
929 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
930 /* detach and destroy the discipline */
931 error = altq_remove(altq);
932 }
933 free(altq, M_PFALTQ);
934 }
935 TAILQ_INIT(V_pf_altq_ifs_inactive);
936 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
937 pf_qid_unref(altq->qid);
938 free(altq, M_PFALTQ);
939 }
940 TAILQ_INIT(V_pf_altqs_inactive);
941 if (error)
942 return (error);
943 *ticket = ++V_ticket_altqs_inactive;
944 V_altqs_inactive_open = 1;
945 return (0);
946 }
947
948 static int
pf_rollback_altq(u_int32_t ticket)949 pf_rollback_altq(u_int32_t ticket)
950 {
951 struct pf_altq *altq, *tmp;
952 int error = 0;
953
954 PF_RULES_WASSERT();
955
956 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
957 return (0);
958 /* Purge the old altq lists */
959 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
960 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
961 /* detach and destroy the discipline */
962 error = altq_remove(altq);
963 }
964 free(altq, M_PFALTQ);
965 }
966 TAILQ_INIT(V_pf_altq_ifs_inactive);
967 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
968 pf_qid_unref(altq->qid);
969 free(altq, M_PFALTQ);
970 }
971 TAILQ_INIT(V_pf_altqs_inactive);
972 V_altqs_inactive_open = 0;
973 return (error);
974 }
975
976 static int
pf_commit_altq(u_int32_t ticket)977 pf_commit_altq(u_int32_t ticket)
978 {
979 struct pf_altqqueue *old_altqs, *old_altq_ifs;
980 struct pf_altq *altq, *tmp;
981 int err, error = 0;
982
983 PF_RULES_WASSERT();
984
985 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
986 return (EBUSY);
987
988 /* swap altqs, keep the old. */
989 old_altqs = V_pf_altqs_active;
990 old_altq_ifs = V_pf_altq_ifs_active;
991 V_pf_altqs_active = V_pf_altqs_inactive;
992 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
993 V_pf_altqs_inactive = old_altqs;
994 V_pf_altq_ifs_inactive = old_altq_ifs;
995 V_ticket_altqs_active = V_ticket_altqs_inactive;
996
997 /* Attach new disciplines */
998 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
999 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1000 /* attach the discipline */
1001 error = altq_pfattach(altq);
1002 if (error == 0 && V_pf_altq_running)
1003 error = pf_enable_altq(altq);
1004 if (error != 0)
1005 return (error);
1006 }
1007 }
1008
1009 /* Purge the old altq lists */
1010 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1011 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1012 /* detach and destroy the discipline */
1013 if (V_pf_altq_running)
1014 error = pf_disable_altq(altq);
1015 err = altq_pfdetach(altq);
1016 if (err != 0 && error == 0)
1017 error = err;
1018 err = altq_remove(altq);
1019 if (err != 0 && error == 0)
1020 error = err;
1021 }
1022 free(altq, M_PFALTQ);
1023 }
1024 TAILQ_INIT(V_pf_altq_ifs_inactive);
1025 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1026 pf_qid_unref(altq->qid);
1027 free(altq, M_PFALTQ);
1028 }
1029 TAILQ_INIT(V_pf_altqs_inactive);
1030
1031 V_altqs_inactive_open = 0;
1032 return (error);
1033 }
1034
1035 static int
pf_enable_altq(struct pf_altq * altq)1036 pf_enable_altq(struct pf_altq *altq)
1037 {
1038 struct ifnet *ifp;
1039 struct tb_profile tb;
1040 int error = 0;
1041
1042 if ((ifp = ifunit(altq->ifname)) == NULL)
1043 return (EINVAL);
1044
1045 if (ifp->if_snd.altq_type != ALTQT_NONE)
1046 error = altq_enable(&ifp->if_snd);
1047
1048 /* set tokenbucket regulator */
1049 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1050 tb.rate = altq->ifbandwidth;
1051 tb.depth = altq->tbrsize;
1052 error = tbr_set(&ifp->if_snd, &tb);
1053 }
1054
1055 return (error);
1056 }
1057
1058 static int
pf_disable_altq(struct pf_altq * altq)1059 pf_disable_altq(struct pf_altq *altq)
1060 {
1061 struct ifnet *ifp;
1062 struct tb_profile tb;
1063 int error;
1064
1065 if ((ifp = ifunit(altq->ifname)) == NULL)
1066 return (EINVAL);
1067
1068 /*
1069 * when the discipline is no longer referenced, it was overridden
1070 * by a new one. if so, just return.
1071 */
1072 if (altq->altq_disc != ifp->if_snd.altq_disc)
1073 return (0);
1074
1075 error = altq_disable(&ifp->if_snd);
1076
1077 if (error == 0) {
1078 /* clear tokenbucket regulator */
1079 tb.rate = 0;
1080 error = tbr_set(&ifp->if_snd, &tb);
1081 }
1082
1083 return (error);
1084 }
1085
1086 static int
pf_altq_ifnet_event_add(struct ifnet * ifp,int remove,u_int32_t ticket,struct pf_altq * altq)1087 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1088 struct pf_altq *altq)
1089 {
1090 struct ifnet *ifp1;
1091 int error = 0;
1092
1093 /* Deactivate the interface in question */
1094 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1095 if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1096 (remove && ifp1 == ifp)) {
1097 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1098 } else {
1099 error = altq_add(ifp1, altq);
1100
1101 if (ticket != V_ticket_altqs_inactive)
1102 error = EBUSY;
1103
1104 if (error)
1105 free(altq, M_PFALTQ);
1106 }
1107
1108 return (error);
1109 }
1110
1111 void
pf_altq_ifnet_event(struct ifnet * ifp,int remove)1112 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1113 {
1114 struct pf_altq *a1, *a2, *a3;
1115 u_int32_t ticket;
1116 int error = 0;
1117
1118 /*
1119 * No need to re-evaluate the configuration for events on interfaces
1120 * that do not support ALTQ, as it's not possible for such
1121 * interfaces to be part of the configuration.
1122 */
1123 if (!ALTQ_IS_READY(&ifp->if_snd))
1124 return;
1125
1126 /* Interrupt userland queue modifications */
1127 if (V_altqs_inactive_open)
1128 pf_rollback_altq(V_ticket_altqs_inactive);
1129
1130 /* Start new altq ruleset */
1131 if (pf_begin_altq(&ticket))
1132 return;
1133
1134 /* Copy the current active set */
1135 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1136 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1137 if (a2 == NULL) {
1138 error = ENOMEM;
1139 break;
1140 }
1141 bcopy(a1, a2, sizeof(struct pf_altq));
1142
1143 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1144 if (error)
1145 break;
1146
1147 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1148 }
1149 if (error)
1150 goto out;
1151 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1152 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1153 if (a2 == NULL) {
1154 error = ENOMEM;
1155 break;
1156 }
1157 bcopy(a1, a2, sizeof(struct pf_altq));
1158
1159 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1160 error = EBUSY;
1161 free(a2, M_PFALTQ);
1162 break;
1163 }
1164 a2->altq_disc = NULL;
1165 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1166 if (strncmp(a3->ifname, a2->ifname,
1167 IFNAMSIZ) == 0) {
1168 a2->altq_disc = a3->altq_disc;
1169 break;
1170 }
1171 }
1172 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1173 if (error)
1174 break;
1175
1176 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1177 }
1178
1179 out:
1180 if (error != 0)
1181 pf_rollback_altq(ticket);
1182 else
1183 pf_commit_altq(ticket);
1184 }
1185 #endif /* ALTQ */
1186
1187 static struct pf_krule_global *
pf_rule_tree_alloc(int flags)1188 pf_rule_tree_alloc(int flags)
1189 {
1190 struct pf_krule_global *tree;
1191
1192 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1193 if (tree == NULL)
1194 return (NULL);
1195 RB_INIT(tree);
1196 return (tree);
1197 }
1198
1199 static void
pf_rule_tree_free(struct pf_krule_global * tree)1200 pf_rule_tree_free(struct pf_krule_global *tree)
1201 {
1202
1203 free(tree, M_TEMP);
1204 }
1205
1206 static int
pf_begin_rules(u_int32_t * ticket,int rs_num,const char * anchor)1207 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1208 {
1209 struct pf_krule_global *tree;
1210 struct pf_kruleset *rs;
1211 struct pf_krule *rule;
1212
1213 PF_RULES_WASSERT();
1214
1215 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1216 return (EINVAL);
1217 tree = pf_rule_tree_alloc(M_NOWAIT);
1218 if (tree == NULL)
1219 return (ENOMEM);
1220 rs = pf_find_or_create_kruleset(anchor);
1221 if (rs == NULL) {
1222 free(tree, M_TEMP);
1223 return (EINVAL);
1224 }
1225 pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1226 rs->rules[rs_num].inactive.tree = tree;
1227
1228 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1229 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1230 rs->rules[rs_num].inactive.rcount--;
1231 }
1232 *ticket = ++rs->rules[rs_num].inactive.ticket;
1233 rs->rules[rs_num].inactive.open = 1;
1234 return (0);
1235 }
1236
1237 static int
pf_rollback_rules(u_int32_t ticket,int rs_num,char * anchor)1238 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1239 {
1240 struct pf_kruleset *rs;
1241 struct pf_krule *rule;
1242
1243 PF_RULES_WASSERT();
1244
1245 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1246 return (EINVAL);
1247 rs = pf_find_kruleset(anchor);
1248 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1249 rs->rules[rs_num].inactive.ticket != ticket)
1250 return (0);
1251 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1252 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1253 rs->rules[rs_num].inactive.rcount--;
1254 }
1255 rs->rules[rs_num].inactive.open = 0;
1256 return (0);
1257 }
1258
1259 #define PF_MD5_UPD(st, elm) \
1260 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1261
1262 #define PF_MD5_UPD_STR(st, elm) \
1263 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1264
1265 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \
1266 (stor) = htonl((st)->elm); \
1267 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1268 } while (0)
1269
1270 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \
1271 (stor) = htons((st)->elm); \
1272 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1273 } while (0)
1274
1275 static void
pf_hash_rule_addr(MD5_CTX * ctx,struct pf_rule_addr * pfr)1276 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1277 {
1278 PF_MD5_UPD(pfr, addr.type);
1279 switch (pfr->addr.type) {
1280 case PF_ADDR_DYNIFTL:
1281 PF_MD5_UPD(pfr, addr.v.ifname);
1282 PF_MD5_UPD(pfr, addr.iflags);
1283 break;
1284 case PF_ADDR_TABLE:
1285 PF_MD5_UPD(pfr, addr.v.tblname);
1286 break;
1287 case PF_ADDR_ADDRMASK:
1288 /* XXX ignore af? */
1289 PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1290 PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1291 break;
1292 }
1293
1294 PF_MD5_UPD(pfr, port[0]);
1295 PF_MD5_UPD(pfr, port[1]);
1296 PF_MD5_UPD(pfr, neg);
1297 PF_MD5_UPD(pfr, port_op);
1298 }
1299
1300 static void
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)1301 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1302 {
1303 u_int16_t x;
1304 u_int32_t y;
1305
1306 pf_hash_rule_addr(ctx, &rule->src);
1307 pf_hash_rule_addr(ctx, &rule->dst);
1308 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1309 PF_MD5_UPD_STR(rule, label[i]);
1310 PF_MD5_UPD_STR(rule, ifname);
1311 PF_MD5_UPD_STR(rule, rcv_ifname);
1312 PF_MD5_UPD_STR(rule, match_tagname);
1313 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1314 PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1315 PF_MD5_UPD_HTONL(rule, prob, y);
1316 PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1317 PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1318 PF_MD5_UPD(rule, uid.op);
1319 PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1320 PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1321 PF_MD5_UPD(rule, gid.op);
1322 PF_MD5_UPD_HTONL(rule, rule_flag, y);
1323 PF_MD5_UPD(rule, action);
1324 PF_MD5_UPD(rule, direction);
1325 PF_MD5_UPD(rule, af);
1326 PF_MD5_UPD(rule, quick);
1327 PF_MD5_UPD(rule, ifnot);
1328 PF_MD5_UPD(rule, match_tag_not);
1329 PF_MD5_UPD(rule, natpass);
1330 PF_MD5_UPD(rule, keep_state);
1331 PF_MD5_UPD(rule, proto);
1332 PF_MD5_UPD(rule, type);
1333 PF_MD5_UPD(rule, code);
1334 PF_MD5_UPD(rule, flags);
1335 PF_MD5_UPD(rule, flagset);
1336 PF_MD5_UPD(rule, allow_opts);
1337 PF_MD5_UPD(rule, rt);
1338 PF_MD5_UPD(rule, tos);
1339 PF_MD5_UPD(rule, scrub_flags);
1340 PF_MD5_UPD(rule, min_ttl);
1341 PF_MD5_UPD(rule, set_tos);
1342 if (rule->anchor != NULL)
1343 PF_MD5_UPD_STR(rule, anchor->path);
1344 }
1345
1346 static void
pf_hash_rule(struct pf_krule * rule)1347 pf_hash_rule(struct pf_krule *rule)
1348 {
1349 MD5_CTX ctx;
1350
1351 MD5Init(&ctx);
1352 pf_hash_rule_rolling(&ctx, rule);
1353 MD5Final(rule->md5sum, &ctx);
1354 }
1355
1356 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1357 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1358 {
1359
1360 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1361 }
1362
1363 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1364 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1365 {
1366 struct pf_kruleset *rs;
1367 struct pf_krule *rule, **old_array, *old_rule;
1368 struct pf_krulequeue *old_rules;
1369 struct pf_krule_global *old_tree;
1370 int error;
1371 u_int32_t old_rcount;
1372
1373 PF_RULES_WASSERT();
1374
1375 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1376 return (EINVAL);
1377 rs = pf_find_kruleset(anchor);
1378 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1379 ticket != rs->rules[rs_num].inactive.ticket)
1380 return (EBUSY);
1381
1382 /* Calculate checksum for the main ruleset */
1383 if (rs == &pf_main_ruleset) {
1384 error = pf_setup_pfsync_matching(rs);
1385 if (error != 0)
1386 return (error);
1387 }
1388
1389 /* Swap rules, keep the old. */
1390 old_rules = rs->rules[rs_num].active.ptr;
1391 old_rcount = rs->rules[rs_num].active.rcount;
1392 old_array = rs->rules[rs_num].active.ptr_array;
1393 old_tree = rs->rules[rs_num].active.tree;
1394
1395 rs->rules[rs_num].active.ptr =
1396 rs->rules[rs_num].inactive.ptr;
1397 rs->rules[rs_num].active.ptr_array =
1398 rs->rules[rs_num].inactive.ptr_array;
1399 rs->rules[rs_num].active.tree =
1400 rs->rules[rs_num].inactive.tree;
1401 rs->rules[rs_num].active.rcount =
1402 rs->rules[rs_num].inactive.rcount;
1403
1404 /* Attempt to preserve counter information. */
1405 if (V_pf_status.keep_counters && old_tree != NULL) {
1406 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1407 entries) {
1408 old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1409 if (old_rule == NULL) {
1410 continue;
1411 }
1412 pf_counter_u64_critical_enter();
1413 pf_counter_u64_rollup_protected(&rule->evaluations,
1414 pf_counter_u64_fetch(&old_rule->evaluations));
1415 pf_counter_u64_rollup_protected(&rule->packets[0],
1416 pf_counter_u64_fetch(&old_rule->packets[0]));
1417 pf_counter_u64_rollup_protected(&rule->packets[1],
1418 pf_counter_u64_fetch(&old_rule->packets[1]));
1419 pf_counter_u64_rollup_protected(&rule->bytes[0],
1420 pf_counter_u64_fetch(&old_rule->bytes[0]));
1421 pf_counter_u64_rollup_protected(&rule->bytes[1],
1422 pf_counter_u64_fetch(&old_rule->bytes[1]));
1423 pf_counter_u64_critical_exit();
1424 }
1425 }
1426
1427 rs->rules[rs_num].inactive.ptr = old_rules;
1428 rs->rules[rs_num].inactive.ptr_array = old_array;
1429 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1430 rs->rules[rs_num].inactive.rcount = old_rcount;
1431
1432 rs->rules[rs_num].active.ticket =
1433 rs->rules[rs_num].inactive.ticket;
1434 pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1435
1436 /* Purge the old rule list. */
1437 PF_UNLNKDRULES_LOCK();
1438 while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1439 pf_unlink_rule_locked(old_rules, rule);
1440 PF_UNLNKDRULES_UNLOCK();
1441 if (rs->rules[rs_num].inactive.ptr_array)
1442 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1443 rs->rules[rs_num].inactive.ptr_array = NULL;
1444 rs->rules[rs_num].inactive.rcount = 0;
1445 rs->rules[rs_num].inactive.open = 0;
1446 pf_remove_if_empty_kruleset(rs);
1447 free(old_tree, M_TEMP);
1448
1449 return (0);
1450 }
1451
1452 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1453 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1454 {
1455 MD5_CTX ctx;
1456 struct pf_krule *rule;
1457 int rs_cnt;
1458 u_int8_t digest[PF_MD5_DIGEST_LENGTH];
1459
1460 MD5Init(&ctx);
1461 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1462 /* XXX PF_RULESET_SCRUB as well? */
1463 if (rs_cnt == PF_RULESET_SCRUB)
1464 continue;
1465
1466 if (rs->rules[rs_cnt].inactive.ptr_array)
1467 free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1468 rs->rules[rs_cnt].inactive.ptr_array = NULL;
1469
1470 if (rs->rules[rs_cnt].inactive.rcount) {
1471 rs->rules[rs_cnt].inactive.ptr_array =
1472 mallocarray(rs->rules[rs_cnt].inactive.rcount,
1473 sizeof(struct pf_rule **),
1474 M_TEMP, M_NOWAIT);
1475
1476 if (!rs->rules[rs_cnt].inactive.ptr_array)
1477 return (ENOMEM);
1478 }
1479
1480 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1481 entries) {
1482 pf_hash_rule_rolling(&ctx, rule);
1483 (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1484 }
1485 }
1486
1487 MD5Final(digest, &ctx);
1488 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1489 return (0);
1490 }
1491
1492 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1493 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1494 {
1495 int error = 0;
1496
1497 switch (addr->type) {
1498 case PF_ADDR_TABLE:
1499 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1500 if (addr->p.tbl == NULL)
1501 error = ENOMEM;
1502 break;
1503 default:
1504 error = EINVAL;
1505 }
1506
1507 return (error);
1508 }
1509
1510 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1511 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1512 sa_family_t af)
1513 {
1514 int error = 0;
1515
1516 switch (addr->type) {
1517 case PF_ADDR_TABLE:
1518 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1519 if (addr->p.tbl == NULL)
1520 error = ENOMEM;
1521 break;
1522 case PF_ADDR_DYNIFTL:
1523 error = pfi_dynaddr_setup(addr, af);
1524 break;
1525 }
1526
1527 return (error);
1528 }
1529
1530 void
pf_addr_copyout(struct pf_addr_wrap * addr)1531 pf_addr_copyout(struct pf_addr_wrap *addr)
1532 {
1533
1534 switch (addr->type) {
1535 case PF_ADDR_DYNIFTL:
1536 pfi_dynaddr_copyout(addr);
1537 break;
1538 case PF_ADDR_TABLE:
1539 pf_tbladdr_copyout(addr);
1540 break;
1541 }
1542 }
1543
1544 static void
pf_src_node_copy(const struct pf_ksrc_node * in,struct pf_src_node * out)1545 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1546 {
1547 int secs = time_uptime, diff;
1548
1549 bzero(out, sizeof(struct pf_src_node));
1550
1551 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1552 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1553
1554 if (in->rule.ptr != NULL)
1555 out->rule.nr = in->rule.ptr->nr;
1556
1557 for (int i = 0; i < 2; i++) {
1558 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1559 out->packets[i] = counter_u64_fetch(in->packets[i]);
1560 }
1561
1562 out->states = in->states;
1563 out->conn = in->conn;
1564 out->af = in->af;
1565 out->ruletype = in->ruletype;
1566
1567 out->creation = secs - in->creation;
1568 if (out->expire > secs)
1569 out->expire -= secs;
1570 else
1571 out->expire = 0;
1572
1573 /* Adjust the connection rate estimate. */
1574 diff = secs - in->conn_rate.last;
1575 if (diff >= in->conn_rate.seconds)
1576 out->conn_rate.count = 0;
1577 else
1578 out->conn_rate.count -=
1579 in->conn_rate.count * diff /
1580 in->conn_rate.seconds;
1581 }
1582
1583 #ifdef ALTQ
1584 /*
1585 * Handle export of struct pf_kaltq to user binaries that may be using any
1586 * version of struct pf_altq.
1587 */
1588 static int
pf_export_kaltq(struct pf_altq * q,struct pfioc_altq_v1 * pa,size_t ioc_size)1589 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1590 {
1591 u_int32_t version;
1592
1593 if (ioc_size == sizeof(struct pfioc_altq_v0))
1594 version = 0;
1595 else
1596 version = pa->version;
1597
1598 if (version > PFIOC_ALTQ_VERSION)
1599 return (EINVAL);
1600
1601 #define ASSIGN(x) exported_q->x = q->x
1602 #define COPY(x) \
1603 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1604 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1605 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1606
1607 switch (version) {
1608 case 0: {
1609 struct pf_altq_v0 *exported_q =
1610 &((struct pfioc_altq_v0 *)pa)->altq;
1611
1612 COPY(ifname);
1613
1614 ASSIGN(scheduler);
1615 ASSIGN(tbrsize);
1616 exported_q->tbrsize = SATU16(q->tbrsize);
1617 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1618
1619 COPY(qname);
1620 COPY(parent);
1621 ASSIGN(parent_qid);
1622 exported_q->bandwidth = SATU32(q->bandwidth);
1623 ASSIGN(priority);
1624 ASSIGN(local_flags);
1625
1626 ASSIGN(qlimit);
1627 ASSIGN(flags);
1628
1629 if (q->scheduler == ALTQT_HFSC) {
1630 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1631 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1632 SATU32(q->pq_u.hfsc_opts.x)
1633
1634 ASSIGN_OPT_SATU32(rtsc_m1);
1635 ASSIGN_OPT(rtsc_d);
1636 ASSIGN_OPT_SATU32(rtsc_m2);
1637
1638 ASSIGN_OPT_SATU32(lssc_m1);
1639 ASSIGN_OPT(lssc_d);
1640 ASSIGN_OPT_SATU32(lssc_m2);
1641
1642 ASSIGN_OPT_SATU32(ulsc_m1);
1643 ASSIGN_OPT(ulsc_d);
1644 ASSIGN_OPT_SATU32(ulsc_m2);
1645
1646 ASSIGN_OPT(flags);
1647
1648 #undef ASSIGN_OPT
1649 #undef ASSIGN_OPT_SATU32
1650 } else
1651 COPY(pq_u);
1652
1653 ASSIGN(qid);
1654 break;
1655 }
1656 case 1: {
1657 struct pf_altq_v1 *exported_q =
1658 &((struct pfioc_altq_v1 *)pa)->altq;
1659
1660 COPY(ifname);
1661
1662 ASSIGN(scheduler);
1663 ASSIGN(tbrsize);
1664 ASSIGN(ifbandwidth);
1665
1666 COPY(qname);
1667 COPY(parent);
1668 ASSIGN(parent_qid);
1669 ASSIGN(bandwidth);
1670 ASSIGN(priority);
1671 ASSIGN(local_flags);
1672
1673 ASSIGN(qlimit);
1674 ASSIGN(flags);
1675 COPY(pq_u);
1676
1677 ASSIGN(qid);
1678 break;
1679 }
1680 default:
1681 panic("%s: unhandled struct pfioc_altq version", __func__);
1682 break;
1683 }
1684
1685 #undef ASSIGN
1686 #undef COPY
1687 #undef SATU16
1688 #undef SATU32
1689
1690 return (0);
1691 }
1692
1693 /*
1694 * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1695 * that may be using any version of it.
1696 */
1697 static int
pf_import_kaltq(struct pfioc_altq_v1 * pa,struct pf_altq * q,size_t ioc_size)1698 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1699 {
1700 u_int32_t version;
1701
1702 if (ioc_size == sizeof(struct pfioc_altq_v0))
1703 version = 0;
1704 else
1705 version = pa->version;
1706
1707 if (version > PFIOC_ALTQ_VERSION)
1708 return (EINVAL);
1709
1710 #define ASSIGN(x) q->x = imported_q->x
1711 #define COPY(x) \
1712 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1713
1714 switch (version) {
1715 case 0: {
1716 struct pf_altq_v0 *imported_q =
1717 &((struct pfioc_altq_v0 *)pa)->altq;
1718
1719 COPY(ifname);
1720
1721 ASSIGN(scheduler);
1722 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1723 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1724
1725 COPY(qname);
1726 COPY(parent);
1727 ASSIGN(parent_qid);
1728 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1729 ASSIGN(priority);
1730 ASSIGN(local_flags);
1731
1732 ASSIGN(qlimit);
1733 ASSIGN(flags);
1734
1735 if (imported_q->scheduler == ALTQT_HFSC) {
1736 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1737
1738 /*
1739 * The m1 and m2 parameters are being copied from
1740 * 32-bit to 64-bit.
1741 */
1742 ASSIGN_OPT(rtsc_m1);
1743 ASSIGN_OPT(rtsc_d);
1744 ASSIGN_OPT(rtsc_m2);
1745
1746 ASSIGN_OPT(lssc_m1);
1747 ASSIGN_OPT(lssc_d);
1748 ASSIGN_OPT(lssc_m2);
1749
1750 ASSIGN_OPT(ulsc_m1);
1751 ASSIGN_OPT(ulsc_d);
1752 ASSIGN_OPT(ulsc_m2);
1753
1754 ASSIGN_OPT(flags);
1755
1756 #undef ASSIGN_OPT
1757 } else
1758 COPY(pq_u);
1759
1760 ASSIGN(qid);
1761 break;
1762 }
1763 case 1: {
1764 struct pf_altq_v1 *imported_q =
1765 &((struct pfioc_altq_v1 *)pa)->altq;
1766
1767 COPY(ifname);
1768
1769 ASSIGN(scheduler);
1770 ASSIGN(tbrsize);
1771 ASSIGN(ifbandwidth);
1772
1773 COPY(qname);
1774 COPY(parent);
1775 ASSIGN(parent_qid);
1776 ASSIGN(bandwidth);
1777 ASSIGN(priority);
1778 ASSIGN(local_flags);
1779
1780 ASSIGN(qlimit);
1781 ASSIGN(flags);
1782 COPY(pq_u);
1783
1784 ASSIGN(qid);
1785 break;
1786 }
1787 default:
1788 panic("%s: unhandled struct pfioc_altq version", __func__);
1789 break;
1790 }
1791
1792 #undef ASSIGN
1793 #undef COPY
1794
1795 return (0);
1796 }
1797
1798 static struct pf_altq *
pf_altq_get_nth_active(u_int32_t n)1799 pf_altq_get_nth_active(u_int32_t n)
1800 {
1801 struct pf_altq *altq;
1802 u_int32_t nr;
1803
1804 nr = 0;
1805 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1806 if (nr == n)
1807 return (altq);
1808 nr++;
1809 }
1810
1811 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1812 if (nr == n)
1813 return (altq);
1814 nr++;
1815 }
1816
1817 return (NULL);
1818 }
1819 #endif /* ALTQ */
1820
1821 struct pf_krule *
pf_krule_alloc(void)1822 pf_krule_alloc(void)
1823 {
1824 struct pf_krule *rule;
1825
1826 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1827 mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1828 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1829 M_WAITOK | M_ZERO);
1830 return (rule);
1831 }
1832
1833 void
pf_krule_free(struct pf_krule * rule)1834 pf_krule_free(struct pf_krule *rule)
1835 {
1836 #ifdef PF_WANT_32_TO_64_COUNTER
1837 bool wowned;
1838 #endif
1839
1840 if (rule == NULL)
1841 return;
1842
1843 #ifdef PF_WANT_32_TO_64_COUNTER
1844 if (rule->allrulelinked) {
1845 wowned = PF_RULES_WOWNED();
1846 if (!wowned)
1847 PF_RULES_WLOCK();
1848 LIST_REMOVE(rule, allrulelist);
1849 V_pf_allrulecount--;
1850 if (!wowned)
1851 PF_RULES_WUNLOCK();
1852 }
1853 #endif
1854
1855 pf_counter_u64_deinit(&rule->evaluations);
1856 for (int i = 0; i < 2; i++) {
1857 pf_counter_u64_deinit(&rule->packets[i]);
1858 pf_counter_u64_deinit(&rule->bytes[i]);
1859 }
1860 counter_u64_free(rule->states_cur);
1861 counter_u64_free(rule->states_tot);
1862 counter_u64_free(rule->src_nodes);
1863 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1864
1865 mtx_destroy(&rule->rpool.mtx);
1866 free(rule, M_PFRULE);
1867 }
1868
1869 void
pf_krule_clear_counters(struct pf_krule * rule)1870 pf_krule_clear_counters(struct pf_krule *rule)
1871 {
1872 pf_counter_u64_zero(&rule->evaluations);
1873 for (int i = 0; i < 2; i++) {
1874 pf_counter_u64_zero(&rule->packets[i]);
1875 pf_counter_u64_zero(&rule->bytes[i]);
1876 }
1877 counter_u64_zero(rule->states_tot);
1878 }
1879
1880 static void
pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr * kpool,struct pf_pooladdr * pool)1881 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1882 struct pf_pooladdr *pool)
1883 {
1884
1885 bzero(pool, sizeof(*pool));
1886 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1887 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1888 }
1889
1890 static int
pf_pooladdr_to_kpooladdr(const struct pf_pooladdr * pool,struct pf_kpooladdr * kpool)1891 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1892 struct pf_kpooladdr *kpool)
1893 {
1894 int ret;
1895
1896 bzero(kpool, sizeof(*kpool));
1897 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1898 ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1899 sizeof(kpool->ifname));
1900 return (ret);
1901 }
1902
1903 static void
pf_pool_to_kpool(const struct pf_pool * pool,struct pf_kpool * kpool)1904 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1905 {
1906 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1907 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1908
1909 bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1910 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1911
1912 kpool->tblidx = pool->tblidx;
1913 kpool->proxy_port[0] = pool->proxy_port[0];
1914 kpool->proxy_port[1] = pool->proxy_port[1];
1915 kpool->opts = pool->opts;
1916 }
1917
1918 static int
pf_rule_to_krule(const struct pf_rule * rule,struct pf_krule * krule)1919 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1920 {
1921 int ret;
1922
1923 #ifndef INET
1924 if (rule->af == AF_INET) {
1925 return (EAFNOSUPPORT);
1926 }
1927 #endif /* INET */
1928 #ifndef INET6
1929 if (rule->af == AF_INET6) {
1930 return (EAFNOSUPPORT);
1931 }
1932 #endif /* INET6 */
1933
1934 ret = pf_check_rule_addr(&rule->src);
1935 if (ret != 0)
1936 return (ret);
1937 ret = pf_check_rule_addr(&rule->dst);
1938 if (ret != 0)
1939 return (ret);
1940
1941 bcopy(&rule->src, &krule->src, sizeof(rule->src));
1942 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1943
1944 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1945 if (ret != 0)
1946 return (ret);
1947 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1948 if (ret != 0)
1949 return (ret);
1950 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1951 if (ret != 0)
1952 return (ret);
1953 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1954 if (ret != 0)
1955 return (ret);
1956 ret = pf_user_strcpy(krule->tagname, rule->tagname,
1957 sizeof(rule->tagname));
1958 if (ret != 0)
1959 return (ret);
1960 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1961 sizeof(rule->match_tagname));
1962 if (ret != 0)
1963 return (ret);
1964 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1965 sizeof(rule->overload_tblname));
1966 if (ret != 0)
1967 return (ret);
1968
1969 pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1970
1971 /* Don't allow userspace to set evaluations, packets or bytes. */
1972 /* kif, anchor, overload_tbl are not copied over. */
1973
1974 krule->os_fingerprint = rule->os_fingerprint;
1975
1976 krule->rtableid = rule->rtableid;
1977 /* pf_rule->timeout is smaller than pf_krule->timeout */
1978 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1979 krule->max_states = rule->max_states;
1980 krule->max_src_nodes = rule->max_src_nodes;
1981 krule->max_src_states = rule->max_src_states;
1982 krule->max_src_conn = rule->max_src_conn;
1983 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1984 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1985 krule->qid = rule->qid;
1986 krule->pqid = rule->pqid;
1987 krule->nr = rule->nr;
1988 krule->prob = rule->prob;
1989 krule->cuid = rule->cuid;
1990 krule->cpid = rule->cpid;
1991
1992 krule->return_icmp = rule->return_icmp;
1993 krule->return_icmp6 = rule->return_icmp6;
1994 krule->max_mss = rule->max_mss;
1995 krule->tag = rule->tag;
1996 krule->match_tag = rule->match_tag;
1997 krule->scrub_flags = rule->scrub_flags;
1998
1999 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2000 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2001
2002 krule->rule_flag = rule->rule_flag;
2003 krule->action = rule->action;
2004 krule->direction = rule->direction;
2005 krule->log = rule->log;
2006 krule->logif = rule->logif;
2007 krule->quick = rule->quick;
2008 krule->ifnot = rule->ifnot;
2009 krule->match_tag_not = rule->match_tag_not;
2010 krule->natpass = rule->natpass;
2011
2012 krule->keep_state = rule->keep_state;
2013 krule->af = rule->af;
2014 krule->proto = rule->proto;
2015 krule->type = rule->type;
2016 krule->code = rule->code;
2017 krule->flags = rule->flags;
2018 krule->flagset = rule->flagset;
2019 krule->min_ttl = rule->min_ttl;
2020 krule->allow_opts = rule->allow_opts;
2021 krule->rt = rule->rt;
2022 krule->return_ttl = rule->return_ttl;
2023 krule->tos = rule->tos;
2024 krule->set_tos = rule->set_tos;
2025
2026 krule->flush = rule->flush;
2027 krule->prio = rule->prio;
2028 krule->set_prio[0] = rule->set_prio[0];
2029 krule->set_prio[1] = rule->set_prio[1];
2030
2031 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2032
2033 return (0);
2034 }
2035
2036 int
pf_ioctl_getrules(struct pfioc_rule * pr)2037 pf_ioctl_getrules(struct pfioc_rule *pr)
2038 {
2039 struct pf_kruleset *ruleset;
2040 struct pf_krule *tail;
2041 int rs_num;
2042
2043 PF_RULES_WLOCK();
2044 ruleset = pf_find_kruleset(pr->anchor);
2045 if (ruleset == NULL) {
2046 PF_RULES_WUNLOCK();
2047 return (EINVAL);
2048 }
2049 rs_num = pf_get_ruleset_number(pr->rule.action);
2050 if (rs_num >= PF_RULESET_MAX) {
2051 PF_RULES_WUNLOCK();
2052 return (EINVAL);
2053 }
2054 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2055 pf_krulequeue);
2056 if (tail)
2057 pr->nr = tail->nr + 1;
2058 else
2059 pr->nr = 0;
2060 pr->ticket = ruleset->rules[rs_num].active.ticket;
2061 PF_RULES_WUNLOCK();
2062
2063 return (0);
2064 }
2065
2066 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)2067 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2068 uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2069 uid_t uid, pid_t pid)
2070 {
2071 struct pf_kruleset *ruleset;
2072 struct pf_krule *tail;
2073 struct pf_kpooladdr *pa;
2074 struct pfi_kkif *kif = NULL, *rcv_kif = NULL;
2075 int rs_num;
2076 int error = 0;
2077
2078 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2079 error = EINVAL;
2080 goto errout_unlocked;
2081 }
2082
2083 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
2084
2085 if (rule->ifname[0])
2086 kif = pf_kkif_create(M_WAITOK);
2087 if (rule->rcv_ifname[0])
2088 rcv_kif = pf_kkif_create(M_WAITOK);
2089 pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2090 for (int i = 0; i < 2; i++) {
2091 pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2092 pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2093 }
2094 rule->states_cur = counter_u64_alloc(M_WAITOK);
2095 rule->states_tot = counter_u64_alloc(M_WAITOK);
2096 rule->src_nodes = counter_u64_alloc(M_WAITOK);
2097 rule->cuid = uid;
2098 rule->cpid = pid;
2099 TAILQ_INIT(&rule->rpool.list);
2100
2101 PF_CONFIG_LOCK();
2102 PF_RULES_WLOCK();
2103 #ifdef PF_WANT_32_TO_64_COUNTER
2104 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2105 MPASS(!rule->allrulelinked);
2106 rule->allrulelinked = true;
2107 V_pf_allrulecount++;
2108 #endif
2109 ruleset = pf_find_kruleset(anchor);
2110 if (ruleset == NULL)
2111 ERROUT(EINVAL);
2112 rs_num = pf_get_ruleset_number(rule->action);
2113 if (rs_num >= PF_RULESET_MAX)
2114 ERROUT(EINVAL);
2115 if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2116 DPFPRINTF(PF_DEBUG_MISC,
2117 ("ticket: %d != [%d]%d\n", ticket, rs_num,
2118 ruleset->rules[rs_num].inactive.ticket));
2119 ERROUT(EBUSY);
2120 }
2121 if (pool_ticket != V_ticket_pabuf) {
2122 DPFPRINTF(PF_DEBUG_MISC,
2123 ("pool_ticket: %d != %d\n", pool_ticket,
2124 V_ticket_pabuf));
2125 ERROUT(EBUSY);
2126 }
2127 /*
2128 * XXXMJG hack: there is no mechanism to ensure they started the
2129 * transaction. Ticket checked above may happen to match by accident,
2130 * even if nobody called DIOCXBEGIN, let alone this process.
2131 * Partially work around it by checking if the RB tree got allocated,
2132 * see pf_begin_rules.
2133 */
2134 if (ruleset->rules[rs_num].inactive.tree == NULL) {
2135 ERROUT(EINVAL);
2136 }
2137
2138 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2139 pf_krulequeue);
2140 if (tail)
2141 rule->nr = tail->nr + 1;
2142 else
2143 rule->nr = 0;
2144 if (rule->ifname[0]) {
2145 rule->kif = pfi_kkif_attach(kif, rule->ifname);
2146 kif = NULL;
2147 pfi_kkif_ref(rule->kif);
2148 } else
2149 rule->kif = NULL;
2150
2151 if (rule->rcv_ifname[0]) {
2152 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname);
2153 rcv_kif = NULL;
2154 pfi_kkif_ref(rule->rcv_kif);
2155 } else
2156 rule->rcv_kif = NULL;
2157
2158 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2159 error = EBUSY;
2160
2161 #ifdef ALTQ
2162 /* set queue IDs */
2163 if (rule->qname[0] != 0) {
2164 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2165 error = EBUSY;
2166 else if (rule->pqname[0] != 0) {
2167 if ((rule->pqid =
2168 pf_qname2qid(rule->pqname)) == 0)
2169 error = 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 error = EBUSY;
2177 if (rule->match_tagname[0])
2178 if ((rule->match_tag =
2179 pf_tagname2tag(rule->match_tagname)) == 0)
2180 error = EBUSY;
2181 if (rule->rt && !rule->direction)
2182 error = EINVAL;
2183 if (!rule->log)
2184 rule->logif = 0;
2185 if (rule->logif >= PFLOGIFS_MAX)
2186 error = EINVAL;
2187 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2188 error = ENOMEM;
2189 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2190 error = ENOMEM;
2191 if (pf_kanchor_setup(rule, ruleset, anchor_call))
2192 error = EINVAL;
2193 if (rule->scrub_flags & PFSTATE_SETPRIO &&
2194 (rule->set_prio[0] > PF_PRIO_MAX ||
2195 rule->set_prio[1] > PF_PRIO_MAX))
2196 error = EINVAL;
2197 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2198 if (pa->addr.type == PF_ADDR_TABLE) {
2199 pa->addr.p.tbl = pfr_attach_table(ruleset,
2200 pa->addr.v.tblname);
2201 if (pa->addr.p.tbl == NULL)
2202 error = ENOMEM;
2203 }
2204
2205 rule->overload_tbl = NULL;
2206 if (rule->overload_tblname[0]) {
2207 if ((rule->overload_tbl = pfr_attach_table(ruleset,
2208 rule->overload_tblname)) == NULL)
2209 error = EINVAL;
2210 else
2211 rule->overload_tbl->pfrkt_flags |=
2212 PFR_TFLAG_ACTIVE;
2213 }
2214
2215 pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2216 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2217 (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2218 (rule->rt > PF_NOPFROUTE)) &&
2219 (TAILQ_FIRST(&rule->rpool.list) == NULL))
2220 error = EINVAL;
2221
2222 if (rule->action == PF_PASS && rule->rpool.opts & PF_POOL_STICKYADDR &&
2223 !rule->keep_state) {
2224 error = EINVAL;
2225 }
2226
2227 if (error) {
2228 pf_free_rule(rule);
2229 rule = NULL;
2230 ERROUT(error);
2231 }
2232
2233 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2234 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2235 rule, entries);
2236 ruleset->rules[rs_num].inactive.rcount++;
2237
2238 PF_RULES_WUNLOCK();
2239 pf_hash_rule(rule);
2240 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2241 PF_RULES_WLOCK();
2242 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2243 ruleset->rules[rs_num].inactive.rcount--;
2244 pf_free_rule(rule);
2245 rule = NULL;
2246 ERROUT(EEXIST);
2247 }
2248 PF_CONFIG_UNLOCK();
2249
2250 return (0);
2251
2252 #undef ERROUT
2253 errout:
2254 PF_RULES_WUNLOCK();
2255 PF_CONFIG_UNLOCK();
2256 errout_unlocked:
2257 pf_kkif_free(rcv_kif);
2258 pf_kkif_free(kif);
2259 pf_krule_free(rule);
2260 return (error);
2261 }
2262
2263 static bool
pf_label_match(const struct pf_krule * rule,const char * label)2264 pf_label_match(const struct pf_krule *rule, const char *label)
2265 {
2266 int i = 0;
2267
2268 while (*rule->label[i]) {
2269 if (strcmp(rule->label[i], label) == 0)
2270 return (true);
2271 i++;
2272 }
2273
2274 return (false);
2275 }
2276
2277 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)2278 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2279 {
2280 struct pf_kstate *s;
2281 int more = 0;
2282
2283 s = pf_find_state_all(key, dir, &more);
2284 if (s == NULL)
2285 return (0);
2286
2287 if (more) {
2288 PF_STATE_UNLOCK(s);
2289 return (0);
2290 }
2291
2292 pf_unlink_state(s);
2293 return (1);
2294 }
2295
2296 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)2297 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2298 {
2299 struct pf_kstate *s;
2300 struct pf_state_key *sk;
2301 struct pf_addr *srcaddr, *dstaddr;
2302 struct pf_state_key_cmp match_key;
2303 int idx, killed = 0;
2304 unsigned int dir;
2305 u_int16_t srcport, dstport;
2306 struct pfi_kkif *kif;
2307
2308 relock_DIOCKILLSTATES:
2309 PF_HASHROW_LOCK(ih);
2310 LIST_FOREACH(s, &ih->states, entry) {
2311 /* For floating states look at the original kif. */
2312 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2313
2314 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2315 if (s->direction == PF_OUT) {
2316 srcaddr = &sk->addr[1];
2317 dstaddr = &sk->addr[0];
2318 srcport = sk->port[1];
2319 dstport = sk->port[0];
2320 } else {
2321 srcaddr = &sk->addr[0];
2322 dstaddr = &sk->addr[1];
2323 srcport = sk->port[0];
2324 dstport = sk->port[1];
2325 }
2326
2327 if (psk->psk_af && sk->af != psk->psk_af)
2328 continue;
2329
2330 if (psk->psk_proto && psk->psk_proto != sk->proto)
2331 continue;
2332
2333 if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2334 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2335 continue;
2336
2337 if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2338 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2339 continue;
2340
2341 if (! PF_MATCHA(psk->psk_rt_addr.neg,
2342 &psk->psk_rt_addr.addr.v.a.addr,
2343 &psk->psk_rt_addr.addr.v.a.mask,
2344 &s->rt_addr, sk->af))
2345 continue;
2346
2347 if (psk->psk_src.port_op != 0 &&
2348 ! pf_match_port(psk->psk_src.port_op,
2349 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2350 continue;
2351
2352 if (psk->psk_dst.port_op != 0 &&
2353 ! pf_match_port(psk->psk_dst.port_op,
2354 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2355 continue;
2356
2357 if (psk->psk_label[0] &&
2358 ! pf_label_match(s->rule.ptr, psk->psk_label))
2359 continue;
2360
2361 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2362 kif->pfik_name))
2363 continue;
2364
2365 if (psk->psk_kill_match) {
2366 /* Create the key to find matching states, with lock
2367 * held. */
2368
2369 bzero(&match_key, sizeof(match_key));
2370
2371 if (s->direction == PF_OUT) {
2372 dir = PF_IN;
2373 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2374 } else {
2375 dir = PF_OUT;
2376 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2377 }
2378
2379 match_key.af = s->key[idx]->af;
2380 match_key.proto = s->key[idx]->proto;
2381 PF_ACPY(&match_key.addr[0],
2382 &s->key[idx]->addr[1], match_key.af);
2383 match_key.port[0] = s->key[idx]->port[1];
2384 PF_ACPY(&match_key.addr[1],
2385 &s->key[idx]->addr[0], match_key.af);
2386 match_key.port[1] = s->key[idx]->port[0];
2387 }
2388
2389 pf_unlink_state(s);
2390 killed++;
2391
2392 if (psk->psk_kill_match)
2393 killed += pf_kill_matching_state(&match_key, dir);
2394
2395 goto relock_DIOCKILLSTATES;
2396 }
2397 PF_HASHROW_UNLOCK(ih);
2398
2399 return (killed);
2400 }
2401
2402 int
pf_start(void)2403 pf_start(void)
2404 {
2405 int error = 0;
2406
2407 sx_xlock(&V_pf_ioctl_lock);
2408 if (V_pf_status.running)
2409 error = EEXIST;
2410 else {
2411 hook_pf();
2412 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2413 hook_pf_eth();
2414 V_pf_status.running = 1;
2415 V_pf_status.since = time_second;
2416 new_unrhdr64(&V_pf_stateid, time_second);
2417
2418 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2419 }
2420 sx_xunlock(&V_pf_ioctl_lock);
2421
2422 return (error);
2423 }
2424
2425 int
pf_stop(void)2426 pf_stop(void)
2427 {
2428 int error = 0;
2429
2430 sx_xlock(&V_pf_ioctl_lock);
2431 if (!V_pf_status.running)
2432 error = ENOENT;
2433 else {
2434 V_pf_status.running = 0;
2435 dehook_pf();
2436 dehook_pf_eth();
2437 V_pf_status.since = time_second;
2438 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2439 }
2440 sx_xunlock(&V_pf_ioctl_lock);
2441
2442 return (error);
2443 }
2444
2445 void
pf_ioctl_clear_status(void)2446 pf_ioctl_clear_status(void)
2447 {
2448 PF_RULES_WLOCK();
2449 for (int i = 0; i < PFRES_MAX; i++)
2450 counter_u64_zero(V_pf_status.counters[i]);
2451 for (int i = 0; i < FCNT_MAX; i++)
2452 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2453 for (int i = 0; i < SCNT_MAX; i++)
2454 counter_u64_zero(V_pf_status.scounters[i]);
2455 for (int i = 0; i < KLCNT_MAX; i++)
2456 counter_u64_zero(V_pf_status.lcounters[i]);
2457 V_pf_status.since = time_second;
2458 if (*V_pf_status.ifname)
2459 pfi_update_status(V_pf_status.ifname, NULL);
2460 PF_RULES_WUNLOCK();
2461 }
2462
2463 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2464 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2465 {
2466 uint32_t old;
2467
2468 if (timeout < 0 || timeout >= PFTM_MAX ||
2469 seconds < 0)
2470 return (EINVAL);
2471
2472 PF_RULES_WLOCK();
2473 old = V_pf_default_rule.timeout[timeout];
2474 if (timeout == PFTM_INTERVAL && seconds == 0)
2475 seconds = 1;
2476 V_pf_default_rule.timeout[timeout] = seconds;
2477 if (timeout == PFTM_INTERVAL && seconds < old)
2478 wakeup(pf_purge_thread);
2479
2480 if (prev_seconds != NULL)
2481 *prev_seconds = old;
2482
2483 PF_RULES_WUNLOCK();
2484
2485 return (0);
2486 }
2487
2488 int
pf_ioctl_get_timeout(int timeout,int * seconds)2489 pf_ioctl_get_timeout(int timeout, int *seconds)
2490 {
2491 PF_RULES_RLOCK_TRACKER;
2492
2493 if (timeout < 0 || timeout >= PFTM_MAX)
2494 return (EINVAL);
2495
2496 PF_RULES_RLOCK();
2497 *seconds = V_pf_default_rule.timeout[timeout];
2498 PF_RULES_RUNLOCK();
2499
2500 return (0);
2501 }
2502
2503 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2504 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2505 {
2506
2507 PF_RULES_WLOCK();
2508 if (index < 0 || index >= PF_LIMIT_MAX ||
2509 V_pf_limits[index].zone == NULL) {
2510 PF_RULES_WUNLOCK();
2511 return (EINVAL);
2512 }
2513 uma_zone_set_max(V_pf_limits[index].zone, limit);
2514 if (old_limit != NULL)
2515 *old_limit = V_pf_limits[index].limit;
2516 V_pf_limits[index].limit = limit;
2517 PF_RULES_WUNLOCK();
2518
2519 return (0);
2520 }
2521
2522 int
pf_ioctl_get_limit(int index,unsigned int * limit)2523 pf_ioctl_get_limit(int index, unsigned int *limit)
2524 {
2525 PF_RULES_RLOCK_TRACKER;
2526
2527 if (index < 0 || index >= PF_LIMIT_MAX)
2528 return (EINVAL);
2529
2530 PF_RULES_RLOCK();
2531 *limit = V_pf_limits[index].limit;
2532 PF_RULES_RUNLOCK();
2533
2534 return (0);
2535 }
2536
2537 int
pf_ioctl_begin_addrs(uint32_t * ticket)2538 pf_ioctl_begin_addrs(uint32_t *ticket)
2539 {
2540 PF_RULES_WLOCK();
2541 pf_empty_kpool(&V_pf_pabuf);
2542 *ticket = ++V_ticket_pabuf;
2543 PF_RULES_WUNLOCK();
2544
2545 return (0);
2546 }
2547
2548 int
pf_ioctl_add_addr(struct pfioc_pooladdr * pp)2549 pf_ioctl_add_addr(struct pfioc_pooladdr *pp)
2550 {
2551 struct pf_kpooladdr *pa = NULL;
2552 struct pfi_kkif *kif = NULL;
2553 int error;
2554
2555 #ifndef INET
2556 if (pp->af == AF_INET)
2557 return (EAFNOSUPPORT);
2558 #endif /* INET */
2559 #ifndef INET6
2560 if (pp->af == AF_INET6)
2561 return (EAFNOSUPPORT);
2562 #endif /* INET6 */
2563
2564 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2565 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2566 pp->addr.addr.type != PF_ADDR_TABLE)
2567 return (EINVAL);
2568
2569 if (pp->addr.addr.p.dyn != NULL)
2570 return (EINVAL);
2571
2572 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2573 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2574 if (error != 0)
2575 goto out;
2576 if (pa->ifname[0])
2577 kif = pf_kkif_create(M_WAITOK);
2578 PF_RULES_WLOCK();
2579 if (pp->ticket != V_ticket_pabuf) {
2580 PF_RULES_WUNLOCK();
2581 if (pa->ifname[0])
2582 pf_kkif_free(kif);
2583 error = EBUSY;
2584 goto out;
2585 }
2586 if (pa->ifname[0]) {
2587 pa->kif = pfi_kkif_attach(kif, pa->ifname);
2588 kif = NULL;
2589 pfi_kkif_ref(pa->kif);
2590 } else
2591 pa->kif = NULL;
2592 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2593 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2594 if (pa->ifname[0])
2595 pfi_kkif_unref(pa->kif);
2596 PF_RULES_WUNLOCK();
2597 goto out;
2598 }
2599 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2600 PF_RULES_WUNLOCK();
2601
2602 return (0);
2603
2604 out:
2605 free(pa, M_PFRULE);
2606 return (error);
2607 }
2608
2609 int
pf_ioctl_get_addrs(struct pfioc_pooladdr * pp)2610 pf_ioctl_get_addrs(struct pfioc_pooladdr *pp)
2611 {
2612 struct pf_kpool *pool;
2613 struct pf_kpooladdr *pa;
2614
2615 PF_RULES_RLOCK_TRACKER;
2616
2617 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2618 pp->nr = 0;
2619
2620 PF_RULES_RLOCK();
2621 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2622 pp->r_num, 0, 1, 0);
2623 if (pool == NULL) {
2624 PF_RULES_RUNLOCK();
2625 return (EBUSY);
2626 }
2627 TAILQ_FOREACH(pa, &pool->list, entries)
2628 pp->nr++;
2629 PF_RULES_RUNLOCK();
2630
2631 return (0);
2632 }
2633
2634 int
pf_ioctl_get_addr(struct pfioc_pooladdr * pp)2635 pf_ioctl_get_addr(struct pfioc_pooladdr *pp)
2636 {
2637 struct pf_kpool *pool;
2638 struct pf_kpooladdr *pa;
2639 u_int32_t nr = 0;
2640
2641 PF_RULES_RLOCK_TRACKER;
2642
2643 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2644
2645 PF_RULES_RLOCK();
2646 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2647 pp->r_num, 0, 1, 1);
2648 if (pool == NULL) {
2649 PF_RULES_RUNLOCK();
2650 return (EBUSY);
2651 }
2652 pa = TAILQ_FIRST(&pool->list);
2653 while ((pa != NULL) && (nr < pp->nr)) {
2654 pa = TAILQ_NEXT(pa, entries);
2655 nr++;
2656 }
2657 if (pa == NULL) {
2658 PF_RULES_RUNLOCK();
2659 return (EBUSY);
2660 }
2661 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2662 pf_addr_copyout(&pp->addr.addr);
2663 PF_RULES_RUNLOCK();
2664
2665 return (0);
2666 }
2667
2668 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2669 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2670 {
2671 int error = 0;
2672 PF_RULES_RLOCK_TRACKER;
2673
2674 #define ERROUT_IOCTL(target, x) \
2675 do { \
2676 error = (x); \
2677 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
2678 goto target; \
2679 } while (0)
2680
2681
2682 /* XXX keep in sync with switch() below */
2683 if (securelevel_gt(td->td_ucred, 2))
2684 switch (cmd) {
2685 case DIOCGETRULES:
2686 case DIOCGETRULENV:
2687 case DIOCGETADDRS:
2688 case DIOCGETADDR:
2689 case DIOCGETSTATE:
2690 case DIOCGETSTATENV:
2691 case DIOCSETSTATUSIF:
2692 case DIOCGETSTATUSNV:
2693 case DIOCCLRSTATUS:
2694 case DIOCNATLOOK:
2695 case DIOCSETDEBUG:
2696 #ifdef COMPAT_FREEBSD14
2697 case DIOCGETSTATES:
2698 case DIOCGETSTATESV2:
2699 #endif
2700 case DIOCGETTIMEOUT:
2701 case DIOCCLRRULECTRS:
2702 case DIOCGETLIMIT:
2703 case DIOCGETALTQSV0:
2704 case DIOCGETALTQSV1:
2705 case DIOCGETALTQV0:
2706 case DIOCGETALTQV1:
2707 case DIOCGETQSTATSV0:
2708 case DIOCGETQSTATSV1:
2709 case DIOCGETRULESETS:
2710 case DIOCGETRULESET:
2711 case DIOCRGETTABLES:
2712 case DIOCRGETTSTATS:
2713 case DIOCRCLRTSTATS:
2714 case DIOCRCLRADDRS:
2715 case DIOCRADDADDRS:
2716 case DIOCRDELADDRS:
2717 case DIOCRSETADDRS:
2718 case DIOCRGETADDRS:
2719 case DIOCRGETASTATS:
2720 case DIOCRCLRASTATS:
2721 case DIOCRTSTADDRS:
2722 case DIOCOSFPGET:
2723 case DIOCGETSRCNODES:
2724 case DIOCCLRSRCNODES:
2725 case DIOCGETSYNCOOKIES:
2726 case DIOCIGETIFACES:
2727 case DIOCGIFSPEEDV0:
2728 case DIOCGIFSPEEDV1:
2729 case DIOCSETIFFLAG:
2730 case DIOCCLRIFFLAG:
2731 case DIOCGETETHRULES:
2732 case DIOCGETETHRULE:
2733 case DIOCGETETHRULESETS:
2734 case DIOCGETETHRULESET:
2735 break;
2736 case DIOCRCLRTABLES:
2737 case DIOCRADDTABLES:
2738 case DIOCRDELTABLES:
2739 case DIOCRSETTFLAGS:
2740 if (((struct pfioc_table *)addr)->pfrio_flags &
2741 PFR_FLAG_DUMMY)
2742 break; /* dummy operation ok */
2743 return (EPERM);
2744 default:
2745 return (EPERM);
2746 }
2747
2748 if (!(flags & FWRITE))
2749 switch (cmd) {
2750 case DIOCGETRULES:
2751 case DIOCGETADDRS:
2752 case DIOCGETADDR:
2753 case DIOCGETSTATE:
2754 case DIOCGETSTATENV:
2755 case DIOCGETSTATUSNV:
2756 #ifdef COMPAT_FREEBSD14
2757 case DIOCGETSTATES:
2758 case DIOCGETSTATESV2:
2759 #endif
2760 case DIOCGETTIMEOUT:
2761 case DIOCGETLIMIT:
2762 case DIOCGETALTQSV0:
2763 case DIOCGETALTQSV1:
2764 case DIOCGETALTQV0:
2765 case DIOCGETALTQV1:
2766 case DIOCGETQSTATSV0:
2767 case DIOCGETQSTATSV1:
2768 case DIOCGETRULESETS:
2769 case DIOCGETRULESET:
2770 case DIOCNATLOOK:
2771 case DIOCRGETTABLES:
2772 case DIOCRGETTSTATS:
2773 case DIOCRGETADDRS:
2774 case DIOCRGETASTATS:
2775 case DIOCRTSTADDRS:
2776 case DIOCOSFPGET:
2777 case DIOCGETSRCNODES:
2778 case DIOCGETSYNCOOKIES:
2779 case DIOCIGETIFACES:
2780 case DIOCGIFSPEEDV1:
2781 case DIOCGIFSPEEDV0:
2782 case DIOCGETRULENV:
2783 case DIOCGETETHRULES:
2784 case DIOCGETETHRULE:
2785 case DIOCGETETHRULESETS:
2786 case DIOCGETETHRULESET:
2787 break;
2788 case DIOCRCLRTABLES:
2789 case DIOCRADDTABLES:
2790 case DIOCRDELTABLES:
2791 case DIOCRCLRTSTATS:
2792 case DIOCRCLRADDRS:
2793 case DIOCRADDADDRS:
2794 case DIOCRDELADDRS:
2795 case DIOCRSETADDRS:
2796 case DIOCRSETTFLAGS:
2797 if (((struct pfioc_table *)addr)->pfrio_flags &
2798 PFR_FLAG_DUMMY) {
2799 flags |= FWRITE; /* need write lock for dummy */
2800 break; /* dummy operation ok */
2801 }
2802 return (EACCES);
2803 default:
2804 return (EACCES);
2805 }
2806
2807 CURVNET_SET(TD_TO_VNET(td));
2808
2809 switch (cmd) {
2810 #ifdef COMPAT_FREEBSD14
2811 case DIOCSTART:
2812 error = pf_start();
2813 break;
2814
2815 case DIOCSTOP:
2816 error = pf_stop();
2817 break;
2818 #endif
2819
2820 case DIOCGETETHRULES: {
2821 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2822 nvlist_t *nvl;
2823 void *packed;
2824 struct pf_keth_rule *tail;
2825 struct pf_keth_ruleset *rs;
2826 u_int32_t ticket, nr;
2827 const char *anchor = "";
2828
2829 nvl = NULL;
2830 packed = NULL;
2831
2832 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2833
2834 if (nv->len > pf_ioctl_maxcount)
2835 ERROUT(ENOMEM);
2836
2837 /* Copy the request in */
2838 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2839 error = copyin(nv->data, packed, nv->len);
2840 if (error)
2841 ERROUT(error);
2842
2843 nvl = nvlist_unpack(packed, nv->len, 0);
2844 if (nvl == NULL)
2845 ERROUT(EBADMSG);
2846
2847 if (! nvlist_exists_string(nvl, "anchor"))
2848 ERROUT(EBADMSG);
2849
2850 anchor = nvlist_get_string(nvl, "anchor");
2851
2852 rs = pf_find_keth_ruleset(anchor);
2853
2854 nvlist_destroy(nvl);
2855 nvl = NULL;
2856 free(packed, M_NVLIST);
2857 packed = NULL;
2858
2859 if (rs == NULL)
2860 ERROUT(ENOENT);
2861
2862 /* Reply */
2863 nvl = nvlist_create(0);
2864 if (nvl == NULL)
2865 ERROUT(ENOMEM);
2866
2867 PF_RULES_RLOCK();
2868
2869 ticket = rs->active.ticket;
2870 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2871 if (tail)
2872 nr = tail->nr + 1;
2873 else
2874 nr = 0;
2875
2876 PF_RULES_RUNLOCK();
2877
2878 nvlist_add_number(nvl, "ticket", ticket);
2879 nvlist_add_number(nvl, "nr", nr);
2880
2881 packed = nvlist_pack(nvl, &nv->len);
2882 if (packed == NULL)
2883 ERROUT(ENOMEM);
2884
2885 if (nv->size == 0)
2886 ERROUT(0);
2887 else if (nv->size < nv->len)
2888 ERROUT(ENOSPC);
2889
2890 error = copyout(packed, nv->data, nv->len);
2891
2892 #undef ERROUT
2893 DIOCGETETHRULES_error:
2894 free(packed, M_NVLIST);
2895 nvlist_destroy(nvl);
2896 break;
2897 }
2898
2899 case DIOCGETETHRULE: {
2900 struct epoch_tracker et;
2901 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2902 nvlist_t *nvl = NULL;
2903 void *nvlpacked = NULL;
2904 struct pf_keth_rule *rule = NULL;
2905 struct pf_keth_ruleset *rs;
2906 u_int32_t ticket, nr;
2907 bool clear = false;
2908 const char *anchor;
2909
2910 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2911
2912 if (nv->len > pf_ioctl_maxcount)
2913 ERROUT(ENOMEM);
2914
2915 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2916 error = copyin(nv->data, nvlpacked, nv->len);
2917 if (error)
2918 ERROUT(error);
2919
2920 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2921 if (nvl == NULL)
2922 ERROUT(EBADMSG);
2923 if (! nvlist_exists_number(nvl, "ticket"))
2924 ERROUT(EBADMSG);
2925 ticket = nvlist_get_number(nvl, "ticket");
2926 if (! nvlist_exists_string(nvl, "anchor"))
2927 ERROUT(EBADMSG);
2928 anchor = nvlist_get_string(nvl, "anchor");
2929
2930 if (nvlist_exists_bool(nvl, "clear"))
2931 clear = nvlist_get_bool(nvl, "clear");
2932
2933 if (clear && !(flags & FWRITE))
2934 ERROUT(EACCES);
2935
2936 if (! nvlist_exists_number(nvl, "nr"))
2937 ERROUT(EBADMSG);
2938 nr = nvlist_get_number(nvl, "nr");
2939
2940 PF_RULES_RLOCK();
2941 rs = pf_find_keth_ruleset(anchor);
2942 if (rs == NULL) {
2943 PF_RULES_RUNLOCK();
2944 ERROUT(ENOENT);
2945 }
2946 if (ticket != rs->active.ticket) {
2947 PF_RULES_RUNLOCK();
2948 ERROUT(EBUSY);
2949 }
2950
2951 nvlist_destroy(nvl);
2952 nvl = NULL;
2953 free(nvlpacked, M_NVLIST);
2954 nvlpacked = NULL;
2955
2956 rule = TAILQ_FIRST(rs->active.rules);
2957 while ((rule != NULL) && (rule->nr != nr))
2958 rule = TAILQ_NEXT(rule, entries);
2959 if (rule == NULL) {
2960 PF_RULES_RUNLOCK();
2961 ERROUT(ENOENT);
2962 }
2963 /* Make sure rule can't go away. */
2964 NET_EPOCH_ENTER(et);
2965 PF_RULES_RUNLOCK();
2966 nvl = pf_keth_rule_to_nveth_rule(rule);
2967 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
2968 NET_EPOCH_EXIT(et);
2969 ERROUT(EBUSY);
2970 }
2971 NET_EPOCH_EXIT(et);
2972 if (nvl == NULL)
2973 ERROUT(ENOMEM);
2974
2975 nvlpacked = nvlist_pack(nvl, &nv->len);
2976 if (nvlpacked == NULL)
2977 ERROUT(ENOMEM);
2978
2979 if (nv->size == 0)
2980 ERROUT(0);
2981 else if (nv->size < nv->len)
2982 ERROUT(ENOSPC);
2983
2984 error = copyout(nvlpacked, nv->data, nv->len);
2985 if (error == 0 && clear) {
2986 counter_u64_zero(rule->evaluations);
2987 for (int i = 0; i < 2; i++) {
2988 counter_u64_zero(rule->packets[i]);
2989 counter_u64_zero(rule->bytes[i]);
2990 }
2991 }
2992
2993 #undef ERROUT
2994 DIOCGETETHRULE_error:
2995 free(nvlpacked, M_NVLIST);
2996 nvlist_destroy(nvl);
2997 break;
2998 }
2999
3000 case DIOCADDETHRULE: {
3001 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3002 nvlist_t *nvl = NULL;
3003 void *nvlpacked = NULL;
3004 struct pf_keth_rule *rule = NULL, *tail = NULL;
3005 struct pf_keth_ruleset *ruleset = NULL;
3006 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
3007 const char *anchor = "", *anchor_call = "";
3008
3009 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3010
3011 if (nv->len > pf_ioctl_maxcount)
3012 ERROUT(ENOMEM);
3013
3014 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3015 error = copyin(nv->data, nvlpacked, nv->len);
3016 if (error)
3017 ERROUT(error);
3018
3019 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3020 if (nvl == NULL)
3021 ERROUT(EBADMSG);
3022
3023 if (! nvlist_exists_number(nvl, "ticket"))
3024 ERROUT(EBADMSG);
3025
3026 if (nvlist_exists_string(nvl, "anchor"))
3027 anchor = nvlist_get_string(nvl, "anchor");
3028 if (nvlist_exists_string(nvl, "anchor_call"))
3029 anchor_call = nvlist_get_string(nvl, "anchor_call");
3030
3031 ruleset = pf_find_keth_ruleset(anchor);
3032 if (ruleset == NULL)
3033 ERROUT(EINVAL);
3034
3035 if (nvlist_get_number(nvl, "ticket") !=
3036 ruleset->inactive.ticket) {
3037 DPFPRINTF(PF_DEBUG_MISC,
3038 ("ticket: %d != %d\n",
3039 (u_int32_t)nvlist_get_number(nvl, "ticket"),
3040 ruleset->inactive.ticket));
3041 ERROUT(EBUSY);
3042 }
3043
3044 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3045 rule->timestamp = NULL;
3046
3047 error = pf_nveth_rule_to_keth_rule(nvl, rule);
3048 if (error != 0)
3049 ERROUT(error);
3050
3051 if (rule->ifname[0])
3052 kif = pf_kkif_create(M_WAITOK);
3053 if (rule->bridge_to_name[0])
3054 bridge_to_kif = pf_kkif_create(M_WAITOK);
3055 rule->evaluations = counter_u64_alloc(M_WAITOK);
3056 for (int i = 0; i < 2; i++) {
3057 rule->packets[i] = counter_u64_alloc(M_WAITOK);
3058 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3059 }
3060 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3061 M_WAITOK | M_ZERO);
3062
3063 PF_RULES_WLOCK();
3064
3065 if (rule->ifname[0]) {
3066 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3067 pfi_kkif_ref(rule->kif);
3068 } else
3069 rule->kif = NULL;
3070 if (rule->bridge_to_name[0]) {
3071 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3072 rule->bridge_to_name);
3073 pfi_kkif_ref(rule->bridge_to);
3074 } else
3075 rule->bridge_to = NULL;
3076
3077 #ifdef ALTQ
3078 /* set queue IDs */
3079 if (rule->qname[0] != 0) {
3080 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3081 error = EBUSY;
3082 else
3083 rule->qid = rule->qid;
3084 }
3085 #endif
3086 if (rule->tagname[0])
3087 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3088 error = EBUSY;
3089 if (rule->match_tagname[0])
3090 if ((rule->match_tag = pf_tagname2tag(
3091 rule->match_tagname)) == 0)
3092 error = EBUSY;
3093
3094 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3095 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3096 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3097 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3098
3099 if (error) {
3100 pf_free_eth_rule(rule);
3101 PF_RULES_WUNLOCK();
3102 ERROUT(error);
3103 }
3104
3105 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3106 pf_free_eth_rule(rule);
3107 PF_RULES_WUNLOCK();
3108 ERROUT(EINVAL);
3109 }
3110
3111 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3112 if (tail)
3113 rule->nr = tail->nr + 1;
3114 else
3115 rule->nr = 0;
3116
3117 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3118
3119 PF_RULES_WUNLOCK();
3120
3121 #undef ERROUT
3122 DIOCADDETHRULE_error:
3123 nvlist_destroy(nvl);
3124 free(nvlpacked, M_NVLIST);
3125 break;
3126 }
3127
3128 case DIOCGETETHRULESETS: {
3129 struct epoch_tracker et;
3130 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3131 nvlist_t *nvl = NULL;
3132 void *nvlpacked = NULL;
3133 struct pf_keth_ruleset *ruleset;
3134 struct pf_keth_anchor *anchor;
3135 int nr = 0;
3136
3137 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3138
3139 if (nv->len > pf_ioctl_maxcount)
3140 ERROUT(ENOMEM);
3141
3142 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3143 error = copyin(nv->data, nvlpacked, nv->len);
3144 if (error)
3145 ERROUT(error);
3146
3147 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3148 if (nvl == NULL)
3149 ERROUT(EBADMSG);
3150 if (! nvlist_exists_string(nvl, "path"))
3151 ERROUT(EBADMSG);
3152
3153 NET_EPOCH_ENTER(et);
3154
3155 if ((ruleset = pf_find_keth_ruleset(
3156 nvlist_get_string(nvl, "path"))) == NULL) {
3157 NET_EPOCH_EXIT(et);
3158 ERROUT(ENOENT);
3159 }
3160
3161 if (ruleset->anchor == NULL) {
3162 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3163 if (anchor->parent == NULL)
3164 nr++;
3165 } else {
3166 RB_FOREACH(anchor, pf_keth_anchor_node,
3167 &ruleset->anchor->children)
3168 nr++;
3169 }
3170
3171 NET_EPOCH_EXIT(et);
3172
3173 nvlist_destroy(nvl);
3174 nvl = NULL;
3175 free(nvlpacked, M_NVLIST);
3176 nvlpacked = NULL;
3177
3178 nvl = nvlist_create(0);
3179 if (nvl == NULL)
3180 ERROUT(ENOMEM);
3181
3182 nvlist_add_number(nvl, "nr", nr);
3183
3184 nvlpacked = nvlist_pack(nvl, &nv->len);
3185 if (nvlpacked == NULL)
3186 ERROUT(ENOMEM);
3187
3188 if (nv->size == 0)
3189 ERROUT(0);
3190 else if (nv->size < nv->len)
3191 ERROUT(ENOSPC);
3192
3193 error = copyout(nvlpacked, nv->data, nv->len);
3194
3195 #undef ERROUT
3196 DIOCGETETHRULESETS_error:
3197 free(nvlpacked, M_NVLIST);
3198 nvlist_destroy(nvl);
3199 break;
3200 }
3201
3202 case DIOCGETETHRULESET: {
3203 struct epoch_tracker et;
3204 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3205 nvlist_t *nvl = NULL;
3206 void *nvlpacked = NULL;
3207 struct pf_keth_ruleset *ruleset;
3208 struct pf_keth_anchor *anchor;
3209 int nr = 0, req_nr = 0;
3210 bool found = false;
3211
3212 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3213
3214 if (nv->len > pf_ioctl_maxcount)
3215 ERROUT(ENOMEM);
3216
3217 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3218 error = copyin(nv->data, nvlpacked, nv->len);
3219 if (error)
3220 ERROUT(error);
3221
3222 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3223 if (nvl == NULL)
3224 ERROUT(EBADMSG);
3225 if (! nvlist_exists_string(nvl, "path"))
3226 ERROUT(EBADMSG);
3227 if (! nvlist_exists_number(nvl, "nr"))
3228 ERROUT(EBADMSG);
3229
3230 req_nr = nvlist_get_number(nvl, "nr");
3231
3232 NET_EPOCH_ENTER(et);
3233
3234 if ((ruleset = pf_find_keth_ruleset(
3235 nvlist_get_string(nvl, "path"))) == NULL) {
3236 NET_EPOCH_EXIT(et);
3237 ERROUT(ENOENT);
3238 }
3239
3240 nvlist_destroy(nvl);
3241 nvl = NULL;
3242 free(nvlpacked, M_NVLIST);
3243 nvlpacked = NULL;
3244
3245 nvl = nvlist_create(0);
3246 if (nvl == NULL) {
3247 NET_EPOCH_EXIT(et);
3248 ERROUT(ENOMEM);
3249 }
3250
3251 if (ruleset->anchor == NULL) {
3252 RB_FOREACH(anchor, pf_keth_anchor_global,
3253 &V_pf_keth_anchors) {
3254 if (anchor->parent == NULL && nr++ == req_nr) {
3255 found = true;
3256 break;
3257 }
3258 }
3259 } else {
3260 RB_FOREACH(anchor, pf_keth_anchor_node,
3261 &ruleset->anchor->children) {
3262 if (nr++ == req_nr) {
3263 found = true;
3264 break;
3265 }
3266 }
3267 }
3268
3269 NET_EPOCH_EXIT(et);
3270 if (found) {
3271 nvlist_add_number(nvl, "nr", nr);
3272 nvlist_add_string(nvl, "name", anchor->name);
3273 if (ruleset->anchor)
3274 nvlist_add_string(nvl, "path",
3275 ruleset->anchor->path);
3276 else
3277 nvlist_add_string(nvl, "path", "");
3278 } else {
3279 ERROUT(EBUSY);
3280 }
3281
3282 nvlpacked = nvlist_pack(nvl, &nv->len);
3283 if (nvlpacked == NULL)
3284 ERROUT(ENOMEM);
3285
3286 if (nv->size == 0)
3287 ERROUT(0);
3288 else if (nv->size < nv->len)
3289 ERROUT(ENOSPC);
3290
3291 error = copyout(nvlpacked, nv->data, nv->len);
3292
3293 #undef ERROUT
3294 DIOCGETETHRULESET_error:
3295 free(nvlpacked, M_NVLIST);
3296 nvlist_destroy(nvl);
3297 break;
3298 }
3299
3300 case DIOCADDRULENV: {
3301 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3302 nvlist_t *nvl = NULL;
3303 void *nvlpacked = NULL;
3304 struct pf_krule *rule = NULL;
3305 const char *anchor = "", *anchor_call = "";
3306 uint32_t ticket = 0, pool_ticket = 0;
3307
3308 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
3309
3310 if (nv->len > pf_ioctl_maxcount)
3311 ERROUT(ENOMEM);
3312
3313 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3314 error = copyin(nv->data, nvlpacked, nv->len);
3315 if (error)
3316 ERROUT(error);
3317
3318 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3319 if (nvl == NULL)
3320 ERROUT(EBADMSG);
3321
3322 if (! nvlist_exists_number(nvl, "ticket"))
3323 ERROUT(EINVAL);
3324 ticket = nvlist_get_number(nvl, "ticket");
3325
3326 if (! nvlist_exists_number(nvl, "pool_ticket"))
3327 ERROUT(EINVAL);
3328 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3329
3330 if (! nvlist_exists_nvlist(nvl, "rule"))
3331 ERROUT(EINVAL);
3332
3333 rule = pf_krule_alloc();
3334 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3335 rule);
3336 if (error)
3337 ERROUT(error);
3338
3339 if (nvlist_exists_string(nvl, "anchor"))
3340 anchor = nvlist_get_string(nvl, "anchor");
3341 if (nvlist_exists_string(nvl, "anchor_call"))
3342 anchor_call = nvlist_get_string(nvl, "anchor_call");
3343
3344 if ((error = nvlist_error(nvl)))
3345 ERROUT(error);
3346
3347 /* Frees rule on error */
3348 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3349 anchor_call, td->td_ucred->cr_ruid,
3350 td->td_proc ? td->td_proc->p_pid : 0);
3351
3352 nvlist_destroy(nvl);
3353 free(nvlpacked, M_NVLIST);
3354 break;
3355 #undef ERROUT
3356 DIOCADDRULENV_error:
3357 pf_krule_free(rule);
3358 nvlist_destroy(nvl);
3359 free(nvlpacked, M_NVLIST);
3360
3361 break;
3362 }
3363 case DIOCADDRULE: {
3364 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3365 struct pf_krule *rule;
3366
3367 rule = pf_krule_alloc();
3368 error = pf_rule_to_krule(&pr->rule, rule);
3369 if (error != 0) {
3370 pf_krule_free(rule);
3371 break;
3372 }
3373
3374 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3375
3376 /* Frees rule on error */
3377 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3378 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3379 td->td_proc ? td->td_proc->p_pid : 0);
3380 break;
3381 }
3382
3383 case DIOCGETRULES: {
3384 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3385
3386 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3387
3388 error = pf_ioctl_getrules(pr);
3389
3390 break;
3391 }
3392
3393 case DIOCGETRULENV: {
3394 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3395 nvlist_t *nvrule = NULL;
3396 nvlist_t *nvl = NULL;
3397 struct pf_kruleset *ruleset;
3398 struct pf_krule *rule;
3399 void *nvlpacked = NULL;
3400 int rs_num, nr;
3401 bool clear_counter = false;
3402
3403 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
3404
3405 if (nv->len > pf_ioctl_maxcount)
3406 ERROUT(ENOMEM);
3407
3408 /* Copy the request in */
3409 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3410 error = copyin(nv->data, nvlpacked, nv->len);
3411 if (error)
3412 ERROUT(error);
3413
3414 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3415 if (nvl == NULL)
3416 ERROUT(EBADMSG);
3417
3418 if (! nvlist_exists_string(nvl, "anchor"))
3419 ERROUT(EBADMSG);
3420 if (! nvlist_exists_number(nvl, "ruleset"))
3421 ERROUT(EBADMSG);
3422 if (! nvlist_exists_number(nvl, "ticket"))
3423 ERROUT(EBADMSG);
3424 if (! nvlist_exists_number(nvl, "nr"))
3425 ERROUT(EBADMSG);
3426
3427 if (nvlist_exists_bool(nvl, "clear_counter"))
3428 clear_counter = nvlist_get_bool(nvl, "clear_counter");
3429
3430 if (clear_counter && !(flags & FWRITE))
3431 ERROUT(EACCES);
3432
3433 nr = nvlist_get_number(nvl, "nr");
3434
3435 PF_RULES_WLOCK();
3436 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3437 if (ruleset == NULL) {
3438 PF_RULES_WUNLOCK();
3439 ERROUT(ENOENT);
3440 }
3441
3442 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3443 if (rs_num >= PF_RULESET_MAX) {
3444 PF_RULES_WUNLOCK();
3445 ERROUT(EINVAL);
3446 }
3447
3448 if (nvlist_get_number(nvl, "ticket") !=
3449 ruleset->rules[rs_num].active.ticket) {
3450 PF_RULES_WUNLOCK();
3451 ERROUT(EBUSY);
3452 }
3453
3454 if ((error = nvlist_error(nvl))) {
3455 PF_RULES_WUNLOCK();
3456 ERROUT(error);
3457 }
3458
3459 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3460 while ((rule != NULL) && (rule->nr != nr))
3461 rule = TAILQ_NEXT(rule, entries);
3462 if (rule == NULL) {
3463 PF_RULES_WUNLOCK();
3464 ERROUT(EBUSY);
3465 }
3466
3467 nvrule = pf_krule_to_nvrule(rule);
3468
3469 nvlist_destroy(nvl);
3470 nvl = nvlist_create(0);
3471 if (nvl == NULL) {
3472 PF_RULES_WUNLOCK();
3473 ERROUT(ENOMEM);
3474 }
3475 nvlist_add_number(nvl, "nr", nr);
3476 nvlist_add_nvlist(nvl, "rule", nvrule);
3477 nvlist_destroy(nvrule);
3478 nvrule = NULL;
3479 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3480 PF_RULES_WUNLOCK();
3481 ERROUT(EBUSY);
3482 }
3483
3484 free(nvlpacked, M_NVLIST);
3485 nvlpacked = nvlist_pack(nvl, &nv->len);
3486 if (nvlpacked == NULL) {
3487 PF_RULES_WUNLOCK();
3488 ERROUT(ENOMEM);
3489 }
3490
3491 if (nv->size == 0) {
3492 PF_RULES_WUNLOCK();
3493 ERROUT(0);
3494 }
3495 else if (nv->size < nv->len) {
3496 PF_RULES_WUNLOCK();
3497 ERROUT(ENOSPC);
3498 }
3499
3500 if (clear_counter)
3501 pf_krule_clear_counters(rule);
3502
3503 PF_RULES_WUNLOCK();
3504
3505 error = copyout(nvlpacked, nv->data, nv->len);
3506
3507 #undef ERROUT
3508 DIOCGETRULENV_error:
3509 free(nvlpacked, M_NVLIST);
3510 nvlist_destroy(nvrule);
3511 nvlist_destroy(nvl);
3512
3513 break;
3514 }
3515
3516 case DIOCCHANGERULE: {
3517 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
3518 struct pf_kruleset *ruleset;
3519 struct pf_krule *oldrule = NULL, *newrule = NULL;
3520 struct pfi_kkif *kif = NULL;
3521 struct pf_kpooladdr *pa;
3522 u_int32_t nr = 0;
3523 int rs_num;
3524
3525 pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3526
3527 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3528 pcr->action > PF_CHANGE_GET_TICKET) {
3529 error = EINVAL;
3530 break;
3531 }
3532 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3533 error = EINVAL;
3534 break;
3535 }
3536
3537 if (pcr->action != PF_CHANGE_REMOVE) {
3538 newrule = pf_krule_alloc();
3539 error = pf_rule_to_krule(&pcr->rule, newrule);
3540 if (error != 0) {
3541 pf_krule_free(newrule);
3542 break;
3543 }
3544
3545 if (newrule->ifname[0])
3546 kif = pf_kkif_create(M_WAITOK);
3547 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3548 for (int i = 0; i < 2; i++) {
3549 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3550 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3551 }
3552 newrule->states_cur = counter_u64_alloc(M_WAITOK);
3553 newrule->states_tot = counter_u64_alloc(M_WAITOK);
3554 newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3555 newrule->cuid = td->td_ucred->cr_ruid;
3556 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3557 TAILQ_INIT(&newrule->rpool.list);
3558 }
3559 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3560
3561 PF_CONFIG_LOCK();
3562 PF_RULES_WLOCK();
3563 #ifdef PF_WANT_32_TO_64_COUNTER
3564 if (newrule != NULL) {
3565 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3566 newrule->allrulelinked = true;
3567 V_pf_allrulecount++;
3568 }
3569 #endif
3570
3571 if (!(pcr->action == PF_CHANGE_REMOVE ||
3572 pcr->action == PF_CHANGE_GET_TICKET) &&
3573 pcr->pool_ticket != V_ticket_pabuf)
3574 ERROUT(EBUSY);
3575
3576 ruleset = pf_find_kruleset(pcr->anchor);
3577 if (ruleset == NULL)
3578 ERROUT(EINVAL);
3579
3580 rs_num = pf_get_ruleset_number(pcr->rule.action);
3581 if (rs_num >= PF_RULESET_MAX)
3582 ERROUT(EINVAL);
3583
3584 /*
3585 * XXXMJG: there is no guarantee that the ruleset was
3586 * created by the usual route of calling DIOCXBEGIN.
3587 * As a result it is possible the rule tree will not
3588 * be allocated yet. Hack around it by doing it here.
3589 * Note it is fine to let the tree persist in case of
3590 * error as it will be freed down the road on future
3591 * updates (if need be).
3592 */
3593 if (ruleset->rules[rs_num].active.tree == NULL) {
3594 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3595 if (ruleset->rules[rs_num].active.tree == NULL) {
3596 ERROUT(ENOMEM);
3597 }
3598 }
3599
3600 if (pcr->action == PF_CHANGE_GET_TICKET) {
3601 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3602 ERROUT(0);
3603 } else if (pcr->ticket !=
3604 ruleset->rules[rs_num].active.ticket)
3605 ERROUT(EINVAL);
3606
3607 if (pcr->action != PF_CHANGE_REMOVE) {
3608 if (newrule->ifname[0]) {
3609 newrule->kif = pfi_kkif_attach(kif,
3610 newrule->ifname);
3611 kif = NULL;
3612 pfi_kkif_ref(newrule->kif);
3613 } else
3614 newrule->kif = NULL;
3615
3616 if (newrule->rtableid > 0 &&
3617 newrule->rtableid >= rt_numfibs)
3618 error = EBUSY;
3619
3620 #ifdef ALTQ
3621 /* set queue IDs */
3622 if (newrule->qname[0] != 0) {
3623 if ((newrule->qid =
3624 pf_qname2qid(newrule->qname)) == 0)
3625 error = EBUSY;
3626 else if (newrule->pqname[0] != 0) {
3627 if ((newrule->pqid =
3628 pf_qname2qid(newrule->pqname)) == 0)
3629 error = EBUSY;
3630 } else
3631 newrule->pqid = newrule->qid;
3632 }
3633 #endif /* ALTQ */
3634 if (newrule->tagname[0])
3635 if ((newrule->tag =
3636 pf_tagname2tag(newrule->tagname)) == 0)
3637 error = EBUSY;
3638 if (newrule->match_tagname[0])
3639 if ((newrule->match_tag = pf_tagname2tag(
3640 newrule->match_tagname)) == 0)
3641 error = EBUSY;
3642 if (newrule->rt && !newrule->direction)
3643 error = EINVAL;
3644 if (!newrule->log)
3645 newrule->logif = 0;
3646 if (newrule->logif >= PFLOGIFS_MAX)
3647 error = EINVAL;
3648 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3649 error = ENOMEM;
3650 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3651 error = ENOMEM;
3652 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3653 error = EINVAL;
3654 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3655 if (pa->addr.type == PF_ADDR_TABLE) {
3656 pa->addr.p.tbl =
3657 pfr_attach_table(ruleset,
3658 pa->addr.v.tblname);
3659 if (pa->addr.p.tbl == NULL)
3660 error = ENOMEM;
3661 }
3662
3663 newrule->overload_tbl = NULL;
3664 if (newrule->overload_tblname[0]) {
3665 if ((newrule->overload_tbl = pfr_attach_table(
3666 ruleset, newrule->overload_tblname)) ==
3667 NULL)
3668 error = EINVAL;
3669 else
3670 newrule->overload_tbl->pfrkt_flags |=
3671 PFR_TFLAG_ACTIVE;
3672 }
3673
3674 pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3675 if (((((newrule->action == PF_NAT) ||
3676 (newrule->action == PF_RDR) ||
3677 (newrule->action == PF_BINAT) ||
3678 (newrule->rt > PF_NOPFROUTE)) &&
3679 !newrule->anchor)) &&
3680 (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3681 error = EINVAL;
3682
3683 if (error) {
3684 pf_free_rule(newrule);
3685 PF_RULES_WUNLOCK();
3686 PF_CONFIG_UNLOCK();
3687 break;
3688 }
3689
3690 newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3691 }
3692 pf_empty_kpool(&V_pf_pabuf);
3693
3694 if (pcr->action == PF_CHANGE_ADD_HEAD)
3695 oldrule = TAILQ_FIRST(
3696 ruleset->rules[rs_num].active.ptr);
3697 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3698 oldrule = TAILQ_LAST(
3699 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3700 else {
3701 oldrule = TAILQ_FIRST(
3702 ruleset->rules[rs_num].active.ptr);
3703 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3704 oldrule = TAILQ_NEXT(oldrule, entries);
3705 if (oldrule == NULL) {
3706 if (newrule != NULL)
3707 pf_free_rule(newrule);
3708 PF_RULES_WUNLOCK();
3709 PF_CONFIG_UNLOCK();
3710 error = EINVAL;
3711 break;
3712 }
3713 }
3714
3715 if (pcr->action == PF_CHANGE_REMOVE) {
3716 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3717 oldrule);
3718 RB_REMOVE(pf_krule_global,
3719 ruleset->rules[rs_num].active.tree, oldrule);
3720 ruleset->rules[rs_num].active.rcount--;
3721 } else {
3722 pf_hash_rule(newrule);
3723 if (RB_INSERT(pf_krule_global,
3724 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3725 pf_free_rule(newrule);
3726 PF_RULES_WUNLOCK();
3727 PF_CONFIG_UNLOCK();
3728 error = EEXIST;
3729 break;
3730 }
3731
3732 if (oldrule == NULL)
3733 TAILQ_INSERT_TAIL(
3734 ruleset->rules[rs_num].active.ptr,
3735 newrule, entries);
3736 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3737 pcr->action == PF_CHANGE_ADD_BEFORE)
3738 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3739 else
3740 TAILQ_INSERT_AFTER(
3741 ruleset->rules[rs_num].active.ptr,
3742 oldrule, newrule, entries);
3743 ruleset->rules[rs_num].active.rcount++;
3744 }
3745
3746 nr = 0;
3747 TAILQ_FOREACH(oldrule,
3748 ruleset->rules[rs_num].active.ptr, entries)
3749 oldrule->nr = nr++;
3750
3751 ruleset->rules[rs_num].active.ticket++;
3752
3753 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3754 pf_remove_if_empty_kruleset(ruleset);
3755
3756 PF_RULES_WUNLOCK();
3757 PF_CONFIG_UNLOCK();
3758 break;
3759
3760 #undef ERROUT
3761 DIOCCHANGERULE_error:
3762 PF_RULES_WUNLOCK();
3763 PF_CONFIG_UNLOCK();
3764 pf_krule_free(newrule);
3765 pf_kkif_free(kif);
3766 break;
3767 }
3768
3769 case DIOCCLRSTATESNV: {
3770 error = pf_clearstates_nv((struct pfioc_nv *)addr);
3771 break;
3772 }
3773
3774 case DIOCKILLSTATESNV: {
3775 error = pf_killstates_nv((struct pfioc_nv *)addr);
3776 break;
3777 }
3778
3779 case DIOCADDSTATE: {
3780 struct pfioc_state *ps = (struct pfioc_state *)addr;
3781 struct pfsync_state_1301 *sp = &ps->state;
3782
3783 if (sp->timeout >= PFTM_MAX) {
3784 error = EINVAL;
3785 break;
3786 }
3787 if (V_pfsync_state_import_ptr != NULL) {
3788 PF_RULES_RLOCK();
3789 error = V_pfsync_state_import_ptr(
3790 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3791 PFSYNC_MSG_VERSION_1301);
3792 PF_RULES_RUNLOCK();
3793 } else
3794 error = EOPNOTSUPP;
3795 break;
3796 }
3797
3798 case DIOCGETSTATE: {
3799 struct pfioc_state *ps = (struct pfioc_state *)addr;
3800 struct pf_kstate *s;
3801
3802 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3803 if (s == NULL) {
3804 error = ENOENT;
3805 break;
3806 }
3807
3808 pfsync_state_export((union pfsync_state_union*)&ps->state,
3809 s, PFSYNC_MSG_VERSION_1301);
3810 PF_STATE_UNLOCK(s);
3811 break;
3812 }
3813
3814 case DIOCGETSTATENV: {
3815 error = pf_getstate((struct pfioc_nv *)addr);
3816 break;
3817 }
3818
3819 #ifdef COMPAT_FREEBSD14
3820 case DIOCGETSTATES: {
3821 struct pfioc_states *ps = (struct pfioc_states *)addr;
3822 struct pf_kstate *s;
3823 struct pfsync_state_1301 *pstore, *p;
3824 int i, nr;
3825 size_t slice_count = 16, count;
3826 void *out;
3827
3828 if (ps->ps_len <= 0) {
3829 nr = uma_zone_get_cur(V_pf_state_z);
3830 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3831 break;
3832 }
3833
3834 out = ps->ps_states;
3835 pstore = mallocarray(slice_count,
3836 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3837 nr = 0;
3838
3839 for (i = 0; i <= V_pf_hashmask; i++) {
3840 struct pf_idhash *ih = &V_pf_idhash[i];
3841
3842 DIOCGETSTATES_retry:
3843 p = pstore;
3844
3845 if (LIST_EMPTY(&ih->states))
3846 continue;
3847
3848 PF_HASHROW_LOCK(ih);
3849 count = 0;
3850 LIST_FOREACH(s, &ih->states, entry) {
3851 if (s->timeout == PFTM_UNLINKED)
3852 continue;
3853 count++;
3854 }
3855
3856 if (count > slice_count) {
3857 PF_HASHROW_UNLOCK(ih);
3858 free(pstore, M_TEMP);
3859 slice_count = count * 2;
3860 pstore = mallocarray(slice_count,
3861 sizeof(struct pfsync_state_1301), M_TEMP,
3862 M_WAITOK | M_ZERO);
3863 goto DIOCGETSTATES_retry;
3864 }
3865
3866 if ((nr+count) * sizeof(*p) > ps->ps_len) {
3867 PF_HASHROW_UNLOCK(ih);
3868 goto DIOCGETSTATES_full;
3869 }
3870
3871 LIST_FOREACH(s, &ih->states, entry) {
3872 if (s->timeout == PFTM_UNLINKED)
3873 continue;
3874
3875 pfsync_state_export((union pfsync_state_union*)p,
3876 s, PFSYNC_MSG_VERSION_1301);
3877 p++;
3878 nr++;
3879 }
3880 PF_HASHROW_UNLOCK(ih);
3881 error = copyout(pstore, out,
3882 sizeof(struct pfsync_state_1301) * count);
3883 if (error)
3884 break;
3885 out = ps->ps_states + nr;
3886 }
3887 DIOCGETSTATES_full:
3888 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3889 free(pstore, M_TEMP);
3890
3891 break;
3892 }
3893
3894 case DIOCGETSTATESV2: {
3895 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
3896 struct pf_kstate *s;
3897 struct pf_state_export *pstore, *p;
3898 int i, nr;
3899 size_t slice_count = 16, count;
3900 void *out;
3901
3902 if (ps->ps_req_version > PF_STATE_VERSION) {
3903 error = ENOTSUP;
3904 break;
3905 }
3906
3907 if (ps->ps_len <= 0) {
3908 nr = uma_zone_get_cur(V_pf_state_z);
3909 ps->ps_len = sizeof(struct pf_state_export) * nr;
3910 break;
3911 }
3912
3913 out = ps->ps_states;
3914 pstore = mallocarray(slice_count,
3915 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3916 nr = 0;
3917
3918 for (i = 0; i <= V_pf_hashmask; i++) {
3919 struct pf_idhash *ih = &V_pf_idhash[i];
3920
3921 DIOCGETSTATESV2_retry:
3922 p = pstore;
3923
3924 if (LIST_EMPTY(&ih->states))
3925 continue;
3926
3927 PF_HASHROW_LOCK(ih);
3928 count = 0;
3929 LIST_FOREACH(s, &ih->states, entry) {
3930 if (s->timeout == PFTM_UNLINKED)
3931 continue;
3932 count++;
3933 }
3934
3935 if (count > slice_count) {
3936 PF_HASHROW_UNLOCK(ih);
3937 free(pstore, M_TEMP);
3938 slice_count = count * 2;
3939 pstore = mallocarray(slice_count,
3940 sizeof(struct pf_state_export), M_TEMP,
3941 M_WAITOK | M_ZERO);
3942 goto DIOCGETSTATESV2_retry;
3943 }
3944
3945 if ((nr+count) * sizeof(*p) > ps->ps_len) {
3946 PF_HASHROW_UNLOCK(ih);
3947 goto DIOCGETSTATESV2_full;
3948 }
3949
3950 LIST_FOREACH(s, &ih->states, entry) {
3951 if (s->timeout == PFTM_UNLINKED)
3952 continue;
3953
3954 pf_state_export(p, s);
3955 p++;
3956 nr++;
3957 }
3958 PF_HASHROW_UNLOCK(ih);
3959 error = copyout(pstore, out,
3960 sizeof(struct pf_state_export) * count);
3961 if (error)
3962 break;
3963 out = ps->ps_states + nr;
3964 }
3965 DIOCGETSTATESV2_full:
3966 ps->ps_len = nr * sizeof(struct pf_state_export);
3967 free(pstore, M_TEMP);
3968
3969 break;
3970 }
3971 #endif
3972 case DIOCGETSTATUSNV: {
3973 error = pf_getstatus((struct pfioc_nv *)addr);
3974 break;
3975 }
3976
3977 case DIOCSETSTATUSIF: {
3978 struct pfioc_if *pi = (struct pfioc_if *)addr;
3979
3980 if (pi->ifname[0] == 0) {
3981 bzero(V_pf_status.ifname, IFNAMSIZ);
3982 break;
3983 }
3984 PF_RULES_WLOCK();
3985 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3986 PF_RULES_WUNLOCK();
3987 break;
3988 }
3989
3990 case DIOCCLRSTATUS: {
3991 pf_ioctl_clear_status();
3992 break;
3993 }
3994
3995 case DIOCNATLOOK: {
3996 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
3997 struct pf_state_key *sk;
3998 struct pf_kstate *state;
3999 struct pf_state_key_cmp key;
4000 int m = 0, direction = pnl->direction;
4001 int sidx, didx;
4002
4003 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
4004 sidx = (direction == PF_IN) ? 1 : 0;
4005 didx = (direction == PF_IN) ? 0 : 1;
4006
4007 if (!pnl->proto ||
4008 PF_AZERO(&pnl->saddr, pnl->af) ||
4009 PF_AZERO(&pnl->daddr, pnl->af) ||
4010 ((pnl->proto == IPPROTO_TCP ||
4011 pnl->proto == IPPROTO_UDP) &&
4012 (!pnl->dport || !pnl->sport)))
4013 error = EINVAL;
4014 else {
4015 bzero(&key, sizeof(key));
4016 key.af = pnl->af;
4017 key.proto = pnl->proto;
4018 PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4019 key.port[sidx] = pnl->sport;
4020 PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4021 key.port[didx] = pnl->dport;
4022
4023 state = pf_find_state_all(&key, direction, &m);
4024 if (state == NULL) {
4025 error = ENOENT;
4026 } else {
4027 if (m > 1) {
4028 PF_STATE_UNLOCK(state);
4029 error = E2BIG; /* more than one state */
4030 } else {
4031 sk = state->key[sidx];
4032 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4033 pnl->rsport = sk->port[sidx];
4034 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4035 pnl->rdport = sk->port[didx];
4036 PF_STATE_UNLOCK(state);
4037 }
4038 }
4039 }
4040 break;
4041 }
4042
4043 case DIOCSETTIMEOUT: {
4044 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4045
4046 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4047 &pt->seconds);
4048 break;
4049 }
4050
4051 case DIOCGETTIMEOUT: {
4052 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4053
4054 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4055 break;
4056 }
4057
4058 case DIOCGETLIMIT: {
4059 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4060
4061 error = pf_ioctl_get_limit(pl->index, &pl->limit);
4062 break;
4063 }
4064
4065 case DIOCSETLIMIT: {
4066 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4067 unsigned int old_limit;
4068
4069 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4070 pl->limit = old_limit;
4071 break;
4072 }
4073
4074 case DIOCSETDEBUG: {
4075 u_int32_t *level = (u_int32_t *)addr;
4076
4077 PF_RULES_WLOCK();
4078 V_pf_status.debug = *level;
4079 PF_RULES_WUNLOCK();
4080 break;
4081 }
4082
4083 case DIOCCLRRULECTRS: {
4084 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4085 struct pf_kruleset *ruleset = &pf_main_ruleset;
4086 struct pf_krule *rule;
4087
4088 PF_RULES_WLOCK();
4089 TAILQ_FOREACH(rule,
4090 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4091 pf_counter_u64_zero(&rule->evaluations);
4092 for (int i = 0; i < 2; i++) {
4093 pf_counter_u64_zero(&rule->packets[i]);
4094 pf_counter_u64_zero(&rule->bytes[i]);
4095 }
4096 }
4097 PF_RULES_WUNLOCK();
4098 break;
4099 }
4100
4101 case DIOCGIFSPEEDV0:
4102 case DIOCGIFSPEEDV1: {
4103 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
4104 struct pf_ifspeed_v1 ps;
4105 struct ifnet *ifp;
4106
4107 if (psp->ifname[0] == '\0') {
4108 error = EINVAL;
4109 break;
4110 }
4111
4112 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4113 if (error != 0)
4114 break;
4115 ifp = ifunit(ps.ifname);
4116 if (ifp != NULL) {
4117 psp->baudrate32 =
4118 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4119 if (cmd == DIOCGIFSPEEDV1)
4120 psp->baudrate = ifp->if_baudrate;
4121 } else {
4122 error = EINVAL;
4123 }
4124 break;
4125 }
4126
4127 #ifdef ALTQ
4128 case DIOCSTARTALTQ: {
4129 struct pf_altq *altq;
4130
4131 PF_RULES_WLOCK();
4132 /* enable all altq interfaces on active list */
4133 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4134 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4135 error = pf_enable_altq(altq);
4136 if (error != 0)
4137 break;
4138 }
4139 }
4140 if (error == 0)
4141 V_pf_altq_running = 1;
4142 PF_RULES_WUNLOCK();
4143 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4144 break;
4145 }
4146
4147 case DIOCSTOPALTQ: {
4148 struct pf_altq *altq;
4149
4150 PF_RULES_WLOCK();
4151 /* disable all altq interfaces on active list */
4152 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4153 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4154 error = pf_disable_altq(altq);
4155 if (error != 0)
4156 break;
4157 }
4158 }
4159 if (error == 0)
4160 V_pf_altq_running = 0;
4161 PF_RULES_WUNLOCK();
4162 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4163 break;
4164 }
4165
4166 case DIOCADDALTQV0:
4167 case DIOCADDALTQV1: {
4168 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4169 struct pf_altq *altq, *a;
4170 struct ifnet *ifp;
4171
4172 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4173 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4174 if (error)
4175 break;
4176 altq->local_flags = 0;
4177
4178 PF_RULES_WLOCK();
4179 if (pa->ticket != V_ticket_altqs_inactive) {
4180 PF_RULES_WUNLOCK();
4181 free(altq, M_PFALTQ);
4182 error = EBUSY;
4183 break;
4184 }
4185
4186 /*
4187 * if this is for a queue, find the discipline and
4188 * copy the necessary fields
4189 */
4190 if (altq->qname[0] != 0) {
4191 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4192 PF_RULES_WUNLOCK();
4193 error = EBUSY;
4194 free(altq, M_PFALTQ);
4195 break;
4196 }
4197 altq->altq_disc = NULL;
4198 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4199 if (strncmp(a->ifname, altq->ifname,
4200 IFNAMSIZ) == 0) {
4201 altq->altq_disc = a->altq_disc;
4202 break;
4203 }
4204 }
4205 }
4206
4207 if ((ifp = ifunit(altq->ifname)) == NULL)
4208 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4209 else
4210 error = altq_add(ifp, altq);
4211
4212 if (error) {
4213 PF_RULES_WUNLOCK();
4214 free(altq, M_PFALTQ);
4215 break;
4216 }
4217
4218 if (altq->qname[0] != 0)
4219 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4220 else
4221 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4222 /* version error check done on import above */
4223 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4224 PF_RULES_WUNLOCK();
4225 break;
4226 }
4227
4228 case DIOCGETALTQSV0:
4229 case DIOCGETALTQSV1: {
4230 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4231 struct pf_altq *altq;
4232
4233 PF_RULES_RLOCK();
4234 pa->nr = 0;
4235 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4236 pa->nr++;
4237 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4238 pa->nr++;
4239 pa->ticket = V_ticket_altqs_active;
4240 PF_RULES_RUNLOCK();
4241 break;
4242 }
4243
4244 case DIOCGETALTQV0:
4245 case DIOCGETALTQV1: {
4246 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4247 struct pf_altq *altq;
4248
4249 PF_RULES_RLOCK();
4250 if (pa->ticket != V_ticket_altqs_active) {
4251 PF_RULES_RUNLOCK();
4252 error = EBUSY;
4253 break;
4254 }
4255 altq = pf_altq_get_nth_active(pa->nr);
4256 if (altq == NULL) {
4257 PF_RULES_RUNLOCK();
4258 error = EBUSY;
4259 break;
4260 }
4261 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4262 PF_RULES_RUNLOCK();
4263 break;
4264 }
4265
4266 case DIOCCHANGEALTQV0:
4267 case DIOCCHANGEALTQV1:
4268 /* CHANGEALTQ not supported yet! */
4269 error = ENODEV;
4270 break;
4271
4272 case DIOCGETQSTATSV0:
4273 case DIOCGETQSTATSV1: {
4274 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
4275 struct pf_altq *altq;
4276 int nbytes;
4277 u_int32_t version;
4278
4279 PF_RULES_RLOCK();
4280 if (pq->ticket != V_ticket_altqs_active) {
4281 PF_RULES_RUNLOCK();
4282 error = EBUSY;
4283 break;
4284 }
4285 nbytes = pq->nbytes;
4286 altq = pf_altq_get_nth_active(pq->nr);
4287 if (altq == NULL) {
4288 PF_RULES_RUNLOCK();
4289 error = EBUSY;
4290 break;
4291 }
4292
4293 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4294 PF_RULES_RUNLOCK();
4295 error = ENXIO;
4296 break;
4297 }
4298 PF_RULES_RUNLOCK();
4299 if (cmd == DIOCGETQSTATSV0)
4300 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
4301 else
4302 version = pq->version;
4303 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4304 if (error == 0) {
4305 pq->scheduler = altq->scheduler;
4306 pq->nbytes = nbytes;
4307 }
4308 break;
4309 }
4310 #endif /* ALTQ */
4311
4312 case DIOCBEGINADDRS: {
4313 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4314
4315 error = pf_ioctl_begin_addrs(&pp->ticket);
4316 break;
4317 }
4318
4319 case DIOCADDADDR: {
4320 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4321
4322 error = pf_ioctl_add_addr(pp);
4323 break;
4324 }
4325
4326 case DIOCGETADDRS: {
4327 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4328
4329 error = pf_ioctl_get_addrs(pp);
4330 break;
4331 }
4332
4333 case DIOCGETADDR: {
4334 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4335
4336 error = pf_ioctl_get_addr(pp);
4337 break;
4338 }
4339
4340 case DIOCCHANGEADDR: {
4341 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
4342 struct pf_kpool *pool;
4343 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
4344 struct pf_kruleset *ruleset;
4345 struct pfi_kkif *kif = NULL;
4346
4347 pca->anchor[sizeof(pca->anchor) - 1] = 0;
4348
4349 if (pca->action < PF_CHANGE_ADD_HEAD ||
4350 pca->action > PF_CHANGE_REMOVE) {
4351 error = EINVAL;
4352 break;
4353 }
4354 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4355 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4356 pca->addr.addr.type != PF_ADDR_TABLE) {
4357 error = EINVAL;
4358 break;
4359 }
4360 if (pca->addr.addr.p.dyn != NULL) {
4361 error = EINVAL;
4362 break;
4363 }
4364
4365 if (pca->action != PF_CHANGE_REMOVE) {
4366 #ifndef INET
4367 if (pca->af == AF_INET) {
4368 error = EAFNOSUPPORT;
4369 break;
4370 }
4371 #endif /* INET */
4372 #ifndef INET6
4373 if (pca->af == AF_INET6) {
4374 error = EAFNOSUPPORT;
4375 break;
4376 }
4377 #endif /* INET6 */
4378 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4379 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4380 if (newpa->ifname[0])
4381 kif = pf_kkif_create(M_WAITOK);
4382 newpa->kif = NULL;
4383 }
4384 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4385 PF_RULES_WLOCK();
4386 ruleset = pf_find_kruleset(pca->anchor);
4387 if (ruleset == NULL)
4388 ERROUT(EBUSY);
4389
4390 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4391 pca->r_num, pca->r_last, 1, 1);
4392 if (pool == NULL)
4393 ERROUT(EBUSY);
4394
4395 if (pca->action != PF_CHANGE_REMOVE) {
4396 if (newpa->ifname[0]) {
4397 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4398 pfi_kkif_ref(newpa->kif);
4399 kif = NULL;
4400 }
4401
4402 switch (newpa->addr.type) {
4403 case PF_ADDR_DYNIFTL:
4404 error = pfi_dynaddr_setup(&newpa->addr,
4405 pca->af);
4406 break;
4407 case PF_ADDR_TABLE:
4408 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4409 newpa->addr.v.tblname);
4410 if (newpa->addr.p.tbl == NULL)
4411 error = ENOMEM;
4412 break;
4413 }
4414 if (error)
4415 goto DIOCCHANGEADDR_error;
4416 }
4417
4418 switch (pca->action) {
4419 case PF_CHANGE_ADD_HEAD:
4420 oldpa = TAILQ_FIRST(&pool->list);
4421 break;
4422 case PF_CHANGE_ADD_TAIL:
4423 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4424 break;
4425 default:
4426 oldpa = TAILQ_FIRST(&pool->list);
4427 for (int i = 0; oldpa && i < pca->nr; i++)
4428 oldpa = TAILQ_NEXT(oldpa, entries);
4429
4430 if (oldpa == NULL)
4431 ERROUT(EINVAL);
4432 }
4433
4434 if (pca->action == PF_CHANGE_REMOVE) {
4435 TAILQ_REMOVE(&pool->list, oldpa, entries);
4436 switch (oldpa->addr.type) {
4437 case PF_ADDR_DYNIFTL:
4438 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4439 break;
4440 case PF_ADDR_TABLE:
4441 pfr_detach_table(oldpa->addr.p.tbl);
4442 break;
4443 }
4444 if (oldpa->kif)
4445 pfi_kkif_unref(oldpa->kif);
4446 free(oldpa, M_PFRULE);
4447 } else {
4448 if (oldpa == NULL)
4449 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4450 else if (pca->action == PF_CHANGE_ADD_HEAD ||
4451 pca->action == PF_CHANGE_ADD_BEFORE)
4452 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4453 else
4454 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4455 newpa, entries);
4456 }
4457
4458 pool->cur = TAILQ_FIRST(&pool->list);
4459 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4460 PF_RULES_WUNLOCK();
4461 break;
4462
4463 #undef ERROUT
4464 DIOCCHANGEADDR_error:
4465 if (newpa != NULL) {
4466 if (newpa->kif)
4467 pfi_kkif_unref(newpa->kif);
4468 free(newpa, M_PFRULE);
4469 }
4470 PF_RULES_WUNLOCK();
4471 pf_kkif_free(kif);
4472 break;
4473 }
4474
4475 case DIOCGETRULESETS: {
4476 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4477 struct pf_kruleset *ruleset;
4478 struct pf_kanchor *anchor;
4479
4480 pr->path[sizeof(pr->path) - 1] = 0;
4481
4482 PF_RULES_RLOCK();
4483 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4484 PF_RULES_RUNLOCK();
4485 error = ENOENT;
4486 break;
4487 }
4488 pr->nr = 0;
4489 if (ruleset->anchor == NULL) {
4490 /* XXX kludge for pf_main_ruleset */
4491 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4492 if (anchor->parent == NULL)
4493 pr->nr++;
4494 } else {
4495 RB_FOREACH(anchor, pf_kanchor_node,
4496 &ruleset->anchor->children)
4497 pr->nr++;
4498 }
4499 PF_RULES_RUNLOCK();
4500 break;
4501 }
4502
4503 case DIOCGETRULESET: {
4504 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4505 struct pf_kruleset *ruleset;
4506 struct pf_kanchor *anchor;
4507 u_int32_t nr = 0;
4508
4509 pr->path[sizeof(pr->path) - 1] = 0;
4510
4511 PF_RULES_RLOCK();
4512 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4513 PF_RULES_RUNLOCK();
4514 error = ENOENT;
4515 break;
4516 }
4517 pr->name[0] = 0;
4518 if (ruleset->anchor == NULL) {
4519 /* XXX kludge for pf_main_ruleset */
4520 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4521 if (anchor->parent == NULL && nr++ == pr->nr) {
4522 strlcpy(pr->name, anchor->name,
4523 sizeof(pr->name));
4524 break;
4525 }
4526 } else {
4527 RB_FOREACH(anchor, pf_kanchor_node,
4528 &ruleset->anchor->children)
4529 if (nr++ == pr->nr) {
4530 strlcpy(pr->name, anchor->name,
4531 sizeof(pr->name));
4532 break;
4533 }
4534 }
4535 if (!pr->name[0])
4536 error = EBUSY;
4537 PF_RULES_RUNLOCK();
4538 break;
4539 }
4540
4541 case DIOCRCLRTABLES: {
4542 struct pfioc_table *io = (struct pfioc_table *)addr;
4543
4544 if (io->pfrio_esize != 0) {
4545 error = ENODEV;
4546 break;
4547 }
4548 PF_RULES_WLOCK();
4549 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4550 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4551 PF_RULES_WUNLOCK();
4552 break;
4553 }
4554
4555 case DIOCRADDTABLES: {
4556 struct pfioc_table *io = (struct pfioc_table *)addr;
4557 struct pfr_table *pfrts;
4558 size_t totlen;
4559
4560 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4561 error = ENODEV;
4562 break;
4563 }
4564
4565 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4566 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4567 error = ENOMEM;
4568 break;
4569 }
4570
4571 totlen = io->pfrio_size * sizeof(struct pfr_table);
4572 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4573 M_TEMP, M_WAITOK);
4574 error = copyin(io->pfrio_buffer, pfrts, totlen);
4575 if (error) {
4576 free(pfrts, M_TEMP);
4577 break;
4578 }
4579 PF_RULES_WLOCK();
4580 error = pfr_add_tables(pfrts, io->pfrio_size,
4581 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4582 PF_RULES_WUNLOCK();
4583 free(pfrts, M_TEMP);
4584 break;
4585 }
4586
4587 case DIOCRDELTABLES: {
4588 struct pfioc_table *io = (struct pfioc_table *)addr;
4589 struct pfr_table *pfrts;
4590 size_t totlen;
4591
4592 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4593 error = ENODEV;
4594 break;
4595 }
4596
4597 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4598 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4599 error = ENOMEM;
4600 break;
4601 }
4602
4603 totlen = io->pfrio_size * sizeof(struct pfr_table);
4604 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4605 M_TEMP, M_WAITOK);
4606 error = copyin(io->pfrio_buffer, pfrts, totlen);
4607 if (error) {
4608 free(pfrts, M_TEMP);
4609 break;
4610 }
4611 PF_RULES_WLOCK();
4612 error = pfr_del_tables(pfrts, io->pfrio_size,
4613 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4614 PF_RULES_WUNLOCK();
4615 free(pfrts, M_TEMP);
4616 break;
4617 }
4618
4619 case DIOCRGETTABLES: {
4620 struct pfioc_table *io = (struct pfioc_table *)addr;
4621 struct pfr_table *pfrts;
4622 size_t totlen;
4623 int n;
4624
4625 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4626 error = ENODEV;
4627 break;
4628 }
4629 PF_RULES_RLOCK();
4630 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4631 if (n < 0) {
4632 PF_RULES_RUNLOCK();
4633 error = EINVAL;
4634 break;
4635 }
4636 io->pfrio_size = min(io->pfrio_size, n);
4637
4638 totlen = io->pfrio_size * sizeof(struct pfr_table);
4639
4640 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4641 M_TEMP, M_NOWAIT | M_ZERO);
4642 if (pfrts == NULL) {
4643 error = ENOMEM;
4644 PF_RULES_RUNLOCK();
4645 break;
4646 }
4647 error = pfr_get_tables(&io->pfrio_table, pfrts,
4648 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4649 PF_RULES_RUNLOCK();
4650 if (error == 0)
4651 error = copyout(pfrts, io->pfrio_buffer, totlen);
4652 free(pfrts, M_TEMP);
4653 break;
4654 }
4655
4656 case DIOCRGETTSTATS: {
4657 struct pfioc_table *io = (struct pfioc_table *)addr;
4658 struct pfr_tstats *pfrtstats;
4659 size_t totlen;
4660 int n;
4661
4662 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4663 error = ENODEV;
4664 break;
4665 }
4666 PF_TABLE_STATS_LOCK();
4667 PF_RULES_RLOCK();
4668 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4669 if (n < 0) {
4670 PF_RULES_RUNLOCK();
4671 PF_TABLE_STATS_UNLOCK();
4672 error = EINVAL;
4673 break;
4674 }
4675 io->pfrio_size = min(io->pfrio_size, n);
4676
4677 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4678 pfrtstats = mallocarray(io->pfrio_size,
4679 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4680 if (pfrtstats == NULL) {
4681 error = ENOMEM;
4682 PF_RULES_RUNLOCK();
4683 PF_TABLE_STATS_UNLOCK();
4684 break;
4685 }
4686 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4687 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4688 PF_RULES_RUNLOCK();
4689 PF_TABLE_STATS_UNLOCK();
4690 if (error == 0)
4691 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4692 free(pfrtstats, M_TEMP);
4693 break;
4694 }
4695
4696 case DIOCRCLRTSTATS: {
4697 struct pfioc_table *io = (struct pfioc_table *)addr;
4698 struct pfr_table *pfrts;
4699 size_t totlen;
4700
4701 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4702 error = ENODEV;
4703 break;
4704 }
4705
4706 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4707 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4708 /* We used to count tables and use the minimum required
4709 * size, so we didn't fail on overly large requests.
4710 * Keep doing so. */
4711 io->pfrio_size = pf_ioctl_maxcount;
4712 break;
4713 }
4714
4715 totlen = io->pfrio_size * sizeof(struct pfr_table);
4716 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4717 M_TEMP, M_WAITOK);
4718 error = copyin(io->pfrio_buffer, pfrts, totlen);
4719 if (error) {
4720 free(pfrts, M_TEMP);
4721 break;
4722 }
4723
4724 PF_TABLE_STATS_LOCK();
4725 PF_RULES_RLOCK();
4726 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4727 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4728 PF_RULES_RUNLOCK();
4729 PF_TABLE_STATS_UNLOCK();
4730 free(pfrts, M_TEMP);
4731 break;
4732 }
4733
4734 case DIOCRSETTFLAGS: {
4735 struct pfioc_table *io = (struct pfioc_table *)addr;
4736 struct pfr_table *pfrts;
4737 size_t totlen;
4738 int n;
4739
4740 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4741 error = ENODEV;
4742 break;
4743 }
4744
4745 PF_RULES_RLOCK();
4746 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4747 if (n < 0) {
4748 PF_RULES_RUNLOCK();
4749 error = EINVAL;
4750 break;
4751 }
4752
4753 io->pfrio_size = min(io->pfrio_size, n);
4754 PF_RULES_RUNLOCK();
4755
4756 totlen = io->pfrio_size * sizeof(struct pfr_table);
4757 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4758 M_TEMP, M_WAITOK);
4759 error = copyin(io->pfrio_buffer, pfrts, totlen);
4760 if (error) {
4761 free(pfrts, M_TEMP);
4762 break;
4763 }
4764 PF_RULES_WLOCK();
4765 error = pfr_set_tflags(pfrts, io->pfrio_size,
4766 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4767 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4768 PF_RULES_WUNLOCK();
4769 free(pfrts, M_TEMP);
4770 break;
4771 }
4772
4773 case DIOCRCLRADDRS: {
4774 struct pfioc_table *io = (struct pfioc_table *)addr;
4775
4776 if (io->pfrio_esize != 0) {
4777 error = ENODEV;
4778 break;
4779 }
4780 PF_RULES_WLOCK();
4781 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4782 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4783 PF_RULES_WUNLOCK();
4784 break;
4785 }
4786
4787 case DIOCRADDADDRS: {
4788 struct pfioc_table *io = (struct pfioc_table *)addr;
4789 struct pfr_addr *pfras;
4790 size_t totlen;
4791
4792 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4793 error = ENODEV;
4794 break;
4795 }
4796 if (io->pfrio_size < 0 ||
4797 io->pfrio_size > pf_ioctl_maxcount ||
4798 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4799 error = EINVAL;
4800 break;
4801 }
4802 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4803 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4804 M_TEMP, M_WAITOK);
4805 error = copyin(io->pfrio_buffer, pfras, totlen);
4806 if (error) {
4807 free(pfras, M_TEMP);
4808 break;
4809 }
4810 PF_RULES_WLOCK();
4811 error = pfr_add_addrs(&io->pfrio_table, pfras,
4812 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4813 PFR_FLAG_USERIOCTL);
4814 PF_RULES_WUNLOCK();
4815 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4816 error = copyout(pfras, io->pfrio_buffer, totlen);
4817 free(pfras, M_TEMP);
4818 break;
4819 }
4820
4821 case DIOCRDELADDRS: {
4822 struct pfioc_table *io = (struct pfioc_table *)addr;
4823 struct pfr_addr *pfras;
4824 size_t totlen;
4825
4826 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4827 error = ENODEV;
4828 break;
4829 }
4830 if (io->pfrio_size < 0 ||
4831 io->pfrio_size > pf_ioctl_maxcount ||
4832 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4833 error = EINVAL;
4834 break;
4835 }
4836 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4837 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4838 M_TEMP, M_WAITOK);
4839 error = copyin(io->pfrio_buffer, pfras, totlen);
4840 if (error) {
4841 free(pfras, M_TEMP);
4842 break;
4843 }
4844 PF_RULES_WLOCK();
4845 error = pfr_del_addrs(&io->pfrio_table, pfras,
4846 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4847 PFR_FLAG_USERIOCTL);
4848 PF_RULES_WUNLOCK();
4849 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4850 error = copyout(pfras, io->pfrio_buffer, totlen);
4851 free(pfras, M_TEMP);
4852 break;
4853 }
4854
4855 case DIOCRSETADDRS: {
4856 struct pfioc_table *io = (struct pfioc_table *)addr;
4857 struct pfr_addr *pfras;
4858 size_t totlen, count;
4859
4860 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4861 error = ENODEV;
4862 break;
4863 }
4864 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4865 error = EINVAL;
4866 break;
4867 }
4868 count = max(io->pfrio_size, io->pfrio_size2);
4869 if (count > pf_ioctl_maxcount ||
4870 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4871 error = EINVAL;
4872 break;
4873 }
4874 totlen = count * sizeof(struct pfr_addr);
4875 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4876 M_WAITOK);
4877 error = copyin(io->pfrio_buffer, pfras, totlen);
4878 if (error) {
4879 free(pfras, M_TEMP);
4880 break;
4881 }
4882 PF_RULES_WLOCK();
4883 error = pfr_set_addrs(&io->pfrio_table, pfras,
4884 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4885 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4886 PFR_FLAG_USERIOCTL, 0);
4887 PF_RULES_WUNLOCK();
4888 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4889 error = copyout(pfras, io->pfrio_buffer, totlen);
4890 free(pfras, M_TEMP);
4891 break;
4892 }
4893
4894 case DIOCRGETADDRS: {
4895 struct pfioc_table *io = (struct pfioc_table *)addr;
4896 struct pfr_addr *pfras;
4897 size_t totlen;
4898
4899 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4900 error = ENODEV;
4901 break;
4902 }
4903 if (io->pfrio_size < 0 ||
4904 io->pfrio_size > pf_ioctl_maxcount ||
4905 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4906 error = EINVAL;
4907 break;
4908 }
4909 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4910 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4911 M_TEMP, M_WAITOK | M_ZERO);
4912 PF_RULES_RLOCK();
4913 error = pfr_get_addrs(&io->pfrio_table, pfras,
4914 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4915 PF_RULES_RUNLOCK();
4916 if (error == 0)
4917 error = copyout(pfras, io->pfrio_buffer, totlen);
4918 free(pfras, M_TEMP);
4919 break;
4920 }
4921
4922 case DIOCRGETASTATS: {
4923 struct pfioc_table *io = (struct pfioc_table *)addr;
4924 struct pfr_astats *pfrastats;
4925 size_t totlen;
4926
4927 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4928 error = ENODEV;
4929 break;
4930 }
4931 if (io->pfrio_size < 0 ||
4932 io->pfrio_size > pf_ioctl_maxcount ||
4933 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4934 error = EINVAL;
4935 break;
4936 }
4937 totlen = io->pfrio_size * sizeof(struct pfr_astats);
4938 pfrastats = mallocarray(io->pfrio_size,
4939 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4940 PF_RULES_RLOCK();
4941 error = pfr_get_astats(&io->pfrio_table, pfrastats,
4942 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4943 PF_RULES_RUNLOCK();
4944 if (error == 0)
4945 error = copyout(pfrastats, io->pfrio_buffer, totlen);
4946 free(pfrastats, M_TEMP);
4947 break;
4948 }
4949
4950 case DIOCRCLRASTATS: {
4951 struct pfioc_table *io = (struct pfioc_table *)addr;
4952 struct pfr_addr *pfras;
4953 size_t totlen;
4954
4955 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4956 error = ENODEV;
4957 break;
4958 }
4959 if (io->pfrio_size < 0 ||
4960 io->pfrio_size > pf_ioctl_maxcount ||
4961 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4962 error = EINVAL;
4963 break;
4964 }
4965 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4966 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4967 M_TEMP, M_WAITOK);
4968 error = copyin(io->pfrio_buffer, pfras, totlen);
4969 if (error) {
4970 free(pfras, M_TEMP);
4971 break;
4972 }
4973 PF_RULES_WLOCK();
4974 error = pfr_clr_astats(&io->pfrio_table, pfras,
4975 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4976 PFR_FLAG_USERIOCTL);
4977 PF_RULES_WUNLOCK();
4978 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4979 error = copyout(pfras, io->pfrio_buffer, totlen);
4980 free(pfras, M_TEMP);
4981 break;
4982 }
4983
4984 case DIOCRTSTADDRS: {
4985 struct pfioc_table *io = (struct pfioc_table *)addr;
4986 struct pfr_addr *pfras;
4987 size_t totlen;
4988
4989 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4990 error = ENODEV;
4991 break;
4992 }
4993 if (io->pfrio_size < 0 ||
4994 io->pfrio_size > pf_ioctl_maxcount ||
4995 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4996 error = EINVAL;
4997 break;
4998 }
4999 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5000 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5001 M_TEMP, M_WAITOK);
5002 error = copyin(io->pfrio_buffer, pfras, totlen);
5003 if (error) {
5004 free(pfras, M_TEMP);
5005 break;
5006 }
5007 PF_RULES_RLOCK();
5008 error = pfr_tst_addrs(&io->pfrio_table, pfras,
5009 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5010 PFR_FLAG_USERIOCTL);
5011 PF_RULES_RUNLOCK();
5012 if (error == 0)
5013 error = copyout(pfras, io->pfrio_buffer, totlen);
5014 free(pfras, M_TEMP);
5015 break;
5016 }
5017
5018 case DIOCRINADEFINE: {
5019 struct pfioc_table *io = (struct pfioc_table *)addr;
5020 struct pfr_addr *pfras;
5021 size_t totlen;
5022
5023 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5024 error = ENODEV;
5025 break;
5026 }
5027 if (io->pfrio_size < 0 ||
5028 io->pfrio_size > pf_ioctl_maxcount ||
5029 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5030 error = EINVAL;
5031 break;
5032 }
5033 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5034 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5035 M_TEMP, M_WAITOK);
5036 error = copyin(io->pfrio_buffer, pfras, totlen);
5037 if (error) {
5038 free(pfras, M_TEMP);
5039 break;
5040 }
5041 PF_RULES_WLOCK();
5042 error = pfr_ina_define(&io->pfrio_table, pfras,
5043 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5044 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5045 PF_RULES_WUNLOCK();
5046 free(pfras, M_TEMP);
5047 break;
5048 }
5049
5050 case DIOCOSFPADD: {
5051 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5052 PF_RULES_WLOCK();
5053 error = pf_osfp_add(io);
5054 PF_RULES_WUNLOCK();
5055 break;
5056 }
5057
5058 case DIOCOSFPGET: {
5059 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5060 PF_RULES_RLOCK();
5061 error = pf_osfp_get(io);
5062 PF_RULES_RUNLOCK();
5063 break;
5064 }
5065
5066 case DIOCXBEGIN: {
5067 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5068 struct pfioc_trans_e *ioes, *ioe;
5069 size_t totlen;
5070 int i;
5071
5072 if (io->esize != sizeof(*ioe)) {
5073 error = ENODEV;
5074 break;
5075 }
5076 if (io->size < 0 ||
5077 io->size > pf_ioctl_maxcount ||
5078 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5079 error = EINVAL;
5080 break;
5081 }
5082 totlen = sizeof(struct pfioc_trans_e) * io->size;
5083 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5084 M_TEMP, M_WAITOK);
5085 error = copyin(io->array, ioes, totlen);
5086 if (error) {
5087 free(ioes, M_TEMP);
5088 break;
5089 }
5090 /* Ensure there's no more ethernet rules to clean up. */
5091 NET_EPOCH_DRAIN_CALLBACKS();
5092 PF_RULES_WLOCK();
5093 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5094 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5095 switch (ioe->rs_num) {
5096 case PF_RULESET_ETH:
5097 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5098 PF_RULES_WUNLOCK();
5099 free(ioes, M_TEMP);
5100 goto fail;
5101 }
5102 break;
5103 #ifdef ALTQ
5104 case PF_RULESET_ALTQ:
5105 if (ioe->anchor[0]) {
5106 PF_RULES_WUNLOCK();
5107 free(ioes, M_TEMP);
5108 error = EINVAL;
5109 goto fail;
5110 }
5111 if ((error = pf_begin_altq(&ioe->ticket))) {
5112 PF_RULES_WUNLOCK();
5113 free(ioes, M_TEMP);
5114 goto fail;
5115 }
5116 break;
5117 #endif /* ALTQ */
5118 case PF_RULESET_TABLE:
5119 {
5120 struct pfr_table table;
5121
5122 bzero(&table, sizeof(table));
5123 strlcpy(table.pfrt_anchor, ioe->anchor,
5124 sizeof(table.pfrt_anchor));
5125 if ((error = pfr_ina_begin(&table,
5126 &ioe->ticket, NULL, 0))) {
5127 PF_RULES_WUNLOCK();
5128 free(ioes, M_TEMP);
5129 goto fail;
5130 }
5131 break;
5132 }
5133 default:
5134 if ((error = pf_begin_rules(&ioe->ticket,
5135 ioe->rs_num, ioe->anchor))) {
5136 PF_RULES_WUNLOCK();
5137 free(ioes, M_TEMP);
5138 goto fail;
5139 }
5140 break;
5141 }
5142 }
5143 PF_RULES_WUNLOCK();
5144 error = copyout(ioes, io->array, totlen);
5145 free(ioes, M_TEMP);
5146 break;
5147 }
5148
5149 case DIOCXROLLBACK: {
5150 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5151 struct pfioc_trans_e *ioe, *ioes;
5152 size_t totlen;
5153 int i;
5154
5155 if (io->esize != sizeof(*ioe)) {
5156 error = ENODEV;
5157 break;
5158 }
5159 if (io->size < 0 ||
5160 io->size > pf_ioctl_maxcount ||
5161 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5162 error = EINVAL;
5163 break;
5164 }
5165 totlen = sizeof(struct pfioc_trans_e) * io->size;
5166 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5167 M_TEMP, M_WAITOK);
5168 error = copyin(io->array, ioes, totlen);
5169 if (error) {
5170 free(ioes, M_TEMP);
5171 break;
5172 }
5173 PF_RULES_WLOCK();
5174 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5175 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5176 switch (ioe->rs_num) {
5177 case PF_RULESET_ETH:
5178 if ((error = pf_rollback_eth(ioe->ticket,
5179 ioe->anchor))) {
5180 PF_RULES_WUNLOCK();
5181 free(ioes, M_TEMP);
5182 goto fail; /* really bad */
5183 }
5184 break;
5185 #ifdef ALTQ
5186 case PF_RULESET_ALTQ:
5187 if (ioe->anchor[0]) {
5188 PF_RULES_WUNLOCK();
5189 free(ioes, M_TEMP);
5190 error = EINVAL;
5191 goto fail;
5192 }
5193 if ((error = pf_rollback_altq(ioe->ticket))) {
5194 PF_RULES_WUNLOCK();
5195 free(ioes, M_TEMP);
5196 goto fail; /* really bad */
5197 }
5198 break;
5199 #endif /* ALTQ */
5200 case PF_RULESET_TABLE:
5201 {
5202 struct pfr_table table;
5203
5204 bzero(&table, sizeof(table));
5205 strlcpy(table.pfrt_anchor, ioe->anchor,
5206 sizeof(table.pfrt_anchor));
5207 if ((error = pfr_ina_rollback(&table,
5208 ioe->ticket, NULL, 0))) {
5209 PF_RULES_WUNLOCK();
5210 free(ioes, M_TEMP);
5211 goto fail; /* really bad */
5212 }
5213 break;
5214 }
5215 default:
5216 if ((error = pf_rollback_rules(ioe->ticket,
5217 ioe->rs_num, ioe->anchor))) {
5218 PF_RULES_WUNLOCK();
5219 free(ioes, M_TEMP);
5220 goto fail; /* really bad */
5221 }
5222 break;
5223 }
5224 }
5225 PF_RULES_WUNLOCK();
5226 free(ioes, M_TEMP);
5227 break;
5228 }
5229
5230 case DIOCXCOMMIT: {
5231 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5232 struct pfioc_trans_e *ioe, *ioes;
5233 struct pf_kruleset *rs;
5234 struct pf_keth_ruleset *ers;
5235 size_t totlen;
5236 int i;
5237
5238 if (io->esize != sizeof(*ioe)) {
5239 error = ENODEV;
5240 break;
5241 }
5242
5243 if (io->size < 0 ||
5244 io->size > pf_ioctl_maxcount ||
5245 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5246 error = EINVAL;
5247 break;
5248 }
5249
5250 totlen = sizeof(struct pfioc_trans_e) * io->size;
5251 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5252 M_TEMP, M_WAITOK);
5253 error = copyin(io->array, ioes, totlen);
5254 if (error) {
5255 free(ioes, M_TEMP);
5256 break;
5257 }
5258 PF_RULES_WLOCK();
5259 /* First makes sure everything will succeed. */
5260 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5261 ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5262 switch (ioe->rs_num) {
5263 case PF_RULESET_ETH:
5264 ers = pf_find_keth_ruleset(ioe->anchor);
5265 if (ers == NULL || ioe->ticket == 0 ||
5266 ioe->ticket != ers->inactive.ticket) {
5267 PF_RULES_WUNLOCK();
5268 free(ioes, M_TEMP);
5269 error = EINVAL;
5270 goto fail;
5271 }
5272 break;
5273 #ifdef ALTQ
5274 case PF_RULESET_ALTQ:
5275 if (ioe->anchor[0]) {
5276 PF_RULES_WUNLOCK();
5277 free(ioes, M_TEMP);
5278 error = EINVAL;
5279 goto fail;
5280 }
5281 if (!V_altqs_inactive_open || ioe->ticket !=
5282 V_ticket_altqs_inactive) {
5283 PF_RULES_WUNLOCK();
5284 free(ioes, M_TEMP);
5285 error = EBUSY;
5286 goto fail;
5287 }
5288 break;
5289 #endif /* ALTQ */
5290 case PF_RULESET_TABLE:
5291 rs = pf_find_kruleset(ioe->anchor);
5292 if (rs == NULL || !rs->topen || ioe->ticket !=
5293 rs->tticket) {
5294 PF_RULES_WUNLOCK();
5295 free(ioes, M_TEMP);
5296 error = EBUSY;
5297 goto fail;
5298 }
5299 break;
5300 default:
5301 if (ioe->rs_num < 0 || ioe->rs_num >=
5302 PF_RULESET_MAX) {
5303 PF_RULES_WUNLOCK();
5304 free(ioes, M_TEMP);
5305 error = EINVAL;
5306 goto fail;
5307 }
5308 rs = pf_find_kruleset(ioe->anchor);
5309 if (rs == NULL ||
5310 !rs->rules[ioe->rs_num].inactive.open ||
5311 rs->rules[ioe->rs_num].inactive.ticket !=
5312 ioe->ticket) {
5313 PF_RULES_WUNLOCK();
5314 free(ioes, M_TEMP);
5315 error = EBUSY;
5316 goto fail;
5317 }
5318 break;
5319 }
5320 }
5321 /* Now do the commit - no errors should happen here. */
5322 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5323 switch (ioe->rs_num) {
5324 case PF_RULESET_ETH:
5325 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5326 PF_RULES_WUNLOCK();
5327 free(ioes, M_TEMP);
5328 goto fail; /* really bad */
5329 }
5330 break;
5331 #ifdef ALTQ
5332 case PF_RULESET_ALTQ:
5333 if ((error = pf_commit_altq(ioe->ticket))) {
5334 PF_RULES_WUNLOCK();
5335 free(ioes, M_TEMP);
5336 goto fail; /* really bad */
5337 }
5338 break;
5339 #endif /* ALTQ */
5340 case PF_RULESET_TABLE:
5341 {
5342 struct pfr_table table;
5343
5344 bzero(&table, sizeof(table));
5345 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
5346 sizeof(table.pfrt_anchor));
5347 if ((error = pfr_ina_commit(&table,
5348 ioe->ticket, NULL, NULL, 0))) {
5349 PF_RULES_WUNLOCK();
5350 free(ioes, M_TEMP);
5351 goto fail; /* really bad */
5352 }
5353 break;
5354 }
5355 default:
5356 if ((error = pf_commit_rules(ioe->ticket,
5357 ioe->rs_num, ioe->anchor))) {
5358 PF_RULES_WUNLOCK();
5359 free(ioes, M_TEMP);
5360 goto fail; /* really bad */
5361 }
5362 break;
5363 }
5364 }
5365 PF_RULES_WUNLOCK();
5366
5367 /* Only hook into EtherNet taffic if we've got rules for it. */
5368 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5369 hook_pf_eth();
5370 else
5371 dehook_pf_eth();
5372
5373 free(ioes, M_TEMP);
5374 break;
5375 }
5376
5377 case DIOCGETSRCNODES: {
5378 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
5379 struct pf_srchash *sh;
5380 struct pf_ksrc_node *n;
5381 struct pf_src_node *p, *pstore;
5382 uint32_t i, nr = 0;
5383
5384 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5385 i++, sh++) {
5386 PF_HASHROW_LOCK(sh);
5387 LIST_FOREACH(n, &sh->nodes, entry)
5388 nr++;
5389 PF_HASHROW_UNLOCK(sh);
5390 }
5391
5392 psn->psn_len = min(psn->psn_len,
5393 sizeof(struct pf_src_node) * nr);
5394
5395 if (psn->psn_len == 0) {
5396 psn->psn_len = sizeof(struct pf_src_node) * nr;
5397 break;
5398 }
5399
5400 nr = 0;
5401
5402 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5403 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5404 i++, sh++) {
5405 PF_HASHROW_LOCK(sh);
5406 LIST_FOREACH(n, &sh->nodes, entry) {
5407
5408 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5409 break;
5410
5411 pf_src_node_copy(n, p);
5412
5413 p++;
5414 nr++;
5415 }
5416 PF_HASHROW_UNLOCK(sh);
5417 }
5418 error = copyout(pstore, psn->psn_src_nodes,
5419 sizeof(struct pf_src_node) * nr);
5420 if (error) {
5421 free(pstore, M_TEMP);
5422 break;
5423 }
5424 psn->psn_len = sizeof(struct pf_src_node) * nr;
5425 free(pstore, M_TEMP);
5426 break;
5427 }
5428
5429 case DIOCCLRSRCNODES: {
5430 pf_clear_srcnodes(NULL);
5431 pf_purge_expired_src_nodes();
5432 break;
5433 }
5434
5435 case DIOCKILLSRCNODES:
5436 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5437 break;
5438
5439 #ifdef COMPAT_FREEBSD13
5440 case DIOCKEEPCOUNTERS_FREEBSD13:
5441 #endif
5442 case DIOCKEEPCOUNTERS:
5443 error = pf_keepcounters((struct pfioc_nv *)addr);
5444 break;
5445
5446 case DIOCGETSYNCOOKIES:
5447 error = pf_get_syncookies((struct pfioc_nv *)addr);
5448 break;
5449
5450 case DIOCSETSYNCOOKIES:
5451 error = pf_set_syncookies((struct pfioc_nv *)addr);
5452 break;
5453
5454 case DIOCSETHOSTID: {
5455 u_int32_t *hostid = (u_int32_t *)addr;
5456
5457 PF_RULES_WLOCK();
5458 if (*hostid == 0)
5459 V_pf_status.hostid = arc4random();
5460 else
5461 V_pf_status.hostid = *hostid;
5462 PF_RULES_WUNLOCK();
5463 break;
5464 }
5465
5466 case DIOCOSFPFLUSH:
5467 PF_RULES_WLOCK();
5468 pf_osfp_flush();
5469 PF_RULES_WUNLOCK();
5470 break;
5471
5472 case DIOCIGETIFACES: {
5473 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5474 struct pfi_kif *ifstore;
5475 size_t bufsiz;
5476
5477 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5478 error = ENODEV;
5479 break;
5480 }
5481
5482 if (io->pfiio_size < 0 ||
5483 io->pfiio_size > pf_ioctl_maxcount ||
5484 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5485 error = EINVAL;
5486 break;
5487 }
5488
5489 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5490
5491 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5492 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5493 M_TEMP, M_WAITOK | M_ZERO);
5494
5495 PF_RULES_RLOCK();
5496 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5497 PF_RULES_RUNLOCK();
5498 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5499 free(ifstore, M_TEMP);
5500 break;
5501 }
5502
5503 case DIOCSETIFFLAG: {
5504 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5505
5506 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5507
5508 PF_RULES_WLOCK();
5509 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5510 PF_RULES_WUNLOCK();
5511 break;
5512 }
5513
5514 case DIOCCLRIFFLAG: {
5515 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5516
5517 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5518
5519 PF_RULES_WLOCK();
5520 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5521 PF_RULES_WUNLOCK();
5522 break;
5523 }
5524
5525 case DIOCSETREASS: {
5526 u_int32_t *reass = (u_int32_t *)addr;
5527
5528 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5529 /* Removal of DF flag without reassembly enabled is not a
5530 * valid combination. Disable reassembly in such case. */
5531 if (!(V_pf_status.reass & PF_REASS_ENABLED))
5532 V_pf_status.reass = 0;
5533 break;
5534 }
5535
5536 default:
5537 error = ENODEV;
5538 break;
5539 }
5540 fail:
5541 CURVNET_RESTORE();
5542
5543 #undef ERROUT_IOCTL
5544
5545 return (error);
5546 }
5547
5548 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5549 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5550 {
5551 bzero(sp, sizeof(union pfsync_state_union));
5552
5553 /* copy from state key */
5554 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5555 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5556 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5557 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5558 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5559 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5560 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5561 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5562 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5563 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5564
5565 /* copy from state */
5566 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5567 bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5568 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5569 sp->pfs_1301.expire = pf_state_expires(st);
5570 if (sp->pfs_1301.expire <= time_uptime)
5571 sp->pfs_1301.expire = htonl(0);
5572 else
5573 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5574
5575 sp->pfs_1301.direction = st->direction;
5576 sp->pfs_1301.log = st->act.log;
5577 sp->pfs_1301.timeout = st->timeout;
5578
5579 switch (msg_version) {
5580 case PFSYNC_MSG_VERSION_1301:
5581 sp->pfs_1301.state_flags = st->state_flags;
5582 break;
5583 case PFSYNC_MSG_VERSION_1400:
5584 sp->pfs_1400.state_flags = htons(st->state_flags);
5585 sp->pfs_1400.qid = htons(st->act.qid);
5586 sp->pfs_1400.pqid = htons(st->act.pqid);
5587 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5588 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5589 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5590 sp->pfs_1400.min_ttl = st->act.min_ttl;
5591 sp->pfs_1400.set_tos = st->act.set_tos;
5592 sp->pfs_1400.max_mss = htons(st->act.max_mss);
5593 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5594 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5595 sp->pfs_1400.rt = st->rt;
5596 if (st->rt_kif)
5597 strlcpy(sp->pfs_1400.rt_ifname,
5598 st->rt_kif->pfik_name,
5599 sizeof(sp->pfs_1400.rt_ifname));
5600 break;
5601 default:
5602 panic("%s: Unsupported pfsync_msg_version %d",
5603 __func__, msg_version);
5604 }
5605
5606 if (st->src_node)
5607 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5608 if (st->nat_src_node)
5609 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5610
5611 sp->pfs_1301.id = st->id;
5612 sp->pfs_1301.creatorid = st->creatorid;
5613 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5614 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5615
5616 if (st->rule.ptr == NULL)
5617 sp->pfs_1301.rule = htonl(-1);
5618 else
5619 sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5620 if (st->anchor.ptr == NULL)
5621 sp->pfs_1301.anchor = htonl(-1);
5622 else
5623 sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5624 if (st->nat_rule.ptr == NULL)
5625 sp->pfs_1301.nat_rule = htonl(-1);
5626 else
5627 sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5628
5629 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5630 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5631 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5632 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5633 }
5634
5635 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5636 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5637 {
5638 bzero(sp, sizeof(*sp));
5639
5640 sp->version = PF_STATE_VERSION;
5641
5642 /* copy from state key */
5643 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5644 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5645 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5646 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5647 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5648 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5649 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5650 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5651 sp->proto = st->key[PF_SK_WIRE]->proto;
5652 sp->af = st->key[PF_SK_WIRE]->af;
5653
5654 /* copy from state */
5655 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5656 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5657 sizeof(sp->orig_ifname));
5658 bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5659 sp->creation = htonl(time_uptime - (st->creation / 1000));
5660 sp->expire = pf_state_expires(st);
5661 if (sp->expire <= time_uptime)
5662 sp->expire = htonl(0);
5663 else
5664 sp->expire = htonl(sp->expire - time_uptime);
5665
5666 sp->direction = st->direction;
5667 sp->log = st->act.log;
5668 sp->timeout = st->timeout;
5669 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5670 sp->state_flags_compat = st->state_flags;
5671 sp->state_flags = htons(st->state_flags);
5672 if (st->src_node)
5673 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5674 if (st->nat_src_node)
5675 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5676
5677 sp->id = st->id;
5678 sp->creatorid = st->creatorid;
5679 pf_state_peer_hton(&st->src, &sp->src);
5680 pf_state_peer_hton(&st->dst, &sp->dst);
5681
5682 if (st->rule.ptr == NULL)
5683 sp->rule = htonl(-1);
5684 else
5685 sp->rule = htonl(st->rule.ptr->nr);
5686 if (st->anchor.ptr == NULL)
5687 sp->anchor = htonl(-1);
5688 else
5689 sp->anchor = htonl(st->anchor.ptr->nr);
5690 if (st->nat_rule.ptr == NULL)
5691 sp->nat_rule = htonl(-1);
5692 else
5693 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5694
5695 sp->packets[0] = st->packets[0];
5696 sp->packets[1] = st->packets[1];
5697 sp->bytes[0] = st->bytes[0];
5698 sp->bytes[1] = st->bytes[1];
5699
5700 sp->qid = htons(st->act.qid);
5701 sp->pqid = htons(st->act.pqid);
5702 sp->dnpipe = htons(st->act.dnpipe);
5703 sp->dnrpipe = htons(st->act.dnrpipe);
5704 sp->rtableid = htonl(st->act.rtableid);
5705 sp->min_ttl = st->act.min_ttl;
5706 sp->set_tos = st->act.set_tos;
5707 sp->max_mss = htons(st->act.max_mss);
5708 sp->rt = st->rt;
5709 if (st->rt_kif)
5710 strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5711 sizeof(sp->rt_ifname));
5712 sp->set_prio[0] = st->act.set_prio[0];
5713 sp->set_prio[1] = st->act.set_prio[1];
5714
5715 }
5716
5717 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5718 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5719 {
5720 struct pfr_ktable *kt;
5721
5722 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5723
5724 kt = aw->p.tbl;
5725 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5726 kt = kt->pfrkt_root;
5727 aw->p.tbl = NULL;
5728 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5729 kt->pfrkt_cnt : -1;
5730 }
5731
5732 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5733 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5734 size_t number, char **names)
5735 {
5736 nvlist_t *nvc;
5737
5738 nvc = nvlist_create(0);
5739 if (nvc == NULL)
5740 return (ENOMEM);
5741
5742 for (int i = 0; i < number; i++) {
5743 nvlist_append_number_array(nvc, "counters",
5744 counter_u64_fetch(counters[i]));
5745 nvlist_append_string_array(nvc, "names",
5746 names[i]);
5747 nvlist_append_number_array(nvc, "ids",
5748 i);
5749 }
5750 nvlist_add_nvlist(nvl, name, nvc);
5751 nvlist_destroy(nvc);
5752
5753 return (0);
5754 }
5755
5756 static int
pf_getstatus(struct pfioc_nv * nv)5757 pf_getstatus(struct pfioc_nv *nv)
5758 {
5759 nvlist_t *nvl = NULL, *nvc = NULL;
5760 void *nvlpacked = NULL;
5761 int error;
5762 struct pf_status s;
5763 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5764 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5765 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5766 PF_RULES_RLOCK_TRACKER;
5767
5768 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
5769
5770 PF_RULES_RLOCK();
5771
5772 nvl = nvlist_create(0);
5773 if (nvl == NULL)
5774 ERROUT(ENOMEM);
5775
5776 nvlist_add_bool(nvl, "running", V_pf_status.running);
5777 nvlist_add_number(nvl, "since", V_pf_status.since);
5778 nvlist_add_number(nvl, "debug", V_pf_status.debug);
5779 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5780 nvlist_add_number(nvl, "states", V_pf_status.states);
5781 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5782 nvlist_add_number(nvl, "reass", V_pf_status.reass);
5783 nvlist_add_bool(nvl, "syncookies_active",
5784 V_pf_status.syncookies_active);
5785 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5786
5787 /* counters */
5788 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5789 PFRES_MAX, pf_reasons);
5790 if (error != 0)
5791 ERROUT(error);
5792
5793 /* lcounters */
5794 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5795 KLCNT_MAX, pf_lcounter);
5796 if (error != 0)
5797 ERROUT(error);
5798
5799 /* fcounters */
5800 nvc = nvlist_create(0);
5801 if (nvc == NULL)
5802 ERROUT(ENOMEM);
5803
5804 for (int i = 0; i < FCNT_MAX; i++) {
5805 nvlist_append_number_array(nvc, "counters",
5806 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5807 nvlist_append_string_array(nvc, "names",
5808 pf_fcounter[i]);
5809 nvlist_append_number_array(nvc, "ids",
5810 i);
5811 }
5812 nvlist_add_nvlist(nvl, "fcounters", nvc);
5813 nvlist_destroy(nvc);
5814 nvc = NULL;
5815
5816 /* scounters */
5817 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5818 SCNT_MAX, pf_fcounter);
5819 if (error != 0)
5820 ERROUT(error);
5821
5822 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5823 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5824 PF_MD5_DIGEST_LENGTH);
5825
5826 pfi_update_status(V_pf_status.ifname, &s);
5827
5828 /* pcounters / bcounters */
5829 for (int i = 0; i < 2; i++) {
5830 for (int j = 0; j < 2; j++) {
5831 for (int k = 0; k < 2; k++) {
5832 nvlist_append_number_array(nvl, "pcounters",
5833 s.pcounters[i][j][k]);
5834 }
5835 nvlist_append_number_array(nvl, "bcounters",
5836 s.bcounters[i][j]);
5837 }
5838 }
5839
5840 nvlpacked = nvlist_pack(nvl, &nv->len);
5841 if (nvlpacked == NULL)
5842 ERROUT(ENOMEM);
5843
5844 if (nv->size == 0)
5845 ERROUT(0);
5846 else if (nv->size < nv->len)
5847 ERROUT(ENOSPC);
5848
5849 PF_RULES_RUNLOCK();
5850 error = copyout(nvlpacked, nv->data, nv->len);
5851 goto done;
5852
5853 #undef ERROUT
5854 errout:
5855 PF_RULES_RUNLOCK();
5856 done:
5857 free(nvlpacked, M_NVLIST);
5858 nvlist_destroy(nvc);
5859 nvlist_destroy(nvl);
5860
5861 return (error);
5862 }
5863
5864 /*
5865 * XXX - Check for version mismatch!!!
5866 */
5867 static void
pf_clear_all_states(void)5868 pf_clear_all_states(void)
5869 {
5870 struct epoch_tracker et;
5871 struct pf_kstate *s;
5872 u_int i;
5873
5874 NET_EPOCH_ENTER(et);
5875 for (i = 0; i <= V_pf_hashmask; i++) {
5876 struct pf_idhash *ih = &V_pf_idhash[i];
5877 relock:
5878 PF_HASHROW_LOCK(ih);
5879 LIST_FOREACH(s, &ih->states, entry) {
5880 s->timeout = PFTM_PURGE;
5881 /* Don't send out individual delete messages. */
5882 s->state_flags |= PFSTATE_NOSYNC;
5883 pf_unlink_state(s);
5884 goto relock;
5885 }
5886 PF_HASHROW_UNLOCK(ih);
5887 }
5888 NET_EPOCH_EXIT(et);
5889 }
5890
5891 static int
pf_clear_tables(void)5892 pf_clear_tables(void)
5893 {
5894 struct pfioc_table io;
5895 int error;
5896
5897 bzero(&io, sizeof(io));
5898 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5899
5900 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5901 io.pfrio_flags);
5902
5903 return (error);
5904 }
5905
5906 static void
pf_clear_srcnodes(struct pf_ksrc_node * n)5907 pf_clear_srcnodes(struct pf_ksrc_node *n)
5908 {
5909 struct pf_kstate *s;
5910 int i;
5911
5912 for (i = 0; i <= V_pf_hashmask; i++) {
5913 struct pf_idhash *ih = &V_pf_idhash[i];
5914
5915 PF_HASHROW_LOCK(ih);
5916 LIST_FOREACH(s, &ih->states, entry) {
5917 if (n == NULL || n == s->src_node)
5918 s->src_node = NULL;
5919 if (n == NULL || n == s->nat_src_node)
5920 s->nat_src_node = NULL;
5921 }
5922 PF_HASHROW_UNLOCK(ih);
5923 }
5924
5925 if (n == NULL) {
5926 struct pf_srchash *sh;
5927
5928 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5929 i++, sh++) {
5930 PF_HASHROW_LOCK(sh);
5931 LIST_FOREACH(n, &sh->nodes, entry) {
5932 n->expire = 1;
5933 n->states = 0;
5934 }
5935 PF_HASHROW_UNLOCK(sh);
5936 }
5937 } else {
5938 /* XXX: hash slot should already be locked here. */
5939 n->expire = 1;
5940 n->states = 0;
5941 }
5942 }
5943
5944 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)5945 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5946 {
5947 struct pf_ksrc_node_list kill;
5948
5949 LIST_INIT(&kill);
5950 for (int i = 0; i <= V_pf_srchashmask; i++) {
5951 struct pf_srchash *sh = &V_pf_srchash[i];
5952 struct pf_ksrc_node *sn, *tmp;
5953
5954 PF_HASHROW_LOCK(sh);
5955 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5956 if (PF_MATCHA(psnk->psnk_src.neg,
5957 &psnk->psnk_src.addr.v.a.addr,
5958 &psnk->psnk_src.addr.v.a.mask,
5959 &sn->addr, sn->af) &&
5960 PF_MATCHA(psnk->psnk_dst.neg,
5961 &psnk->psnk_dst.addr.v.a.addr,
5962 &psnk->psnk_dst.addr.v.a.mask,
5963 &sn->raddr, sn->af)) {
5964 pf_unlink_src_node(sn);
5965 LIST_INSERT_HEAD(&kill, sn, entry);
5966 sn->expire = 1;
5967 }
5968 PF_HASHROW_UNLOCK(sh);
5969 }
5970
5971 for (int i = 0; i <= V_pf_hashmask; i++) {
5972 struct pf_idhash *ih = &V_pf_idhash[i];
5973 struct pf_kstate *s;
5974
5975 PF_HASHROW_LOCK(ih);
5976 LIST_FOREACH(s, &ih->states, entry) {
5977 if (s->src_node && s->src_node->expire == 1)
5978 s->src_node = NULL;
5979 if (s->nat_src_node && s->nat_src_node->expire == 1)
5980 s->nat_src_node = NULL;
5981 }
5982 PF_HASHROW_UNLOCK(ih);
5983 }
5984
5985 psnk->psnk_killed = pf_free_src_nodes(&kill);
5986 }
5987
5988 static int
pf_keepcounters(struct pfioc_nv * nv)5989 pf_keepcounters(struct pfioc_nv *nv)
5990 {
5991 nvlist_t *nvl = NULL;
5992 void *nvlpacked = NULL;
5993 int error = 0;
5994
5995 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
5996
5997 if (nv->len > pf_ioctl_maxcount)
5998 ERROUT(ENOMEM);
5999
6000 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6001 error = copyin(nv->data, nvlpacked, nv->len);
6002 if (error)
6003 ERROUT(error);
6004
6005 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6006 if (nvl == NULL)
6007 ERROUT(EBADMSG);
6008
6009 if (! nvlist_exists_bool(nvl, "keep_counters"))
6010 ERROUT(EBADMSG);
6011
6012 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6013
6014 on_error:
6015 nvlist_destroy(nvl);
6016 free(nvlpacked, M_NVLIST);
6017 return (error);
6018 }
6019
6020 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6021 pf_clear_states(const struct pf_kstate_kill *kill)
6022 {
6023 struct pf_state_key_cmp match_key;
6024 struct pf_kstate *s;
6025 struct pfi_kkif *kif;
6026 int idx;
6027 unsigned int killed = 0, dir;
6028
6029 NET_EPOCH_ASSERT();
6030
6031 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6032 struct pf_idhash *ih = &V_pf_idhash[i];
6033
6034 relock_DIOCCLRSTATES:
6035 PF_HASHROW_LOCK(ih);
6036 LIST_FOREACH(s, &ih->states, entry) {
6037 /* For floating states look at the original kif. */
6038 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6039
6040 if (kill->psk_ifname[0] &&
6041 strcmp(kill->psk_ifname,
6042 kif->pfik_name))
6043 continue;
6044
6045 if (kill->psk_kill_match) {
6046 bzero(&match_key, sizeof(match_key));
6047
6048 if (s->direction == PF_OUT) {
6049 dir = PF_IN;
6050 idx = PF_SK_STACK;
6051 } else {
6052 dir = PF_OUT;
6053 idx = PF_SK_WIRE;
6054 }
6055
6056 match_key.af = s->key[idx]->af;
6057 match_key.proto = s->key[idx]->proto;
6058 PF_ACPY(&match_key.addr[0],
6059 &s->key[idx]->addr[1], match_key.af);
6060 match_key.port[0] = s->key[idx]->port[1];
6061 PF_ACPY(&match_key.addr[1],
6062 &s->key[idx]->addr[0], match_key.af);
6063 match_key.port[1] = s->key[idx]->port[0];
6064 }
6065
6066 /*
6067 * Don't send out individual
6068 * delete messages.
6069 */
6070 s->state_flags |= PFSTATE_NOSYNC;
6071 pf_unlink_state(s);
6072 killed++;
6073
6074 if (kill->psk_kill_match)
6075 killed += pf_kill_matching_state(&match_key,
6076 dir);
6077
6078 goto relock_DIOCCLRSTATES;
6079 }
6080 PF_HASHROW_UNLOCK(ih);
6081 }
6082
6083 if (V_pfsync_clear_states_ptr != NULL)
6084 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6085
6086 return (killed);
6087 }
6088
6089 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6090 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6091 {
6092 struct pf_kstate *s;
6093
6094 NET_EPOCH_ASSERT();
6095 if (kill->psk_pfcmp.id) {
6096 if (kill->psk_pfcmp.creatorid == 0)
6097 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6098 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6099 kill->psk_pfcmp.creatorid))) {
6100 pf_unlink_state(s);
6101 *killed = 1;
6102 }
6103 return;
6104 }
6105
6106 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6107 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6108 }
6109
6110 static int
pf_killstates_nv(struct pfioc_nv * nv)6111 pf_killstates_nv(struct pfioc_nv *nv)
6112 {
6113 struct pf_kstate_kill kill;
6114 struct epoch_tracker et;
6115 nvlist_t *nvl = NULL;
6116 void *nvlpacked = NULL;
6117 int error = 0;
6118 unsigned int killed = 0;
6119
6120 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6121
6122 if (nv->len > pf_ioctl_maxcount)
6123 ERROUT(ENOMEM);
6124
6125 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6126 error = copyin(nv->data, nvlpacked, nv->len);
6127 if (error)
6128 ERROUT(error);
6129
6130 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6131 if (nvl == NULL)
6132 ERROUT(EBADMSG);
6133
6134 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6135 if (error)
6136 ERROUT(error);
6137
6138 NET_EPOCH_ENTER(et);
6139 pf_killstates(&kill, &killed);
6140 NET_EPOCH_EXIT(et);
6141
6142 free(nvlpacked, M_NVLIST);
6143 nvlpacked = NULL;
6144 nvlist_destroy(nvl);
6145 nvl = nvlist_create(0);
6146 if (nvl == NULL)
6147 ERROUT(ENOMEM);
6148
6149 nvlist_add_number(nvl, "killed", killed);
6150
6151 nvlpacked = nvlist_pack(nvl, &nv->len);
6152 if (nvlpacked == NULL)
6153 ERROUT(ENOMEM);
6154
6155 if (nv->size == 0)
6156 ERROUT(0);
6157 else if (nv->size < nv->len)
6158 ERROUT(ENOSPC);
6159
6160 error = copyout(nvlpacked, nv->data, nv->len);
6161
6162 on_error:
6163 nvlist_destroy(nvl);
6164 free(nvlpacked, M_NVLIST);
6165 return (error);
6166 }
6167
6168 static int
pf_clearstates_nv(struct pfioc_nv * nv)6169 pf_clearstates_nv(struct pfioc_nv *nv)
6170 {
6171 struct pf_kstate_kill kill;
6172 struct epoch_tracker et;
6173 nvlist_t *nvl = NULL;
6174 void *nvlpacked = NULL;
6175 int error = 0;
6176 unsigned int killed;
6177
6178 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6179
6180 if (nv->len > pf_ioctl_maxcount)
6181 ERROUT(ENOMEM);
6182
6183 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6184 error = copyin(nv->data, nvlpacked, nv->len);
6185 if (error)
6186 ERROUT(error);
6187
6188 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6189 if (nvl == NULL)
6190 ERROUT(EBADMSG);
6191
6192 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6193 if (error)
6194 ERROUT(error);
6195
6196 NET_EPOCH_ENTER(et);
6197 killed = pf_clear_states(&kill);
6198 NET_EPOCH_EXIT(et);
6199
6200 free(nvlpacked, M_NVLIST);
6201 nvlpacked = NULL;
6202 nvlist_destroy(nvl);
6203 nvl = nvlist_create(0);
6204 if (nvl == NULL)
6205 ERROUT(ENOMEM);
6206
6207 nvlist_add_number(nvl, "killed", killed);
6208
6209 nvlpacked = nvlist_pack(nvl, &nv->len);
6210 if (nvlpacked == NULL)
6211 ERROUT(ENOMEM);
6212
6213 if (nv->size == 0)
6214 ERROUT(0);
6215 else if (nv->size < nv->len)
6216 ERROUT(ENOSPC);
6217
6218 error = copyout(nvlpacked, nv->data, nv->len);
6219
6220 #undef ERROUT
6221 on_error:
6222 nvlist_destroy(nvl);
6223 free(nvlpacked, M_NVLIST);
6224 return (error);
6225 }
6226
6227 static int
pf_getstate(struct pfioc_nv * nv)6228 pf_getstate(struct pfioc_nv *nv)
6229 {
6230 nvlist_t *nvl = NULL, *nvls;
6231 void *nvlpacked = NULL;
6232 struct pf_kstate *s = NULL;
6233 int error = 0;
6234 uint64_t id, creatorid;
6235
6236 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6237
6238 if (nv->len > pf_ioctl_maxcount)
6239 ERROUT(ENOMEM);
6240
6241 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6242 error = copyin(nv->data, nvlpacked, nv->len);
6243 if (error)
6244 ERROUT(error);
6245
6246 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6247 if (nvl == NULL)
6248 ERROUT(EBADMSG);
6249
6250 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6251 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6252
6253 s = pf_find_state_byid(id, creatorid);
6254 if (s == NULL)
6255 ERROUT(ENOENT);
6256
6257 free(nvlpacked, M_NVLIST);
6258 nvlpacked = NULL;
6259 nvlist_destroy(nvl);
6260 nvl = nvlist_create(0);
6261 if (nvl == NULL)
6262 ERROUT(ENOMEM);
6263
6264 nvls = pf_state_to_nvstate(s);
6265 if (nvls == NULL)
6266 ERROUT(ENOMEM);
6267
6268 nvlist_add_nvlist(nvl, "state", nvls);
6269 nvlist_destroy(nvls);
6270
6271 nvlpacked = nvlist_pack(nvl, &nv->len);
6272 if (nvlpacked == NULL)
6273 ERROUT(ENOMEM);
6274
6275 if (nv->size == 0)
6276 ERROUT(0);
6277 else if (nv->size < nv->len)
6278 ERROUT(ENOSPC);
6279
6280 error = copyout(nvlpacked, nv->data, nv->len);
6281
6282 #undef ERROUT
6283 errout:
6284 if (s != NULL)
6285 PF_STATE_UNLOCK(s);
6286 free(nvlpacked, M_NVLIST);
6287 nvlist_destroy(nvl);
6288 return (error);
6289 }
6290
6291 /*
6292 * XXX - Check for version mismatch!!!
6293 */
6294
6295 /*
6296 * Duplicate pfctl -Fa operation to get rid of as much as we can.
6297 */
6298 static int
shutdown_pf(void)6299 shutdown_pf(void)
6300 {
6301 int error = 0;
6302 u_int32_t t[5];
6303 char nn = '\0';
6304 struct pf_kanchor *anchor;
6305 struct pf_keth_anchor *eth_anchor;
6306 int rs_num;
6307
6308 do {
6309 /* Unlink rules of all user defined anchors */
6310 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6311 /* Wildcard based anchors may not have a respective
6312 * explicit anchor rule or they may be left empty
6313 * without rules. It leads to anchor.refcnt=0, and the
6314 * rest of the logic does not expect it. */
6315 if (anchor->refcnt == 0)
6316 anchor->refcnt = 1;
6317 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6318 if ((error = pf_begin_rules(&t[rs_num], rs_num,
6319 anchor->path)) != 0) {
6320 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6321 "anchor.path=%s rs_num=%d\n",
6322 anchor->path, rs_num));
6323 goto error; /* XXX: rollback? */
6324 }
6325 }
6326 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6327 error = pf_commit_rules(t[rs_num], rs_num,
6328 anchor->path);
6329 MPASS(error == 0);
6330 }
6331 }
6332
6333 /* Unlink rules of all user defined ether anchors */
6334 RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6335 &V_pf_keth_anchors) {
6336 /* Wildcard based anchors may not have a respective
6337 * explicit anchor rule or they may be left empty
6338 * without rules. It leads to anchor.refcnt=0, and the
6339 * rest of the logic does not expect it. */
6340 if (eth_anchor->refcnt == 0)
6341 eth_anchor->refcnt = 1;
6342 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6343 != 0) {
6344 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6345 "anchor.path=%s\n", eth_anchor->path));
6346 goto error;
6347 }
6348 error = pf_commit_eth(t[0], eth_anchor->path);
6349 MPASS(error == 0);
6350 }
6351
6352 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6353 != 0) {
6354 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6355 break;
6356 }
6357 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6358 != 0) {
6359 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6360 break; /* XXX: rollback? */
6361 }
6362 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6363 != 0) {
6364 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6365 break; /* XXX: rollback? */
6366 }
6367 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6368 != 0) {
6369 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6370 break; /* XXX: rollback? */
6371 }
6372 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6373 != 0) {
6374 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6375 break; /* XXX: rollback? */
6376 }
6377
6378 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6379 MPASS(error == 0);
6380 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6381 MPASS(error == 0);
6382 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6383 MPASS(error == 0);
6384 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6385 MPASS(error == 0);
6386 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6387 MPASS(error == 0);
6388
6389 if ((error = pf_clear_tables()) != 0)
6390 break;
6391
6392 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6393 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6394 break;
6395 }
6396 error = pf_commit_eth(t[0], &nn);
6397 MPASS(error == 0);
6398
6399 #ifdef ALTQ
6400 if ((error = pf_begin_altq(&t[0])) != 0) {
6401 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6402 break;
6403 }
6404 pf_commit_altq(t[0]);
6405 #endif
6406
6407 pf_clear_all_states();
6408
6409 pf_clear_srcnodes(NULL);
6410
6411 /* status does not use malloced mem so no need to cleanup */
6412 /* fingerprints and interfaces have their own cleanup code */
6413 } while(0);
6414
6415 error:
6416 return (error);
6417 }
6418
6419 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6420 pf_check_return(int chk, struct mbuf **m)
6421 {
6422
6423 switch (chk) {
6424 case PF_PASS:
6425 if (*m == NULL)
6426 return (PFIL_CONSUMED);
6427 else
6428 return (PFIL_PASS);
6429 break;
6430 default:
6431 if (*m != NULL) {
6432 m_freem(*m);
6433 *m = NULL;
6434 }
6435 return (PFIL_DROPPED);
6436 }
6437 }
6438
6439 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6440 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6441 void *ruleset __unused, struct inpcb *inp)
6442 {
6443 int chk;
6444
6445 CURVNET_ASSERT_SET();
6446
6447 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6448
6449 return (pf_check_return(chk, m));
6450 }
6451
6452 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6453 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6454 void *ruleset __unused, struct inpcb *inp)
6455 {
6456 int chk;
6457
6458 CURVNET_ASSERT_SET();
6459
6460 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6461
6462 return (pf_check_return(chk, m));
6463 }
6464
6465 #ifdef INET
6466 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6467 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6468 void *ruleset __unused, struct inpcb *inp)
6469 {
6470 int chk;
6471
6472 CURVNET_ASSERT_SET();
6473
6474 chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6475
6476 return (pf_check_return(chk, m));
6477 }
6478
6479 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6480 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6481 void *ruleset __unused, struct inpcb *inp)
6482 {
6483 int chk;
6484
6485 CURVNET_ASSERT_SET();
6486
6487 chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6488
6489 return (pf_check_return(chk, m));
6490 }
6491 #endif
6492
6493 #ifdef INET6
6494 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6495 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6496 void *ruleset __unused, struct inpcb *inp)
6497 {
6498 int chk;
6499
6500 CURVNET_ASSERT_SET();
6501
6502 /*
6503 * In case of loopback traffic IPv6 uses the real interface in
6504 * order to support scoped addresses. In order to support stateful
6505 * filtering we have change this to lo0 as it is the case in IPv4.
6506 */
6507 chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6508 m, inp, NULL);
6509
6510 return (pf_check_return(chk, m));
6511 }
6512
6513 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6514 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6515 void *ruleset __unused, struct inpcb *inp)
6516 {
6517 int chk;
6518
6519 CURVNET_ASSERT_SET();
6520
6521 chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6522
6523 return (pf_check_return(chk, m));
6524 }
6525 #endif /* INET6 */
6526
6527 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6528 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6529 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
6530 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
6531
6532 #ifdef INET
6533 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6534 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6535 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
6536 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
6537 #endif
6538 #ifdef INET6
6539 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6540 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6541 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
6542 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
6543 #endif
6544
6545 static void
hook_pf_eth(void)6546 hook_pf_eth(void)
6547 {
6548 struct pfil_hook_args pha = {
6549 .pa_version = PFIL_VERSION,
6550 .pa_modname = "pf",
6551 .pa_type = PFIL_TYPE_ETHERNET,
6552 };
6553 struct pfil_link_args pla = {
6554 .pa_version = PFIL_VERSION,
6555 };
6556 int ret __diagused;
6557
6558 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6559 return;
6560
6561 pha.pa_mbuf_chk = pf_eth_check_in;
6562 pha.pa_flags = PFIL_IN;
6563 pha.pa_rulname = "eth-in";
6564 V_pf_eth_in_hook = pfil_add_hook(&pha);
6565 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6566 pla.pa_head = V_link_pfil_head;
6567 pla.pa_hook = V_pf_eth_in_hook;
6568 ret = pfil_link(&pla);
6569 MPASS(ret == 0);
6570 pha.pa_mbuf_chk = pf_eth_check_out;
6571 pha.pa_flags = PFIL_OUT;
6572 pha.pa_rulname = "eth-out";
6573 V_pf_eth_out_hook = pfil_add_hook(&pha);
6574 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6575 pla.pa_head = V_link_pfil_head;
6576 pla.pa_hook = V_pf_eth_out_hook;
6577 ret = pfil_link(&pla);
6578 MPASS(ret == 0);
6579
6580 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6581 }
6582
6583 static void
hook_pf(void)6584 hook_pf(void)
6585 {
6586 struct pfil_hook_args pha = {
6587 .pa_version = PFIL_VERSION,
6588 .pa_modname = "pf",
6589 };
6590 struct pfil_link_args pla = {
6591 .pa_version = PFIL_VERSION,
6592 };
6593 int ret __diagused;
6594
6595 if (atomic_load_bool(&V_pf_pfil_hooked))
6596 return;
6597
6598 #ifdef INET
6599 pha.pa_type = PFIL_TYPE_IP4;
6600 pha.pa_mbuf_chk = pf_check_in;
6601 pha.pa_flags = PFIL_IN;
6602 pha.pa_rulname = "default-in";
6603 V_pf_ip4_in_hook = pfil_add_hook(&pha);
6604 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6605 pla.pa_head = V_inet_pfil_head;
6606 pla.pa_hook = V_pf_ip4_in_hook;
6607 ret = pfil_link(&pla);
6608 MPASS(ret == 0);
6609 pha.pa_mbuf_chk = pf_check_out;
6610 pha.pa_flags = PFIL_OUT;
6611 pha.pa_rulname = "default-out";
6612 V_pf_ip4_out_hook = pfil_add_hook(&pha);
6613 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6614 pla.pa_head = V_inet_pfil_head;
6615 pla.pa_hook = V_pf_ip4_out_hook;
6616 ret = pfil_link(&pla);
6617 MPASS(ret == 0);
6618 if (V_pf_filter_local) {
6619 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6620 pla.pa_head = V_inet_local_pfil_head;
6621 pla.pa_hook = V_pf_ip4_out_hook;
6622 ret = pfil_link(&pla);
6623 MPASS(ret == 0);
6624 }
6625 #endif
6626 #ifdef INET6
6627 pha.pa_type = PFIL_TYPE_IP6;
6628 pha.pa_mbuf_chk = pf_check6_in;
6629 pha.pa_flags = PFIL_IN;
6630 pha.pa_rulname = "default-in6";
6631 V_pf_ip6_in_hook = pfil_add_hook(&pha);
6632 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6633 pla.pa_head = V_inet6_pfil_head;
6634 pla.pa_hook = V_pf_ip6_in_hook;
6635 ret = pfil_link(&pla);
6636 MPASS(ret == 0);
6637 pha.pa_mbuf_chk = pf_check6_out;
6638 pha.pa_rulname = "default-out6";
6639 pha.pa_flags = PFIL_OUT;
6640 V_pf_ip6_out_hook = pfil_add_hook(&pha);
6641 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6642 pla.pa_head = V_inet6_pfil_head;
6643 pla.pa_hook = V_pf_ip6_out_hook;
6644 ret = pfil_link(&pla);
6645 MPASS(ret == 0);
6646 if (V_pf_filter_local) {
6647 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6648 pla.pa_head = V_inet6_local_pfil_head;
6649 pla.pa_hook = V_pf_ip6_out_hook;
6650 ret = pfil_link(&pla);
6651 MPASS(ret == 0);
6652 }
6653 #endif
6654
6655 atomic_store_bool(&V_pf_pfil_hooked, true);
6656 }
6657
6658 static void
dehook_pf_eth(void)6659 dehook_pf_eth(void)
6660 {
6661
6662 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6663 return;
6664
6665 pfil_remove_hook(V_pf_eth_in_hook);
6666 pfil_remove_hook(V_pf_eth_out_hook);
6667
6668 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6669 }
6670
6671 static void
dehook_pf(void)6672 dehook_pf(void)
6673 {
6674
6675 if (!atomic_load_bool(&V_pf_pfil_hooked))
6676 return;
6677
6678 #ifdef INET
6679 pfil_remove_hook(V_pf_ip4_in_hook);
6680 pfil_remove_hook(V_pf_ip4_out_hook);
6681 #endif
6682 #ifdef INET6
6683 pfil_remove_hook(V_pf_ip6_in_hook);
6684 pfil_remove_hook(V_pf_ip6_out_hook);
6685 #endif
6686
6687 atomic_store_bool(&V_pf_pfil_hooked, false);
6688 }
6689
6690 static void
pf_load_vnet(void)6691 pf_load_vnet(void)
6692 {
6693 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6694 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6695
6696 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6697 sx_init(&V_pf_ioctl_lock, "pf ioctl");
6698
6699 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6700 PF_RULE_TAG_HASH_SIZE_DEFAULT);
6701 #ifdef ALTQ
6702 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6703 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6704 #endif
6705
6706 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6707
6708 pfattach_vnet();
6709 V_pf_vnet_active = 1;
6710 }
6711
6712 static int
pf_load(void)6713 pf_load(void)
6714 {
6715 int error;
6716
6717 sx_init(&pf_end_lock, "pf end thread");
6718
6719 pf_mtag_initialize();
6720
6721 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6722 if (pf_dev == NULL)
6723 return (ENOMEM);
6724
6725 pf_end_threads = 0;
6726 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6727 if (error != 0)
6728 return (error);
6729
6730 pfi_initialize();
6731
6732 return (0);
6733 }
6734
6735 static void
pf_unload_vnet(void)6736 pf_unload_vnet(void)
6737 {
6738 int ret __diagused;
6739
6740 V_pf_vnet_active = 0;
6741 V_pf_status.running = 0;
6742 dehook_pf();
6743 dehook_pf_eth();
6744
6745 PF_RULES_WLOCK();
6746 pf_syncookies_cleanup();
6747 shutdown_pf();
6748 PF_RULES_WUNLOCK();
6749
6750 /* Make sure we've cleaned up ethernet rules before we continue. */
6751 NET_EPOCH_DRAIN_CALLBACKS();
6752
6753 ret = swi_remove(V_pf_swi_cookie);
6754 MPASS(ret == 0);
6755 ret = intr_event_destroy(V_pf_swi_ie);
6756 MPASS(ret == 0);
6757
6758 pf_unload_vnet_purge();
6759
6760 pf_normalize_cleanup();
6761 PF_RULES_WLOCK();
6762 pfi_cleanup_vnet();
6763 PF_RULES_WUNLOCK();
6764 pfr_cleanup();
6765 pf_osfp_flush();
6766 pf_cleanup();
6767 if (IS_DEFAULT_VNET(curvnet))
6768 pf_mtag_cleanup();
6769
6770 pf_cleanup_tagset(&V_pf_tags);
6771 #ifdef ALTQ
6772 pf_cleanup_tagset(&V_pf_qids);
6773 #endif
6774 uma_zdestroy(V_pf_tag_z);
6775
6776 #ifdef PF_WANT_32_TO_64_COUNTER
6777 PF_RULES_WLOCK();
6778 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6779
6780 MPASS(LIST_EMPTY(&V_pf_allkiflist));
6781 MPASS(V_pf_allkifcount == 0);
6782
6783 LIST_REMOVE(&V_pf_default_rule, allrulelist);
6784 V_pf_allrulecount--;
6785 LIST_REMOVE(V_pf_rulemarker, allrulelist);
6786
6787 MPASS(LIST_EMPTY(&V_pf_allrulelist));
6788 MPASS(V_pf_allrulecount == 0);
6789
6790 PF_RULES_WUNLOCK();
6791
6792 free(V_pf_kifmarker, PFI_MTYPE);
6793 free(V_pf_rulemarker, M_PFRULE);
6794 #endif
6795
6796 /* Free counters last as we updated them during shutdown. */
6797 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6798 for (int i = 0; i < 2; i++) {
6799 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6800 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6801 }
6802 counter_u64_free(V_pf_default_rule.states_cur);
6803 counter_u64_free(V_pf_default_rule.states_tot);
6804 counter_u64_free(V_pf_default_rule.src_nodes);
6805 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6806
6807 for (int i = 0; i < PFRES_MAX; i++)
6808 counter_u64_free(V_pf_status.counters[i]);
6809 for (int i = 0; i < KLCNT_MAX; i++)
6810 counter_u64_free(V_pf_status.lcounters[i]);
6811 for (int i = 0; i < FCNT_MAX; i++)
6812 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6813 for (int i = 0; i < SCNT_MAX; i++)
6814 counter_u64_free(V_pf_status.scounters[i]);
6815
6816 rm_destroy(&V_pf_rules_lock);
6817 sx_destroy(&V_pf_ioctl_lock);
6818 }
6819
6820 static void
pf_unload(void)6821 pf_unload(void)
6822 {
6823
6824 sx_xlock(&pf_end_lock);
6825 pf_end_threads = 1;
6826 while (pf_end_threads < 2) {
6827 wakeup_one(pf_purge_thread);
6828 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6829 }
6830 sx_xunlock(&pf_end_lock);
6831
6832 pf_nl_unregister();
6833
6834 if (pf_dev != NULL)
6835 destroy_dev(pf_dev);
6836
6837 pfi_cleanup();
6838
6839 sx_destroy(&pf_end_lock);
6840 }
6841
6842 static void
vnet_pf_init(void * unused __unused)6843 vnet_pf_init(void *unused __unused)
6844 {
6845
6846 pf_load_vnet();
6847 }
6848 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6849 vnet_pf_init, NULL);
6850
6851 static void
vnet_pf_uninit(const void * unused __unused)6852 vnet_pf_uninit(const void *unused __unused)
6853 {
6854
6855 pf_unload_vnet();
6856 }
6857 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6858 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6859 vnet_pf_uninit, NULL);
6860
6861 static int
pf_modevent(module_t mod,int type,void * data)6862 pf_modevent(module_t mod, int type, void *data)
6863 {
6864 int error = 0;
6865
6866 switch(type) {
6867 case MOD_LOAD:
6868 error = pf_load();
6869 pf_nl_register();
6870 break;
6871 case MOD_UNLOAD:
6872 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6873 * the vnet_pf_uninit()s */
6874 break;
6875 default:
6876 error = EINVAL;
6877 break;
6878 }
6879
6880 return (error);
6881 }
6882
6883 static moduledata_t pf_mod = {
6884 "pf",
6885 pf_modevent,
6886 0
6887 };
6888
6889 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6890 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6891 MODULE_VERSION(pf, PF_MODVER);
6892