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_addr_copyout(struct pf_addr_wrap *);
133 static void pf_src_node_copy(const struct pf_ksrc_node *,
134 struct pf_src_node *);
135 #ifdef ALTQ
136 static int pf_export_kaltq(struct pf_altq *,
137 struct pfioc_altq_v1 *, size_t);
138 static int pf_import_kaltq(struct pfioc_altq_v1 *,
139 struct pf_altq *, size_t);
140 #endif /* ALTQ */
141
142 VNET_DEFINE(struct pf_krule, pf_default_rule);
143
144 static __inline int pf_krule_compare(struct pf_krule *,
145 struct pf_krule *);
146
147 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
148
149 #ifdef ALTQ
150 VNET_DEFINE_STATIC(int, pf_altq_running);
151 #define V_pf_altq_running VNET(pf_altq_running)
152 #endif
153
154 #define TAGID_MAX 50000
155 struct pf_tagname {
156 TAILQ_ENTRY(pf_tagname) namehash_entries;
157 TAILQ_ENTRY(pf_tagname) taghash_entries;
158 char name[PF_TAG_NAME_SIZE];
159 uint16_t tag;
160 int ref;
161 };
162
163 struct pf_tagset {
164 TAILQ_HEAD(, pf_tagname) *namehash;
165 TAILQ_HEAD(, pf_tagname) *taghash;
166 unsigned int mask;
167 uint32_t seed;
168 BITSET_DEFINE(, TAGID_MAX) avail;
169 };
170
171 VNET_DEFINE(struct pf_tagset, pf_tags);
172 #define V_pf_tags VNET(pf_tags)
173 static unsigned int pf_rule_tag_hashsize;
174 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128
175 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
176 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
177 "Size of pf(4) rule tag hashtable");
178
179 #ifdef ALTQ
180 VNET_DEFINE(struct pf_tagset, pf_qids);
181 #define V_pf_qids VNET(pf_qids)
182 static unsigned int pf_queue_tag_hashsize;
183 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128
184 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
185 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
186 "Size of pf(4) queue tag hashtable");
187 #endif
188 VNET_DEFINE(uma_zone_t, pf_tag_z);
189 #define V_pf_tag_z VNET(pf_tag_z)
190 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
191 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
192
193 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
194 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
195 #endif
196
197 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
198 #define V_pf_filter_local VNET(pf_filter_local)
199 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
200 &VNET_NAME(pf_filter_local), false,
201 "Enable filtering for packets delivered to local network stack");
202
203 #ifdef PF_DEFAULT_TO_DROP
204 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
205 #else
206 VNET_DEFINE_STATIC(bool, default_to_drop);
207 #endif
208 #define V_default_to_drop VNET(default_to_drop)
209 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
210 &VNET_NAME(default_to_drop), false,
211 "Make the default rule drop all packets.");
212
213 static void pf_init_tagset(struct pf_tagset *, unsigned int *,
214 unsigned int);
215 static void pf_cleanup_tagset(struct pf_tagset *);
216 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *);
217 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t);
218 static u_int16_t tagname2tag(struct pf_tagset *, const char *);
219 static u_int16_t pf_tagname2tag(const char *);
220 static void tag_unref(struct pf_tagset *, u_int16_t);
221
222 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
223
224 struct cdev *pf_dev;
225
226 /*
227 * XXX - These are new and need to be checked when moveing to a new version
228 */
229 static void pf_clear_all_states(void);
230 static int pf_killstates_row(struct pf_kstate_kill *,
231 struct pf_idhash *);
232 static int pf_killstates_nv(struct pfioc_nv *);
233 static int pf_clearstates_nv(struct pfioc_nv *);
234 static int pf_getstate(struct pfioc_nv *);
235 static int pf_getstatus(struct pfioc_nv *);
236 static int pf_clear_tables(void);
237 static void pf_clear_srcnodes(struct pf_ksrc_node *);
238 static void pf_kill_srcnodes(struct pfioc_src_node_kill *);
239 static int pf_keepcounters(struct pfioc_nv *);
240 static void pf_tbladdr_copyout(struct pf_addr_wrap *);
241
242 /*
243 * Wrapper functions for pfil(9) hooks
244 */
245 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
246 int flags, void *ruleset __unused, struct inpcb *inp);
247 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
248 int flags, void *ruleset __unused, struct inpcb *inp);
249 #ifdef INET
250 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
251 int flags, void *ruleset __unused, struct inpcb *inp);
252 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
253 int flags, void *ruleset __unused, struct inpcb *inp);
254 #endif
255 #ifdef INET6
256 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
257 int flags, void *ruleset __unused, struct inpcb *inp);
258 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
259 int flags, void *ruleset __unused, struct inpcb *inp);
260 #endif
261
262 static void hook_pf_eth(void);
263 static void hook_pf(void);
264 static void dehook_pf_eth(void);
265 static void dehook_pf(void);
266 static int shutdown_pf(void);
267 static int pf_load(void);
268 static void pf_unload(void);
269
270 static struct cdevsw pf_cdevsw = {
271 .d_ioctl = pfioctl,
272 .d_name = PF_NAME,
273 .d_version = D_VERSION,
274 };
275
276 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
277 #define V_pf_pfil_hooked VNET(pf_pfil_hooked)
278 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
279 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked)
280
281 /*
282 * We need a flag that is neither hooked nor running to know when
283 * the VNET is "valid". We primarily need this to control (global)
284 * external event, e.g., eventhandlers.
285 */
286 VNET_DEFINE(int, pf_vnet_active);
287 #define V_pf_vnet_active VNET(pf_vnet_active)
288
289 int pf_end_threads;
290 struct proc *pf_purge_proc;
291
292 VNET_DEFINE(struct rmlock, pf_rules_lock);
293 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
294 #define V_pf_ioctl_lock VNET(pf_ioctl_lock)
295 struct sx pf_end_lock;
296
297 /* pfsync */
298 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
299 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
300 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
301 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
302 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
303 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
304 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
305 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
306
307 /* pflog */
308 pflog_packet_t *pflog_packet_ptr = NULL;
309
310 /*
311 * Copy a user-provided string, returning an error if truncation would occur.
312 * Avoid scanning past "sz" bytes in the source string since there's no
313 * guarantee that it's nul-terminated.
314 */
315 static int
pf_user_strcpy(char * dst,const char * src,size_t sz)316 pf_user_strcpy(char *dst, const char *src, size_t sz)
317 {
318 if (strnlen(src, sz) == sz)
319 return (EINVAL);
320 (void)strlcpy(dst, src, sz);
321 return (0);
322 }
323
324 static void
pfattach_vnet(void)325 pfattach_vnet(void)
326 {
327 u_int32_t *my_timeout = V_pf_default_rule.timeout;
328
329 bzero(&V_pf_status, sizeof(V_pf_status));
330
331 pf_initialize();
332 pfr_initialize();
333 pfi_initialize_vnet();
334 pf_normalize_init();
335 pf_syncookies_init();
336
337 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
338 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
339
340 RB_INIT(&V_pf_anchors);
341 pf_init_kruleset(&pf_main_ruleset);
342
343 pf_init_keth(V_pf_keth);
344
345 /* default rule should never be garbage collected */
346 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
347 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
348 V_pf_default_rule.nr = -1;
349 V_pf_default_rule.rtableid = -1;
350
351 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
352 for (int i = 0; i < 2; i++) {
353 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
354 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
355 }
356 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
357 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
358 V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
359
360 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
361 M_WAITOK | M_ZERO);
362
363 #ifdef PF_WANT_32_TO_64_COUNTER
364 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
365 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
366 PF_RULES_WLOCK();
367 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
368 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
369 V_pf_allrulecount++;
370 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
371 PF_RULES_WUNLOCK();
372 #endif
373
374 /* initialize default timeouts */
375 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
376 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
377 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
378 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
379 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
380 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
381 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
382 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
383 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
384 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
385 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
386 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
387 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
388 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
389 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
390 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
391 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
392 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
393 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
394 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
395 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
396 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
397 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
398 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
399 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
400
401 V_pf_status.debug = PF_DEBUG_URGENT;
402 /*
403 * XXX This is different than in OpenBSD where reassembly is enabled by
404 * defult. In FreeBSD we expect people to still use scrub rules and
405 * switch to the new syntax later. Only when they switch they must
406 * explicitly enable reassemle. We could change the default once the
407 * scrub rule functionality is hopefully removed some day in future.
408 */
409 V_pf_status.reass = 0;
410
411 V_pf_pfil_hooked = false;
412 V_pf_pfil_eth_hooked = false;
413
414 /* XXX do our best to avoid a conflict */
415 V_pf_status.hostid = arc4random();
416
417 for (int i = 0; i < PFRES_MAX; i++)
418 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
419 for (int i = 0; i < KLCNT_MAX; i++)
420 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
421 for (int i = 0; i < FCNT_MAX; i++)
422 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
423 for (int i = 0; i < SCNT_MAX; i++)
424 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
425
426 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
427 INTR_MPSAFE, &V_pf_swi_cookie) != 0)
428 /* XXXGL: leaked all above. */
429 return;
430 }
431
432 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)433 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
434 u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
435 u_int8_t check_ticket)
436 {
437 struct pf_kruleset *ruleset;
438 struct pf_krule *rule;
439 int rs_num;
440
441 ruleset = pf_find_kruleset(anchor);
442 if (ruleset == NULL)
443 return (NULL);
444 rs_num = pf_get_ruleset_number(rule_action);
445 if (rs_num >= PF_RULESET_MAX)
446 return (NULL);
447 if (active) {
448 if (check_ticket && ticket !=
449 ruleset->rules[rs_num].active.ticket)
450 return (NULL);
451 if (r_last)
452 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
453 pf_krulequeue);
454 else
455 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
456 } else {
457 if (check_ticket && ticket !=
458 ruleset->rules[rs_num].inactive.ticket)
459 return (NULL);
460 if (r_last)
461 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
462 pf_krulequeue);
463 else
464 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
465 }
466 if (!r_last) {
467 while ((rule != NULL) && (rule->nr != rule_number))
468 rule = TAILQ_NEXT(rule, entries);
469 }
470 if (rule == NULL)
471 return (NULL);
472
473 return (&rule->rpool);
474 }
475
476 static void
pf_mv_kpool(struct pf_kpalist * poola,struct pf_kpalist * poolb)477 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
478 {
479 struct pf_kpooladdr *mv_pool_pa;
480
481 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
482 TAILQ_REMOVE(poola, mv_pool_pa, entries);
483 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
484 }
485 }
486
487 static void
pf_empty_kpool(struct pf_kpalist * poola)488 pf_empty_kpool(struct pf_kpalist *poola)
489 {
490 struct pf_kpooladdr *pa;
491
492 while ((pa = TAILQ_FIRST(poola)) != NULL) {
493 switch (pa->addr.type) {
494 case PF_ADDR_DYNIFTL:
495 pfi_dynaddr_remove(pa->addr.p.dyn);
496 break;
497 case PF_ADDR_TABLE:
498 /* XXX: this could be unfinished pooladdr on pabuf */
499 if (pa->addr.p.tbl != NULL)
500 pfr_detach_table(pa->addr.p.tbl);
501 break;
502 }
503 if (pa->kif)
504 pfi_kkif_unref(pa->kif);
505 TAILQ_REMOVE(poola, pa, entries);
506 free(pa, M_PFRULE);
507 }
508 }
509
510 static void
pf_unlink_rule_locked(struct pf_krulequeue * rulequeue,struct pf_krule * rule)511 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
512 {
513
514 PF_RULES_WASSERT();
515 PF_UNLNKDRULES_ASSERT();
516
517 TAILQ_REMOVE(rulequeue, rule, entries);
518
519 rule->rule_ref |= PFRULE_REFS;
520 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
521 }
522
523 static void
pf_unlink_rule(struct pf_krulequeue * rulequeue,struct pf_krule * rule)524 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
525 {
526
527 PF_RULES_WASSERT();
528
529 PF_UNLNKDRULES_LOCK();
530 pf_unlink_rule_locked(rulequeue, rule);
531 PF_UNLNKDRULES_UNLOCK();
532 }
533
534 static void
pf_free_eth_rule(struct pf_keth_rule * rule)535 pf_free_eth_rule(struct pf_keth_rule *rule)
536 {
537 PF_RULES_WASSERT();
538
539 if (rule == NULL)
540 return;
541
542 if (rule->tag)
543 tag_unref(&V_pf_tags, rule->tag);
544 if (rule->match_tag)
545 tag_unref(&V_pf_tags, rule->match_tag);
546 #ifdef ALTQ
547 pf_qid_unref(rule->qid);
548 #endif
549
550 if (rule->bridge_to)
551 pfi_kkif_unref(rule->bridge_to);
552 if (rule->kif)
553 pfi_kkif_unref(rule->kif);
554
555 if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
556 pfr_detach_table(rule->ipsrc.addr.p.tbl);
557 if (rule->ipdst.addr.type == PF_ADDR_TABLE)
558 pfr_detach_table(rule->ipdst.addr.p.tbl);
559
560 counter_u64_free(rule->evaluations);
561 for (int i = 0; i < 2; i++) {
562 counter_u64_free(rule->packets[i]);
563 counter_u64_free(rule->bytes[i]);
564 }
565 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
566 pf_keth_anchor_remove(rule);
567
568 free(rule, M_PFRULE);
569 }
570
571 void
pf_free_rule(struct pf_krule * rule)572 pf_free_rule(struct pf_krule *rule)
573 {
574
575 PF_RULES_WASSERT();
576 PF_CONFIG_ASSERT();
577
578 if (rule->tag)
579 tag_unref(&V_pf_tags, rule->tag);
580 if (rule->match_tag)
581 tag_unref(&V_pf_tags, rule->match_tag);
582 #ifdef ALTQ
583 if (rule->pqid != rule->qid)
584 pf_qid_unref(rule->pqid);
585 pf_qid_unref(rule->qid);
586 #endif
587 switch (rule->src.addr.type) {
588 case PF_ADDR_DYNIFTL:
589 pfi_dynaddr_remove(rule->src.addr.p.dyn);
590 break;
591 case PF_ADDR_TABLE:
592 pfr_detach_table(rule->src.addr.p.tbl);
593 break;
594 }
595 switch (rule->dst.addr.type) {
596 case PF_ADDR_DYNIFTL:
597 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
598 break;
599 case PF_ADDR_TABLE:
600 pfr_detach_table(rule->dst.addr.p.tbl);
601 break;
602 }
603 if (rule->overload_tbl)
604 pfr_detach_table(rule->overload_tbl);
605 if (rule->kif)
606 pfi_kkif_unref(rule->kif);
607 pf_kanchor_remove(rule);
608 pf_empty_kpool(&rule->rpool.list);
609
610 pf_krule_free(rule);
611 }
612
613 static void
pf_init_tagset(struct pf_tagset * ts,unsigned int * tunable_size,unsigned int default_size)614 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
615 unsigned int default_size)
616 {
617 unsigned int i;
618 unsigned int hashsize;
619
620 if (*tunable_size == 0 || !powerof2(*tunable_size))
621 *tunable_size = default_size;
622
623 hashsize = *tunable_size;
624 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
625 M_WAITOK);
626 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
627 M_WAITOK);
628 ts->mask = hashsize - 1;
629 ts->seed = arc4random();
630 for (i = 0; i < hashsize; i++) {
631 TAILQ_INIT(&ts->namehash[i]);
632 TAILQ_INIT(&ts->taghash[i]);
633 }
634 BIT_FILL(TAGID_MAX, &ts->avail);
635 }
636
637 static void
pf_cleanup_tagset(struct pf_tagset * ts)638 pf_cleanup_tagset(struct pf_tagset *ts)
639 {
640 unsigned int i;
641 unsigned int hashsize;
642 struct pf_tagname *t, *tmp;
643
644 /*
645 * Only need to clean up one of the hashes as each tag is hashed
646 * into each table.
647 */
648 hashsize = ts->mask + 1;
649 for (i = 0; i < hashsize; i++)
650 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
651 uma_zfree(V_pf_tag_z, t);
652
653 free(ts->namehash, M_PFHASH);
654 free(ts->taghash, M_PFHASH);
655 }
656
657 static uint16_t
tagname2hashindex(const struct pf_tagset * ts,const char * tagname)658 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
659 {
660 size_t len;
661
662 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
663 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
664 }
665
666 static uint16_t
tag2hashindex(const struct pf_tagset * ts,uint16_t tag)667 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
668 {
669
670 return (tag & ts->mask);
671 }
672
673 static u_int16_t
tagname2tag(struct pf_tagset * ts,const char * tagname)674 tagname2tag(struct pf_tagset *ts, const char *tagname)
675 {
676 struct pf_tagname *tag;
677 u_int32_t index;
678 u_int16_t new_tagid;
679
680 PF_RULES_WASSERT();
681
682 index = tagname2hashindex(ts, tagname);
683 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
684 if (strcmp(tagname, tag->name) == 0) {
685 tag->ref++;
686 return (tag->tag);
687 }
688
689 /*
690 * new entry
691 *
692 * to avoid fragmentation, we do a linear search from the beginning
693 * and take the first free slot we find.
694 */
695 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
696 /*
697 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
698 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
699 * set. It may also return a bit number greater than TAGID_MAX due
700 * to rounding of the number of bits in the vector up to a multiple
701 * of the vector word size at declaration/allocation time.
702 */
703 if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
704 return (0);
705
706 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */
707 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
708
709 /* allocate and fill new struct pf_tagname */
710 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
711 if (tag == NULL)
712 return (0);
713 strlcpy(tag->name, tagname, sizeof(tag->name));
714 tag->tag = new_tagid;
715 tag->ref = 1;
716
717 /* Insert into namehash */
718 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
719
720 /* Insert into taghash */
721 index = tag2hashindex(ts, new_tagid);
722 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
723
724 return (tag->tag);
725 }
726
727 static void
tag_unref(struct pf_tagset * ts,u_int16_t tag)728 tag_unref(struct pf_tagset *ts, u_int16_t tag)
729 {
730 struct pf_tagname *t;
731 uint16_t index;
732
733 PF_RULES_WASSERT();
734
735 index = tag2hashindex(ts, tag);
736 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
737 if (tag == t->tag) {
738 if (--t->ref == 0) {
739 TAILQ_REMOVE(&ts->taghash[index], t,
740 taghash_entries);
741 index = tagname2hashindex(ts, t->name);
742 TAILQ_REMOVE(&ts->namehash[index], t,
743 namehash_entries);
744 /* Bits are 0-based for BIT_SET() */
745 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
746 uma_zfree(V_pf_tag_z, t);
747 }
748 break;
749 }
750 }
751
752 static uint16_t
pf_tagname2tag(const char * tagname)753 pf_tagname2tag(const char *tagname)
754 {
755 return (tagname2tag(&V_pf_tags, tagname));
756 }
757
758 static int
pf_begin_eth(uint32_t * ticket,const char * anchor)759 pf_begin_eth(uint32_t *ticket, const char *anchor)
760 {
761 struct pf_keth_rule *rule, *tmp;
762 struct pf_keth_ruleset *rs;
763
764 PF_RULES_WASSERT();
765
766 rs = pf_find_or_create_keth_ruleset(anchor);
767 if (rs == NULL)
768 return (EINVAL);
769
770 /* Purge old inactive rules. */
771 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
772 tmp) {
773 TAILQ_REMOVE(rs->inactive.rules, rule,
774 entries);
775 pf_free_eth_rule(rule);
776 }
777
778 *ticket = ++rs->inactive.ticket;
779 rs->inactive.open = 1;
780
781 return (0);
782 }
783
784 static void
pf_rollback_eth_cb(struct epoch_context * ctx)785 pf_rollback_eth_cb(struct epoch_context *ctx)
786 {
787 struct pf_keth_ruleset *rs;
788
789 rs = __containerof(ctx, struct pf_keth_ruleset, epoch_ctx);
790
791 CURVNET_SET(rs->vnet);
792
793 PF_RULES_WLOCK();
794 pf_rollback_eth(rs->inactive.ticket,
795 rs->anchor ? rs->anchor->path : "");
796 PF_RULES_WUNLOCK();
797
798 CURVNET_RESTORE();
799 }
800
801 static int
pf_rollback_eth(uint32_t ticket,const char * anchor)802 pf_rollback_eth(uint32_t ticket, const char *anchor)
803 {
804 struct pf_keth_rule *rule, *tmp;
805 struct pf_keth_ruleset *rs;
806
807 PF_RULES_WASSERT();
808
809 rs = pf_find_keth_ruleset(anchor);
810 if (rs == NULL)
811 return (EINVAL);
812
813 if (!rs->inactive.open ||
814 ticket != rs->inactive.ticket)
815 return (0);
816
817 /* Purge old inactive rules. */
818 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
819 tmp) {
820 TAILQ_REMOVE(rs->inactive.rules, rule, entries);
821 pf_free_eth_rule(rule);
822 }
823
824 rs->inactive.open = 0;
825
826 pf_remove_if_empty_keth_ruleset(rs);
827
828 return (0);
829 }
830
831 #define PF_SET_SKIP_STEPS(i) \
832 do { \
833 while (head[i] != cur) { \
834 head[i]->skip[i].ptr = cur; \
835 head[i] = TAILQ_NEXT(head[i], entries); \
836 } \
837 } while (0)
838
839 static void
pf_eth_calc_skip_steps(struct pf_keth_ruleq * rules)840 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
841 {
842 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
843 int i;
844
845 cur = TAILQ_FIRST(rules);
846 prev = cur;
847 for (i = 0; i < PFE_SKIP_COUNT; ++i)
848 head[i] = cur;
849 while (cur != NULL) {
850 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
851 PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
852 if (cur->direction != prev->direction)
853 PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
854 if (cur->proto != prev->proto)
855 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
856 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
857 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
858 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
859 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
860 if (cur->ipsrc.neg != prev->ipsrc.neg ||
861 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
862 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
863 if (cur->ipdst.neg != prev->ipdst.neg ||
864 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
865 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
866
867 prev = cur;
868 cur = TAILQ_NEXT(cur, entries);
869 }
870 for (i = 0; i < PFE_SKIP_COUNT; ++i)
871 PF_SET_SKIP_STEPS(i);
872 }
873
874 static int
pf_commit_eth(uint32_t ticket,const char * anchor)875 pf_commit_eth(uint32_t ticket, const char *anchor)
876 {
877 struct pf_keth_ruleq *rules;
878 struct pf_keth_ruleset *rs;
879
880 rs = pf_find_keth_ruleset(anchor);
881 if (rs == NULL) {
882 return (EINVAL);
883 }
884
885 if (!rs->inactive.open ||
886 ticket != rs->inactive.ticket)
887 return (EBUSY);
888
889 PF_RULES_WASSERT();
890
891 pf_eth_calc_skip_steps(rs->inactive.rules);
892
893 rules = rs->active.rules;
894 ck_pr_store_ptr(&rs->active.rules, rs->inactive.rules);
895 rs->inactive.rules = rules;
896 rs->inactive.ticket = rs->active.ticket;
897
898 /* Clean up inactive rules (i.e. previously active rules), only when
899 * we're sure they're no longer used. */
900 NET_EPOCH_CALL(pf_rollback_eth_cb, &rs->epoch_ctx);
901
902 return (0);
903 }
904
905 #ifdef ALTQ
906 static uint16_t
pf_qname2qid(const char * qname)907 pf_qname2qid(const char *qname)
908 {
909 return (tagname2tag(&V_pf_qids, qname));
910 }
911
912 static void
pf_qid_unref(uint16_t qid)913 pf_qid_unref(uint16_t qid)
914 {
915 tag_unref(&V_pf_qids, qid);
916 }
917
918 static int
pf_begin_altq(u_int32_t * ticket)919 pf_begin_altq(u_int32_t *ticket)
920 {
921 struct pf_altq *altq, *tmp;
922 int error = 0;
923
924 PF_RULES_WASSERT();
925
926 /* Purge the old altq lists */
927 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
928 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
929 /* detach and destroy the discipline */
930 error = altq_remove(altq);
931 }
932 free(altq, M_PFALTQ);
933 }
934 TAILQ_INIT(V_pf_altq_ifs_inactive);
935 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
936 pf_qid_unref(altq->qid);
937 free(altq, M_PFALTQ);
938 }
939 TAILQ_INIT(V_pf_altqs_inactive);
940 if (error)
941 return (error);
942 *ticket = ++V_ticket_altqs_inactive;
943 V_altqs_inactive_open = 1;
944 return (0);
945 }
946
947 static int
pf_rollback_altq(u_int32_t ticket)948 pf_rollback_altq(u_int32_t ticket)
949 {
950 struct pf_altq *altq, *tmp;
951 int error = 0;
952
953 PF_RULES_WASSERT();
954
955 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
956 return (0);
957 /* Purge the old altq lists */
958 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
959 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
960 /* detach and destroy the discipline */
961 error = altq_remove(altq);
962 }
963 free(altq, M_PFALTQ);
964 }
965 TAILQ_INIT(V_pf_altq_ifs_inactive);
966 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
967 pf_qid_unref(altq->qid);
968 free(altq, M_PFALTQ);
969 }
970 TAILQ_INIT(V_pf_altqs_inactive);
971 V_altqs_inactive_open = 0;
972 return (error);
973 }
974
975 static int
pf_commit_altq(u_int32_t ticket)976 pf_commit_altq(u_int32_t ticket)
977 {
978 struct pf_altqqueue *old_altqs, *old_altq_ifs;
979 struct pf_altq *altq, *tmp;
980 int err, error = 0;
981
982 PF_RULES_WASSERT();
983
984 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
985 return (EBUSY);
986
987 /* swap altqs, keep the old. */
988 old_altqs = V_pf_altqs_active;
989 old_altq_ifs = V_pf_altq_ifs_active;
990 V_pf_altqs_active = V_pf_altqs_inactive;
991 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
992 V_pf_altqs_inactive = old_altqs;
993 V_pf_altq_ifs_inactive = old_altq_ifs;
994 V_ticket_altqs_active = V_ticket_altqs_inactive;
995
996 /* Attach new disciplines */
997 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
998 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
999 /* attach the discipline */
1000 error = altq_pfattach(altq);
1001 if (error == 0 && V_pf_altq_running)
1002 error = pf_enable_altq(altq);
1003 if (error != 0)
1004 return (error);
1005 }
1006 }
1007
1008 /* Purge the old altq lists */
1009 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1010 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1011 /* detach and destroy the discipline */
1012 if (V_pf_altq_running)
1013 error = pf_disable_altq(altq);
1014 err = altq_pfdetach(altq);
1015 if (err != 0 && error == 0)
1016 error = err;
1017 err = altq_remove(altq);
1018 if (err != 0 && error == 0)
1019 error = err;
1020 }
1021 free(altq, M_PFALTQ);
1022 }
1023 TAILQ_INIT(V_pf_altq_ifs_inactive);
1024 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1025 pf_qid_unref(altq->qid);
1026 free(altq, M_PFALTQ);
1027 }
1028 TAILQ_INIT(V_pf_altqs_inactive);
1029
1030 V_altqs_inactive_open = 0;
1031 return (error);
1032 }
1033
1034 static int
pf_enable_altq(struct pf_altq * altq)1035 pf_enable_altq(struct pf_altq *altq)
1036 {
1037 struct ifnet *ifp;
1038 struct tb_profile tb;
1039 int error = 0;
1040
1041 if ((ifp = ifunit(altq->ifname)) == NULL)
1042 return (EINVAL);
1043
1044 if (ifp->if_snd.altq_type != ALTQT_NONE)
1045 error = altq_enable(&ifp->if_snd);
1046
1047 /* set tokenbucket regulator */
1048 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1049 tb.rate = altq->ifbandwidth;
1050 tb.depth = altq->tbrsize;
1051 error = tbr_set(&ifp->if_snd, &tb);
1052 }
1053
1054 return (error);
1055 }
1056
1057 static int
pf_disable_altq(struct pf_altq * altq)1058 pf_disable_altq(struct pf_altq *altq)
1059 {
1060 struct ifnet *ifp;
1061 struct tb_profile tb;
1062 int error;
1063
1064 if ((ifp = ifunit(altq->ifname)) == NULL)
1065 return (EINVAL);
1066
1067 /*
1068 * when the discipline is no longer referenced, it was overridden
1069 * by a new one. if so, just return.
1070 */
1071 if (altq->altq_disc != ifp->if_snd.altq_disc)
1072 return (0);
1073
1074 error = altq_disable(&ifp->if_snd);
1075
1076 if (error == 0) {
1077 /* clear tokenbucket regulator */
1078 tb.rate = 0;
1079 error = tbr_set(&ifp->if_snd, &tb);
1080 }
1081
1082 return (error);
1083 }
1084
1085 static int
pf_altq_ifnet_event_add(struct ifnet * ifp,int remove,u_int32_t ticket,struct pf_altq * altq)1086 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1087 struct pf_altq *altq)
1088 {
1089 struct ifnet *ifp1;
1090 int error = 0;
1091
1092 /* Deactivate the interface in question */
1093 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1094 if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1095 (remove && ifp1 == ifp)) {
1096 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1097 } else {
1098 error = altq_add(ifp1, altq);
1099
1100 if (ticket != V_ticket_altqs_inactive)
1101 error = EBUSY;
1102
1103 if (error)
1104 free(altq, M_PFALTQ);
1105 }
1106
1107 return (error);
1108 }
1109
1110 void
pf_altq_ifnet_event(struct ifnet * ifp,int remove)1111 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1112 {
1113 struct pf_altq *a1, *a2, *a3;
1114 u_int32_t ticket;
1115 int error = 0;
1116
1117 /*
1118 * No need to re-evaluate the configuration for events on interfaces
1119 * that do not support ALTQ, as it's not possible for such
1120 * interfaces to be part of the configuration.
1121 */
1122 if (!ALTQ_IS_READY(&ifp->if_snd))
1123 return;
1124
1125 /* Interrupt userland queue modifications */
1126 if (V_altqs_inactive_open)
1127 pf_rollback_altq(V_ticket_altqs_inactive);
1128
1129 /* Start new altq ruleset */
1130 if (pf_begin_altq(&ticket))
1131 return;
1132
1133 /* Copy the current active set */
1134 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1135 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1136 if (a2 == NULL) {
1137 error = ENOMEM;
1138 break;
1139 }
1140 bcopy(a1, a2, sizeof(struct pf_altq));
1141
1142 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1143 if (error)
1144 break;
1145
1146 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1147 }
1148 if (error)
1149 goto out;
1150 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1151 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1152 if (a2 == NULL) {
1153 error = ENOMEM;
1154 break;
1155 }
1156 bcopy(a1, a2, sizeof(struct pf_altq));
1157
1158 if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
1159 error = EBUSY;
1160 free(a2, M_PFALTQ);
1161 break;
1162 }
1163 a2->altq_disc = NULL;
1164 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1165 if (strncmp(a3->ifname, a2->ifname,
1166 IFNAMSIZ) == 0) {
1167 a2->altq_disc = a3->altq_disc;
1168 break;
1169 }
1170 }
1171 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1172 if (error)
1173 break;
1174
1175 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1176 }
1177
1178 out:
1179 if (error != 0)
1180 pf_rollback_altq(ticket);
1181 else
1182 pf_commit_altq(ticket);
1183 }
1184 #endif /* ALTQ */
1185
1186 static struct pf_krule_global *
pf_rule_tree_alloc(int flags)1187 pf_rule_tree_alloc(int flags)
1188 {
1189 struct pf_krule_global *tree;
1190
1191 tree = malloc(sizeof(struct pf_krule_global), M_TEMP, flags);
1192 if (tree == NULL)
1193 return (NULL);
1194 RB_INIT(tree);
1195 return (tree);
1196 }
1197
1198 static void
pf_rule_tree_free(struct pf_krule_global * tree)1199 pf_rule_tree_free(struct pf_krule_global *tree)
1200 {
1201
1202 free(tree, M_TEMP);
1203 }
1204
1205 static int
pf_begin_rules(u_int32_t * ticket,int rs_num,const char * anchor)1206 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1207 {
1208 struct pf_krule_global *tree;
1209 struct pf_kruleset *rs;
1210 struct pf_krule *rule;
1211
1212 PF_RULES_WASSERT();
1213
1214 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1215 return (EINVAL);
1216 tree = pf_rule_tree_alloc(M_NOWAIT);
1217 if (tree == NULL)
1218 return (ENOMEM);
1219 rs = pf_find_or_create_kruleset(anchor);
1220 if (rs == NULL) {
1221 free(tree, M_TEMP);
1222 return (EINVAL);
1223 }
1224 pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1225 rs->rules[rs_num].inactive.tree = tree;
1226
1227 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1228 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1229 rs->rules[rs_num].inactive.rcount--;
1230 }
1231 *ticket = ++rs->rules[rs_num].inactive.ticket;
1232 rs->rules[rs_num].inactive.open = 1;
1233 return (0);
1234 }
1235
1236 static int
pf_rollback_rules(u_int32_t ticket,int rs_num,char * anchor)1237 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1238 {
1239 struct pf_kruleset *rs;
1240 struct pf_krule *rule;
1241
1242 PF_RULES_WASSERT();
1243
1244 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1245 return (EINVAL);
1246 rs = pf_find_kruleset(anchor);
1247 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1248 rs->rules[rs_num].inactive.ticket != ticket)
1249 return (0);
1250 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1251 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1252 rs->rules[rs_num].inactive.rcount--;
1253 }
1254 rs->rules[rs_num].inactive.open = 0;
1255 return (0);
1256 }
1257
1258 #define PF_MD5_UPD(st, elm) \
1259 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1260
1261 #define PF_MD5_UPD_STR(st, elm) \
1262 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1263
1264 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \
1265 (stor) = htonl((st)->elm); \
1266 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1267 } while (0)
1268
1269 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \
1270 (stor) = htons((st)->elm); \
1271 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1272 } while (0)
1273
1274 static void
pf_hash_rule_addr(MD5_CTX * ctx,struct pf_rule_addr * pfr)1275 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1276 {
1277 PF_MD5_UPD(pfr, addr.type);
1278 switch (pfr->addr.type) {
1279 case PF_ADDR_DYNIFTL:
1280 PF_MD5_UPD(pfr, addr.v.ifname);
1281 PF_MD5_UPD(pfr, addr.iflags);
1282 break;
1283 case PF_ADDR_TABLE:
1284 PF_MD5_UPD(pfr, addr.v.tblname);
1285 break;
1286 case PF_ADDR_ADDRMASK:
1287 /* XXX ignore af? */
1288 PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1289 PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1290 break;
1291 }
1292
1293 PF_MD5_UPD(pfr, port[0]);
1294 PF_MD5_UPD(pfr, port[1]);
1295 PF_MD5_UPD(pfr, neg);
1296 PF_MD5_UPD(pfr, port_op);
1297 }
1298
1299 static void
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)1300 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1301 {
1302 u_int16_t x;
1303 u_int32_t y;
1304
1305 pf_hash_rule_addr(ctx, &rule->src);
1306 pf_hash_rule_addr(ctx, &rule->dst);
1307 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1308 PF_MD5_UPD_STR(rule, label[i]);
1309 PF_MD5_UPD_STR(rule, ifname);
1310 PF_MD5_UPD_STR(rule, match_tagname);
1311 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1312 PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1313 PF_MD5_UPD_HTONL(rule, prob, y);
1314 PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1315 PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1316 PF_MD5_UPD(rule, uid.op);
1317 PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1318 PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1319 PF_MD5_UPD(rule, gid.op);
1320 PF_MD5_UPD_HTONL(rule, rule_flag, y);
1321 PF_MD5_UPD(rule, action);
1322 PF_MD5_UPD(rule, direction);
1323 PF_MD5_UPD(rule, af);
1324 PF_MD5_UPD(rule, quick);
1325 PF_MD5_UPD(rule, ifnot);
1326 PF_MD5_UPD(rule, match_tag_not);
1327 PF_MD5_UPD(rule, natpass);
1328 PF_MD5_UPD(rule, keep_state);
1329 PF_MD5_UPD(rule, proto);
1330 PF_MD5_UPD(rule, type);
1331 PF_MD5_UPD(rule, code);
1332 PF_MD5_UPD(rule, flags);
1333 PF_MD5_UPD(rule, flagset);
1334 PF_MD5_UPD(rule, allow_opts);
1335 PF_MD5_UPD(rule, rt);
1336 PF_MD5_UPD(rule, tos);
1337 PF_MD5_UPD(rule, scrub_flags);
1338 PF_MD5_UPD(rule, min_ttl);
1339 PF_MD5_UPD(rule, set_tos);
1340 if (rule->anchor != NULL)
1341 PF_MD5_UPD_STR(rule, anchor->path);
1342 }
1343
1344 static void
pf_hash_rule(struct pf_krule * rule)1345 pf_hash_rule(struct pf_krule *rule)
1346 {
1347 MD5_CTX ctx;
1348
1349 MD5Init(&ctx);
1350 pf_hash_rule_rolling(&ctx, rule);
1351 MD5Final(rule->md5sum, &ctx);
1352 }
1353
1354 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1355 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1356 {
1357
1358 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1359 }
1360
1361 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1362 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1363 {
1364 struct pf_kruleset *rs;
1365 struct pf_krule *rule, **old_array, *old_rule;
1366 struct pf_krulequeue *old_rules;
1367 struct pf_krule_global *old_tree;
1368 int error;
1369 u_int32_t old_rcount;
1370
1371 PF_RULES_WASSERT();
1372
1373 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1374 return (EINVAL);
1375 rs = pf_find_kruleset(anchor);
1376 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1377 ticket != rs->rules[rs_num].inactive.ticket)
1378 return (EBUSY);
1379
1380 /* Calculate checksum for the main ruleset */
1381 if (rs == &pf_main_ruleset) {
1382 error = pf_setup_pfsync_matching(rs);
1383 if (error != 0)
1384 return (error);
1385 }
1386
1387 /* Swap rules, keep the old. */
1388 old_rules = rs->rules[rs_num].active.ptr;
1389 old_rcount = rs->rules[rs_num].active.rcount;
1390 old_array = rs->rules[rs_num].active.ptr_array;
1391 old_tree = rs->rules[rs_num].active.tree;
1392
1393 rs->rules[rs_num].active.ptr =
1394 rs->rules[rs_num].inactive.ptr;
1395 rs->rules[rs_num].active.ptr_array =
1396 rs->rules[rs_num].inactive.ptr_array;
1397 rs->rules[rs_num].active.tree =
1398 rs->rules[rs_num].inactive.tree;
1399 rs->rules[rs_num].active.rcount =
1400 rs->rules[rs_num].inactive.rcount;
1401
1402 /* Attempt to preserve counter information. */
1403 if (V_pf_status.keep_counters && old_tree != NULL) {
1404 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1405 entries) {
1406 old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1407 if (old_rule == NULL) {
1408 continue;
1409 }
1410 pf_counter_u64_critical_enter();
1411 pf_counter_u64_rollup_protected(&rule->evaluations,
1412 pf_counter_u64_fetch(&old_rule->evaluations));
1413 pf_counter_u64_rollup_protected(&rule->packets[0],
1414 pf_counter_u64_fetch(&old_rule->packets[0]));
1415 pf_counter_u64_rollup_protected(&rule->packets[1],
1416 pf_counter_u64_fetch(&old_rule->packets[1]));
1417 pf_counter_u64_rollup_protected(&rule->bytes[0],
1418 pf_counter_u64_fetch(&old_rule->bytes[0]));
1419 pf_counter_u64_rollup_protected(&rule->bytes[1],
1420 pf_counter_u64_fetch(&old_rule->bytes[1]));
1421 pf_counter_u64_critical_exit();
1422 }
1423 }
1424
1425 rs->rules[rs_num].inactive.ptr = old_rules;
1426 rs->rules[rs_num].inactive.ptr_array = old_array;
1427 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1428 rs->rules[rs_num].inactive.rcount = old_rcount;
1429
1430 rs->rules[rs_num].active.ticket =
1431 rs->rules[rs_num].inactive.ticket;
1432 pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1433
1434 /* Purge the old rule list. */
1435 PF_UNLNKDRULES_LOCK();
1436 while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1437 pf_unlink_rule_locked(old_rules, rule);
1438 PF_UNLNKDRULES_UNLOCK();
1439 if (rs->rules[rs_num].inactive.ptr_array)
1440 free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1441 rs->rules[rs_num].inactive.ptr_array = NULL;
1442 rs->rules[rs_num].inactive.rcount = 0;
1443 rs->rules[rs_num].inactive.open = 0;
1444 pf_remove_if_empty_kruleset(rs);
1445 free(old_tree, M_TEMP);
1446
1447 return (0);
1448 }
1449
1450 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1451 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1452 {
1453 MD5_CTX ctx;
1454 struct pf_krule *rule;
1455 int rs_cnt;
1456 u_int8_t digest[PF_MD5_DIGEST_LENGTH];
1457
1458 MD5Init(&ctx);
1459 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1460 /* XXX PF_RULESET_SCRUB as well? */
1461 if (rs_cnt == PF_RULESET_SCRUB)
1462 continue;
1463
1464 if (rs->rules[rs_cnt].inactive.ptr_array)
1465 free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1466 rs->rules[rs_cnt].inactive.ptr_array = NULL;
1467
1468 if (rs->rules[rs_cnt].inactive.rcount) {
1469 rs->rules[rs_cnt].inactive.ptr_array =
1470 mallocarray(rs->rules[rs_cnt].inactive.rcount,
1471 sizeof(struct pf_rule **),
1472 M_TEMP, M_NOWAIT);
1473
1474 if (!rs->rules[rs_cnt].inactive.ptr_array)
1475 return (ENOMEM);
1476 }
1477
1478 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1479 entries) {
1480 pf_hash_rule_rolling(&ctx, rule);
1481 (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1482 }
1483 }
1484
1485 MD5Final(digest, &ctx);
1486 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1487 return (0);
1488 }
1489
1490 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1491 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1492 {
1493 int error = 0;
1494
1495 switch (addr->type) {
1496 case PF_ADDR_TABLE:
1497 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1498 if (addr->p.tbl == NULL)
1499 error = ENOMEM;
1500 break;
1501 default:
1502 error = EINVAL;
1503 }
1504
1505 return (error);
1506 }
1507
1508 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1509 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1510 sa_family_t af)
1511 {
1512 int error = 0;
1513
1514 switch (addr->type) {
1515 case PF_ADDR_TABLE:
1516 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1517 if (addr->p.tbl == NULL)
1518 error = ENOMEM;
1519 break;
1520 case PF_ADDR_DYNIFTL:
1521 error = pfi_dynaddr_setup(addr, af);
1522 break;
1523 }
1524
1525 return (error);
1526 }
1527
1528 static void
pf_addr_copyout(struct pf_addr_wrap * addr)1529 pf_addr_copyout(struct pf_addr_wrap *addr)
1530 {
1531
1532 switch (addr->type) {
1533 case PF_ADDR_DYNIFTL:
1534 pfi_dynaddr_copyout(addr);
1535 break;
1536 case PF_ADDR_TABLE:
1537 pf_tbladdr_copyout(addr);
1538 break;
1539 }
1540 }
1541
1542 static void
pf_src_node_copy(const struct pf_ksrc_node * in,struct pf_src_node * out)1543 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1544 {
1545 int secs = time_uptime, diff;
1546
1547 bzero(out, sizeof(struct pf_src_node));
1548
1549 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1550 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1551
1552 if (in->rule.ptr != NULL)
1553 out->rule.nr = in->rule.ptr->nr;
1554
1555 for (int i = 0; i < 2; i++) {
1556 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1557 out->packets[i] = counter_u64_fetch(in->packets[i]);
1558 }
1559
1560 out->states = in->states;
1561 out->conn = in->conn;
1562 out->af = in->af;
1563 out->ruletype = in->ruletype;
1564
1565 out->creation = secs - in->creation;
1566 if (out->expire > secs)
1567 out->expire -= secs;
1568 else
1569 out->expire = 0;
1570
1571 /* Adjust the connection rate estimate. */
1572 diff = secs - in->conn_rate.last;
1573 if (diff >= in->conn_rate.seconds)
1574 out->conn_rate.count = 0;
1575 else
1576 out->conn_rate.count -=
1577 in->conn_rate.count * diff /
1578 in->conn_rate.seconds;
1579 }
1580
1581 #ifdef ALTQ
1582 /*
1583 * Handle export of struct pf_kaltq to user binaries that may be using any
1584 * version of struct pf_altq.
1585 */
1586 static int
pf_export_kaltq(struct pf_altq * q,struct pfioc_altq_v1 * pa,size_t ioc_size)1587 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1588 {
1589 u_int32_t version;
1590
1591 if (ioc_size == sizeof(struct pfioc_altq_v0))
1592 version = 0;
1593 else
1594 version = pa->version;
1595
1596 if (version > PFIOC_ALTQ_VERSION)
1597 return (EINVAL);
1598
1599 #define ASSIGN(x) exported_q->x = q->x
1600 #define COPY(x) \
1601 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1602 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1603 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1604
1605 switch (version) {
1606 case 0: {
1607 struct pf_altq_v0 *exported_q =
1608 &((struct pfioc_altq_v0 *)pa)->altq;
1609
1610 COPY(ifname);
1611
1612 ASSIGN(scheduler);
1613 ASSIGN(tbrsize);
1614 exported_q->tbrsize = SATU16(q->tbrsize);
1615 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1616
1617 COPY(qname);
1618 COPY(parent);
1619 ASSIGN(parent_qid);
1620 exported_q->bandwidth = SATU32(q->bandwidth);
1621 ASSIGN(priority);
1622 ASSIGN(local_flags);
1623
1624 ASSIGN(qlimit);
1625 ASSIGN(flags);
1626
1627 if (q->scheduler == ALTQT_HFSC) {
1628 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1629 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1630 SATU32(q->pq_u.hfsc_opts.x)
1631
1632 ASSIGN_OPT_SATU32(rtsc_m1);
1633 ASSIGN_OPT(rtsc_d);
1634 ASSIGN_OPT_SATU32(rtsc_m2);
1635
1636 ASSIGN_OPT_SATU32(lssc_m1);
1637 ASSIGN_OPT(lssc_d);
1638 ASSIGN_OPT_SATU32(lssc_m2);
1639
1640 ASSIGN_OPT_SATU32(ulsc_m1);
1641 ASSIGN_OPT(ulsc_d);
1642 ASSIGN_OPT_SATU32(ulsc_m2);
1643
1644 ASSIGN_OPT(flags);
1645
1646 #undef ASSIGN_OPT
1647 #undef ASSIGN_OPT_SATU32
1648 } else
1649 COPY(pq_u);
1650
1651 ASSIGN(qid);
1652 break;
1653 }
1654 case 1: {
1655 struct pf_altq_v1 *exported_q =
1656 &((struct pfioc_altq_v1 *)pa)->altq;
1657
1658 COPY(ifname);
1659
1660 ASSIGN(scheduler);
1661 ASSIGN(tbrsize);
1662 ASSIGN(ifbandwidth);
1663
1664 COPY(qname);
1665 COPY(parent);
1666 ASSIGN(parent_qid);
1667 ASSIGN(bandwidth);
1668 ASSIGN(priority);
1669 ASSIGN(local_flags);
1670
1671 ASSIGN(qlimit);
1672 ASSIGN(flags);
1673 COPY(pq_u);
1674
1675 ASSIGN(qid);
1676 break;
1677 }
1678 default:
1679 panic("%s: unhandled struct pfioc_altq version", __func__);
1680 break;
1681 }
1682
1683 #undef ASSIGN
1684 #undef COPY
1685 #undef SATU16
1686 #undef SATU32
1687
1688 return (0);
1689 }
1690
1691 /*
1692 * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1693 * that may be using any version of it.
1694 */
1695 static int
pf_import_kaltq(struct pfioc_altq_v1 * pa,struct pf_altq * q,size_t ioc_size)1696 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1697 {
1698 u_int32_t version;
1699
1700 if (ioc_size == sizeof(struct pfioc_altq_v0))
1701 version = 0;
1702 else
1703 version = pa->version;
1704
1705 if (version > PFIOC_ALTQ_VERSION)
1706 return (EINVAL);
1707
1708 #define ASSIGN(x) q->x = imported_q->x
1709 #define COPY(x) \
1710 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1711
1712 switch (version) {
1713 case 0: {
1714 struct pf_altq_v0 *imported_q =
1715 &((struct pfioc_altq_v0 *)pa)->altq;
1716
1717 COPY(ifname);
1718
1719 ASSIGN(scheduler);
1720 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1721 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1722
1723 COPY(qname);
1724 COPY(parent);
1725 ASSIGN(parent_qid);
1726 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1727 ASSIGN(priority);
1728 ASSIGN(local_flags);
1729
1730 ASSIGN(qlimit);
1731 ASSIGN(flags);
1732
1733 if (imported_q->scheduler == ALTQT_HFSC) {
1734 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1735
1736 /*
1737 * The m1 and m2 parameters are being copied from
1738 * 32-bit to 64-bit.
1739 */
1740 ASSIGN_OPT(rtsc_m1);
1741 ASSIGN_OPT(rtsc_d);
1742 ASSIGN_OPT(rtsc_m2);
1743
1744 ASSIGN_OPT(lssc_m1);
1745 ASSIGN_OPT(lssc_d);
1746 ASSIGN_OPT(lssc_m2);
1747
1748 ASSIGN_OPT(ulsc_m1);
1749 ASSIGN_OPT(ulsc_d);
1750 ASSIGN_OPT(ulsc_m2);
1751
1752 ASSIGN_OPT(flags);
1753
1754 #undef ASSIGN_OPT
1755 } else
1756 COPY(pq_u);
1757
1758 ASSIGN(qid);
1759 break;
1760 }
1761 case 1: {
1762 struct pf_altq_v1 *imported_q =
1763 &((struct pfioc_altq_v1 *)pa)->altq;
1764
1765 COPY(ifname);
1766
1767 ASSIGN(scheduler);
1768 ASSIGN(tbrsize);
1769 ASSIGN(ifbandwidth);
1770
1771 COPY(qname);
1772 COPY(parent);
1773 ASSIGN(parent_qid);
1774 ASSIGN(bandwidth);
1775 ASSIGN(priority);
1776 ASSIGN(local_flags);
1777
1778 ASSIGN(qlimit);
1779 ASSIGN(flags);
1780 COPY(pq_u);
1781
1782 ASSIGN(qid);
1783 break;
1784 }
1785 default:
1786 panic("%s: unhandled struct pfioc_altq version", __func__);
1787 break;
1788 }
1789
1790 #undef ASSIGN
1791 #undef COPY
1792
1793 return (0);
1794 }
1795
1796 static struct pf_altq *
pf_altq_get_nth_active(u_int32_t n)1797 pf_altq_get_nth_active(u_int32_t n)
1798 {
1799 struct pf_altq *altq;
1800 u_int32_t nr;
1801
1802 nr = 0;
1803 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1804 if (nr == n)
1805 return (altq);
1806 nr++;
1807 }
1808
1809 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1810 if (nr == n)
1811 return (altq);
1812 nr++;
1813 }
1814
1815 return (NULL);
1816 }
1817 #endif /* ALTQ */
1818
1819 struct pf_krule *
pf_krule_alloc(void)1820 pf_krule_alloc(void)
1821 {
1822 struct pf_krule *rule;
1823
1824 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
1825 mtx_init(&rule->rpool.mtx, "pf_krule_pool", NULL, MTX_DEF);
1826 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
1827 M_WAITOK | M_ZERO);
1828 return (rule);
1829 }
1830
1831 void
pf_krule_free(struct pf_krule * rule)1832 pf_krule_free(struct pf_krule *rule)
1833 {
1834 #ifdef PF_WANT_32_TO_64_COUNTER
1835 bool wowned;
1836 #endif
1837
1838 if (rule == NULL)
1839 return;
1840
1841 #ifdef PF_WANT_32_TO_64_COUNTER
1842 if (rule->allrulelinked) {
1843 wowned = PF_RULES_WOWNED();
1844 if (!wowned)
1845 PF_RULES_WLOCK();
1846 LIST_REMOVE(rule, allrulelist);
1847 V_pf_allrulecount--;
1848 if (!wowned)
1849 PF_RULES_WUNLOCK();
1850 }
1851 #endif
1852
1853 pf_counter_u64_deinit(&rule->evaluations);
1854 for (int i = 0; i < 2; i++) {
1855 pf_counter_u64_deinit(&rule->packets[i]);
1856 pf_counter_u64_deinit(&rule->bytes[i]);
1857 }
1858 counter_u64_free(rule->states_cur);
1859 counter_u64_free(rule->states_tot);
1860 counter_u64_free(rule->src_nodes);
1861 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
1862
1863 mtx_destroy(&rule->rpool.mtx);
1864 free(rule, M_PFRULE);
1865 }
1866
1867 void
pf_krule_clear_counters(struct pf_krule * rule)1868 pf_krule_clear_counters(struct pf_krule *rule)
1869 {
1870 pf_counter_u64_zero(&rule->evaluations);
1871 for (int i = 0; i < 2; i++) {
1872 pf_counter_u64_zero(&rule->packets[i]);
1873 pf_counter_u64_zero(&rule->bytes[i]);
1874 }
1875 counter_u64_zero(rule->states_tot);
1876 }
1877
1878 static void
pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr * kpool,struct pf_pooladdr * pool)1879 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1880 struct pf_pooladdr *pool)
1881 {
1882
1883 bzero(pool, sizeof(*pool));
1884 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1885 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1886 }
1887
1888 static int
pf_pooladdr_to_kpooladdr(const struct pf_pooladdr * pool,struct pf_kpooladdr * kpool)1889 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1890 struct pf_kpooladdr *kpool)
1891 {
1892 int ret;
1893
1894 bzero(kpool, sizeof(*kpool));
1895 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1896 ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1897 sizeof(kpool->ifname));
1898 return (ret);
1899 }
1900
1901 static void
pf_pool_to_kpool(const struct pf_pool * pool,struct pf_kpool * kpool)1902 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1903 {
1904 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1905 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1906
1907 bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1908 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1909
1910 kpool->tblidx = pool->tblidx;
1911 kpool->proxy_port[0] = pool->proxy_port[0];
1912 kpool->proxy_port[1] = pool->proxy_port[1];
1913 kpool->opts = pool->opts;
1914 }
1915
1916 static int
pf_rule_to_krule(const struct pf_rule * rule,struct pf_krule * krule)1917 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1918 {
1919 int ret;
1920
1921 #ifndef INET
1922 if (rule->af == AF_INET) {
1923 return (EAFNOSUPPORT);
1924 }
1925 #endif /* INET */
1926 #ifndef INET6
1927 if (rule->af == AF_INET6) {
1928 return (EAFNOSUPPORT);
1929 }
1930 #endif /* INET6 */
1931
1932 ret = pf_check_rule_addr(&rule->src);
1933 if (ret != 0)
1934 return (ret);
1935 ret = pf_check_rule_addr(&rule->dst);
1936 if (ret != 0)
1937 return (ret);
1938
1939 bcopy(&rule->src, &krule->src, sizeof(rule->src));
1940 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1941
1942 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1943 if (ret != 0)
1944 return (ret);
1945 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1946 if (ret != 0)
1947 return (ret);
1948 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1949 if (ret != 0)
1950 return (ret);
1951 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1952 if (ret != 0)
1953 return (ret);
1954 ret = pf_user_strcpy(krule->tagname, rule->tagname,
1955 sizeof(rule->tagname));
1956 if (ret != 0)
1957 return (ret);
1958 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1959 sizeof(rule->match_tagname));
1960 if (ret != 0)
1961 return (ret);
1962 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1963 sizeof(rule->overload_tblname));
1964 if (ret != 0)
1965 return (ret);
1966
1967 pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1968
1969 /* Don't allow userspace to set evaluations, packets or bytes. */
1970 /* kif, anchor, overload_tbl are not copied over. */
1971
1972 krule->os_fingerprint = rule->os_fingerprint;
1973
1974 krule->rtableid = rule->rtableid;
1975 /* pf_rule->timeout is smaller than pf_krule->timeout */
1976 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
1977 krule->max_states = rule->max_states;
1978 krule->max_src_nodes = rule->max_src_nodes;
1979 krule->max_src_states = rule->max_src_states;
1980 krule->max_src_conn = rule->max_src_conn;
1981 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1982 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1983 krule->qid = rule->qid;
1984 krule->pqid = rule->pqid;
1985 krule->nr = rule->nr;
1986 krule->prob = rule->prob;
1987 krule->cuid = rule->cuid;
1988 krule->cpid = rule->cpid;
1989
1990 krule->return_icmp = rule->return_icmp;
1991 krule->return_icmp6 = rule->return_icmp6;
1992 krule->max_mss = rule->max_mss;
1993 krule->tag = rule->tag;
1994 krule->match_tag = rule->match_tag;
1995 krule->scrub_flags = rule->scrub_flags;
1996
1997 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1998 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1999
2000 krule->rule_flag = rule->rule_flag;
2001 krule->action = rule->action;
2002 krule->direction = rule->direction;
2003 krule->log = rule->log;
2004 krule->logif = rule->logif;
2005 krule->quick = rule->quick;
2006 krule->ifnot = rule->ifnot;
2007 krule->match_tag_not = rule->match_tag_not;
2008 krule->natpass = rule->natpass;
2009
2010 krule->keep_state = rule->keep_state;
2011 krule->af = rule->af;
2012 krule->proto = rule->proto;
2013 krule->type = rule->type;
2014 krule->code = rule->code;
2015 krule->flags = rule->flags;
2016 krule->flagset = rule->flagset;
2017 krule->min_ttl = rule->min_ttl;
2018 krule->allow_opts = rule->allow_opts;
2019 krule->rt = rule->rt;
2020 krule->return_ttl = rule->return_ttl;
2021 krule->tos = rule->tos;
2022 krule->set_tos = rule->set_tos;
2023
2024 krule->flush = rule->flush;
2025 krule->prio = rule->prio;
2026 krule->set_prio[0] = rule->set_prio[0];
2027 krule->set_prio[1] = rule->set_prio[1];
2028
2029 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2030
2031 return (0);
2032 }
2033
2034 int
pf_ioctl_getrules(struct pfioc_rule * pr)2035 pf_ioctl_getrules(struct pfioc_rule *pr)
2036 {
2037 struct pf_kruleset *ruleset;
2038 struct pf_krule *tail;
2039 int rs_num;
2040
2041 PF_RULES_WLOCK();
2042 ruleset = pf_find_kruleset(pr->anchor);
2043 if (ruleset == NULL) {
2044 PF_RULES_WUNLOCK();
2045 return (EINVAL);
2046 }
2047 rs_num = pf_get_ruleset_number(pr->rule.action);
2048 if (rs_num >= PF_RULESET_MAX) {
2049 PF_RULES_WUNLOCK();
2050 return (EINVAL);
2051 }
2052 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2053 pf_krulequeue);
2054 if (tail)
2055 pr->nr = tail->nr + 1;
2056 else
2057 pr->nr = 0;
2058 pr->ticket = ruleset->rules[rs_num].active.ticket;
2059 PF_RULES_WUNLOCK();
2060
2061 return (0);
2062 }
2063
2064 int
pf_ioctl_addrule(struct pf_krule * rule,uint32_t ticket,uint32_t pool_ticket,const char * anchor,const char * anchor_call,uid_t uid,pid_t pid)2065 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
2066 uint32_t pool_ticket, const char *anchor, const char *anchor_call,
2067 uid_t uid, pid_t pid)
2068 {
2069 struct pf_kruleset *ruleset;
2070 struct pf_krule *tail;
2071 struct pf_kpooladdr *pa;
2072 struct pfi_kkif *kif = NULL;
2073 int rs_num;
2074 int error = 0;
2075
2076 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
2077 error = EINVAL;
2078 goto errout_unlocked;
2079 }
2080
2081 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
2082
2083 if (rule->ifname[0])
2084 kif = pf_kkif_create(M_WAITOK);
2085 pf_counter_u64_init(&rule->evaluations, M_WAITOK);
2086 for (int i = 0; i < 2; i++) {
2087 pf_counter_u64_init(&rule->packets[i], M_WAITOK);
2088 pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
2089 }
2090 rule->states_cur = counter_u64_alloc(M_WAITOK);
2091 rule->states_tot = counter_u64_alloc(M_WAITOK);
2092 rule->src_nodes = counter_u64_alloc(M_WAITOK);
2093 rule->cuid = uid;
2094 rule->cpid = pid;
2095 TAILQ_INIT(&rule->rpool.list);
2096
2097 PF_CONFIG_LOCK();
2098 PF_RULES_WLOCK();
2099 #ifdef PF_WANT_32_TO_64_COUNTER
2100 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
2101 MPASS(!rule->allrulelinked);
2102 rule->allrulelinked = true;
2103 V_pf_allrulecount++;
2104 #endif
2105 ruleset = pf_find_kruleset(anchor);
2106 if (ruleset == NULL)
2107 ERROUT(EINVAL);
2108 rs_num = pf_get_ruleset_number(rule->action);
2109 if (rs_num >= PF_RULESET_MAX)
2110 ERROUT(EINVAL);
2111 if (ticket != ruleset->rules[rs_num].inactive.ticket) {
2112 DPFPRINTF(PF_DEBUG_MISC,
2113 ("ticket: %d != [%d]%d\n", ticket, rs_num,
2114 ruleset->rules[rs_num].inactive.ticket));
2115 ERROUT(EBUSY);
2116 }
2117 if (pool_ticket != V_ticket_pabuf) {
2118 DPFPRINTF(PF_DEBUG_MISC,
2119 ("pool_ticket: %d != %d\n", pool_ticket,
2120 V_ticket_pabuf));
2121 ERROUT(EBUSY);
2122 }
2123 /*
2124 * XXXMJG hack: there is no mechanism to ensure they started the
2125 * transaction. Ticket checked above may happen to match by accident,
2126 * even if nobody called DIOCXBEGIN, let alone this process.
2127 * Partially work around it by checking if the RB tree got allocated,
2128 * see pf_begin_rules.
2129 */
2130 if (ruleset->rules[rs_num].inactive.tree == NULL) {
2131 ERROUT(EINVAL);
2132 }
2133
2134 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
2135 pf_krulequeue);
2136 if (tail)
2137 rule->nr = tail->nr + 1;
2138 else
2139 rule->nr = 0;
2140 if (rule->ifname[0]) {
2141 rule->kif = pfi_kkif_attach(kif, rule->ifname);
2142 kif = NULL;
2143 pfi_kkif_ref(rule->kif);
2144 } else
2145 rule->kif = NULL;
2146
2147 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
2148 error = EBUSY;
2149
2150 #ifdef ALTQ
2151 /* set queue IDs */
2152 if (rule->qname[0] != 0) {
2153 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
2154 error = EBUSY;
2155 else if (rule->pqname[0] != 0) {
2156 if ((rule->pqid =
2157 pf_qname2qid(rule->pqname)) == 0)
2158 error = EBUSY;
2159 } else
2160 rule->pqid = rule->qid;
2161 }
2162 #endif
2163 if (rule->tagname[0])
2164 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
2165 error = EBUSY;
2166 if (rule->match_tagname[0])
2167 if ((rule->match_tag =
2168 pf_tagname2tag(rule->match_tagname)) == 0)
2169 error = EBUSY;
2170 if (rule->rt && !rule->direction)
2171 error = EINVAL;
2172 if (!rule->log)
2173 rule->logif = 0;
2174 if (rule->logif >= PFLOGIFS_MAX)
2175 error = EINVAL;
2176 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
2177 error = ENOMEM;
2178 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
2179 error = ENOMEM;
2180 if (pf_kanchor_setup(rule, ruleset, anchor_call))
2181 error = EINVAL;
2182 if (rule->scrub_flags & PFSTATE_SETPRIO &&
2183 (rule->set_prio[0] > PF_PRIO_MAX ||
2184 rule->set_prio[1] > PF_PRIO_MAX))
2185 error = EINVAL;
2186 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2187 if (pa->addr.type == PF_ADDR_TABLE) {
2188 pa->addr.p.tbl = pfr_attach_table(ruleset,
2189 pa->addr.v.tblname);
2190 if (pa->addr.p.tbl == NULL)
2191 error = ENOMEM;
2192 }
2193
2194 rule->overload_tbl = NULL;
2195 if (rule->overload_tblname[0]) {
2196 if ((rule->overload_tbl = pfr_attach_table(ruleset,
2197 rule->overload_tblname)) == NULL)
2198 error = EINVAL;
2199 else
2200 rule->overload_tbl->pfrkt_flags |=
2201 PFR_TFLAG_ACTIVE;
2202 }
2203
2204 pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
2205 if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
2206 (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
2207 (rule->rt > PF_NOPFROUTE)) &&
2208 (TAILQ_FIRST(&rule->rpool.list) == NULL))
2209 error = EINVAL;
2210
2211 if (error) {
2212 pf_free_rule(rule);
2213 rule = NULL;
2214 ERROUT(error);
2215 }
2216
2217 rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
2218 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2219 rule, entries);
2220 ruleset->rules[rs_num].inactive.rcount++;
2221
2222 PF_RULES_WUNLOCK();
2223 pf_hash_rule(rule);
2224 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
2225 PF_RULES_WLOCK();
2226 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
2227 ruleset->rules[rs_num].inactive.rcount--;
2228 pf_free_rule(rule);
2229 rule = NULL;
2230 ERROUT(EEXIST);
2231 }
2232 PF_CONFIG_UNLOCK();
2233
2234 return (0);
2235
2236 #undef ERROUT
2237 errout:
2238 PF_RULES_WUNLOCK();
2239 PF_CONFIG_UNLOCK();
2240 errout_unlocked:
2241 pf_kkif_free(kif);
2242 pf_krule_free(rule);
2243 return (error);
2244 }
2245
2246 static bool
pf_label_match(const struct pf_krule * rule,const char * label)2247 pf_label_match(const struct pf_krule *rule, const char *label)
2248 {
2249 int i = 0;
2250
2251 while (*rule->label[i]) {
2252 if (strcmp(rule->label[i], label) == 0)
2253 return (true);
2254 i++;
2255 }
2256
2257 return (false);
2258 }
2259
2260 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)2261 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2262 {
2263 struct pf_kstate *s;
2264 int more = 0;
2265
2266 s = pf_find_state_all(key, dir, &more);
2267 if (s == NULL)
2268 return (0);
2269
2270 if (more) {
2271 PF_STATE_UNLOCK(s);
2272 return (0);
2273 }
2274
2275 pf_unlink_state(s);
2276 return (1);
2277 }
2278
2279 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)2280 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2281 {
2282 struct pf_kstate *s;
2283 struct pf_state_key *sk;
2284 struct pf_addr *srcaddr, *dstaddr;
2285 struct pf_state_key_cmp match_key;
2286 int idx, killed = 0;
2287 unsigned int dir;
2288 u_int16_t srcport, dstport;
2289 struct pfi_kkif *kif;
2290
2291 relock_DIOCKILLSTATES:
2292 PF_HASHROW_LOCK(ih);
2293 LIST_FOREACH(s, &ih->states, entry) {
2294 /* For floating states look at the original kif. */
2295 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2296
2297 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
2298 if (s->direction == PF_OUT) {
2299 srcaddr = &sk->addr[1];
2300 dstaddr = &sk->addr[0];
2301 srcport = sk->port[1];
2302 dstport = sk->port[0];
2303 } else {
2304 srcaddr = &sk->addr[0];
2305 dstaddr = &sk->addr[1];
2306 srcport = sk->port[0];
2307 dstport = sk->port[1];
2308 }
2309
2310 if (psk->psk_af && sk->af != psk->psk_af)
2311 continue;
2312
2313 if (psk->psk_proto && psk->psk_proto != sk->proto)
2314 continue;
2315
2316 if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2317 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2318 continue;
2319
2320 if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2321 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2322 continue;
2323
2324 if (! PF_MATCHA(psk->psk_rt_addr.neg,
2325 &psk->psk_rt_addr.addr.v.a.addr,
2326 &psk->psk_rt_addr.addr.v.a.mask,
2327 &s->rt_addr, sk->af))
2328 continue;
2329
2330 if (psk->psk_src.port_op != 0 &&
2331 ! pf_match_port(psk->psk_src.port_op,
2332 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2333 continue;
2334
2335 if (psk->psk_dst.port_op != 0 &&
2336 ! pf_match_port(psk->psk_dst.port_op,
2337 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2338 continue;
2339
2340 if (psk->psk_label[0] &&
2341 ! pf_label_match(s->rule.ptr, psk->psk_label))
2342 continue;
2343
2344 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2345 kif->pfik_name))
2346 continue;
2347
2348 if (psk->psk_kill_match) {
2349 /* Create the key to find matching states, with lock
2350 * held. */
2351
2352 bzero(&match_key, sizeof(match_key));
2353
2354 if (s->direction == PF_OUT) {
2355 dir = PF_IN;
2356 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
2357 } else {
2358 dir = PF_OUT;
2359 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
2360 }
2361
2362 match_key.af = s->key[idx]->af;
2363 match_key.proto = s->key[idx]->proto;
2364 PF_ACPY(&match_key.addr[0],
2365 &s->key[idx]->addr[1], match_key.af);
2366 match_key.port[0] = s->key[idx]->port[1];
2367 PF_ACPY(&match_key.addr[1],
2368 &s->key[idx]->addr[0], match_key.af);
2369 match_key.port[1] = s->key[idx]->port[0];
2370 }
2371
2372 pf_unlink_state(s);
2373 killed++;
2374
2375 if (psk->psk_kill_match)
2376 killed += pf_kill_matching_state(&match_key, dir);
2377
2378 goto relock_DIOCKILLSTATES;
2379 }
2380 PF_HASHROW_UNLOCK(ih);
2381
2382 return (killed);
2383 }
2384
2385 int
pf_start(void)2386 pf_start(void)
2387 {
2388 int error = 0;
2389
2390 sx_xlock(&V_pf_ioctl_lock);
2391 if (V_pf_status.running)
2392 error = EEXIST;
2393 else {
2394 hook_pf();
2395 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2396 hook_pf_eth();
2397 V_pf_status.running = 1;
2398 V_pf_status.since = time_second;
2399 new_unrhdr64(&V_pf_stateid, time_second);
2400
2401 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2402 }
2403 sx_xunlock(&V_pf_ioctl_lock);
2404
2405 return (error);
2406 }
2407
2408 int
pf_stop(void)2409 pf_stop(void)
2410 {
2411 int error = 0;
2412
2413 sx_xlock(&V_pf_ioctl_lock);
2414 if (!V_pf_status.running)
2415 error = ENOENT;
2416 else {
2417 V_pf_status.running = 0;
2418 dehook_pf();
2419 dehook_pf_eth();
2420 V_pf_status.since = time_second;
2421 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2422 }
2423 sx_xunlock(&V_pf_ioctl_lock);
2424
2425 return (error);
2426 }
2427
2428 void
pf_ioctl_clear_status(void)2429 pf_ioctl_clear_status(void)
2430 {
2431 PF_RULES_WLOCK();
2432 for (int i = 0; i < PFRES_MAX; i++)
2433 counter_u64_zero(V_pf_status.counters[i]);
2434 for (int i = 0; i < FCNT_MAX; i++)
2435 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2436 for (int i = 0; i < SCNT_MAX; i++)
2437 counter_u64_zero(V_pf_status.scounters[i]);
2438 for (int i = 0; i < KLCNT_MAX; i++)
2439 counter_u64_zero(V_pf_status.lcounters[i]);
2440 V_pf_status.since = time_second;
2441 if (*V_pf_status.ifname)
2442 pfi_update_status(V_pf_status.ifname, NULL);
2443 PF_RULES_WUNLOCK();
2444 }
2445
2446 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2447 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2448 {
2449 uint32_t old;
2450
2451 if (timeout < 0 || timeout >= PFTM_MAX ||
2452 seconds < 0)
2453 return (EINVAL);
2454
2455 PF_RULES_WLOCK();
2456 old = V_pf_default_rule.timeout[timeout];
2457 if (timeout == PFTM_INTERVAL && seconds == 0)
2458 seconds = 1;
2459 V_pf_default_rule.timeout[timeout] = seconds;
2460 if (timeout == PFTM_INTERVAL && seconds < old)
2461 wakeup(pf_purge_thread);
2462
2463 if (prev_seconds != NULL)
2464 *prev_seconds = old;
2465
2466 PF_RULES_WUNLOCK();
2467
2468 return (0);
2469 }
2470
2471 int
pf_ioctl_get_timeout(int timeout,int * seconds)2472 pf_ioctl_get_timeout(int timeout, int *seconds)
2473 {
2474 PF_RULES_RLOCK_TRACKER;
2475
2476 if (timeout < 0 || timeout >= PFTM_MAX)
2477 return (EINVAL);
2478
2479 PF_RULES_RLOCK();
2480 *seconds = V_pf_default_rule.timeout[timeout];
2481 PF_RULES_RUNLOCK();
2482
2483 return (0);
2484 }
2485
2486 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2487 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2488 {
2489
2490 PF_RULES_WLOCK();
2491 if (index < 0 || index >= PF_LIMIT_MAX ||
2492 V_pf_limits[index].zone == NULL) {
2493 PF_RULES_WUNLOCK();
2494 return (EINVAL);
2495 }
2496 uma_zone_set_max(V_pf_limits[index].zone, limit);
2497 if (old_limit != NULL)
2498 *old_limit = V_pf_limits[index].limit;
2499 V_pf_limits[index].limit = limit;
2500 PF_RULES_WUNLOCK();
2501
2502 return (0);
2503 }
2504
2505 int
pf_ioctl_get_limit(int index,unsigned int * limit)2506 pf_ioctl_get_limit(int index, unsigned int *limit)
2507 {
2508 PF_RULES_RLOCK_TRACKER;
2509
2510 if (index < 0 || index >= PF_LIMIT_MAX)
2511 return (EINVAL);
2512
2513 PF_RULES_RLOCK();
2514 *limit = V_pf_limits[index].limit;
2515 PF_RULES_RUNLOCK();
2516
2517 return (0);
2518 }
2519
2520 int
pf_ioctl_begin_addrs(uint32_t * ticket)2521 pf_ioctl_begin_addrs(uint32_t *ticket)
2522 {
2523 PF_RULES_WLOCK();
2524 pf_empty_kpool(&V_pf_pabuf);
2525 *ticket = ++V_ticket_pabuf;
2526 PF_RULES_WUNLOCK();
2527
2528 return (0);
2529 }
2530
2531 int
pf_ioctl_add_addr(struct pfioc_pooladdr * pp)2532 pf_ioctl_add_addr(struct pfioc_pooladdr *pp)
2533 {
2534 struct pf_kpooladdr *pa = NULL;
2535 struct pfi_kkif *kif = NULL;
2536 int error;
2537
2538 #ifndef INET
2539 if (pp->af == AF_INET)
2540 return (EAFNOSUPPORT);
2541 #endif /* INET */
2542 #ifndef INET6
2543 if (pp->af == AF_INET6)
2544 return (EAFNOSUPPORT);
2545 #endif /* INET6 */
2546
2547 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2548 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2549 pp->addr.addr.type != PF_ADDR_TABLE)
2550 return (EINVAL);
2551
2552 if (pp->addr.addr.p.dyn != NULL)
2553 return (EINVAL);
2554
2555 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2556 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2557 if (error != 0)
2558 goto out;
2559 if (pa->ifname[0])
2560 kif = pf_kkif_create(M_WAITOK);
2561 PF_RULES_WLOCK();
2562 if (pp->ticket != V_ticket_pabuf) {
2563 PF_RULES_WUNLOCK();
2564 if (pa->ifname[0])
2565 pf_kkif_free(kif);
2566 error = EBUSY;
2567 goto out;
2568 }
2569 if (pa->ifname[0]) {
2570 pa->kif = pfi_kkif_attach(kif, pa->ifname);
2571 kif = NULL;
2572 pfi_kkif_ref(pa->kif);
2573 } else
2574 pa->kif = NULL;
2575 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2576 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2577 if (pa->ifname[0])
2578 pfi_kkif_unref(pa->kif);
2579 PF_RULES_WUNLOCK();
2580 goto out;
2581 }
2582 TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2583 PF_RULES_WUNLOCK();
2584
2585 return (0);
2586
2587 out:
2588 free(pa, M_PFRULE);
2589 return (error);
2590 }
2591
2592 int
pf_ioctl_get_addrs(struct pfioc_pooladdr * pp)2593 pf_ioctl_get_addrs(struct pfioc_pooladdr *pp)
2594 {
2595 struct pf_kpool *pool;
2596 struct pf_kpooladdr *pa;
2597
2598 PF_RULES_RLOCK_TRACKER;
2599
2600 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2601 pp->nr = 0;
2602
2603 PF_RULES_RLOCK();
2604 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2605 pp->r_num, 0, 1, 0);
2606 if (pool == NULL) {
2607 PF_RULES_RUNLOCK();
2608 return (EBUSY);
2609 }
2610 TAILQ_FOREACH(pa, &pool->list, entries)
2611 pp->nr++;
2612 PF_RULES_RUNLOCK();
2613
2614 return (0);
2615 }
2616
2617 int
pf_ioctl_get_addr(struct pfioc_pooladdr * pp)2618 pf_ioctl_get_addr(struct pfioc_pooladdr *pp)
2619 {
2620 struct pf_kpool *pool;
2621 struct pf_kpooladdr *pa;
2622 u_int32_t nr = 0;
2623
2624 PF_RULES_RLOCK_TRACKER;
2625
2626 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2627
2628 PF_RULES_RLOCK();
2629 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2630 pp->r_num, 0, 1, 1);
2631 if (pool == NULL) {
2632 PF_RULES_RUNLOCK();
2633 return (EBUSY);
2634 }
2635 pa = TAILQ_FIRST(&pool->list);
2636 while ((pa != NULL) && (nr < pp->nr)) {
2637 pa = TAILQ_NEXT(pa, entries);
2638 nr++;
2639 }
2640 if (pa == NULL) {
2641 PF_RULES_RUNLOCK();
2642 return (EBUSY);
2643 }
2644 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2645 pf_addr_copyout(&pp->addr.addr);
2646 PF_RULES_RUNLOCK();
2647
2648 return (0);
2649 }
2650
2651 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2652 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2653 {
2654 int error = 0;
2655 PF_RULES_RLOCK_TRACKER;
2656
2657 #define ERROUT_IOCTL(target, x) \
2658 do { \
2659 error = (x); \
2660 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
2661 goto target; \
2662 } while (0)
2663
2664
2665 /* XXX keep in sync with switch() below */
2666 if (securelevel_gt(td->td_ucred, 2))
2667 switch (cmd) {
2668 case DIOCGETRULES:
2669 case DIOCGETRULENV:
2670 case DIOCGETADDRS:
2671 case DIOCGETADDR:
2672 case DIOCGETSTATE:
2673 case DIOCGETSTATENV:
2674 case DIOCSETSTATUSIF:
2675 case DIOCGETSTATUSNV:
2676 case DIOCCLRSTATUS:
2677 case DIOCNATLOOK:
2678 case DIOCSETDEBUG:
2679 #ifdef COMPAT_FREEBSD14
2680 case DIOCGETSTATES:
2681 case DIOCGETSTATESV2:
2682 #endif
2683 case DIOCGETTIMEOUT:
2684 case DIOCCLRRULECTRS:
2685 case DIOCGETLIMIT:
2686 case DIOCGETALTQSV0:
2687 case DIOCGETALTQSV1:
2688 case DIOCGETALTQV0:
2689 case DIOCGETALTQV1:
2690 case DIOCGETQSTATSV0:
2691 case DIOCGETQSTATSV1:
2692 case DIOCGETRULESETS:
2693 case DIOCGETRULESET:
2694 case DIOCRGETTABLES:
2695 case DIOCRGETTSTATS:
2696 case DIOCRCLRTSTATS:
2697 case DIOCRCLRADDRS:
2698 case DIOCRADDADDRS:
2699 case DIOCRDELADDRS:
2700 case DIOCRSETADDRS:
2701 case DIOCRGETADDRS:
2702 case DIOCRGETASTATS:
2703 case DIOCRCLRASTATS:
2704 case DIOCRTSTADDRS:
2705 case DIOCOSFPGET:
2706 case DIOCGETSRCNODES:
2707 case DIOCCLRSRCNODES:
2708 case DIOCGETSYNCOOKIES:
2709 case DIOCIGETIFACES:
2710 case DIOCGIFSPEEDV0:
2711 case DIOCGIFSPEEDV1:
2712 case DIOCSETIFFLAG:
2713 case DIOCCLRIFFLAG:
2714 case DIOCGETETHRULES:
2715 case DIOCGETETHRULE:
2716 case DIOCGETETHRULESETS:
2717 case DIOCGETETHRULESET:
2718 break;
2719 case DIOCRCLRTABLES:
2720 case DIOCRADDTABLES:
2721 case DIOCRDELTABLES:
2722 case DIOCRSETTFLAGS:
2723 if (((struct pfioc_table *)addr)->pfrio_flags &
2724 PFR_FLAG_DUMMY)
2725 break; /* dummy operation ok */
2726 return (EPERM);
2727 default:
2728 return (EPERM);
2729 }
2730
2731 if (!(flags & FWRITE))
2732 switch (cmd) {
2733 case DIOCGETRULES:
2734 case DIOCGETADDRS:
2735 case DIOCGETADDR:
2736 case DIOCGETSTATE:
2737 case DIOCGETSTATENV:
2738 case DIOCGETSTATUSNV:
2739 #ifdef COMPAT_FREEBSD14
2740 case DIOCGETSTATES:
2741 case DIOCGETSTATESV2:
2742 #endif
2743 case DIOCGETTIMEOUT:
2744 case DIOCGETLIMIT:
2745 case DIOCGETALTQSV0:
2746 case DIOCGETALTQSV1:
2747 case DIOCGETALTQV0:
2748 case DIOCGETALTQV1:
2749 case DIOCGETQSTATSV0:
2750 case DIOCGETQSTATSV1:
2751 case DIOCGETRULESETS:
2752 case DIOCGETRULESET:
2753 case DIOCNATLOOK:
2754 case DIOCRGETTABLES:
2755 case DIOCRGETTSTATS:
2756 case DIOCRGETADDRS:
2757 case DIOCRGETASTATS:
2758 case DIOCRTSTADDRS:
2759 case DIOCOSFPGET:
2760 case DIOCGETSRCNODES:
2761 case DIOCGETSYNCOOKIES:
2762 case DIOCIGETIFACES:
2763 case DIOCGIFSPEEDV1:
2764 case DIOCGIFSPEEDV0:
2765 case DIOCGETRULENV:
2766 case DIOCGETETHRULES:
2767 case DIOCGETETHRULE:
2768 case DIOCGETETHRULESETS:
2769 case DIOCGETETHRULESET:
2770 break;
2771 case DIOCRCLRTABLES:
2772 case DIOCRADDTABLES:
2773 case DIOCRDELTABLES:
2774 case DIOCRCLRTSTATS:
2775 case DIOCRCLRADDRS:
2776 case DIOCRADDADDRS:
2777 case DIOCRDELADDRS:
2778 case DIOCRSETADDRS:
2779 case DIOCRSETTFLAGS:
2780 if (((struct pfioc_table *)addr)->pfrio_flags &
2781 PFR_FLAG_DUMMY) {
2782 flags |= FWRITE; /* need write lock for dummy */
2783 break; /* dummy operation ok */
2784 }
2785 return (EACCES);
2786 default:
2787 return (EACCES);
2788 }
2789
2790 CURVNET_SET(TD_TO_VNET(td));
2791
2792 switch (cmd) {
2793 #ifdef COMPAT_FREEBSD14
2794 case DIOCSTART:
2795 error = pf_start();
2796 break;
2797
2798 case DIOCSTOP:
2799 error = pf_stop();
2800 break;
2801 #endif
2802
2803 case DIOCGETETHRULES: {
2804 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2805 nvlist_t *nvl;
2806 void *packed;
2807 struct pf_keth_rule *tail;
2808 struct pf_keth_ruleset *rs;
2809 u_int32_t ticket, nr;
2810 const char *anchor = "";
2811
2812 nvl = NULL;
2813 packed = NULL;
2814
2815 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2816
2817 if (nv->len > pf_ioctl_maxcount)
2818 ERROUT(ENOMEM);
2819
2820 /* Copy the request in */
2821 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2822 if (packed == NULL)
2823 ERROUT(ENOMEM);
2824
2825 error = copyin(nv->data, packed, nv->len);
2826 if (error)
2827 ERROUT(error);
2828
2829 nvl = nvlist_unpack(packed, nv->len, 0);
2830 if (nvl == NULL)
2831 ERROUT(EBADMSG);
2832
2833 if (! nvlist_exists_string(nvl, "anchor"))
2834 ERROUT(EBADMSG);
2835
2836 anchor = nvlist_get_string(nvl, "anchor");
2837
2838 rs = pf_find_keth_ruleset(anchor);
2839
2840 nvlist_destroy(nvl);
2841 nvl = NULL;
2842 free(packed, M_NVLIST);
2843 packed = NULL;
2844
2845 if (rs == NULL)
2846 ERROUT(ENOENT);
2847
2848 /* Reply */
2849 nvl = nvlist_create(0);
2850 if (nvl == NULL)
2851 ERROUT(ENOMEM);
2852
2853 PF_RULES_RLOCK();
2854
2855 ticket = rs->active.ticket;
2856 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2857 if (tail)
2858 nr = tail->nr + 1;
2859 else
2860 nr = 0;
2861
2862 PF_RULES_RUNLOCK();
2863
2864 nvlist_add_number(nvl, "ticket", ticket);
2865 nvlist_add_number(nvl, "nr", nr);
2866
2867 packed = nvlist_pack(nvl, &nv->len);
2868 if (packed == NULL)
2869 ERROUT(ENOMEM);
2870
2871 if (nv->size == 0)
2872 ERROUT(0);
2873 else if (nv->size < nv->len)
2874 ERROUT(ENOSPC);
2875
2876 error = copyout(packed, nv->data, nv->len);
2877
2878 #undef ERROUT
2879 DIOCGETETHRULES_error:
2880 free(packed, M_NVLIST);
2881 nvlist_destroy(nvl);
2882 break;
2883 }
2884
2885 case DIOCGETETHRULE: {
2886 struct epoch_tracker et;
2887 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2888 nvlist_t *nvl = NULL;
2889 void *nvlpacked = NULL;
2890 struct pf_keth_rule *rule = NULL;
2891 struct pf_keth_ruleset *rs;
2892 u_int32_t ticket, nr;
2893 bool clear = false;
2894 const char *anchor;
2895
2896 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2897
2898 if (nv->len > pf_ioctl_maxcount)
2899 ERROUT(ENOMEM);
2900
2901 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2902 if (nvlpacked == NULL)
2903 ERROUT(ENOMEM);
2904
2905 error = copyin(nv->data, nvlpacked, nv->len);
2906 if (error)
2907 ERROUT(error);
2908
2909 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2910 if (nvl == NULL)
2911 ERROUT(EBADMSG);
2912 if (! nvlist_exists_number(nvl, "ticket"))
2913 ERROUT(EBADMSG);
2914 ticket = nvlist_get_number(nvl, "ticket");
2915 if (! nvlist_exists_string(nvl, "anchor"))
2916 ERROUT(EBADMSG);
2917 anchor = nvlist_get_string(nvl, "anchor");
2918
2919 if (nvlist_exists_bool(nvl, "clear"))
2920 clear = nvlist_get_bool(nvl, "clear");
2921
2922 if (clear && !(flags & FWRITE))
2923 ERROUT(EACCES);
2924
2925 if (! nvlist_exists_number(nvl, "nr"))
2926 ERROUT(EBADMSG);
2927 nr = nvlist_get_number(nvl, "nr");
2928
2929 PF_RULES_RLOCK();
2930 rs = pf_find_keth_ruleset(anchor);
2931 if (rs == NULL) {
2932 PF_RULES_RUNLOCK();
2933 ERROUT(ENOENT);
2934 }
2935 if (ticket != rs->active.ticket) {
2936 PF_RULES_RUNLOCK();
2937 ERROUT(EBUSY);
2938 }
2939
2940 nvlist_destroy(nvl);
2941 nvl = NULL;
2942 free(nvlpacked, M_NVLIST);
2943 nvlpacked = NULL;
2944
2945 rule = TAILQ_FIRST(rs->active.rules);
2946 while ((rule != NULL) && (rule->nr != nr))
2947 rule = TAILQ_NEXT(rule, entries);
2948 if (rule == NULL) {
2949 PF_RULES_RUNLOCK();
2950 ERROUT(ENOENT);
2951 }
2952 /* Make sure rule can't go away. */
2953 NET_EPOCH_ENTER(et);
2954 PF_RULES_RUNLOCK();
2955 nvl = pf_keth_rule_to_nveth_rule(rule);
2956 if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2957 ERROUT(EBUSY);
2958 NET_EPOCH_EXIT(et);
2959 if (nvl == NULL)
2960 ERROUT(ENOMEM);
2961
2962 nvlpacked = nvlist_pack(nvl, &nv->len);
2963 if (nvlpacked == NULL)
2964 ERROUT(ENOMEM);
2965
2966 if (nv->size == 0)
2967 ERROUT(0);
2968 else if (nv->size < nv->len)
2969 ERROUT(ENOSPC);
2970
2971 error = copyout(nvlpacked, nv->data, nv->len);
2972 if (error == 0 && clear) {
2973 counter_u64_zero(rule->evaluations);
2974 for (int i = 0; i < 2; i++) {
2975 counter_u64_zero(rule->packets[i]);
2976 counter_u64_zero(rule->bytes[i]);
2977 }
2978 }
2979
2980 #undef ERROUT
2981 DIOCGETETHRULE_error:
2982 free(nvlpacked, M_NVLIST);
2983 nvlist_destroy(nvl);
2984 break;
2985 }
2986
2987 case DIOCADDETHRULE: {
2988 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2989 nvlist_t *nvl = NULL;
2990 void *nvlpacked = NULL;
2991 struct pf_keth_rule *rule = NULL, *tail = NULL;
2992 struct pf_keth_ruleset *ruleset = NULL;
2993 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
2994 const char *anchor = "", *anchor_call = "";
2995
2996 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2997
2998 if (nv->len > pf_ioctl_maxcount)
2999 ERROUT(ENOMEM);
3000
3001 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3002 if (nvlpacked == NULL)
3003 ERROUT(ENOMEM);
3004
3005 error = copyin(nv->data, nvlpacked, nv->len);
3006 if (error)
3007 ERROUT(error);
3008
3009 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3010 if (nvl == NULL)
3011 ERROUT(EBADMSG);
3012
3013 if (! nvlist_exists_number(nvl, "ticket"))
3014 ERROUT(EBADMSG);
3015
3016 if (nvlist_exists_string(nvl, "anchor"))
3017 anchor = nvlist_get_string(nvl, "anchor");
3018 if (nvlist_exists_string(nvl, "anchor_call"))
3019 anchor_call = nvlist_get_string(nvl, "anchor_call");
3020
3021 ruleset = pf_find_keth_ruleset(anchor);
3022 if (ruleset == NULL)
3023 ERROUT(EINVAL);
3024
3025 if (nvlist_get_number(nvl, "ticket") !=
3026 ruleset->inactive.ticket) {
3027 DPFPRINTF(PF_DEBUG_MISC,
3028 ("ticket: %d != %d\n",
3029 (u_int32_t)nvlist_get_number(nvl, "ticket"),
3030 ruleset->inactive.ticket));
3031 ERROUT(EBUSY);
3032 }
3033
3034 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3035 if (rule == NULL)
3036 ERROUT(ENOMEM);
3037 rule->timestamp = NULL;
3038
3039 error = pf_nveth_rule_to_keth_rule(nvl, rule);
3040 if (error != 0)
3041 ERROUT(error);
3042
3043 if (rule->ifname[0])
3044 kif = pf_kkif_create(M_WAITOK);
3045 if (rule->bridge_to_name[0])
3046 bridge_to_kif = pf_kkif_create(M_WAITOK);
3047 rule->evaluations = counter_u64_alloc(M_WAITOK);
3048 for (int i = 0; i < 2; i++) {
3049 rule->packets[i] = counter_u64_alloc(M_WAITOK);
3050 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3051 }
3052 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3053 M_WAITOK | M_ZERO);
3054
3055 PF_RULES_WLOCK();
3056
3057 if (rule->ifname[0]) {
3058 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3059 pfi_kkif_ref(rule->kif);
3060 } else
3061 rule->kif = NULL;
3062 if (rule->bridge_to_name[0]) {
3063 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3064 rule->bridge_to_name);
3065 pfi_kkif_ref(rule->bridge_to);
3066 } else
3067 rule->bridge_to = NULL;
3068
3069 #ifdef ALTQ
3070 /* set queue IDs */
3071 if (rule->qname[0] != 0) {
3072 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3073 error = EBUSY;
3074 else
3075 rule->qid = rule->qid;
3076 }
3077 #endif
3078 if (rule->tagname[0])
3079 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3080 error = EBUSY;
3081 if (rule->match_tagname[0])
3082 if ((rule->match_tag = pf_tagname2tag(
3083 rule->match_tagname)) == 0)
3084 error = EBUSY;
3085
3086 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3087 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3088 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3089 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3090
3091 if (error) {
3092 pf_free_eth_rule(rule);
3093 PF_RULES_WUNLOCK();
3094 ERROUT(error);
3095 }
3096
3097 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3098 pf_free_eth_rule(rule);
3099 PF_RULES_WUNLOCK();
3100 ERROUT(EINVAL);
3101 }
3102
3103 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3104 if (tail)
3105 rule->nr = tail->nr + 1;
3106 else
3107 rule->nr = 0;
3108
3109 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3110
3111 PF_RULES_WUNLOCK();
3112
3113 #undef ERROUT
3114 DIOCADDETHRULE_error:
3115 nvlist_destroy(nvl);
3116 free(nvlpacked, M_NVLIST);
3117 break;
3118 }
3119
3120 case DIOCGETETHRULESETS: {
3121 struct epoch_tracker et;
3122 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3123 nvlist_t *nvl = NULL;
3124 void *nvlpacked = NULL;
3125 struct pf_keth_ruleset *ruleset;
3126 struct pf_keth_anchor *anchor;
3127 int nr = 0;
3128
3129 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3130
3131 if (nv->len > pf_ioctl_maxcount)
3132 ERROUT(ENOMEM);
3133
3134 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3135 if (nvlpacked == NULL)
3136 ERROUT(ENOMEM);
3137
3138 error = copyin(nv->data, nvlpacked, nv->len);
3139 if (error)
3140 ERROUT(error);
3141
3142 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3143 if (nvl == NULL)
3144 ERROUT(EBADMSG);
3145 if (! nvlist_exists_string(nvl, "path"))
3146 ERROUT(EBADMSG);
3147
3148 NET_EPOCH_ENTER(et);
3149
3150 if ((ruleset = pf_find_keth_ruleset(
3151 nvlist_get_string(nvl, "path"))) == NULL) {
3152 NET_EPOCH_EXIT(et);
3153 ERROUT(ENOENT);
3154 }
3155
3156 if (ruleset->anchor == NULL) {
3157 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3158 if (anchor->parent == NULL)
3159 nr++;
3160 } else {
3161 RB_FOREACH(anchor, pf_keth_anchor_node,
3162 &ruleset->anchor->children)
3163 nr++;
3164 }
3165
3166 NET_EPOCH_EXIT(et);
3167
3168 nvlist_destroy(nvl);
3169 nvl = NULL;
3170 free(nvlpacked, M_NVLIST);
3171 nvlpacked = NULL;
3172
3173 nvl = nvlist_create(0);
3174 if (nvl == NULL)
3175 ERROUT(ENOMEM);
3176
3177 nvlist_add_number(nvl, "nr", nr);
3178
3179 nvlpacked = nvlist_pack(nvl, &nv->len);
3180 if (nvlpacked == NULL)
3181 ERROUT(ENOMEM);
3182
3183 if (nv->size == 0)
3184 ERROUT(0);
3185 else if (nv->size < nv->len)
3186 ERROUT(ENOSPC);
3187
3188 error = copyout(nvlpacked, nv->data, nv->len);
3189
3190 #undef ERROUT
3191 DIOCGETETHRULESETS_error:
3192 free(nvlpacked, M_NVLIST);
3193 nvlist_destroy(nvl);
3194 break;
3195 }
3196
3197 case DIOCGETETHRULESET: {
3198 struct epoch_tracker et;
3199 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3200 nvlist_t *nvl = NULL;
3201 void *nvlpacked = NULL;
3202 struct pf_keth_ruleset *ruleset;
3203 struct pf_keth_anchor *anchor;
3204 int nr = 0, req_nr = 0;
3205 bool found = false;
3206
3207 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3208
3209 if (nv->len > pf_ioctl_maxcount)
3210 ERROUT(ENOMEM);
3211
3212 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3213 if (nvlpacked == NULL)
3214 ERROUT(ENOMEM);
3215
3216 error = copyin(nv->data, nvlpacked, nv->len);
3217 if (error)
3218 ERROUT(error);
3219
3220 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3221 if (nvl == NULL)
3222 ERROUT(EBADMSG);
3223 if (! nvlist_exists_string(nvl, "path"))
3224 ERROUT(EBADMSG);
3225 if (! nvlist_exists_number(nvl, "nr"))
3226 ERROUT(EBADMSG);
3227
3228 req_nr = nvlist_get_number(nvl, "nr");
3229
3230 NET_EPOCH_ENTER(et);
3231
3232 if ((ruleset = pf_find_keth_ruleset(
3233 nvlist_get_string(nvl, "path"))) == NULL) {
3234 NET_EPOCH_EXIT(et);
3235 ERROUT(ENOENT);
3236 }
3237
3238 nvlist_destroy(nvl);
3239 nvl = NULL;
3240 free(nvlpacked, M_NVLIST);
3241 nvlpacked = NULL;
3242
3243 nvl = nvlist_create(0);
3244 if (nvl == NULL) {
3245 NET_EPOCH_EXIT(et);
3246 ERROUT(ENOMEM);
3247 }
3248
3249 if (ruleset->anchor == NULL) {
3250 RB_FOREACH(anchor, pf_keth_anchor_global,
3251 &V_pf_keth_anchors) {
3252 if (anchor->parent == NULL && nr++ == req_nr) {
3253 found = true;
3254 break;
3255 }
3256 }
3257 } else {
3258 RB_FOREACH(anchor, pf_keth_anchor_node,
3259 &ruleset->anchor->children) {
3260 if (nr++ == req_nr) {
3261 found = true;
3262 break;
3263 }
3264 }
3265 }
3266
3267 NET_EPOCH_EXIT(et);
3268 if (found) {
3269 nvlist_add_number(nvl, "nr", nr);
3270 nvlist_add_string(nvl, "name", anchor->name);
3271 if (ruleset->anchor)
3272 nvlist_add_string(nvl, "path",
3273 ruleset->anchor->path);
3274 else
3275 nvlist_add_string(nvl, "path", "");
3276 } else {
3277 ERROUT(EBUSY);
3278 }
3279
3280 nvlpacked = nvlist_pack(nvl, &nv->len);
3281 if (nvlpacked == NULL)
3282 ERROUT(ENOMEM);
3283
3284 if (nv->size == 0)
3285 ERROUT(0);
3286 else if (nv->size < nv->len)
3287 ERROUT(ENOSPC);
3288
3289 error = copyout(nvlpacked, nv->data, nv->len);
3290
3291 #undef ERROUT
3292 DIOCGETETHRULESET_error:
3293 free(nvlpacked, M_NVLIST);
3294 nvlist_destroy(nvl);
3295 break;
3296 }
3297
3298 case DIOCADDRULENV: {
3299 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3300 nvlist_t *nvl = NULL;
3301 void *nvlpacked = NULL;
3302 struct pf_krule *rule = NULL;
3303 const char *anchor = "", *anchor_call = "";
3304 uint32_t ticket = 0, pool_ticket = 0;
3305
3306 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
3307
3308 if (nv->len > pf_ioctl_maxcount)
3309 ERROUT(ENOMEM);
3310
3311 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3312 error = copyin(nv->data, nvlpacked, nv->len);
3313 if (error)
3314 ERROUT(error);
3315
3316 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3317 if (nvl == NULL)
3318 ERROUT(EBADMSG);
3319
3320 if (! nvlist_exists_number(nvl, "ticket"))
3321 ERROUT(EINVAL);
3322 ticket = nvlist_get_number(nvl, "ticket");
3323
3324 if (! nvlist_exists_number(nvl, "pool_ticket"))
3325 ERROUT(EINVAL);
3326 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3327
3328 if (! nvlist_exists_nvlist(nvl, "rule"))
3329 ERROUT(EINVAL);
3330
3331 rule = pf_krule_alloc();
3332 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3333 rule);
3334 if (error)
3335 ERROUT(error);
3336
3337 if (nvlist_exists_string(nvl, "anchor"))
3338 anchor = nvlist_get_string(nvl, "anchor");
3339 if (nvlist_exists_string(nvl, "anchor_call"))
3340 anchor_call = nvlist_get_string(nvl, "anchor_call");
3341
3342 if ((error = nvlist_error(nvl)))
3343 ERROUT(error);
3344
3345 /* Frees rule on error */
3346 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3347 anchor_call, td->td_ucred->cr_ruid,
3348 td->td_proc ? td->td_proc->p_pid : 0);
3349
3350 nvlist_destroy(nvl);
3351 free(nvlpacked, M_NVLIST);
3352 break;
3353 #undef ERROUT
3354 DIOCADDRULENV_error:
3355 pf_krule_free(rule);
3356 nvlist_destroy(nvl);
3357 free(nvlpacked, M_NVLIST);
3358
3359 break;
3360 }
3361 case DIOCADDRULE: {
3362 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3363 struct pf_krule *rule;
3364
3365 rule = pf_krule_alloc();
3366 error = pf_rule_to_krule(&pr->rule, rule);
3367 if (error != 0) {
3368 pf_krule_free(rule);
3369 break;
3370 }
3371
3372 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3373
3374 /* Frees rule on error */
3375 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3376 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3377 td->td_proc ? td->td_proc->p_pid : 0);
3378 break;
3379 }
3380
3381 case DIOCGETRULES: {
3382 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3383
3384 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3385
3386 error = pf_ioctl_getrules(pr);
3387
3388 break;
3389 }
3390
3391 case DIOCGETRULENV: {
3392 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3393 nvlist_t *nvrule = NULL;
3394 nvlist_t *nvl = NULL;
3395 struct pf_kruleset *ruleset;
3396 struct pf_krule *rule;
3397 void *nvlpacked = NULL;
3398 int rs_num, nr;
3399 bool clear_counter = false;
3400
3401 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
3402
3403 if (nv->len > pf_ioctl_maxcount)
3404 ERROUT(ENOMEM);
3405
3406 /* Copy the request in */
3407 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3408 if (nvlpacked == NULL)
3409 ERROUT(ENOMEM);
3410
3411 error = copyin(nv->data, nvlpacked, nv->len);
3412 if (error)
3413 ERROUT(error);
3414
3415 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3416 if (nvl == NULL)
3417 ERROUT(EBADMSG);
3418
3419 if (! nvlist_exists_string(nvl, "anchor"))
3420 ERROUT(EBADMSG);
3421 if (! nvlist_exists_number(nvl, "ruleset"))
3422 ERROUT(EBADMSG);
3423 if (! nvlist_exists_number(nvl, "ticket"))
3424 ERROUT(EBADMSG);
3425 if (! nvlist_exists_number(nvl, "nr"))
3426 ERROUT(EBADMSG);
3427
3428 if (nvlist_exists_bool(nvl, "clear_counter"))
3429 clear_counter = nvlist_get_bool(nvl, "clear_counter");
3430
3431 if (clear_counter && !(flags & FWRITE))
3432 ERROUT(EACCES);
3433
3434 nr = nvlist_get_number(nvl, "nr");
3435
3436 PF_RULES_WLOCK();
3437 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3438 if (ruleset == NULL) {
3439 PF_RULES_WUNLOCK();
3440 ERROUT(ENOENT);
3441 }
3442
3443 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3444 if (rs_num >= PF_RULESET_MAX) {
3445 PF_RULES_WUNLOCK();
3446 ERROUT(EINVAL);
3447 }
3448
3449 if (nvlist_get_number(nvl, "ticket") !=
3450 ruleset->rules[rs_num].active.ticket) {
3451 PF_RULES_WUNLOCK();
3452 ERROUT(EBUSY);
3453 }
3454
3455 if ((error = nvlist_error(nvl))) {
3456 PF_RULES_WUNLOCK();
3457 ERROUT(error);
3458 }
3459
3460 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3461 while ((rule != NULL) && (rule->nr != nr))
3462 rule = TAILQ_NEXT(rule, entries);
3463 if (rule == NULL) {
3464 PF_RULES_WUNLOCK();
3465 ERROUT(EBUSY);
3466 }
3467
3468 nvrule = pf_krule_to_nvrule(rule);
3469
3470 nvlist_destroy(nvl);
3471 nvl = nvlist_create(0);
3472 if (nvl == NULL) {
3473 PF_RULES_WUNLOCK();
3474 ERROUT(ENOMEM);
3475 }
3476 nvlist_add_number(nvl, "nr", nr);
3477 nvlist_add_nvlist(nvl, "rule", nvrule);
3478 nvlist_destroy(nvrule);
3479 nvrule = NULL;
3480 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3481 PF_RULES_WUNLOCK();
3482 ERROUT(EBUSY);
3483 }
3484
3485 free(nvlpacked, M_NVLIST);
3486 nvlpacked = nvlist_pack(nvl, &nv->len);
3487 if (nvlpacked == NULL) {
3488 PF_RULES_WUNLOCK();
3489 ERROUT(ENOMEM);
3490 }
3491
3492 if (nv->size == 0) {
3493 PF_RULES_WUNLOCK();
3494 ERROUT(0);
3495 }
3496 else if (nv->size < nv->len) {
3497 PF_RULES_WUNLOCK();
3498 ERROUT(ENOSPC);
3499 }
3500
3501 if (clear_counter)
3502 pf_krule_clear_counters(rule);
3503
3504 PF_RULES_WUNLOCK();
3505
3506 error = copyout(nvlpacked, nv->data, nv->len);
3507
3508 #undef ERROUT
3509 DIOCGETRULENV_error:
3510 free(nvlpacked, M_NVLIST);
3511 nvlist_destroy(nvrule);
3512 nvlist_destroy(nvl);
3513
3514 break;
3515 }
3516
3517 case DIOCCHANGERULE: {
3518 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
3519 struct pf_kruleset *ruleset;
3520 struct pf_krule *oldrule = NULL, *newrule = NULL;
3521 struct pfi_kkif *kif = NULL;
3522 struct pf_kpooladdr *pa;
3523 u_int32_t nr = 0;
3524 int rs_num;
3525
3526 pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3527
3528 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3529 pcr->action > PF_CHANGE_GET_TICKET) {
3530 error = EINVAL;
3531 break;
3532 }
3533 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3534 error = EINVAL;
3535 break;
3536 }
3537
3538 if (pcr->action != PF_CHANGE_REMOVE) {
3539 newrule = pf_krule_alloc();
3540 error = pf_rule_to_krule(&pcr->rule, newrule);
3541 if (error != 0) {
3542 pf_krule_free(newrule);
3543 break;
3544 }
3545
3546 if (newrule->ifname[0])
3547 kif = pf_kkif_create(M_WAITOK);
3548 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3549 for (int i = 0; i < 2; i++) {
3550 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3551 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3552 }
3553 newrule->states_cur = counter_u64_alloc(M_WAITOK);
3554 newrule->states_tot = counter_u64_alloc(M_WAITOK);
3555 newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3556 newrule->cuid = td->td_ucred->cr_ruid;
3557 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3558 TAILQ_INIT(&newrule->rpool.list);
3559 }
3560 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3561
3562 PF_CONFIG_LOCK();
3563 PF_RULES_WLOCK();
3564 #ifdef PF_WANT_32_TO_64_COUNTER
3565 if (newrule != NULL) {
3566 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3567 newrule->allrulelinked = true;
3568 V_pf_allrulecount++;
3569 }
3570 #endif
3571
3572 if (!(pcr->action == PF_CHANGE_REMOVE ||
3573 pcr->action == PF_CHANGE_GET_TICKET) &&
3574 pcr->pool_ticket != V_ticket_pabuf)
3575 ERROUT(EBUSY);
3576
3577 ruleset = pf_find_kruleset(pcr->anchor);
3578 if (ruleset == NULL)
3579 ERROUT(EINVAL);
3580
3581 rs_num = pf_get_ruleset_number(pcr->rule.action);
3582 if (rs_num >= PF_RULESET_MAX)
3583 ERROUT(EINVAL);
3584
3585 /*
3586 * XXXMJG: there is no guarantee that the ruleset was
3587 * created by the usual route of calling DIOCXBEGIN.
3588 * As a result it is possible the rule tree will not
3589 * be allocated yet. Hack around it by doing it here.
3590 * Note it is fine to let the tree persist in case of
3591 * error as it will be freed down the road on future
3592 * updates (if need be).
3593 */
3594 if (ruleset->rules[rs_num].active.tree == NULL) {
3595 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3596 if (ruleset->rules[rs_num].active.tree == NULL) {
3597 ERROUT(ENOMEM);
3598 }
3599 }
3600
3601 if (pcr->action == PF_CHANGE_GET_TICKET) {
3602 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3603 ERROUT(0);
3604 } else if (pcr->ticket !=
3605 ruleset->rules[rs_num].active.ticket)
3606 ERROUT(EINVAL);
3607
3608 if (pcr->action != PF_CHANGE_REMOVE) {
3609 if (newrule->ifname[0]) {
3610 newrule->kif = pfi_kkif_attach(kif,
3611 newrule->ifname);
3612 kif = NULL;
3613 pfi_kkif_ref(newrule->kif);
3614 } else
3615 newrule->kif = NULL;
3616
3617 if (newrule->rtableid > 0 &&
3618 newrule->rtableid >= rt_numfibs)
3619 error = EBUSY;
3620
3621 #ifdef ALTQ
3622 /* set queue IDs */
3623 if (newrule->qname[0] != 0) {
3624 if ((newrule->qid =
3625 pf_qname2qid(newrule->qname)) == 0)
3626 error = EBUSY;
3627 else if (newrule->pqname[0] != 0) {
3628 if ((newrule->pqid =
3629 pf_qname2qid(newrule->pqname)) == 0)
3630 error = EBUSY;
3631 } else
3632 newrule->pqid = newrule->qid;
3633 }
3634 #endif /* ALTQ */
3635 if (newrule->tagname[0])
3636 if ((newrule->tag =
3637 pf_tagname2tag(newrule->tagname)) == 0)
3638 error = EBUSY;
3639 if (newrule->match_tagname[0])
3640 if ((newrule->match_tag = pf_tagname2tag(
3641 newrule->match_tagname)) == 0)
3642 error = EBUSY;
3643 if (newrule->rt && !newrule->direction)
3644 error = EINVAL;
3645 if (!newrule->log)
3646 newrule->logif = 0;
3647 if (newrule->logif >= PFLOGIFS_MAX)
3648 error = EINVAL;
3649 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3650 error = ENOMEM;
3651 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3652 error = ENOMEM;
3653 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3654 error = EINVAL;
3655 TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3656 if (pa->addr.type == PF_ADDR_TABLE) {
3657 pa->addr.p.tbl =
3658 pfr_attach_table(ruleset,
3659 pa->addr.v.tblname);
3660 if (pa->addr.p.tbl == NULL)
3661 error = ENOMEM;
3662 }
3663
3664 newrule->overload_tbl = NULL;
3665 if (newrule->overload_tblname[0]) {
3666 if ((newrule->overload_tbl = pfr_attach_table(
3667 ruleset, newrule->overload_tblname)) ==
3668 NULL)
3669 error = EINVAL;
3670 else
3671 newrule->overload_tbl->pfrkt_flags |=
3672 PFR_TFLAG_ACTIVE;
3673 }
3674
3675 pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3676 if (((((newrule->action == PF_NAT) ||
3677 (newrule->action == PF_RDR) ||
3678 (newrule->action == PF_BINAT) ||
3679 (newrule->rt > PF_NOPFROUTE)) &&
3680 !newrule->anchor)) &&
3681 (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3682 error = EINVAL;
3683
3684 if (error) {
3685 pf_free_rule(newrule);
3686 PF_RULES_WUNLOCK();
3687 PF_CONFIG_UNLOCK();
3688 break;
3689 }
3690
3691 newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3692 }
3693 pf_empty_kpool(&V_pf_pabuf);
3694
3695 if (pcr->action == PF_CHANGE_ADD_HEAD)
3696 oldrule = TAILQ_FIRST(
3697 ruleset->rules[rs_num].active.ptr);
3698 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3699 oldrule = TAILQ_LAST(
3700 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3701 else {
3702 oldrule = TAILQ_FIRST(
3703 ruleset->rules[rs_num].active.ptr);
3704 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3705 oldrule = TAILQ_NEXT(oldrule, entries);
3706 if (oldrule == NULL) {
3707 if (newrule != NULL)
3708 pf_free_rule(newrule);
3709 PF_RULES_WUNLOCK();
3710 PF_CONFIG_UNLOCK();
3711 error = EINVAL;
3712 break;
3713 }
3714 }
3715
3716 if (pcr->action == PF_CHANGE_REMOVE) {
3717 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3718 oldrule);
3719 RB_REMOVE(pf_krule_global,
3720 ruleset->rules[rs_num].active.tree, oldrule);
3721 ruleset->rules[rs_num].active.rcount--;
3722 } else {
3723 pf_hash_rule(newrule);
3724 if (RB_INSERT(pf_krule_global,
3725 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3726 pf_free_rule(newrule);
3727 PF_RULES_WUNLOCK();
3728 PF_CONFIG_UNLOCK();
3729 error = EEXIST;
3730 break;
3731 }
3732
3733 if (oldrule == NULL)
3734 TAILQ_INSERT_TAIL(
3735 ruleset->rules[rs_num].active.ptr,
3736 newrule, entries);
3737 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3738 pcr->action == PF_CHANGE_ADD_BEFORE)
3739 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3740 else
3741 TAILQ_INSERT_AFTER(
3742 ruleset->rules[rs_num].active.ptr,
3743 oldrule, newrule, entries);
3744 ruleset->rules[rs_num].active.rcount++;
3745 }
3746
3747 nr = 0;
3748 TAILQ_FOREACH(oldrule,
3749 ruleset->rules[rs_num].active.ptr, entries)
3750 oldrule->nr = nr++;
3751
3752 ruleset->rules[rs_num].active.ticket++;
3753
3754 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3755 pf_remove_if_empty_kruleset(ruleset);
3756
3757 PF_RULES_WUNLOCK();
3758 PF_CONFIG_UNLOCK();
3759 break;
3760
3761 #undef ERROUT
3762 DIOCCHANGERULE_error:
3763 PF_RULES_WUNLOCK();
3764 PF_CONFIG_UNLOCK();
3765 pf_krule_free(newrule);
3766 pf_kkif_free(kif);
3767 break;
3768 }
3769
3770 case DIOCCLRSTATESNV: {
3771 error = pf_clearstates_nv((struct pfioc_nv *)addr);
3772 break;
3773 }
3774
3775 case DIOCKILLSTATESNV: {
3776 error = pf_killstates_nv((struct pfioc_nv *)addr);
3777 break;
3778 }
3779
3780 case DIOCADDSTATE: {
3781 struct pfioc_state *ps = (struct pfioc_state *)addr;
3782 struct pfsync_state_1301 *sp = &ps->state;
3783
3784 if (sp->timeout >= PFTM_MAX) {
3785 error = EINVAL;
3786 break;
3787 }
3788 if (V_pfsync_state_import_ptr != NULL) {
3789 PF_RULES_RLOCK();
3790 error = V_pfsync_state_import_ptr(
3791 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3792 PFSYNC_MSG_VERSION_1301);
3793 PF_RULES_RUNLOCK();
3794 } else
3795 error = EOPNOTSUPP;
3796 break;
3797 }
3798
3799 case DIOCGETSTATE: {
3800 struct pfioc_state *ps = (struct pfioc_state *)addr;
3801 struct pf_kstate *s;
3802
3803 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3804 if (s == NULL) {
3805 error = ENOENT;
3806 break;
3807 }
3808
3809 pfsync_state_export((union pfsync_state_union*)&ps->state,
3810 s, PFSYNC_MSG_VERSION_1301);
3811 PF_STATE_UNLOCK(s);
3812 break;
3813 }
3814
3815 case DIOCGETSTATENV: {
3816 error = pf_getstate((struct pfioc_nv *)addr);
3817 break;
3818 }
3819
3820 #ifdef COMPAT_FREEBSD14
3821 case DIOCGETSTATES: {
3822 struct pfioc_states *ps = (struct pfioc_states *)addr;
3823 struct pf_kstate *s;
3824 struct pfsync_state_1301 *pstore, *p;
3825 int i, nr;
3826 size_t slice_count = 16, count;
3827 void *out;
3828
3829 if (ps->ps_len <= 0) {
3830 nr = uma_zone_get_cur(V_pf_state_z);
3831 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3832 break;
3833 }
3834
3835 out = ps->ps_states;
3836 pstore = mallocarray(slice_count,
3837 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3838 nr = 0;
3839
3840 for (i = 0; i <= V_pf_hashmask; i++) {
3841 struct pf_idhash *ih = &V_pf_idhash[i];
3842
3843 DIOCGETSTATES_retry:
3844 p = pstore;
3845
3846 if (LIST_EMPTY(&ih->states))
3847 continue;
3848
3849 PF_HASHROW_LOCK(ih);
3850 count = 0;
3851 LIST_FOREACH(s, &ih->states, entry) {
3852 if (s->timeout == PFTM_UNLINKED)
3853 continue;
3854 count++;
3855 }
3856
3857 if (count > slice_count) {
3858 PF_HASHROW_UNLOCK(ih);
3859 free(pstore, M_TEMP);
3860 slice_count = count * 2;
3861 pstore = mallocarray(slice_count,
3862 sizeof(struct pfsync_state_1301), M_TEMP,
3863 M_WAITOK | M_ZERO);
3864 goto DIOCGETSTATES_retry;
3865 }
3866
3867 if ((nr+count) * sizeof(*p) > ps->ps_len) {
3868 PF_HASHROW_UNLOCK(ih);
3869 goto DIOCGETSTATES_full;
3870 }
3871
3872 LIST_FOREACH(s, &ih->states, entry) {
3873 if (s->timeout == PFTM_UNLINKED)
3874 continue;
3875
3876 pfsync_state_export((union pfsync_state_union*)p,
3877 s, PFSYNC_MSG_VERSION_1301);
3878 p++;
3879 nr++;
3880 }
3881 PF_HASHROW_UNLOCK(ih);
3882 error = copyout(pstore, out,
3883 sizeof(struct pfsync_state_1301) * count);
3884 if (error)
3885 break;
3886 out = ps->ps_states + nr;
3887 }
3888 DIOCGETSTATES_full:
3889 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3890 free(pstore, M_TEMP);
3891
3892 break;
3893 }
3894
3895 case DIOCGETSTATESV2: {
3896 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
3897 struct pf_kstate *s;
3898 struct pf_state_export *pstore, *p;
3899 int i, nr;
3900 size_t slice_count = 16, count;
3901 void *out;
3902
3903 if (ps->ps_req_version > PF_STATE_VERSION) {
3904 error = ENOTSUP;
3905 break;
3906 }
3907
3908 if (ps->ps_len <= 0) {
3909 nr = uma_zone_get_cur(V_pf_state_z);
3910 ps->ps_len = sizeof(struct pf_state_export) * nr;
3911 break;
3912 }
3913
3914 out = ps->ps_states;
3915 pstore = mallocarray(slice_count,
3916 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3917 nr = 0;
3918
3919 for (i = 0; i <= V_pf_hashmask; i++) {
3920 struct pf_idhash *ih = &V_pf_idhash[i];
3921
3922 DIOCGETSTATESV2_retry:
3923 p = pstore;
3924
3925 if (LIST_EMPTY(&ih->states))
3926 continue;
3927
3928 PF_HASHROW_LOCK(ih);
3929 count = 0;
3930 LIST_FOREACH(s, &ih->states, entry) {
3931 if (s->timeout == PFTM_UNLINKED)
3932 continue;
3933 count++;
3934 }
3935
3936 if (count > slice_count) {
3937 PF_HASHROW_UNLOCK(ih);
3938 free(pstore, M_TEMP);
3939 slice_count = count * 2;
3940 pstore = mallocarray(slice_count,
3941 sizeof(struct pf_state_export), M_TEMP,
3942 M_WAITOK | M_ZERO);
3943 goto DIOCGETSTATESV2_retry;
3944 }
3945
3946 if ((nr+count) * sizeof(*p) > ps->ps_len) {
3947 PF_HASHROW_UNLOCK(ih);
3948 goto DIOCGETSTATESV2_full;
3949 }
3950
3951 LIST_FOREACH(s, &ih->states, entry) {
3952 if (s->timeout == PFTM_UNLINKED)
3953 continue;
3954
3955 pf_state_export(p, s);
3956 p++;
3957 nr++;
3958 }
3959 PF_HASHROW_UNLOCK(ih);
3960 error = copyout(pstore, out,
3961 sizeof(struct pf_state_export) * count);
3962 if (error)
3963 break;
3964 out = ps->ps_states + nr;
3965 }
3966 DIOCGETSTATESV2_full:
3967 ps->ps_len = nr * sizeof(struct pf_state_export);
3968 free(pstore, M_TEMP);
3969
3970 break;
3971 }
3972 #endif
3973 case DIOCGETSTATUSNV: {
3974 error = pf_getstatus((struct pfioc_nv *)addr);
3975 break;
3976 }
3977
3978 case DIOCSETSTATUSIF: {
3979 struct pfioc_if *pi = (struct pfioc_if *)addr;
3980
3981 if (pi->ifname[0] == 0) {
3982 bzero(V_pf_status.ifname, IFNAMSIZ);
3983 break;
3984 }
3985 PF_RULES_WLOCK();
3986 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3987 PF_RULES_WUNLOCK();
3988 break;
3989 }
3990
3991 case DIOCCLRSTATUS: {
3992 pf_ioctl_clear_status();
3993 break;
3994 }
3995
3996 case DIOCNATLOOK: {
3997 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
3998 struct pf_state_key *sk;
3999 struct pf_kstate *state;
4000 struct pf_state_key_cmp key;
4001 int m = 0, direction = pnl->direction;
4002 int sidx, didx;
4003
4004 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
4005 sidx = (direction == PF_IN) ? 1 : 0;
4006 didx = (direction == PF_IN) ? 0 : 1;
4007
4008 if (!pnl->proto ||
4009 PF_AZERO(&pnl->saddr, pnl->af) ||
4010 PF_AZERO(&pnl->daddr, pnl->af) ||
4011 ((pnl->proto == IPPROTO_TCP ||
4012 pnl->proto == IPPROTO_UDP) &&
4013 (!pnl->dport || !pnl->sport)))
4014 error = EINVAL;
4015 else {
4016 bzero(&key, sizeof(key));
4017 key.af = pnl->af;
4018 key.proto = pnl->proto;
4019 PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4020 key.port[sidx] = pnl->sport;
4021 PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4022 key.port[didx] = pnl->dport;
4023
4024 state = pf_find_state_all(&key, direction, &m);
4025 if (state == NULL) {
4026 error = ENOENT;
4027 } else {
4028 if (m > 1) {
4029 PF_STATE_UNLOCK(state);
4030 error = E2BIG; /* more than one state */
4031 } else {
4032 sk = state->key[sidx];
4033 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4034 pnl->rsport = sk->port[sidx];
4035 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4036 pnl->rdport = sk->port[didx];
4037 PF_STATE_UNLOCK(state);
4038 }
4039 }
4040 }
4041 break;
4042 }
4043
4044 case DIOCSETTIMEOUT: {
4045 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4046
4047 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4048 &pt->seconds);
4049 break;
4050 }
4051
4052 case DIOCGETTIMEOUT: {
4053 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4054
4055 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4056 break;
4057 }
4058
4059 case DIOCGETLIMIT: {
4060 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4061
4062 error = pf_ioctl_get_limit(pl->index, &pl->limit);
4063 break;
4064 }
4065
4066 case DIOCSETLIMIT: {
4067 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4068 unsigned int old_limit;
4069
4070 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4071 pl->limit = old_limit;
4072 break;
4073 }
4074
4075 case DIOCSETDEBUG: {
4076 u_int32_t *level = (u_int32_t *)addr;
4077
4078 PF_RULES_WLOCK();
4079 V_pf_status.debug = *level;
4080 PF_RULES_WUNLOCK();
4081 break;
4082 }
4083
4084 case DIOCCLRRULECTRS: {
4085 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4086 struct pf_kruleset *ruleset = &pf_main_ruleset;
4087 struct pf_krule *rule;
4088
4089 PF_RULES_WLOCK();
4090 TAILQ_FOREACH(rule,
4091 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4092 pf_counter_u64_zero(&rule->evaluations);
4093 for (int i = 0; i < 2; i++) {
4094 pf_counter_u64_zero(&rule->packets[i]);
4095 pf_counter_u64_zero(&rule->bytes[i]);
4096 }
4097 }
4098 PF_RULES_WUNLOCK();
4099 break;
4100 }
4101
4102 case DIOCGIFSPEEDV0:
4103 case DIOCGIFSPEEDV1: {
4104 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
4105 struct pf_ifspeed_v1 ps;
4106 struct ifnet *ifp;
4107
4108 if (psp->ifname[0] == '\0') {
4109 error = EINVAL;
4110 break;
4111 }
4112
4113 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4114 if (error != 0)
4115 break;
4116 ifp = ifunit(ps.ifname);
4117 if (ifp != NULL) {
4118 psp->baudrate32 =
4119 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4120 if (cmd == DIOCGIFSPEEDV1)
4121 psp->baudrate = ifp->if_baudrate;
4122 } else {
4123 error = EINVAL;
4124 }
4125 break;
4126 }
4127
4128 #ifdef ALTQ
4129 case DIOCSTARTALTQ: {
4130 struct pf_altq *altq;
4131
4132 PF_RULES_WLOCK();
4133 /* enable all altq interfaces on active list */
4134 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4135 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4136 error = pf_enable_altq(altq);
4137 if (error != 0)
4138 break;
4139 }
4140 }
4141 if (error == 0)
4142 V_pf_altq_running = 1;
4143 PF_RULES_WUNLOCK();
4144 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4145 break;
4146 }
4147
4148 case DIOCSTOPALTQ: {
4149 struct pf_altq *altq;
4150
4151 PF_RULES_WLOCK();
4152 /* disable all altq interfaces on active list */
4153 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4154 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4155 error = pf_disable_altq(altq);
4156 if (error != 0)
4157 break;
4158 }
4159 }
4160 if (error == 0)
4161 V_pf_altq_running = 0;
4162 PF_RULES_WUNLOCK();
4163 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4164 break;
4165 }
4166
4167 case DIOCADDALTQV0:
4168 case DIOCADDALTQV1: {
4169 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4170 struct pf_altq *altq, *a;
4171 struct ifnet *ifp;
4172
4173 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4174 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4175 if (error)
4176 break;
4177 altq->local_flags = 0;
4178
4179 PF_RULES_WLOCK();
4180 if (pa->ticket != V_ticket_altqs_inactive) {
4181 PF_RULES_WUNLOCK();
4182 free(altq, M_PFALTQ);
4183 error = EBUSY;
4184 break;
4185 }
4186
4187 /*
4188 * if this is for a queue, find the discipline and
4189 * copy the necessary fields
4190 */
4191 if (altq->qname[0] != 0) {
4192 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4193 PF_RULES_WUNLOCK();
4194 error = EBUSY;
4195 free(altq, M_PFALTQ);
4196 break;
4197 }
4198 altq->altq_disc = NULL;
4199 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4200 if (strncmp(a->ifname, altq->ifname,
4201 IFNAMSIZ) == 0) {
4202 altq->altq_disc = a->altq_disc;
4203 break;
4204 }
4205 }
4206 }
4207
4208 if ((ifp = ifunit(altq->ifname)) == NULL)
4209 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4210 else
4211 error = altq_add(ifp, altq);
4212
4213 if (error) {
4214 PF_RULES_WUNLOCK();
4215 free(altq, M_PFALTQ);
4216 break;
4217 }
4218
4219 if (altq->qname[0] != 0)
4220 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4221 else
4222 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4223 /* version error check done on import above */
4224 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4225 PF_RULES_WUNLOCK();
4226 break;
4227 }
4228
4229 case DIOCGETALTQSV0:
4230 case DIOCGETALTQSV1: {
4231 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4232 struct pf_altq *altq;
4233
4234 PF_RULES_RLOCK();
4235 pa->nr = 0;
4236 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4237 pa->nr++;
4238 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4239 pa->nr++;
4240 pa->ticket = V_ticket_altqs_active;
4241 PF_RULES_RUNLOCK();
4242 break;
4243 }
4244
4245 case DIOCGETALTQV0:
4246 case DIOCGETALTQV1: {
4247 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4248 struct pf_altq *altq;
4249
4250 PF_RULES_RLOCK();
4251 if (pa->ticket != V_ticket_altqs_active) {
4252 PF_RULES_RUNLOCK();
4253 error = EBUSY;
4254 break;
4255 }
4256 altq = pf_altq_get_nth_active(pa->nr);
4257 if (altq == NULL) {
4258 PF_RULES_RUNLOCK();
4259 error = EBUSY;
4260 break;
4261 }
4262 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4263 PF_RULES_RUNLOCK();
4264 break;
4265 }
4266
4267 case DIOCCHANGEALTQV0:
4268 case DIOCCHANGEALTQV1:
4269 /* CHANGEALTQ not supported yet! */
4270 error = ENODEV;
4271 break;
4272
4273 case DIOCGETQSTATSV0:
4274 case DIOCGETQSTATSV1: {
4275 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
4276 struct pf_altq *altq;
4277 int nbytes;
4278 u_int32_t version;
4279
4280 PF_RULES_RLOCK();
4281 if (pq->ticket != V_ticket_altqs_active) {
4282 PF_RULES_RUNLOCK();
4283 error = EBUSY;
4284 break;
4285 }
4286 nbytes = pq->nbytes;
4287 altq = pf_altq_get_nth_active(pq->nr);
4288 if (altq == NULL) {
4289 PF_RULES_RUNLOCK();
4290 error = EBUSY;
4291 break;
4292 }
4293
4294 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4295 PF_RULES_RUNLOCK();
4296 error = ENXIO;
4297 break;
4298 }
4299 PF_RULES_RUNLOCK();
4300 if (cmd == DIOCGETQSTATSV0)
4301 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
4302 else
4303 version = pq->version;
4304 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4305 if (error == 0) {
4306 pq->scheduler = altq->scheduler;
4307 pq->nbytes = nbytes;
4308 }
4309 break;
4310 }
4311 #endif /* ALTQ */
4312
4313 case DIOCBEGINADDRS: {
4314 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4315
4316 error = pf_ioctl_begin_addrs(&pp->ticket);
4317 break;
4318 }
4319
4320 case DIOCADDADDR: {
4321 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4322
4323 error = pf_ioctl_add_addr(pp);
4324 break;
4325 }
4326
4327 case DIOCGETADDRS: {
4328 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4329
4330 error = pf_ioctl_get_addrs(pp);
4331 break;
4332 }
4333
4334 case DIOCGETADDR: {
4335 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4336
4337 error = pf_ioctl_get_addr(pp);
4338 break;
4339 }
4340
4341 case DIOCCHANGEADDR: {
4342 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
4343 struct pf_kpool *pool;
4344 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
4345 struct pf_kruleset *ruleset;
4346 struct pfi_kkif *kif = NULL;
4347
4348 pca->anchor[sizeof(pca->anchor) - 1] = 0;
4349
4350 if (pca->action < PF_CHANGE_ADD_HEAD ||
4351 pca->action > PF_CHANGE_REMOVE) {
4352 error = EINVAL;
4353 break;
4354 }
4355 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4356 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4357 pca->addr.addr.type != PF_ADDR_TABLE) {
4358 error = EINVAL;
4359 break;
4360 }
4361 if (pca->addr.addr.p.dyn != NULL) {
4362 error = EINVAL;
4363 break;
4364 }
4365
4366 if (pca->action != PF_CHANGE_REMOVE) {
4367 #ifndef INET
4368 if (pca->af == AF_INET) {
4369 error = EAFNOSUPPORT;
4370 break;
4371 }
4372 #endif /* INET */
4373 #ifndef INET6
4374 if (pca->af == AF_INET6) {
4375 error = EAFNOSUPPORT;
4376 break;
4377 }
4378 #endif /* INET6 */
4379 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4380 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4381 if (newpa->ifname[0])
4382 kif = pf_kkif_create(M_WAITOK);
4383 newpa->kif = NULL;
4384 }
4385 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4386 PF_RULES_WLOCK();
4387 ruleset = pf_find_kruleset(pca->anchor);
4388 if (ruleset == NULL)
4389 ERROUT(EBUSY);
4390
4391 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4392 pca->r_num, pca->r_last, 1, 1);
4393 if (pool == NULL)
4394 ERROUT(EBUSY);
4395
4396 if (pca->action != PF_CHANGE_REMOVE) {
4397 if (newpa->ifname[0]) {
4398 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4399 pfi_kkif_ref(newpa->kif);
4400 kif = NULL;
4401 }
4402
4403 switch (newpa->addr.type) {
4404 case PF_ADDR_DYNIFTL:
4405 error = pfi_dynaddr_setup(&newpa->addr,
4406 pca->af);
4407 break;
4408 case PF_ADDR_TABLE:
4409 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4410 newpa->addr.v.tblname);
4411 if (newpa->addr.p.tbl == NULL)
4412 error = ENOMEM;
4413 break;
4414 }
4415 if (error)
4416 goto DIOCCHANGEADDR_error;
4417 }
4418
4419 switch (pca->action) {
4420 case PF_CHANGE_ADD_HEAD:
4421 oldpa = TAILQ_FIRST(&pool->list);
4422 break;
4423 case PF_CHANGE_ADD_TAIL:
4424 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4425 break;
4426 default:
4427 oldpa = TAILQ_FIRST(&pool->list);
4428 for (int i = 0; oldpa && i < pca->nr; i++)
4429 oldpa = TAILQ_NEXT(oldpa, entries);
4430
4431 if (oldpa == NULL)
4432 ERROUT(EINVAL);
4433 }
4434
4435 if (pca->action == PF_CHANGE_REMOVE) {
4436 TAILQ_REMOVE(&pool->list, oldpa, entries);
4437 switch (oldpa->addr.type) {
4438 case PF_ADDR_DYNIFTL:
4439 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4440 break;
4441 case PF_ADDR_TABLE:
4442 pfr_detach_table(oldpa->addr.p.tbl);
4443 break;
4444 }
4445 if (oldpa->kif)
4446 pfi_kkif_unref(oldpa->kif);
4447 free(oldpa, M_PFRULE);
4448 } else {
4449 if (oldpa == NULL)
4450 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4451 else if (pca->action == PF_CHANGE_ADD_HEAD ||
4452 pca->action == PF_CHANGE_ADD_BEFORE)
4453 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4454 else
4455 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4456 newpa, entries);
4457 }
4458
4459 pool->cur = TAILQ_FIRST(&pool->list);
4460 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4461 PF_RULES_WUNLOCK();
4462 break;
4463
4464 #undef ERROUT
4465 DIOCCHANGEADDR_error:
4466 if (newpa != NULL) {
4467 if (newpa->kif)
4468 pfi_kkif_unref(newpa->kif);
4469 free(newpa, M_PFRULE);
4470 }
4471 PF_RULES_WUNLOCK();
4472 pf_kkif_free(kif);
4473 break;
4474 }
4475
4476 case DIOCGETRULESETS: {
4477 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4478 struct pf_kruleset *ruleset;
4479 struct pf_kanchor *anchor;
4480
4481 pr->path[sizeof(pr->path) - 1] = 0;
4482
4483 PF_RULES_RLOCK();
4484 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4485 PF_RULES_RUNLOCK();
4486 error = ENOENT;
4487 break;
4488 }
4489 pr->nr = 0;
4490 if (ruleset->anchor == NULL) {
4491 /* XXX kludge for pf_main_ruleset */
4492 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4493 if (anchor->parent == NULL)
4494 pr->nr++;
4495 } else {
4496 RB_FOREACH(anchor, pf_kanchor_node,
4497 &ruleset->anchor->children)
4498 pr->nr++;
4499 }
4500 PF_RULES_RUNLOCK();
4501 break;
4502 }
4503
4504 case DIOCGETRULESET: {
4505 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4506 struct pf_kruleset *ruleset;
4507 struct pf_kanchor *anchor;
4508 u_int32_t nr = 0;
4509
4510 pr->path[sizeof(pr->path) - 1] = 0;
4511
4512 PF_RULES_RLOCK();
4513 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4514 PF_RULES_RUNLOCK();
4515 error = ENOENT;
4516 break;
4517 }
4518 pr->name[0] = 0;
4519 if (ruleset->anchor == NULL) {
4520 /* XXX kludge for pf_main_ruleset */
4521 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4522 if (anchor->parent == NULL && nr++ == pr->nr) {
4523 strlcpy(pr->name, anchor->name,
4524 sizeof(pr->name));
4525 break;
4526 }
4527 } else {
4528 RB_FOREACH(anchor, pf_kanchor_node,
4529 &ruleset->anchor->children)
4530 if (nr++ == pr->nr) {
4531 strlcpy(pr->name, anchor->name,
4532 sizeof(pr->name));
4533 break;
4534 }
4535 }
4536 if (!pr->name[0])
4537 error = EBUSY;
4538 PF_RULES_RUNLOCK();
4539 break;
4540 }
4541
4542 case DIOCRCLRTABLES: {
4543 struct pfioc_table *io = (struct pfioc_table *)addr;
4544
4545 if (io->pfrio_esize != 0) {
4546 error = ENODEV;
4547 break;
4548 }
4549 PF_RULES_WLOCK();
4550 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4551 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4552 PF_RULES_WUNLOCK();
4553 break;
4554 }
4555
4556 case DIOCRADDTABLES: {
4557 struct pfioc_table *io = (struct pfioc_table *)addr;
4558 struct pfr_table *pfrts;
4559 size_t totlen;
4560
4561 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4562 error = ENODEV;
4563 break;
4564 }
4565
4566 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4567 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4568 error = ENOMEM;
4569 break;
4570 }
4571
4572 totlen = io->pfrio_size * sizeof(struct pfr_table);
4573 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4574 M_TEMP, M_WAITOK);
4575 error = copyin(io->pfrio_buffer, pfrts, totlen);
4576 if (error) {
4577 free(pfrts, M_TEMP);
4578 break;
4579 }
4580 PF_RULES_WLOCK();
4581 error = pfr_add_tables(pfrts, io->pfrio_size,
4582 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4583 PF_RULES_WUNLOCK();
4584 free(pfrts, M_TEMP);
4585 break;
4586 }
4587
4588 case DIOCRDELTABLES: {
4589 struct pfioc_table *io = (struct pfioc_table *)addr;
4590 struct pfr_table *pfrts;
4591 size_t totlen;
4592
4593 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4594 error = ENODEV;
4595 break;
4596 }
4597
4598 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4599 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4600 error = ENOMEM;
4601 break;
4602 }
4603
4604 totlen = io->pfrio_size * sizeof(struct pfr_table);
4605 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4606 M_TEMP, M_WAITOK);
4607 error = copyin(io->pfrio_buffer, pfrts, totlen);
4608 if (error) {
4609 free(pfrts, M_TEMP);
4610 break;
4611 }
4612 PF_RULES_WLOCK();
4613 error = pfr_del_tables(pfrts, io->pfrio_size,
4614 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4615 PF_RULES_WUNLOCK();
4616 free(pfrts, M_TEMP);
4617 break;
4618 }
4619
4620 case DIOCRGETTABLES: {
4621 struct pfioc_table *io = (struct pfioc_table *)addr;
4622 struct pfr_table *pfrts;
4623 size_t totlen;
4624 int n;
4625
4626 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4627 error = ENODEV;
4628 break;
4629 }
4630 PF_RULES_RLOCK();
4631 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4632 if (n < 0) {
4633 PF_RULES_RUNLOCK();
4634 error = EINVAL;
4635 break;
4636 }
4637 io->pfrio_size = min(io->pfrio_size, n);
4638
4639 totlen = io->pfrio_size * sizeof(struct pfr_table);
4640
4641 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4642 M_TEMP, M_NOWAIT | M_ZERO);
4643 if (pfrts == NULL) {
4644 error = ENOMEM;
4645 PF_RULES_RUNLOCK();
4646 break;
4647 }
4648 error = pfr_get_tables(&io->pfrio_table, pfrts,
4649 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4650 PF_RULES_RUNLOCK();
4651 if (error == 0)
4652 error = copyout(pfrts, io->pfrio_buffer, totlen);
4653 free(pfrts, M_TEMP);
4654 break;
4655 }
4656
4657 case DIOCRGETTSTATS: {
4658 struct pfioc_table *io = (struct pfioc_table *)addr;
4659 struct pfr_tstats *pfrtstats;
4660 size_t totlen;
4661 int n;
4662
4663 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4664 error = ENODEV;
4665 break;
4666 }
4667 PF_TABLE_STATS_LOCK();
4668 PF_RULES_RLOCK();
4669 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4670 if (n < 0) {
4671 PF_RULES_RUNLOCK();
4672 PF_TABLE_STATS_UNLOCK();
4673 error = EINVAL;
4674 break;
4675 }
4676 io->pfrio_size = min(io->pfrio_size, n);
4677
4678 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4679 pfrtstats = mallocarray(io->pfrio_size,
4680 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4681 if (pfrtstats == NULL) {
4682 error = ENOMEM;
4683 PF_RULES_RUNLOCK();
4684 PF_TABLE_STATS_UNLOCK();
4685 break;
4686 }
4687 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4688 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4689 PF_RULES_RUNLOCK();
4690 PF_TABLE_STATS_UNLOCK();
4691 if (error == 0)
4692 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4693 free(pfrtstats, M_TEMP);
4694 break;
4695 }
4696
4697 case DIOCRCLRTSTATS: {
4698 struct pfioc_table *io = (struct pfioc_table *)addr;
4699 struct pfr_table *pfrts;
4700 size_t totlen;
4701
4702 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4703 error = ENODEV;
4704 break;
4705 }
4706
4707 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4708 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4709 /* We used to count tables and use the minimum required
4710 * size, so we didn't fail on overly large requests.
4711 * Keep doing so. */
4712 io->pfrio_size = pf_ioctl_maxcount;
4713 break;
4714 }
4715
4716 totlen = io->pfrio_size * sizeof(struct pfr_table);
4717 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4718 M_TEMP, M_WAITOK);
4719 error = copyin(io->pfrio_buffer, pfrts, totlen);
4720 if (error) {
4721 free(pfrts, M_TEMP);
4722 break;
4723 }
4724
4725 PF_TABLE_STATS_LOCK();
4726 PF_RULES_RLOCK();
4727 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4728 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4729 PF_RULES_RUNLOCK();
4730 PF_TABLE_STATS_UNLOCK();
4731 free(pfrts, M_TEMP);
4732 break;
4733 }
4734
4735 case DIOCRSETTFLAGS: {
4736 struct pfioc_table *io = (struct pfioc_table *)addr;
4737 struct pfr_table *pfrts;
4738 size_t totlen;
4739 int n;
4740
4741 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4742 error = ENODEV;
4743 break;
4744 }
4745
4746 PF_RULES_RLOCK();
4747 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4748 if (n < 0) {
4749 PF_RULES_RUNLOCK();
4750 error = EINVAL;
4751 break;
4752 }
4753
4754 io->pfrio_size = min(io->pfrio_size, n);
4755 PF_RULES_RUNLOCK();
4756
4757 totlen = io->pfrio_size * sizeof(struct pfr_table);
4758 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4759 M_TEMP, M_WAITOK);
4760 error = copyin(io->pfrio_buffer, pfrts, totlen);
4761 if (error) {
4762 free(pfrts, M_TEMP);
4763 break;
4764 }
4765 PF_RULES_WLOCK();
4766 error = pfr_set_tflags(pfrts, io->pfrio_size,
4767 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4768 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4769 PF_RULES_WUNLOCK();
4770 free(pfrts, M_TEMP);
4771 break;
4772 }
4773
4774 case DIOCRCLRADDRS: {
4775 struct pfioc_table *io = (struct pfioc_table *)addr;
4776
4777 if (io->pfrio_esize != 0) {
4778 error = ENODEV;
4779 break;
4780 }
4781 PF_RULES_WLOCK();
4782 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4783 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4784 PF_RULES_WUNLOCK();
4785 break;
4786 }
4787
4788 case DIOCRADDADDRS: {
4789 struct pfioc_table *io = (struct pfioc_table *)addr;
4790 struct pfr_addr *pfras;
4791 size_t totlen;
4792
4793 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4794 error = ENODEV;
4795 break;
4796 }
4797 if (io->pfrio_size < 0 ||
4798 io->pfrio_size > pf_ioctl_maxcount ||
4799 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4800 error = EINVAL;
4801 break;
4802 }
4803 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4804 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4805 M_TEMP, M_WAITOK);
4806 error = copyin(io->pfrio_buffer, pfras, totlen);
4807 if (error) {
4808 free(pfras, M_TEMP);
4809 break;
4810 }
4811 PF_RULES_WLOCK();
4812 error = pfr_add_addrs(&io->pfrio_table, pfras,
4813 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4814 PFR_FLAG_USERIOCTL);
4815 PF_RULES_WUNLOCK();
4816 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4817 error = copyout(pfras, io->pfrio_buffer, totlen);
4818 free(pfras, M_TEMP);
4819 break;
4820 }
4821
4822 case DIOCRDELADDRS: {
4823 struct pfioc_table *io = (struct pfioc_table *)addr;
4824 struct pfr_addr *pfras;
4825 size_t totlen;
4826
4827 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4828 error = ENODEV;
4829 break;
4830 }
4831 if (io->pfrio_size < 0 ||
4832 io->pfrio_size > pf_ioctl_maxcount ||
4833 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4834 error = EINVAL;
4835 break;
4836 }
4837 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4838 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4839 M_TEMP, M_WAITOK);
4840 error = copyin(io->pfrio_buffer, pfras, totlen);
4841 if (error) {
4842 free(pfras, M_TEMP);
4843 break;
4844 }
4845 PF_RULES_WLOCK();
4846 error = pfr_del_addrs(&io->pfrio_table, pfras,
4847 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4848 PFR_FLAG_USERIOCTL);
4849 PF_RULES_WUNLOCK();
4850 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4851 error = copyout(pfras, io->pfrio_buffer, totlen);
4852 free(pfras, M_TEMP);
4853 break;
4854 }
4855
4856 case DIOCRSETADDRS: {
4857 struct pfioc_table *io = (struct pfioc_table *)addr;
4858 struct pfr_addr *pfras;
4859 size_t totlen, count;
4860
4861 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4862 error = ENODEV;
4863 break;
4864 }
4865 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4866 error = EINVAL;
4867 break;
4868 }
4869 count = max(io->pfrio_size, io->pfrio_size2);
4870 if (count > pf_ioctl_maxcount ||
4871 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4872 error = EINVAL;
4873 break;
4874 }
4875 totlen = count * sizeof(struct pfr_addr);
4876 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4877 M_WAITOK);
4878 error = copyin(io->pfrio_buffer, pfras, totlen);
4879 if (error) {
4880 free(pfras, M_TEMP);
4881 break;
4882 }
4883 PF_RULES_WLOCK();
4884 error = pfr_set_addrs(&io->pfrio_table, pfras,
4885 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4886 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4887 PFR_FLAG_USERIOCTL, 0);
4888 PF_RULES_WUNLOCK();
4889 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4890 error = copyout(pfras, io->pfrio_buffer, totlen);
4891 free(pfras, M_TEMP);
4892 break;
4893 }
4894
4895 case DIOCRGETADDRS: {
4896 struct pfioc_table *io = (struct pfioc_table *)addr;
4897 struct pfr_addr *pfras;
4898 size_t totlen;
4899
4900 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4901 error = ENODEV;
4902 break;
4903 }
4904 if (io->pfrio_size < 0 ||
4905 io->pfrio_size > pf_ioctl_maxcount ||
4906 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4907 error = EINVAL;
4908 break;
4909 }
4910 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4911 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4912 M_TEMP, M_WAITOK | M_ZERO);
4913 PF_RULES_RLOCK();
4914 error = pfr_get_addrs(&io->pfrio_table, pfras,
4915 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4916 PF_RULES_RUNLOCK();
4917 if (error == 0)
4918 error = copyout(pfras, io->pfrio_buffer, totlen);
4919 free(pfras, M_TEMP);
4920 break;
4921 }
4922
4923 case DIOCRGETASTATS: {
4924 struct pfioc_table *io = (struct pfioc_table *)addr;
4925 struct pfr_astats *pfrastats;
4926 size_t totlen;
4927
4928 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4929 error = ENODEV;
4930 break;
4931 }
4932 if (io->pfrio_size < 0 ||
4933 io->pfrio_size > pf_ioctl_maxcount ||
4934 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4935 error = EINVAL;
4936 break;
4937 }
4938 totlen = io->pfrio_size * sizeof(struct pfr_astats);
4939 pfrastats = mallocarray(io->pfrio_size,
4940 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4941 PF_RULES_RLOCK();
4942 error = pfr_get_astats(&io->pfrio_table, pfrastats,
4943 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4944 PF_RULES_RUNLOCK();
4945 if (error == 0)
4946 error = copyout(pfrastats, io->pfrio_buffer, totlen);
4947 free(pfrastats, M_TEMP);
4948 break;
4949 }
4950
4951 case DIOCRCLRASTATS: {
4952 struct pfioc_table *io = (struct pfioc_table *)addr;
4953 struct pfr_addr *pfras;
4954 size_t totlen;
4955
4956 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4957 error = ENODEV;
4958 break;
4959 }
4960 if (io->pfrio_size < 0 ||
4961 io->pfrio_size > pf_ioctl_maxcount ||
4962 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4963 error = EINVAL;
4964 break;
4965 }
4966 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4967 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4968 M_TEMP, M_WAITOK);
4969 error = copyin(io->pfrio_buffer, pfras, totlen);
4970 if (error) {
4971 free(pfras, M_TEMP);
4972 break;
4973 }
4974 PF_RULES_WLOCK();
4975 error = pfr_clr_astats(&io->pfrio_table, pfras,
4976 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4977 PFR_FLAG_USERIOCTL);
4978 PF_RULES_WUNLOCK();
4979 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4980 error = copyout(pfras, io->pfrio_buffer, totlen);
4981 free(pfras, M_TEMP);
4982 break;
4983 }
4984
4985 case DIOCRTSTADDRS: {
4986 struct pfioc_table *io = (struct pfioc_table *)addr;
4987 struct pfr_addr *pfras;
4988 size_t totlen;
4989
4990 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4991 error = ENODEV;
4992 break;
4993 }
4994 if (io->pfrio_size < 0 ||
4995 io->pfrio_size > pf_ioctl_maxcount ||
4996 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4997 error = EINVAL;
4998 break;
4999 }
5000 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5001 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5002 M_TEMP, M_WAITOK);
5003 error = copyin(io->pfrio_buffer, pfras, totlen);
5004 if (error) {
5005 free(pfras, M_TEMP);
5006 break;
5007 }
5008 PF_RULES_RLOCK();
5009 error = pfr_tst_addrs(&io->pfrio_table, pfras,
5010 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5011 PFR_FLAG_USERIOCTL);
5012 PF_RULES_RUNLOCK();
5013 if (error == 0)
5014 error = copyout(pfras, io->pfrio_buffer, totlen);
5015 free(pfras, M_TEMP);
5016 break;
5017 }
5018
5019 case DIOCRINADEFINE: {
5020 struct pfioc_table *io = (struct pfioc_table *)addr;
5021 struct pfr_addr *pfras;
5022 size_t totlen;
5023
5024 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5025 error = ENODEV;
5026 break;
5027 }
5028 if (io->pfrio_size < 0 ||
5029 io->pfrio_size > pf_ioctl_maxcount ||
5030 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5031 error = EINVAL;
5032 break;
5033 }
5034 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5035 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5036 M_TEMP, M_WAITOK);
5037 error = copyin(io->pfrio_buffer, pfras, totlen);
5038 if (error) {
5039 free(pfras, M_TEMP);
5040 break;
5041 }
5042 PF_RULES_WLOCK();
5043 error = pfr_ina_define(&io->pfrio_table, pfras,
5044 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5045 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5046 PF_RULES_WUNLOCK();
5047 free(pfras, M_TEMP);
5048 break;
5049 }
5050
5051 case DIOCOSFPADD: {
5052 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5053 PF_RULES_WLOCK();
5054 error = pf_osfp_add(io);
5055 PF_RULES_WUNLOCK();
5056 break;
5057 }
5058
5059 case DIOCOSFPGET: {
5060 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5061 PF_RULES_RLOCK();
5062 error = pf_osfp_get(io);
5063 PF_RULES_RUNLOCK();
5064 break;
5065 }
5066
5067 case DIOCXBEGIN: {
5068 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5069 struct pfioc_trans_e *ioes, *ioe;
5070 size_t totlen;
5071 int i;
5072
5073 if (io->esize != sizeof(*ioe)) {
5074 error = ENODEV;
5075 break;
5076 }
5077 if (io->size < 0 ||
5078 io->size > pf_ioctl_maxcount ||
5079 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5080 error = EINVAL;
5081 break;
5082 }
5083 totlen = sizeof(struct pfioc_trans_e) * io->size;
5084 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5085 M_TEMP, M_WAITOK);
5086 error = copyin(io->array, ioes, totlen);
5087 if (error) {
5088 free(ioes, M_TEMP);
5089 break;
5090 }
5091 /* Ensure there's no more ethernet rules to clean up. */
5092 NET_EPOCH_DRAIN_CALLBACKS();
5093 PF_RULES_WLOCK();
5094 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5095 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5096 switch (ioe->rs_num) {
5097 case PF_RULESET_ETH:
5098 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5099 PF_RULES_WUNLOCK();
5100 free(ioes, M_TEMP);
5101 goto fail;
5102 }
5103 break;
5104 #ifdef ALTQ
5105 case PF_RULESET_ALTQ:
5106 if (ioe->anchor[0]) {
5107 PF_RULES_WUNLOCK();
5108 free(ioes, M_TEMP);
5109 error = EINVAL;
5110 goto fail;
5111 }
5112 if ((error = pf_begin_altq(&ioe->ticket))) {
5113 PF_RULES_WUNLOCK();
5114 free(ioes, M_TEMP);
5115 goto fail;
5116 }
5117 break;
5118 #endif /* ALTQ */
5119 case PF_RULESET_TABLE:
5120 {
5121 struct pfr_table table;
5122
5123 bzero(&table, sizeof(table));
5124 strlcpy(table.pfrt_anchor, ioe->anchor,
5125 sizeof(table.pfrt_anchor));
5126 if ((error = pfr_ina_begin(&table,
5127 &ioe->ticket, NULL, 0))) {
5128 PF_RULES_WUNLOCK();
5129 free(ioes, M_TEMP);
5130 goto fail;
5131 }
5132 break;
5133 }
5134 default:
5135 if ((error = pf_begin_rules(&ioe->ticket,
5136 ioe->rs_num, ioe->anchor))) {
5137 PF_RULES_WUNLOCK();
5138 free(ioes, M_TEMP);
5139 goto fail;
5140 }
5141 break;
5142 }
5143 }
5144 PF_RULES_WUNLOCK();
5145 error = copyout(ioes, io->array, totlen);
5146 free(ioes, M_TEMP);
5147 break;
5148 }
5149
5150 case DIOCXROLLBACK: {
5151 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5152 struct pfioc_trans_e *ioe, *ioes;
5153 size_t totlen;
5154 int i;
5155
5156 if (io->esize != sizeof(*ioe)) {
5157 error = ENODEV;
5158 break;
5159 }
5160 if (io->size < 0 ||
5161 io->size > pf_ioctl_maxcount ||
5162 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5163 error = EINVAL;
5164 break;
5165 }
5166 totlen = sizeof(struct pfioc_trans_e) * io->size;
5167 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5168 M_TEMP, M_WAITOK);
5169 error = copyin(io->array, ioes, totlen);
5170 if (error) {
5171 free(ioes, M_TEMP);
5172 break;
5173 }
5174 PF_RULES_WLOCK();
5175 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5176 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5177 switch (ioe->rs_num) {
5178 case PF_RULESET_ETH:
5179 if ((error = pf_rollback_eth(ioe->ticket,
5180 ioe->anchor))) {
5181 PF_RULES_WUNLOCK();
5182 free(ioes, M_TEMP);
5183 goto fail; /* really bad */
5184 }
5185 break;
5186 #ifdef ALTQ
5187 case PF_RULESET_ALTQ:
5188 if (ioe->anchor[0]) {
5189 PF_RULES_WUNLOCK();
5190 free(ioes, M_TEMP);
5191 error = EINVAL;
5192 goto fail;
5193 }
5194 if ((error = pf_rollback_altq(ioe->ticket))) {
5195 PF_RULES_WUNLOCK();
5196 free(ioes, M_TEMP);
5197 goto fail; /* really bad */
5198 }
5199 break;
5200 #endif /* ALTQ */
5201 case PF_RULESET_TABLE:
5202 {
5203 struct pfr_table table;
5204
5205 bzero(&table, sizeof(table));
5206 strlcpy(table.pfrt_anchor, ioe->anchor,
5207 sizeof(table.pfrt_anchor));
5208 if ((error = pfr_ina_rollback(&table,
5209 ioe->ticket, NULL, 0))) {
5210 PF_RULES_WUNLOCK();
5211 free(ioes, M_TEMP);
5212 goto fail; /* really bad */
5213 }
5214 break;
5215 }
5216 default:
5217 if ((error = pf_rollback_rules(ioe->ticket,
5218 ioe->rs_num, ioe->anchor))) {
5219 PF_RULES_WUNLOCK();
5220 free(ioes, M_TEMP);
5221 goto fail; /* really bad */
5222 }
5223 break;
5224 }
5225 }
5226 PF_RULES_WUNLOCK();
5227 free(ioes, M_TEMP);
5228 break;
5229 }
5230
5231 case DIOCXCOMMIT: {
5232 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5233 struct pfioc_trans_e *ioe, *ioes;
5234 struct pf_kruleset *rs;
5235 struct pf_keth_ruleset *ers;
5236 size_t totlen;
5237 int i;
5238
5239 if (io->esize != sizeof(*ioe)) {
5240 error = ENODEV;
5241 break;
5242 }
5243
5244 if (io->size < 0 ||
5245 io->size > pf_ioctl_maxcount ||
5246 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5247 error = EINVAL;
5248 break;
5249 }
5250
5251 totlen = sizeof(struct pfioc_trans_e) * io->size;
5252 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5253 M_TEMP, M_WAITOK);
5254 error = copyin(io->array, ioes, totlen);
5255 if (error) {
5256 free(ioes, M_TEMP);
5257 break;
5258 }
5259 PF_RULES_WLOCK();
5260 /* First makes sure everything will succeed. */
5261 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5262 ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5263 switch (ioe->rs_num) {
5264 case PF_RULESET_ETH:
5265 ers = pf_find_keth_ruleset(ioe->anchor);
5266 if (ers == NULL || ioe->ticket == 0 ||
5267 ioe->ticket != ers->inactive.ticket) {
5268 PF_RULES_WUNLOCK();
5269 free(ioes, M_TEMP);
5270 error = EINVAL;
5271 goto fail;
5272 }
5273 break;
5274 #ifdef ALTQ
5275 case PF_RULESET_ALTQ:
5276 if (ioe->anchor[0]) {
5277 PF_RULES_WUNLOCK();
5278 free(ioes, M_TEMP);
5279 error = EINVAL;
5280 goto fail;
5281 }
5282 if (!V_altqs_inactive_open || ioe->ticket !=
5283 V_ticket_altqs_inactive) {
5284 PF_RULES_WUNLOCK();
5285 free(ioes, M_TEMP);
5286 error = EBUSY;
5287 goto fail;
5288 }
5289 break;
5290 #endif /* ALTQ */
5291 case PF_RULESET_TABLE:
5292 rs = pf_find_kruleset(ioe->anchor);
5293 if (rs == NULL || !rs->topen || ioe->ticket !=
5294 rs->tticket) {
5295 PF_RULES_WUNLOCK();
5296 free(ioes, M_TEMP);
5297 error = EBUSY;
5298 goto fail;
5299 }
5300 break;
5301 default:
5302 if (ioe->rs_num < 0 || ioe->rs_num >=
5303 PF_RULESET_MAX) {
5304 PF_RULES_WUNLOCK();
5305 free(ioes, M_TEMP);
5306 error = EINVAL;
5307 goto fail;
5308 }
5309 rs = pf_find_kruleset(ioe->anchor);
5310 if (rs == NULL ||
5311 !rs->rules[ioe->rs_num].inactive.open ||
5312 rs->rules[ioe->rs_num].inactive.ticket !=
5313 ioe->ticket) {
5314 PF_RULES_WUNLOCK();
5315 free(ioes, M_TEMP);
5316 error = EBUSY;
5317 goto fail;
5318 }
5319 break;
5320 }
5321 }
5322 /* Now do the commit - no errors should happen here. */
5323 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5324 switch (ioe->rs_num) {
5325 case PF_RULESET_ETH:
5326 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5327 PF_RULES_WUNLOCK();
5328 free(ioes, M_TEMP);
5329 goto fail; /* really bad */
5330 }
5331 break;
5332 #ifdef ALTQ
5333 case PF_RULESET_ALTQ:
5334 if ((error = pf_commit_altq(ioe->ticket))) {
5335 PF_RULES_WUNLOCK();
5336 free(ioes, M_TEMP);
5337 goto fail; /* really bad */
5338 }
5339 break;
5340 #endif /* ALTQ */
5341 case PF_RULESET_TABLE:
5342 {
5343 struct pfr_table table;
5344
5345 bzero(&table, sizeof(table));
5346 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
5347 sizeof(table.pfrt_anchor));
5348 if ((error = pfr_ina_commit(&table,
5349 ioe->ticket, NULL, NULL, 0))) {
5350 PF_RULES_WUNLOCK();
5351 free(ioes, M_TEMP);
5352 goto fail; /* really bad */
5353 }
5354 break;
5355 }
5356 default:
5357 if ((error = pf_commit_rules(ioe->ticket,
5358 ioe->rs_num, ioe->anchor))) {
5359 PF_RULES_WUNLOCK();
5360 free(ioes, M_TEMP);
5361 goto fail; /* really bad */
5362 }
5363 break;
5364 }
5365 }
5366 PF_RULES_WUNLOCK();
5367
5368 /* Only hook into EtherNet taffic if we've got rules for it. */
5369 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5370 hook_pf_eth();
5371 else
5372 dehook_pf_eth();
5373
5374 free(ioes, M_TEMP);
5375 break;
5376 }
5377
5378 case DIOCGETSRCNODES: {
5379 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
5380 struct pf_srchash *sh;
5381 struct pf_ksrc_node *n;
5382 struct pf_src_node *p, *pstore;
5383 uint32_t i, nr = 0;
5384
5385 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5386 i++, sh++) {
5387 PF_HASHROW_LOCK(sh);
5388 LIST_FOREACH(n, &sh->nodes, entry)
5389 nr++;
5390 PF_HASHROW_UNLOCK(sh);
5391 }
5392
5393 psn->psn_len = min(psn->psn_len,
5394 sizeof(struct pf_src_node) * nr);
5395
5396 if (psn->psn_len == 0) {
5397 psn->psn_len = sizeof(struct pf_src_node) * nr;
5398 break;
5399 }
5400
5401 nr = 0;
5402
5403 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5404 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5405 i++, sh++) {
5406 PF_HASHROW_LOCK(sh);
5407 LIST_FOREACH(n, &sh->nodes, entry) {
5408
5409 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5410 break;
5411
5412 pf_src_node_copy(n, p);
5413
5414 p++;
5415 nr++;
5416 }
5417 PF_HASHROW_UNLOCK(sh);
5418 }
5419 error = copyout(pstore, psn->psn_src_nodes,
5420 sizeof(struct pf_src_node) * nr);
5421 if (error) {
5422 free(pstore, M_TEMP);
5423 break;
5424 }
5425 psn->psn_len = sizeof(struct pf_src_node) * nr;
5426 free(pstore, M_TEMP);
5427 break;
5428 }
5429
5430 case DIOCCLRSRCNODES: {
5431 pf_clear_srcnodes(NULL);
5432 pf_purge_expired_src_nodes();
5433 break;
5434 }
5435
5436 case DIOCKILLSRCNODES:
5437 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5438 break;
5439
5440 #ifdef COMPAT_FREEBSD13
5441 case DIOCKEEPCOUNTERS_FREEBSD13:
5442 #endif
5443 case DIOCKEEPCOUNTERS:
5444 error = pf_keepcounters((struct pfioc_nv *)addr);
5445 break;
5446
5447 case DIOCGETSYNCOOKIES:
5448 error = pf_get_syncookies((struct pfioc_nv *)addr);
5449 break;
5450
5451 case DIOCSETSYNCOOKIES:
5452 error = pf_set_syncookies((struct pfioc_nv *)addr);
5453 break;
5454
5455 case DIOCSETHOSTID: {
5456 u_int32_t *hostid = (u_int32_t *)addr;
5457
5458 PF_RULES_WLOCK();
5459 if (*hostid == 0)
5460 V_pf_status.hostid = arc4random();
5461 else
5462 V_pf_status.hostid = *hostid;
5463 PF_RULES_WUNLOCK();
5464 break;
5465 }
5466
5467 case DIOCOSFPFLUSH:
5468 PF_RULES_WLOCK();
5469 pf_osfp_flush();
5470 PF_RULES_WUNLOCK();
5471 break;
5472
5473 case DIOCIGETIFACES: {
5474 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5475 struct pfi_kif *ifstore;
5476 size_t bufsiz;
5477
5478 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5479 error = ENODEV;
5480 break;
5481 }
5482
5483 if (io->pfiio_size < 0 ||
5484 io->pfiio_size > pf_ioctl_maxcount ||
5485 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5486 error = EINVAL;
5487 break;
5488 }
5489
5490 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5491
5492 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5493 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5494 M_TEMP, M_WAITOK | M_ZERO);
5495
5496 PF_RULES_RLOCK();
5497 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5498 PF_RULES_RUNLOCK();
5499 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5500 free(ifstore, M_TEMP);
5501 break;
5502 }
5503
5504 case DIOCSETIFFLAG: {
5505 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5506
5507 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5508
5509 PF_RULES_WLOCK();
5510 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5511 PF_RULES_WUNLOCK();
5512 break;
5513 }
5514
5515 case DIOCCLRIFFLAG: {
5516 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5517
5518 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5519
5520 PF_RULES_WLOCK();
5521 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5522 PF_RULES_WUNLOCK();
5523 break;
5524 }
5525
5526 case DIOCSETREASS: {
5527 u_int32_t *reass = (u_int32_t *)addr;
5528
5529 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5530 /* Removal of DF flag without reassembly enabled is not a
5531 * valid combination. Disable reassembly in such case. */
5532 if (!(V_pf_status.reass & PF_REASS_ENABLED))
5533 V_pf_status.reass = 0;
5534 break;
5535 }
5536
5537 default:
5538 error = ENODEV;
5539 break;
5540 }
5541 fail:
5542 CURVNET_RESTORE();
5543
5544 #undef ERROUT_IOCTL
5545
5546 return (error);
5547 }
5548
5549 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5550 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5551 {
5552 bzero(sp, sizeof(union pfsync_state_union));
5553
5554 /* copy from state key */
5555 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5556 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5557 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5558 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5559 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5560 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5561 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5562 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5563 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5564 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5565
5566 /* copy from state */
5567 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5568 bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5569 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5570 sp->pfs_1301.expire = pf_state_expires(st);
5571 if (sp->pfs_1301.expire <= time_uptime)
5572 sp->pfs_1301.expire = htonl(0);
5573 else
5574 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5575
5576 sp->pfs_1301.direction = st->direction;
5577 sp->pfs_1301.log = st->act.log;
5578 sp->pfs_1301.timeout = st->timeout;
5579
5580 switch (msg_version) {
5581 case PFSYNC_MSG_VERSION_1301:
5582 sp->pfs_1301.state_flags = st->state_flags;
5583 break;
5584 case PFSYNC_MSG_VERSION_1400:
5585 sp->pfs_1400.state_flags = htons(st->state_flags);
5586 sp->pfs_1400.qid = htons(st->act.qid);
5587 sp->pfs_1400.pqid = htons(st->act.pqid);
5588 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5589 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5590 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5591 sp->pfs_1400.min_ttl = st->act.min_ttl;
5592 sp->pfs_1400.set_tos = st->act.set_tos;
5593 sp->pfs_1400.max_mss = htons(st->act.max_mss);
5594 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5595 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5596 sp->pfs_1400.rt = st->rt;
5597 if (st->rt_kif)
5598 strlcpy(sp->pfs_1400.rt_ifname,
5599 st->rt_kif->pfik_name,
5600 sizeof(sp->pfs_1400.rt_ifname));
5601 break;
5602 default:
5603 panic("%s: Unsupported pfsync_msg_version %d",
5604 __func__, msg_version);
5605 }
5606
5607 if (st->src_node)
5608 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5609 if (st->nat_src_node)
5610 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5611
5612 sp->pfs_1301.id = st->id;
5613 sp->pfs_1301.creatorid = st->creatorid;
5614 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5615 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5616
5617 if (st->rule.ptr == NULL)
5618 sp->pfs_1301.rule = htonl(-1);
5619 else
5620 sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5621 if (st->anchor.ptr == NULL)
5622 sp->pfs_1301.anchor = htonl(-1);
5623 else
5624 sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5625 if (st->nat_rule.ptr == NULL)
5626 sp->pfs_1301.nat_rule = htonl(-1);
5627 else
5628 sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5629
5630 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5631 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5632 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5633 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5634 }
5635
5636 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5637 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5638 {
5639 bzero(sp, sizeof(*sp));
5640
5641 sp->version = PF_STATE_VERSION;
5642
5643 /* copy from state key */
5644 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5645 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5646 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5647 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5648 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5649 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5650 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5651 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5652 sp->proto = st->key[PF_SK_WIRE]->proto;
5653 sp->af = st->key[PF_SK_WIRE]->af;
5654
5655 /* copy from state */
5656 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5657 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5658 sizeof(sp->orig_ifname));
5659 bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5660 sp->creation = htonl(time_uptime - (st->creation / 1000));
5661 sp->expire = pf_state_expires(st);
5662 if (sp->expire <= time_uptime)
5663 sp->expire = htonl(0);
5664 else
5665 sp->expire = htonl(sp->expire - time_uptime);
5666
5667 sp->direction = st->direction;
5668 sp->log = st->act.log;
5669 sp->timeout = st->timeout;
5670 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5671 sp->state_flags_compat = st->state_flags;
5672 sp->state_flags = htons(st->state_flags);
5673 if (st->src_node)
5674 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5675 if (st->nat_src_node)
5676 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5677
5678 sp->id = st->id;
5679 sp->creatorid = st->creatorid;
5680 pf_state_peer_hton(&st->src, &sp->src);
5681 pf_state_peer_hton(&st->dst, &sp->dst);
5682
5683 if (st->rule.ptr == NULL)
5684 sp->rule = htonl(-1);
5685 else
5686 sp->rule = htonl(st->rule.ptr->nr);
5687 if (st->anchor.ptr == NULL)
5688 sp->anchor = htonl(-1);
5689 else
5690 sp->anchor = htonl(st->anchor.ptr->nr);
5691 if (st->nat_rule.ptr == NULL)
5692 sp->nat_rule = htonl(-1);
5693 else
5694 sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5695
5696 sp->packets[0] = st->packets[0];
5697 sp->packets[1] = st->packets[1];
5698 sp->bytes[0] = st->bytes[0];
5699 sp->bytes[1] = st->bytes[1];
5700
5701 sp->qid = htons(st->act.qid);
5702 sp->pqid = htons(st->act.pqid);
5703 sp->dnpipe = htons(st->act.dnpipe);
5704 sp->dnrpipe = htons(st->act.dnrpipe);
5705 sp->rtableid = htonl(st->act.rtableid);
5706 sp->min_ttl = st->act.min_ttl;
5707 sp->set_tos = st->act.set_tos;
5708 sp->max_mss = htons(st->act.max_mss);
5709 sp->rt = st->rt;
5710 if (st->rt_kif)
5711 strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5712 sizeof(sp->rt_ifname));
5713 sp->set_prio[0] = st->act.set_prio[0];
5714 sp->set_prio[1] = st->act.set_prio[1];
5715
5716 }
5717
5718 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5719 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5720 {
5721 struct pfr_ktable *kt;
5722
5723 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5724
5725 kt = aw->p.tbl;
5726 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5727 kt = kt->pfrkt_root;
5728 aw->p.tbl = NULL;
5729 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5730 kt->pfrkt_cnt : -1;
5731 }
5732
5733 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5734 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5735 size_t number, char **names)
5736 {
5737 nvlist_t *nvc;
5738
5739 nvc = nvlist_create(0);
5740 if (nvc == NULL)
5741 return (ENOMEM);
5742
5743 for (int i = 0; i < number; i++) {
5744 nvlist_append_number_array(nvc, "counters",
5745 counter_u64_fetch(counters[i]));
5746 nvlist_append_string_array(nvc, "names",
5747 names[i]);
5748 nvlist_append_number_array(nvc, "ids",
5749 i);
5750 }
5751 nvlist_add_nvlist(nvl, name, nvc);
5752 nvlist_destroy(nvc);
5753
5754 return (0);
5755 }
5756
5757 static int
pf_getstatus(struct pfioc_nv * nv)5758 pf_getstatus(struct pfioc_nv *nv)
5759 {
5760 nvlist_t *nvl = NULL, *nvc = NULL;
5761 void *nvlpacked = NULL;
5762 int error;
5763 struct pf_status s;
5764 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5765 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5766 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5767 PF_RULES_RLOCK_TRACKER;
5768
5769 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
5770
5771 PF_RULES_RLOCK();
5772
5773 nvl = nvlist_create(0);
5774 if (nvl == NULL)
5775 ERROUT(ENOMEM);
5776
5777 nvlist_add_bool(nvl, "running", V_pf_status.running);
5778 nvlist_add_number(nvl, "since", V_pf_status.since);
5779 nvlist_add_number(nvl, "debug", V_pf_status.debug);
5780 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5781 nvlist_add_number(nvl, "states", V_pf_status.states);
5782 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5783 nvlist_add_number(nvl, "reass", V_pf_status.reass);
5784 nvlist_add_bool(nvl, "syncookies_active",
5785 V_pf_status.syncookies_active);
5786 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5787
5788 /* counters */
5789 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5790 PFRES_MAX, pf_reasons);
5791 if (error != 0)
5792 ERROUT(error);
5793
5794 /* lcounters */
5795 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5796 KLCNT_MAX, pf_lcounter);
5797 if (error != 0)
5798 ERROUT(error);
5799
5800 /* fcounters */
5801 nvc = nvlist_create(0);
5802 if (nvc == NULL)
5803 ERROUT(ENOMEM);
5804
5805 for (int i = 0; i < FCNT_MAX; i++) {
5806 nvlist_append_number_array(nvc, "counters",
5807 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5808 nvlist_append_string_array(nvc, "names",
5809 pf_fcounter[i]);
5810 nvlist_append_number_array(nvc, "ids",
5811 i);
5812 }
5813 nvlist_add_nvlist(nvl, "fcounters", nvc);
5814 nvlist_destroy(nvc);
5815 nvc = NULL;
5816
5817 /* scounters */
5818 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5819 SCNT_MAX, pf_fcounter);
5820 if (error != 0)
5821 ERROUT(error);
5822
5823 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5824 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5825 PF_MD5_DIGEST_LENGTH);
5826
5827 pfi_update_status(V_pf_status.ifname, &s);
5828
5829 /* pcounters / bcounters */
5830 for (int i = 0; i < 2; i++) {
5831 for (int j = 0; j < 2; j++) {
5832 for (int k = 0; k < 2; k++) {
5833 nvlist_append_number_array(nvl, "pcounters",
5834 s.pcounters[i][j][k]);
5835 }
5836 nvlist_append_number_array(nvl, "bcounters",
5837 s.bcounters[i][j]);
5838 }
5839 }
5840
5841 nvlpacked = nvlist_pack(nvl, &nv->len);
5842 if (nvlpacked == NULL)
5843 ERROUT(ENOMEM);
5844
5845 if (nv->size == 0)
5846 ERROUT(0);
5847 else if (nv->size < nv->len)
5848 ERROUT(ENOSPC);
5849
5850 PF_RULES_RUNLOCK();
5851 error = copyout(nvlpacked, nv->data, nv->len);
5852 goto done;
5853
5854 #undef ERROUT
5855 errout:
5856 PF_RULES_RUNLOCK();
5857 done:
5858 free(nvlpacked, M_NVLIST);
5859 nvlist_destroy(nvc);
5860 nvlist_destroy(nvl);
5861
5862 return (error);
5863 }
5864
5865 /*
5866 * XXX - Check for version mismatch!!!
5867 */
5868 static void
pf_clear_all_states(void)5869 pf_clear_all_states(void)
5870 {
5871 struct epoch_tracker et;
5872 struct pf_kstate *s;
5873 u_int i;
5874
5875 NET_EPOCH_ENTER(et);
5876 for (i = 0; i <= V_pf_hashmask; i++) {
5877 struct pf_idhash *ih = &V_pf_idhash[i];
5878 relock:
5879 PF_HASHROW_LOCK(ih);
5880 LIST_FOREACH(s, &ih->states, entry) {
5881 s->timeout = PFTM_PURGE;
5882 /* Don't send out individual delete messages. */
5883 s->state_flags |= PFSTATE_NOSYNC;
5884 pf_unlink_state(s);
5885 goto relock;
5886 }
5887 PF_HASHROW_UNLOCK(ih);
5888 }
5889 NET_EPOCH_EXIT(et);
5890 }
5891
5892 static int
pf_clear_tables(void)5893 pf_clear_tables(void)
5894 {
5895 struct pfioc_table io;
5896 int error;
5897
5898 bzero(&io, sizeof(io));
5899 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5900
5901 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5902 io.pfrio_flags);
5903
5904 return (error);
5905 }
5906
5907 static void
pf_clear_srcnodes(struct pf_ksrc_node * n)5908 pf_clear_srcnodes(struct pf_ksrc_node *n)
5909 {
5910 struct pf_kstate *s;
5911 int i;
5912
5913 for (i = 0; i <= V_pf_hashmask; i++) {
5914 struct pf_idhash *ih = &V_pf_idhash[i];
5915
5916 PF_HASHROW_LOCK(ih);
5917 LIST_FOREACH(s, &ih->states, entry) {
5918 if (n == NULL || n == s->src_node)
5919 s->src_node = NULL;
5920 if (n == NULL || n == s->nat_src_node)
5921 s->nat_src_node = NULL;
5922 }
5923 PF_HASHROW_UNLOCK(ih);
5924 }
5925
5926 if (n == NULL) {
5927 struct pf_srchash *sh;
5928
5929 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5930 i++, sh++) {
5931 PF_HASHROW_LOCK(sh);
5932 LIST_FOREACH(n, &sh->nodes, entry) {
5933 n->expire = 1;
5934 n->states = 0;
5935 }
5936 PF_HASHROW_UNLOCK(sh);
5937 }
5938 } else {
5939 /* XXX: hash slot should already be locked here. */
5940 n->expire = 1;
5941 n->states = 0;
5942 }
5943 }
5944
5945 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)5946 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5947 {
5948 struct pf_ksrc_node_list kill;
5949
5950 LIST_INIT(&kill);
5951 for (int i = 0; i <= V_pf_srchashmask; i++) {
5952 struct pf_srchash *sh = &V_pf_srchash[i];
5953 struct pf_ksrc_node *sn, *tmp;
5954
5955 PF_HASHROW_LOCK(sh);
5956 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5957 if (PF_MATCHA(psnk->psnk_src.neg,
5958 &psnk->psnk_src.addr.v.a.addr,
5959 &psnk->psnk_src.addr.v.a.mask,
5960 &sn->addr, sn->af) &&
5961 PF_MATCHA(psnk->psnk_dst.neg,
5962 &psnk->psnk_dst.addr.v.a.addr,
5963 &psnk->psnk_dst.addr.v.a.mask,
5964 &sn->raddr, sn->af)) {
5965 pf_unlink_src_node(sn);
5966 LIST_INSERT_HEAD(&kill, sn, entry);
5967 sn->expire = 1;
5968 }
5969 PF_HASHROW_UNLOCK(sh);
5970 }
5971
5972 for (int i = 0; i <= V_pf_hashmask; i++) {
5973 struct pf_idhash *ih = &V_pf_idhash[i];
5974 struct pf_kstate *s;
5975
5976 PF_HASHROW_LOCK(ih);
5977 LIST_FOREACH(s, &ih->states, entry) {
5978 if (s->src_node && s->src_node->expire == 1)
5979 s->src_node = NULL;
5980 if (s->nat_src_node && s->nat_src_node->expire == 1)
5981 s->nat_src_node = NULL;
5982 }
5983 PF_HASHROW_UNLOCK(ih);
5984 }
5985
5986 psnk->psnk_killed = pf_free_src_nodes(&kill);
5987 }
5988
5989 static int
pf_keepcounters(struct pfioc_nv * nv)5990 pf_keepcounters(struct pfioc_nv *nv)
5991 {
5992 nvlist_t *nvl = NULL;
5993 void *nvlpacked = NULL;
5994 int error = 0;
5995
5996 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
5997
5998 if (nv->len > pf_ioctl_maxcount)
5999 ERROUT(ENOMEM);
6000
6001 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6002 if (nvlpacked == NULL)
6003 ERROUT(ENOMEM);
6004
6005 error = copyin(nv->data, nvlpacked, nv->len);
6006 if (error)
6007 ERROUT(error);
6008
6009 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6010 if (nvl == NULL)
6011 ERROUT(EBADMSG);
6012
6013 if (! nvlist_exists_bool(nvl, "keep_counters"))
6014 ERROUT(EBADMSG);
6015
6016 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6017
6018 on_error:
6019 nvlist_destroy(nvl);
6020 free(nvlpacked, M_NVLIST);
6021 return (error);
6022 }
6023
6024 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6025 pf_clear_states(const struct pf_kstate_kill *kill)
6026 {
6027 struct pf_state_key_cmp match_key;
6028 struct pf_kstate *s;
6029 struct pfi_kkif *kif;
6030 int idx;
6031 unsigned int killed = 0, dir;
6032
6033 NET_EPOCH_ASSERT();
6034
6035 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6036 struct pf_idhash *ih = &V_pf_idhash[i];
6037
6038 relock_DIOCCLRSTATES:
6039 PF_HASHROW_LOCK(ih);
6040 LIST_FOREACH(s, &ih->states, entry) {
6041 /* For floating states look at the original kif. */
6042 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6043
6044 if (kill->psk_ifname[0] &&
6045 strcmp(kill->psk_ifname,
6046 kif->pfik_name))
6047 continue;
6048
6049 if (kill->psk_kill_match) {
6050 bzero(&match_key, sizeof(match_key));
6051
6052 if (s->direction == PF_OUT) {
6053 dir = PF_IN;
6054 idx = PF_SK_STACK;
6055 } else {
6056 dir = PF_OUT;
6057 idx = PF_SK_WIRE;
6058 }
6059
6060 match_key.af = s->key[idx]->af;
6061 match_key.proto = s->key[idx]->proto;
6062 PF_ACPY(&match_key.addr[0],
6063 &s->key[idx]->addr[1], match_key.af);
6064 match_key.port[0] = s->key[idx]->port[1];
6065 PF_ACPY(&match_key.addr[1],
6066 &s->key[idx]->addr[0], match_key.af);
6067 match_key.port[1] = s->key[idx]->port[0];
6068 }
6069
6070 /*
6071 * Don't send out individual
6072 * delete messages.
6073 */
6074 s->state_flags |= PFSTATE_NOSYNC;
6075 pf_unlink_state(s);
6076 killed++;
6077
6078 if (kill->psk_kill_match)
6079 killed += pf_kill_matching_state(&match_key,
6080 dir);
6081
6082 goto relock_DIOCCLRSTATES;
6083 }
6084 PF_HASHROW_UNLOCK(ih);
6085 }
6086
6087 if (V_pfsync_clear_states_ptr != NULL)
6088 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6089
6090 return (killed);
6091 }
6092
6093 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6094 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6095 {
6096 struct pf_kstate *s;
6097
6098 NET_EPOCH_ASSERT();
6099 if (kill->psk_pfcmp.id) {
6100 if (kill->psk_pfcmp.creatorid == 0)
6101 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6102 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6103 kill->psk_pfcmp.creatorid))) {
6104 pf_unlink_state(s);
6105 *killed = 1;
6106 }
6107 return;
6108 }
6109
6110 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6111 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6112 }
6113
6114 static int
pf_killstates_nv(struct pfioc_nv * nv)6115 pf_killstates_nv(struct pfioc_nv *nv)
6116 {
6117 struct pf_kstate_kill kill;
6118 struct epoch_tracker et;
6119 nvlist_t *nvl = NULL;
6120 void *nvlpacked = NULL;
6121 int error = 0;
6122 unsigned int killed = 0;
6123
6124 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6125
6126 if (nv->len > pf_ioctl_maxcount)
6127 ERROUT(ENOMEM);
6128
6129 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6130 if (nvlpacked == NULL)
6131 ERROUT(ENOMEM);
6132
6133 error = copyin(nv->data, nvlpacked, nv->len);
6134 if (error)
6135 ERROUT(error);
6136
6137 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6138 if (nvl == NULL)
6139 ERROUT(EBADMSG);
6140
6141 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6142 if (error)
6143 ERROUT(error);
6144
6145 NET_EPOCH_ENTER(et);
6146 pf_killstates(&kill, &killed);
6147 NET_EPOCH_EXIT(et);
6148
6149 free(nvlpacked, M_NVLIST);
6150 nvlpacked = NULL;
6151 nvlist_destroy(nvl);
6152 nvl = nvlist_create(0);
6153 if (nvl == NULL)
6154 ERROUT(ENOMEM);
6155
6156 nvlist_add_number(nvl, "killed", killed);
6157
6158 nvlpacked = nvlist_pack(nvl, &nv->len);
6159 if (nvlpacked == NULL)
6160 ERROUT(ENOMEM);
6161
6162 if (nv->size == 0)
6163 ERROUT(0);
6164 else if (nv->size < nv->len)
6165 ERROUT(ENOSPC);
6166
6167 error = copyout(nvlpacked, nv->data, nv->len);
6168
6169 on_error:
6170 nvlist_destroy(nvl);
6171 free(nvlpacked, M_NVLIST);
6172 return (error);
6173 }
6174
6175 static int
pf_clearstates_nv(struct pfioc_nv * nv)6176 pf_clearstates_nv(struct pfioc_nv *nv)
6177 {
6178 struct pf_kstate_kill kill;
6179 struct epoch_tracker et;
6180 nvlist_t *nvl = NULL;
6181 void *nvlpacked = NULL;
6182 int error = 0;
6183 unsigned int killed;
6184
6185 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6186
6187 if (nv->len > pf_ioctl_maxcount)
6188 ERROUT(ENOMEM);
6189
6190 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6191 if (nvlpacked == NULL)
6192 ERROUT(ENOMEM);
6193
6194 error = copyin(nv->data, nvlpacked, nv->len);
6195 if (error)
6196 ERROUT(error);
6197
6198 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6199 if (nvl == NULL)
6200 ERROUT(EBADMSG);
6201
6202 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6203 if (error)
6204 ERROUT(error);
6205
6206 NET_EPOCH_ENTER(et);
6207 killed = pf_clear_states(&kill);
6208 NET_EPOCH_EXIT(et);
6209
6210 free(nvlpacked, M_NVLIST);
6211 nvlpacked = NULL;
6212 nvlist_destroy(nvl);
6213 nvl = nvlist_create(0);
6214 if (nvl == NULL)
6215 ERROUT(ENOMEM);
6216
6217 nvlist_add_number(nvl, "killed", killed);
6218
6219 nvlpacked = nvlist_pack(nvl, &nv->len);
6220 if (nvlpacked == NULL)
6221 ERROUT(ENOMEM);
6222
6223 if (nv->size == 0)
6224 ERROUT(0);
6225 else if (nv->size < nv->len)
6226 ERROUT(ENOSPC);
6227
6228 error = copyout(nvlpacked, nv->data, nv->len);
6229
6230 #undef ERROUT
6231 on_error:
6232 nvlist_destroy(nvl);
6233 free(nvlpacked, M_NVLIST);
6234 return (error);
6235 }
6236
6237 static int
pf_getstate(struct pfioc_nv * nv)6238 pf_getstate(struct pfioc_nv *nv)
6239 {
6240 nvlist_t *nvl = NULL, *nvls;
6241 void *nvlpacked = NULL;
6242 struct pf_kstate *s = NULL;
6243 int error = 0;
6244 uint64_t id, creatorid;
6245
6246 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6247
6248 if (nv->len > pf_ioctl_maxcount)
6249 ERROUT(ENOMEM);
6250
6251 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6252 if (nvlpacked == NULL)
6253 ERROUT(ENOMEM);
6254
6255 error = copyin(nv->data, nvlpacked, nv->len);
6256 if (error)
6257 ERROUT(error);
6258
6259 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6260 if (nvl == NULL)
6261 ERROUT(EBADMSG);
6262
6263 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6264 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6265
6266 s = pf_find_state_byid(id, creatorid);
6267 if (s == NULL)
6268 ERROUT(ENOENT);
6269
6270 free(nvlpacked, M_NVLIST);
6271 nvlpacked = NULL;
6272 nvlist_destroy(nvl);
6273 nvl = nvlist_create(0);
6274 if (nvl == NULL)
6275 ERROUT(ENOMEM);
6276
6277 nvls = pf_state_to_nvstate(s);
6278 if (nvls == NULL)
6279 ERROUT(ENOMEM);
6280
6281 nvlist_add_nvlist(nvl, "state", nvls);
6282 nvlist_destroy(nvls);
6283
6284 nvlpacked = nvlist_pack(nvl, &nv->len);
6285 if (nvlpacked == NULL)
6286 ERROUT(ENOMEM);
6287
6288 if (nv->size == 0)
6289 ERROUT(0);
6290 else if (nv->size < nv->len)
6291 ERROUT(ENOSPC);
6292
6293 error = copyout(nvlpacked, nv->data, nv->len);
6294
6295 #undef ERROUT
6296 errout:
6297 if (s != NULL)
6298 PF_STATE_UNLOCK(s);
6299 free(nvlpacked, M_NVLIST);
6300 nvlist_destroy(nvl);
6301 return (error);
6302 }
6303
6304 /*
6305 * XXX - Check for version mismatch!!!
6306 */
6307
6308 /*
6309 * Duplicate pfctl -Fa operation to get rid of as much as we can.
6310 */
6311 static int
shutdown_pf(void)6312 shutdown_pf(void)
6313 {
6314 int error = 0;
6315 u_int32_t t[5];
6316 char nn = '\0';
6317 struct pf_kanchor *anchor;
6318 struct pf_keth_anchor *eth_anchor;
6319 int rs_num;
6320
6321 do {
6322 /* Unlink rules of all user defined anchors */
6323 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6324 /* Wildcard based anchors may not have a respective
6325 * explicit anchor rule or they may be left empty
6326 * without rules. It leads to anchor.refcnt=0, and the
6327 * rest of the logic does not expect it. */
6328 if (anchor->refcnt == 0)
6329 anchor->refcnt = 1;
6330 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6331 if ((error = pf_begin_rules(&t[rs_num], rs_num,
6332 anchor->path)) != 0) {
6333 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6334 "anchor.path=%s rs_num=%d\n",
6335 anchor->path, rs_num));
6336 goto error; /* XXX: rollback? */
6337 }
6338 }
6339 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6340 error = pf_commit_rules(t[rs_num], rs_num,
6341 anchor->path);
6342 MPASS(error == 0);
6343 }
6344 }
6345
6346 /* Unlink rules of all user defined ether anchors */
6347 RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6348 &V_pf_keth_anchors) {
6349 /* Wildcard based anchors may not have a respective
6350 * explicit anchor rule or they may be left empty
6351 * without rules. It leads to anchor.refcnt=0, and the
6352 * rest of the logic does not expect it. */
6353 if (eth_anchor->refcnt == 0)
6354 eth_anchor->refcnt = 1;
6355 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6356 != 0) {
6357 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6358 "anchor.path=%s\n", eth_anchor->path));
6359 goto error;
6360 }
6361 error = pf_commit_eth(t[0], eth_anchor->path);
6362 MPASS(error == 0);
6363 }
6364
6365 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6366 != 0) {
6367 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6368 break;
6369 }
6370 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6371 != 0) {
6372 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6373 break; /* XXX: rollback? */
6374 }
6375 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6376 != 0) {
6377 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6378 break; /* XXX: rollback? */
6379 }
6380 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6381 != 0) {
6382 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6383 break; /* XXX: rollback? */
6384 }
6385 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6386 != 0) {
6387 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6388 break; /* XXX: rollback? */
6389 }
6390
6391 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6392 MPASS(error == 0);
6393 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6394 MPASS(error == 0);
6395 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6396 MPASS(error == 0);
6397 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6398 MPASS(error == 0);
6399 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6400 MPASS(error == 0);
6401
6402 if ((error = pf_clear_tables()) != 0)
6403 break;
6404
6405 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6406 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6407 break;
6408 }
6409 error = pf_commit_eth(t[0], &nn);
6410 MPASS(error == 0);
6411
6412 #ifdef ALTQ
6413 if ((error = pf_begin_altq(&t[0])) != 0) {
6414 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6415 break;
6416 }
6417 pf_commit_altq(t[0]);
6418 #endif
6419
6420 pf_clear_all_states();
6421
6422 pf_clear_srcnodes(NULL);
6423
6424 /* status does not use malloced mem so no need to cleanup */
6425 /* fingerprints and interfaces have their own cleanup code */
6426 } while(0);
6427
6428 error:
6429 return (error);
6430 }
6431
6432 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6433 pf_check_return(int chk, struct mbuf **m)
6434 {
6435
6436 switch (chk) {
6437 case PF_PASS:
6438 if (*m == NULL)
6439 return (PFIL_CONSUMED);
6440 else
6441 return (PFIL_PASS);
6442 break;
6443 default:
6444 if (*m != NULL) {
6445 m_freem(*m);
6446 *m = NULL;
6447 }
6448 return (PFIL_DROPPED);
6449 }
6450 }
6451
6452 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6453 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6454 void *ruleset __unused, struct inpcb *inp)
6455 {
6456 int chk;
6457
6458 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6459
6460 return (pf_check_return(chk, m));
6461 }
6462
6463 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6464 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6465 void *ruleset __unused, struct inpcb *inp)
6466 {
6467 int chk;
6468
6469 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6470
6471 return (pf_check_return(chk, m));
6472 }
6473
6474 #ifdef INET
6475 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6476 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6477 void *ruleset __unused, struct inpcb *inp)
6478 {
6479 int chk;
6480
6481 chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6482
6483 return (pf_check_return(chk, m));
6484 }
6485
6486 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6487 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6488 void *ruleset __unused, struct inpcb *inp)
6489 {
6490 int chk;
6491
6492 chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6493
6494 return (pf_check_return(chk, m));
6495 }
6496 #endif
6497
6498 #ifdef INET6
6499 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6500 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6501 void *ruleset __unused, struct inpcb *inp)
6502 {
6503 int chk;
6504
6505 /*
6506 * In case of loopback traffic IPv6 uses the real interface in
6507 * order to support scoped addresses. In order to support stateful
6508 * filtering we have change this to lo0 as it is the case in IPv4.
6509 */
6510 CURVNET_SET(ifp->if_vnet);
6511 chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6512 m, inp, NULL);
6513 CURVNET_RESTORE();
6514
6515 return (pf_check_return(chk, m));
6516 }
6517
6518 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6519 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6520 void *ruleset __unused, struct inpcb *inp)
6521 {
6522 int chk;
6523
6524 CURVNET_SET(ifp->if_vnet);
6525 chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6526 CURVNET_RESTORE();
6527
6528 return (pf_check_return(chk, m));
6529 }
6530 #endif /* INET6 */
6531
6532 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6533 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6534 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
6535 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
6536
6537 #ifdef INET
6538 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6539 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6540 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
6541 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
6542 #endif
6543 #ifdef INET6
6544 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6545 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6546 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
6547 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
6548 #endif
6549
6550 static void
hook_pf_eth(void)6551 hook_pf_eth(void)
6552 {
6553 struct pfil_hook_args pha = {
6554 .pa_version = PFIL_VERSION,
6555 .pa_modname = "pf",
6556 .pa_type = PFIL_TYPE_ETHERNET,
6557 };
6558 struct pfil_link_args pla = {
6559 .pa_version = PFIL_VERSION,
6560 };
6561 int ret __diagused;
6562
6563 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6564 return;
6565
6566 pha.pa_mbuf_chk = pf_eth_check_in;
6567 pha.pa_flags = PFIL_IN;
6568 pha.pa_rulname = "eth-in";
6569 V_pf_eth_in_hook = pfil_add_hook(&pha);
6570 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6571 pla.pa_head = V_link_pfil_head;
6572 pla.pa_hook = V_pf_eth_in_hook;
6573 ret = pfil_link(&pla);
6574 MPASS(ret == 0);
6575 pha.pa_mbuf_chk = pf_eth_check_out;
6576 pha.pa_flags = PFIL_OUT;
6577 pha.pa_rulname = "eth-out";
6578 V_pf_eth_out_hook = pfil_add_hook(&pha);
6579 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6580 pla.pa_head = V_link_pfil_head;
6581 pla.pa_hook = V_pf_eth_out_hook;
6582 ret = pfil_link(&pla);
6583 MPASS(ret == 0);
6584
6585 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6586 }
6587
6588 static void
hook_pf(void)6589 hook_pf(void)
6590 {
6591 struct pfil_hook_args pha = {
6592 .pa_version = PFIL_VERSION,
6593 .pa_modname = "pf",
6594 };
6595 struct pfil_link_args pla = {
6596 .pa_version = PFIL_VERSION,
6597 };
6598 int ret __diagused;
6599
6600 if (atomic_load_bool(&V_pf_pfil_hooked))
6601 return;
6602
6603 #ifdef INET
6604 pha.pa_type = PFIL_TYPE_IP4;
6605 pha.pa_mbuf_chk = pf_check_in;
6606 pha.pa_flags = PFIL_IN;
6607 pha.pa_rulname = "default-in";
6608 V_pf_ip4_in_hook = pfil_add_hook(&pha);
6609 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6610 pla.pa_head = V_inet_pfil_head;
6611 pla.pa_hook = V_pf_ip4_in_hook;
6612 ret = pfil_link(&pla);
6613 MPASS(ret == 0);
6614 pha.pa_mbuf_chk = pf_check_out;
6615 pha.pa_flags = PFIL_OUT;
6616 pha.pa_rulname = "default-out";
6617 V_pf_ip4_out_hook = pfil_add_hook(&pha);
6618 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6619 pla.pa_head = V_inet_pfil_head;
6620 pla.pa_hook = V_pf_ip4_out_hook;
6621 ret = pfil_link(&pla);
6622 MPASS(ret == 0);
6623 if (V_pf_filter_local) {
6624 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6625 pla.pa_head = V_inet_local_pfil_head;
6626 pla.pa_hook = V_pf_ip4_out_hook;
6627 ret = pfil_link(&pla);
6628 MPASS(ret == 0);
6629 }
6630 #endif
6631 #ifdef INET6
6632 pha.pa_type = PFIL_TYPE_IP6;
6633 pha.pa_mbuf_chk = pf_check6_in;
6634 pha.pa_flags = PFIL_IN;
6635 pha.pa_rulname = "default-in6";
6636 V_pf_ip6_in_hook = pfil_add_hook(&pha);
6637 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6638 pla.pa_head = V_inet6_pfil_head;
6639 pla.pa_hook = V_pf_ip6_in_hook;
6640 ret = pfil_link(&pla);
6641 MPASS(ret == 0);
6642 pha.pa_mbuf_chk = pf_check6_out;
6643 pha.pa_rulname = "default-out6";
6644 pha.pa_flags = PFIL_OUT;
6645 V_pf_ip6_out_hook = pfil_add_hook(&pha);
6646 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6647 pla.pa_head = V_inet6_pfil_head;
6648 pla.pa_hook = V_pf_ip6_out_hook;
6649 ret = pfil_link(&pla);
6650 MPASS(ret == 0);
6651 if (V_pf_filter_local) {
6652 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6653 pla.pa_head = V_inet6_local_pfil_head;
6654 pla.pa_hook = V_pf_ip6_out_hook;
6655 ret = pfil_link(&pla);
6656 MPASS(ret == 0);
6657 }
6658 #endif
6659
6660 atomic_store_bool(&V_pf_pfil_hooked, true);
6661 }
6662
6663 static void
dehook_pf_eth(void)6664 dehook_pf_eth(void)
6665 {
6666
6667 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6668 return;
6669
6670 pfil_remove_hook(V_pf_eth_in_hook);
6671 pfil_remove_hook(V_pf_eth_out_hook);
6672
6673 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6674 }
6675
6676 static void
dehook_pf(void)6677 dehook_pf(void)
6678 {
6679
6680 if (!atomic_load_bool(&V_pf_pfil_hooked))
6681 return;
6682
6683 #ifdef INET
6684 pfil_remove_hook(V_pf_ip4_in_hook);
6685 pfil_remove_hook(V_pf_ip4_out_hook);
6686 #endif
6687 #ifdef INET6
6688 pfil_remove_hook(V_pf_ip6_in_hook);
6689 pfil_remove_hook(V_pf_ip6_out_hook);
6690 #endif
6691
6692 atomic_store_bool(&V_pf_pfil_hooked, false);
6693 }
6694
6695 static void
pf_load_vnet(void)6696 pf_load_vnet(void)
6697 {
6698 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6699 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6700
6701 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6702 sx_init(&V_pf_ioctl_lock, "pf ioctl");
6703
6704 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6705 PF_RULE_TAG_HASH_SIZE_DEFAULT);
6706 #ifdef ALTQ
6707 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6708 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6709 #endif
6710
6711 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6712
6713 pfattach_vnet();
6714 V_pf_vnet_active = 1;
6715 }
6716
6717 static int
pf_load(void)6718 pf_load(void)
6719 {
6720 int error;
6721
6722 sx_init(&pf_end_lock, "pf end thread");
6723
6724 pf_mtag_initialize();
6725
6726 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6727 if (pf_dev == NULL)
6728 return (ENOMEM);
6729
6730 pf_end_threads = 0;
6731 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6732 if (error != 0)
6733 return (error);
6734
6735 pfi_initialize();
6736
6737 return (0);
6738 }
6739
6740 static void
pf_unload_vnet(void)6741 pf_unload_vnet(void)
6742 {
6743 int ret __diagused;
6744
6745 V_pf_vnet_active = 0;
6746 V_pf_status.running = 0;
6747 dehook_pf();
6748 dehook_pf_eth();
6749
6750 PF_RULES_WLOCK();
6751 pf_syncookies_cleanup();
6752 shutdown_pf();
6753 PF_RULES_WUNLOCK();
6754
6755 /* Make sure we've cleaned up ethernet rules before we continue. */
6756 NET_EPOCH_DRAIN_CALLBACKS();
6757
6758 ret = swi_remove(V_pf_swi_cookie);
6759 MPASS(ret == 0);
6760 ret = intr_event_destroy(V_pf_swi_ie);
6761 MPASS(ret == 0);
6762
6763 pf_unload_vnet_purge();
6764
6765 pf_normalize_cleanup();
6766 PF_RULES_WLOCK();
6767 pfi_cleanup_vnet();
6768 PF_RULES_WUNLOCK();
6769 pfr_cleanup();
6770 pf_osfp_flush();
6771 pf_cleanup();
6772 if (IS_DEFAULT_VNET(curvnet))
6773 pf_mtag_cleanup();
6774
6775 pf_cleanup_tagset(&V_pf_tags);
6776 #ifdef ALTQ
6777 pf_cleanup_tagset(&V_pf_qids);
6778 #endif
6779 uma_zdestroy(V_pf_tag_z);
6780
6781 #ifdef PF_WANT_32_TO_64_COUNTER
6782 PF_RULES_WLOCK();
6783 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6784
6785 MPASS(LIST_EMPTY(&V_pf_allkiflist));
6786 MPASS(V_pf_allkifcount == 0);
6787
6788 LIST_REMOVE(&V_pf_default_rule, allrulelist);
6789 V_pf_allrulecount--;
6790 LIST_REMOVE(V_pf_rulemarker, allrulelist);
6791
6792 MPASS(LIST_EMPTY(&V_pf_allrulelist));
6793 MPASS(V_pf_allrulecount == 0);
6794
6795 PF_RULES_WUNLOCK();
6796
6797 free(V_pf_kifmarker, PFI_MTYPE);
6798 free(V_pf_rulemarker, M_PFRULE);
6799 #endif
6800
6801 /* Free counters last as we updated them during shutdown. */
6802 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6803 for (int i = 0; i < 2; i++) {
6804 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6805 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6806 }
6807 counter_u64_free(V_pf_default_rule.states_cur);
6808 counter_u64_free(V_pf_default_rule.states_tot);
6809 counter_u64_free(V_pf_default_rule.src_nodes);
6810 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6811
6812 for (int i = 0; i < PFRES_MAX; i++)
6813 counter_u64_free(V_pf_status.counters[i]);
6814 for (int i = 0; i < KLCNT_MAX; i++)
6815 counter_u64_free(V_pf_status.lcounters[i]);
6816 for (int i = 0; i < FCNT_MAX; i++)
6817 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6818 for (int i = 0; i < SCNT_MAX; i++)
6819 counter_u64_free(V_pf_status.scounters[i]);
6820
6821 rm_destroy(&V_pf_rules_lock);
6822 sx_destroy(&V_pf_ioctl_lock);
6823 }
6824
6825 static void
pf_unload(void)6826 pf_unload(void)
6827 {
6828
6829 sx_xlock(&pf_end_lock);
6830 pf_end_threads = 1;
6831 while (pf_end_threads < 2) {
6832 wakeup_one(pf_purge_thread);
6833 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6834 }
6835 sx_xunlock(&pf_end_lock);
6836
6837 pf_nl_unregister();
6838
6839 if (pf_dev != NULL)
6840 destroy_dev(pf_dev);
6841
6842 pfi_cleanup();
6843
6844 sx_destroy(&pf_end_lock);
6845 }
6846
6847 static void
vnet_pf_init(void * unused __unused)6848 vnet_pf_init(void *unused __unused)
6849 {
6850
6851 pf_load_vnet();
6852 }
6853 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6854 vnet_pf_init, NULL);
6855
6856 static void
vnet_pf_uninit(const void * unused __unused)6857 vnet_pf_uninit(const void *unused __unused)
6858 {
6859
6860 pf_unload_vnet();
6861 }
6862 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6863 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6864 vnet_pf_uninit, NULL);
6865
6866 static int
pf_modevent(module_t mod,int type,void * data)6867 pf_modevent(module_t mod, int type, void *data)
6868 {
6869 int error = 0;
6870
6871 switch(type) {
6872 case MOD_LOAD:
6873 error = pf_load();
6874 pf_nl_register();
6875 break;
6876 case MOD_UNLOAD:
6877 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6878 * the vnet_pf_uninit()s */
6879 break;
6880 default:
6881 error = EINVAL;
6882 break;
6883 }
6884
6885 return (error);
6886 }
6887
6888 static moduledata_t pf_mod = {
6889 "pf",
6890 pf_modevent,
6891 0
6892 };
6893
6894 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6895 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6896 MODULE_VERSION(pf, PF_MODVER);
6897