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