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 int
pf_start(void)2410 pf_start(void)
2411 {
2412 int error = 0;
2413
2414 sx_xlock(&V_pf_ioctl_lock);
2415 if (V_pf_status.running)
2416 error = EEXIST;
2417 else {
2418 hook_pf();
2419 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
2420 hook_pf_eth();
2421 V_pf_status.running = 1;
2422 V_pf_status.since = time_second;
2423 new_unrhdr64(&V_pf_stateid, time_second);
2424
2425 DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2426 }
2427 sx_xunlock(&V_pf_ioctl_lock);
2428
2429 return (error);
2430 }
2431
2432 int
pf_stop(void)2433 pf_stop(void)
2434 {
2435 int error = 0;
2436
2437 sx_xlock(&V_pf_ioctl_lock);
2438 if (!V_pf_status.running)
2439 error = ENOENT;
2440 else {
2441 V_pf_status.running = 0;
2442 dehook_pf();
2443 dehook_pf_eth();
2444 V_pf_status.since = time_second;
2445 DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2446 }
2447 sx_xunlock(&V_pf_ioctl_lock);
2448
2449 return (error);
2450 }
2451
2452 void
pf_ioctl_clear_status(void)2453 pf_ioctl_clear_status(void)
2454 {
2455 PF_RULES_WLOCK();
2456 for (int i = 0; i < PFRES_MAX; i++)
2457 counter_u64_zero(V_pf_status.counters[i]);
2458 for (int i = 0; i < FCNT_MAX; i++)
2459 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
2460 for (int i = 0; i < SCNT_MAX; i++)
2461 counter_u64_zero(V_pf_status.scounters[i]);
2462 for (int i = 0; i < KLCNT_MAX; i++)
2463 counter_u64_zero(V_pf_status.lcounters[i]);
2464 V_pf_status.since = time_second;
2465 if (*V_pf_status.ifname)
2466 pfi_update_status(V_pf_status.ifname, NULL);
2467 PF_RULES_WUNLOCK();
2468 }
2469
2470 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)2471 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2472 {
2473 uint32_t old;
2474
2475 if (timeout < 0 || timeout >= PFTM_MAX ||
2476 seconds < 0)
2477 return (EINVAL);
2478
2479 PF_RULES_WLOCK();
2480 old = V_pf_default_rule.timeout[timeout];
2481 if (timeout == PFTM_INTERVAL && seconds == 0)
2482 seconds = 1;
2483 V_pf_default_rule.timeout[timeout] = seconds;
2484 if (timeout == PFTM_INTERVAL && seconds < old)
2485 wakeup(pf_purge_thread);
2486
2487 if (prev_seconds != NULL)
2488 *prev_seconds = old;
2489
2490 PF_RULES_WUNLOCK();
2491
2492 return (0);
2493 }
2494
2495 int
pf_ioctl_get_timeout(int timeout,int * seconds)2496 pf_ioctl_get_timeout(int timeout, int *seconds)
2497 {
2498 PF_RULES_RLOCK_TRACKER;
2499
2500 if (timeout < 0 || timeout >= PFTM_MAX)
2501 return (EINVAL);
2502
2503 PF_RULES_RLOCK();
2504 *seconds = V_pf_default_rule.timeout[timeout];
2505 PF_RULES_RUNLOCK();
2506
2507 return (0);
2508 }
2509
2510 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)2511 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2512 {
2513
2514 PF_RULES_WLOCK();
2515 if (index < 0 || index >= PF_LIMIT_MAX ||
2516 V_pf_limits[index].zone == NULL) {
2517 PF_RULES_WUNLOCK();
2518 return (EINVAL);
2519 }
2520 uma_zone_set_max(V_pf_limits[index].zone, limit);
2521 if (old_limit != NULL)
2522 *old_limit = V_pf_limits[index].limit;
2523 V_pf_limits[index].limit = limit;
2524 PF_RULES_WUNLOCK();
2525
2526 return (0);
2527 }
2528
2529 int
pf_ioctl_get_limit(int index,unsigned int * limit)2530 pf_ioctl_get_limit(int index, unsigned int *limit)
2531 {
2532 PF_RULES_RLOCK_TRACKER;
2533
2534 if (index < 0 || index >= PF_LIMIT_MAX)
2535 return (EINVAL);
2536
2537 PF_RULES_RLOCK();
2538 *limit = V_pf_limits[index].limit;
2539 PF_RULES_RUNLOCK();
2540
2541 return (0);
2542 }
2543
2544 int
pf_ioctl_begin_addrs(uint32_t * ticket)2545 pf_ioctl_begin_addrs(uint32_t *ticket)
2546 {
2547 PF_RULES_WLOCK();
2548 pf_empty_kpool(&V_pf_pabuf[0]);
2549 pf_empty_kpool(&V_pf_pabuf[1]);
2550 pf_empty_kpool(&V_pf_pabuf[2]);
2551 *ticket = ++V_ticket_pabuf;
2552 PF_RULES_WUNLOCK();
2553
2554 return (0);
2555 }
2556
2557 int
pf_ioctl_add_addr(struct pf_nl_pooladdr * pp)2558 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
2559 {
2560 struct pf_kpooladdr *pa = NULL;
2561 struct pfi_kkif *kif = NULL;
2562 int error;
2563
2564 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2565 pp->which != PF_RT)
2566 return (EINVAL);
2567
2568 #ifndef INET
2569 if (pp->af == AF_INET)
2570 return (EAFNOSUPPORT);
2571 #endif /* INET */
2572 #ifndef INET6
2573 if (pp->af == AF_INET6)
2574 return (EAFNOSUPPORT);
2575 #endif /* INET6 */
2576
2577 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2578 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2579 pp->addr.addr.type != PF_ADDR_TABLE)
2580 return (EINVAL);
2581
2582 if (pp->addr.addr.p.dyn != NULL)
2583 return (EINVAL);
2584
2585 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2586 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2587 if (error != 0)
2588 goto out;
2589 if (pa->ifname[0])
2590 kif = pf_kkif_create(M_WAITOK);
2591 PF_RULES_WLOCK();
2592 if (pp->ticket != V_ticket_pabuf) {
2593 PF_RULES_WUNLOCK();
2594 if (pa->ifname[0])
2595 pf_kkif_free(kif);
2596 error = EBUSY;
2597 goto out;
2598 }
2599 if (pa->ifname[0]) {
2600 pa->kif = pfi_kkif_attach(kif, pa->ifname);
2601 kif = NULL;
2602 pfi_kkif_ref(pa->kif);
2603 } else
2604 pa->kif = NULL;
2605 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2606 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2607 if (pa->ifname[0])
2608 pfi_kkif_unref(pa->kif);
2609 PF_RULES_WUNLOCK();
2610 goto out;
2611 }
2612 switch (pp->which) {
2613 case PF_NAT:
2614 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries);
2615 break;
2616 case PF_RDR:
2617 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries);
2618 break;
2619 case PF_RT:
2620 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries);
2621 break;
2622 }
2623 PF_RULES_WUNLOCK();
2624
2625 return (0);
2626
2627 out:
2628 free(pa, M_PFRULE);
2629 return (error);
2630 }
2631
2632 int
pf_ioctl_get_addrs(struct pf_nl_pooladdr * pp)2633 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
2634 {
2635 struct pf_kpool *pool;
2636 struct pf_kpooladdr *pa;
2637
2638 PF_RULES_RLOCK_TRACKER;
2639
2640 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2641 pp->which != PF_RT)
2642 return (EINVAL);
2643
2644 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2645 pp->nr = 0;
2646
2647 PF_RULES_RLOCK();
2648 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2649 pp->r_num, 0, 1, 0, pp->which);
2650 if (pool == NULL) {
2651 PF_RULES_RUNLOCK();
2652 return (EBUSY);
2653 }
2654 TAILQ_FOREACH(pa, &pool->list, entries)
2655 pp->nr++;
2656 PF_RULES_RUNLOCK();
2657
2658 return (0);
2659 }
2660
2661 int
pf_ioctl_get_addr(struct pf_nl_pooladdr * pp)2662 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
2663 {
2664 struct pf_kpool *pool;
2665 struct pf_kpooladdr *pa;
2666 u_int32_t nr = 0;
2667
2668 if (pp->which != PF_RDR && pp->which != PF_NAT &&
2669 pp->which != PF_RT)
2670 return (EINVAL);
2671
2672 PF_RULES_RLOCK_TRACKER;
2673
2674 pp->anchor[sizeof(pp->anchor) - 1] = 0;
2675
2676 PF_RULES_RLOCK();
2677 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2678 pp->r_num, 0, 1, 1, pp->which);
2679 if (pool == NULL) {
2680 PF_RULES_RUNLOCK();
2681 return (EBUSY);
2682 }
2683 pa = TAILQ_FIRST(&pool->list);
2684 while ((pa != NULL) && (nr < pp->nr)) {
2685 pa = TAILQ_NEXT(pa, entries);
2686 nr++;
2687 }
2688 if (pa == NULL) {
2689 PF_RULES_RUNLOCK();
2690 return (EBUSY);
2691 }
2692 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
2693 pf_addr_copyout(&pp->addr.addr);
2694 PF_RULES_RUNLOCK();
2695
2696 return (0);
2697 }
2698
2699 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)2700 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
2701 {
2702 struct pf_kruleset *ruleset;
2703 struct pf_kanchor *anchor;
2704
2705 PF_RULES_RLOCK_TRACKER;
2706
2707 pr->path[sizeof(pr->path) - 1] = 0;
2708
2709 PF_RULES_RLOCK();
2710 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2711 PF_RULES_RUNLOCK();
2712 return (ENOENT);
2713 }
2714 pr->nr = 0;
2715 if (ruleset->anchor == NULL) {
2716 /* XXX kludge for pf_main_ruleset */
2717 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2718 if (anchor->parent == NULL)
2719 pr->nr++;
2720 } else {
2721 RB_FOREACH(anchor, pf_kanchor_node,
2722 &ruleset->anchor->children)
2723 pr->nr++;
2724 }
2725 PF_RULES_RUNLOCK();
2726
2727 return (0);
2728 }
2729
2730 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)2731 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
2732 {
2733 struct pf_kruleset *ruleset;
2734 struct pf_kanchor *anchor;
2735 u_int32_t nr = 0;
2736 int error = 0;
2737
2738 PF_RULES_RLOCK_TRACKER;
2739
2740 PF_RULES_RLOCK();
2741 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
2742 PF_RULES_RUNLOCK();
2743 return (ENOENT);
2744 }
2745
2746 pr->name[0] = 0;
2747 if (ruleset->anchor == NULL) {
2748 /* XXX kludge for pf_main_ruleset */
2749 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
2750 if (anchor->parent == NULL && nr++ == pr->nr) {
2751 strlcpy(pr->name, anchor->name,
2752 sizeof(pr->name));
2753 break;
2754 }
2755 } else {
2756 RB_FOREACH(anchor, pf_kanchor_node,
2757 &ruleset->anchor->children)
2758 if (nr++ == pr->nr) {
2759 strlcpy(pr->name, anchor->name,
2760 sizeof(pr->name));
2761 break;
2762 }
2763 }
2764 if (!pr->name[0])
2765 error = EBUSY;
2766 PF_RULES_RUNLOCK();
2767
2768 return (error);
2769 }
2770
2771 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)2772 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2773 {
2774 int error = 0;
2775 PF_RULES_RLOCK_TRACKER;
2776
2777 #define ERROUT_IOCTL(target, x) \
2778 do { \
2779 error = (x); \
2780 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
2781 goto target; \
2782 } while (0)
2783
2784
2785 /* XXX keep in sync with switch() below */
2786 if (securelevel_gt(td->td_ucred, 2))
2787 switch (cmd) {
2788 case DIOCGETRULES:
2789 case DIOCGETRULENV:
2790 case DIOCGETADDRS:
2791 case DIOCGETADDR:
2792 case DIOCGETSTATE:
2793 case DIOCGETSTATENV:
2794 case DIOCSETSTATUSIF:
2795 case DIOCGETSTATUSNV:
2796 case DIOCCLRSTATUS:
2797 case DIOCNATLOOK:
2798 case DIOCSETDEBUG:
2799 #ifdef COMPAT_FREEBSD14
2800 case DIOCGETSTATES:
2801 case DIOCGETSTATESV2:
2802 #endif
2803 case DIOCGETTIMEOUT:
2804 case DIOCCLRRULECTRS:
2805 case DIOCGETLIMIT:
2806 case DIOCGETALTQSV0:
2807 case DIOCGETALTQSV1:
2808 case DIOCGETALTQV0:
2809 case DIOCGETALTQV1:
2810 case DIOCGETQSTATSV0:
2811 case DIOCGETQSTATSV1:
2812 case DIOCGETRULESETS:
2813 case DIOCGETRULESET:
2814 case DIOCRGETTABLES:
2815 case DIOCRGETTSTATS:
2816 case DIOCRCLRTSTATS:
2817 case DIOCRCLRADDRS:
2818 case DIOCRADDADDRS:
2819 case DIOCRDELADDRS:
2820 case DIOCRSETADDRS:
2821 case DIOCRGETADDRS:
2822 case DIOCRGETASTATS:
2823 case DIOCRCLRASTATS:
2824 case DIOCRTSTADDRS:
2825 case DIOCOSFPGET:
2826 case DIOCGETSRCNODES:
2827 case DIOCCLRSRCNODES:
2828 case DIOCGETSYNCOOKIES:
2829 case DIOCIGETIFACES:
2830 case DIOCGIFSPEEDV0:
2831 case DIOCGIFSPEEDV1:
2832 case DIOCSETIFFLAG:
2833 case DIOCCLRIFFLAG:
2834 case DIOCGETETHRULES:
2835 case DIOCGETETHRULE:
2836 case DIOCGETETHRULESETS:
2837 case DIOCGETETHRULESET:
2838 break;
2839 case DIOCRCLRTABLES:
2840 case DIOCRADDTABLES:
2841 case DIOCRDELTABLES:
2842 case DIOCRSETTFLAGS:
2843 if (((struct pfioc_table *)addr)->pfrio_flags &
2844 PFR_FLAG_DUMMY)
2845 break; /* dummy operation ok */
2846 return (EPERM);
2847 default:
2848 return (EPERM);
2849 }
2850
2851 if (!(flags & FWRITE))
2852 switch (cmd) {
2853 case DIOCGETRULES:
2854 case DIOCGETADDRS:
2855 case DIOCGETADDR:
2856 case DIOCGETSTATE:
2857 case DIOCGETSTATENV:
2858 case DIOCGETSTATUSNV:
2859 #ifdef COMPAT_FREEBSD14
2860 case DIOCGETSTATES:
2861 case DIOCGETSTATESV2:
2862 #endif
2863 case DIOCGETTIMEOUT:
2864 case DIOCGETLIMIT:
2865 case DIOCGETALTQSV0:
2866 case DIOCGETALTQSV1:
2867 case DIOCGETALTQV0:
2868 case DIOCGETALTQV1:
2869 case DIOCGETQSTATSV0:
2870 case DIOCGETQSTATSV1:
2871 case DIOCGETRULESETS:
2872 case DIOCGETRULESET:
2873 case DIOCNATLOOK:
2874 case DIOCRGETTABLES:
2875 case DIOCRGETTSTATS:
2876 case DIOCRGETADDRS:
2877 case DIOCRGETASTATS:
2878 case DIOCRTSTADDRS:
2879 case DIOCOSFPGET:
2880 case DIOCGETSRCNODES:
2881 case DIOCGETSYNCOOKIES:
2882 case DIOCIGETIFACES:
2883 case DIOCGIFSPEEDV1:
2884 case DIOCGIFSPEEDV0:
2885 case DIOCGETRULENV:
2886 case DIOCGETETHRULES:
2887 case DIOCGETETHRULE:
2888 case DIOCGETETHRULESETS:
2889 case DIOCGETETHRULESET:
2890 break;
2891 case DIOCRCLRTABLES:
2892 case DIOCRADDTABLES:
2893 case DIOCRDELTABLES:
2894 case DIOCRCLRTSTATS:
2895 case DIOCRCLRADDRS:
2896 case DIOCRADDADDRS:
2897 case DIOCRDELADDRS:
2898 case DIOCRSETADDRS:
2899 case DIOCRSETTFLAGS:
2900 if (((struct pfioc_table *)addr)->pfrio_flags &
2901 PFR_FLAG_DUMMY) {
2902 flags |= FWRITE; /* need write lock for dummy */
2903 break; /* dummy operation ok */
2904 }
2905 return (EACCES);
2906 default:
2907 return (EACCES);
2908 }
2909
2910 CURVNET_SET(TD_TO_VNET(td));
2911
2912 switch (cmd) {
2913 #ifdef COMPAT_FREEBSD14
2914 case DIOCSTART:
2915 error = pf_start();
2916 break;
2917
2918 case DIOCSTOP:
2919 error = pf_stop();
2920 break;
2921 #endif
2922
2923 case DIOCGETETHRULES: {
2924 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
2925 nvlist_t *nvl;
2926 void *packed;
2927 struct pf_keth_rule *tail;
2928 struct pf_keth_ruleset *rs;
2929 u_int32_t ticket, nr;
2930 const char *anchor = "";
2931
2932 nvl = NULL;
2933 packed = NULL;
2934
2935 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2936
2937 if (nv->len > pf_ioctl_maxcount)
2938 ERROUT(ENOMEM);
2939
2940 /* Copy the request in */
2941 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2942 error = copyin(nv->data, packed, nv->len);
2943 if (error)
2944 ERROUT(error);
2945
2946 nvl = nvlist_unpack(packed, nv->len, 0);
2947 if (nvl == NULL)
2948 ERROUT(EBADMSG);
2949
2950 if (! nvlist_exists_string(nvl, "anchor"))
2951 ERROUT(EBADMSG);
2952
2953 anchor = nvlist_get_string(nvl, "anchor");
2954
2955 rs = pf_find_keth_ruleset(anchor);
2956
2957 nvlist_destroy(nvl);
2958 nvl = NULL;
2959 free(packed, M_NVLIST);
2960 packed = NULL;
2961
2962 if (rs == NULL)
2963 ERROUT(ENOENT);
2964
2965 /* Reply */
2966 nvl = nvlist_create(0);
2967 if (nvl == NULL)
2968 ERROUT(ENOMEM);
2969
2970 PF_RULES_RLOCK();
2971
2972 ticket = rs->active.ticket;
2973 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2974 if (tail)
2975 nr = tail->nr + 1;
2976 else
2977 nr = 0;
2978
2979 PF_RULES_RUNLOCK();
2980
2981 nvlist_add_number(nvl, "ticket", ticket);
2982 nvlist_add_number(nvl, "nr", nr);
2983
2984 packed = nvlist_pack(nvl, &nv->len);
2985 if (packed == NULL)
2986 ERROUT(ENOMEM);
2987
2988 if (nv->size == 0)
2989 ERROUT(0);
2990 else if (nv->size < nv->len)
2991 ERROUT(ENOSPC);
2992
2993 error = copyout(packed, nv->data, nv->len);
2994
2995 #undef ERROUT
2996 DIOCGETETHRULES_error:
2997 free(packed, M_NVLIST);
2998 nvlist_destroy(nvl);
2999 break;
3000 }
3001
3002 case DIOCGETETHRULE: {
3003 struct epoch_tracker et;
3004 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3005 nvlist_t *nvl = NULL;
3006 void *nvlpacked = NULL;
3007 struct pf_keth_rule *rule = NULL;
3008 struct pf_keth_ruleset *rs;
3009 u_int32_t ticket, nr;
3010 bool clear = false;
3011 const char *anchor;
3012
3013 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
3014
3015 if (nv->len > pf_ioctl_maxcount)
3016 ERROUT(ENOMEM);
3017
3018 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3019 error = copyin(nv->data, nvlpacked, nv->len);
3020 if (error)
3021 ERROUT(error);
3022
3023 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3024 if (nvl == NULL)
3025 ERROUT(EBADMSG);
3026 if (! nvlist_exists_number(nvl, "ticket"))
3027 ERROUT(EBADMSG);
3028 ticket = nvlist_get_number(nvl, "ticket");
3029 if (! nvlist_exists_string(nvl, "anchor"))
3030 ERROUT(EBADMSG);
3031 anchor = nvlist_get_string(nvl, "anchor");
3032
3033 if (nvlist_exists_bool(nvl, "clear"))
3034 clear = nvlist_get_bool(nvl, "clear");
3035
3036 if (clear && !(flags & FWRITE))
3037 ERROUT(EACCES);
3038
3039 if (! nvlist_exists_number(nvl, "nr"))
3040 ERROUT(EBADMSG);
3041 nr = nvlist_get_number(nvl, "nr");
3042
3043 PF_RULES_RLOCK();
3044 rs = pf_find_keth_ruleset(anchor);
3045 if (rs == NULL) {
3046 PF_RULES_RUNLOCK();
3047 ERROUT(ENOENT);
3048 }
3049 if (ticket != rs->active.ticket) {
3050 PF_RULES_RUNLOCK();
3051 ERROUT(EBUSY);
3052 }
3053
3054 nvlist_destroy(nvl);
3055 nvl = NULL;
3056 free(nvlpacked, M_NVLIST);
3057 nvlpacked = NULL;
3058
3059 rule = TAILQ_FIRST(rs->active.rules);
3060 while ((rule != NULL) && (rule->nr != nr))
3061 rule = TAILQ_NEXT(rule, entries);
3062 if (rule == NULL) {
3063 PF_RULES_RUNLOCK();
3064 ERROUT(ENOENT);
3065 }
3066 /* Make sure rule can't go away. */
3067 NET_EPOCH_ENTER(et);
3068 PF_RULES_RUNLOCK();
3069 nvl = pf_keth_rule_to_nveth_rule(rule);
3070 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
3071 NET_EPOCH_EXIT(et);
3072 ERROUT(EBUSY);
3073 }
3074 NET_EPOCH_EXIT(et);
3075 if (nvl == NULL)
3076 ERROUT(ENOMEM);
3077
3078 nvlpacked = nvlist_pack(nvl, &nv->len);
3079 if (nvlpacked == NULL)
3080 ERROUT(ENOMEM);
3081
3082 if (nv->size == 0)
3083 ERROUT(0);
3084 else if (nv->size < nv->len)
3085 ERROUT(ENOSPC);
3086
3087 error = copyout(nvlpacked, nv->data, nv->len);
3088 if (error == 0 && clear) {
3089 counter_u64_zero(rule->evaluations);
3090 for (int i = 0; i < 2; i++) {
3091 counter_u64_zero(rule->packets[i]);
3092 counter_u64_zero(rule->bytes[i]);
3093 }
3094 }
3095
3096 #undef ERROUT
3097 DIOCGETETHRULE_error:
3098 free(nvlpacked, M_NVLIST);
3099 nvlist_destroy(nvl);
3100 break;
3101 }
3102
3103 case DIOCADDETHRULE: {
3104 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3105 nvlist_t *nvl = NULL;
3106 void *nvlpacked = NULL;
3107 struct pf_keth_rule *rule = NULL, *tail = NULL;
3108 struct pf_keth_ruleset *ruleset = NULL;
3109 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
3110 const char *anchor = "", *anchor_call = "";
3111
3112 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
3113
3114 if (nv->len > pf_ioctl_maxcount)
3115 ERROUT(ENOMEM);
3116
3117 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3118 error = copyin(nv->data, nvlpacked, nv->len);
3119 if (error)
3120 ERROUT(error);
3121
3122 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3123 if (nvl == NULL)
3124 ERROUT(EBADMSG);
3125
3126 if (! nvlist_exists_number(nvl, "ticket"))
3127 ERROUT(EBADMSG);
3128
3129 if (nvlist_exists_string(nvl, "anchor"))
3130 anchor = nvlist_get_string(nvl, "anchor");
3131 if (nvlist_exists_string(nvl, "anchor_call"))
3132 anchor_call = nvlist_get_string(nvl, "anchor_call");
3133
3134 ruleset = pf_find_keth_ruleset(anchor);
3135 if (ruleset == NULL)
3136 ERROUT(EINVAL);
3137
3138 if (nvlist_get_number(nvl, "ticket") !=
3139 ruleset->inactive.ticket) {
3140 DPFPRINTF(PF_DEBUG_MISC,
3141 ("ticket: %d != %d\n",
3142 (u_int32_t)nvlist_get_number(nvl, "ticket"),
3143 ruleset->inactive.ticket));
3144 ERROUT(EBUSY);
3145 }
3146
3147 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3148 rule->timestamp = NULL;
3149
3150 error = pf_nveth_rule_to_keth_rule(nvl, rule);
3151 if (error != 0)
3152 ERROUT(error);
3153
3154 if (rule->ifname[0])
3155 kif = pf_kkif_create(M_WAITOK);
3156 if (rule->bridge_to_name[0])
3157 bridge_to_kif = pf_kkif_create(M_WAITOK);
3158 rule->evaluations = counter_u64_alloc(M_WAITOK);
3159 for (int i = 0; i < 2; i++) {
3160 rule->packets[i] = counter_u64_alloc(M_WAITOK);
3161 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3162 }
3163 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3164 M_WAITOK | M_ZERO);
3165
3166 PF_RULES_WLOCK();
3167
3168 if (rule->ifname[0]) {
3169 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3170 pfi_kkif_ref(rule->kif);
3171 } else
3172 rule->kif = NULL;
3173 if (rule->bridge_to_name[0]) {
3174 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3175 rule->bridge_to_name);
3176 pfi_kkif_ref(rule->bridge_to);
3177 } else
3178 rule->bridge_to = NULL;
3179
3180 #ifdef ALTQ
3181 /* set queue IDs */
3182 if (rule->qname[0] != 0) {
3183 if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3184 error = EBUSY;
3185 else
3186 rule->qid = rule->qid;
3187 }
3188 #endif
3189 if (rule->tagname[0])
3190 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3191 error = EBUSY;
3192 if (rule->match_tagname[0])
3193 if ((rule->match_tag = pf_tagname2tag(
3194 rule->match_tagname)) == 0)
3195 error = EBUSY;
3196
3197 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3198 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3199 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3200 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3201
3202 if (error) {
3203 pf_free_eth_rule(rule);
3204 PF_RULES_WUNLOCK();
3205 ERROUT(error);
3206 }
3207
3208 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3209 pf_free_eth_rule(rule);
3210 PF_RULES_WUNLOCK();
3211 ERROUT(EINVAL);
3212 }
3213
3214 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3215 if (tail)
3216 rule->nr = tail->nr + 1;
3217 else
3218 rule->nr = 0;
3219
3220 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3221
3222 PF_RULES_WUNLOCK();
3223
3224 #undef ERROUT
3225 DIOCADDETHRULE_error:
3226 nvlist_destroy(nvl);
3227 free(nvlpacked, M_NVLIST);
3228 break;
3229 }
3230
3231 case DIOCGETETHRULESETS: {
3232 struct epoch_tracker et;
3233 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3234 nvlist_t *nvl = NULL;
3235 void *nvlpacked = NULL;
3236 struct pf_keth_ruleset *ruleset;
3237 struct pf_keth_anchor *anchor;
3238 int nr = 0;
3239
3240 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3241
3242 if (nv->len > pf_ioctl_maxcount)
3243 ERROUT(ENOMEM);
3244
3245 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3246 error = copyin(nv->data, nvlpacked, nv->len);
3247 if (error)
3248 ERROUT(error);
3249
3250 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3251 if (nvl == NULL)
3252 ERROUT(EBADMSG);
3253 if (! nvlist_exists_string(nvl, "path"))
3254 ERROUT(EBADMSG);
3255
3256 NET_EPOCH_ENTER(et);
3257
3258 if ((ruleset = pf_find_keth_ruleset(
3259 nvlist_get_string(nvl, "path"))) == NULL) {
3260 NET_EPOCH_EXIT(et);
3261 ERROUT(ENOENT);
3262 }
3263
3264 if (ruleset->anchor == NULL) {
3265 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3266 if (anchor->parent == NULL)
3267 nr++;
3268 } else {
3269 RB_FOREACH(anchor, pf_keth_anchor_node,
3270 &ruleset->anchor->children)
3271 nr++;
3272 }
3273
3274 NET_EPOCH_EXIT(et);
3275
3276 nvlist_destroy(nvl);
3277 nvl = NULL;
3278 free(nvlpacked, M_NVLIST);
3279 nvlpacked = NULL;
3280
3281 nvl = nvlist_create(0);
3282 if (nvl == NULL)
3283 ERROUT(ENOMEM);
3284
3285 nvlist_add_number(nvl, "nr", nr);
3286
3287 nvlpacked = nvlist_pack(nvl, &nv->len);
3288 if (nvlpacked == NULL)
3289 ERROUT(ENOMEM);
3290
3291 if (nv->size == 0)
3292 ERROUT(0);
3293 else if (nv->size < nv->len)
3294 ERROUT(ENOSPC);
3295
3296 error = copyout(nvlpacked, nv->data, nv->len);
3297
3298 #undef ERROUT
3299 DIOCGETETHRULESETS_error:
3300 free(nvlpacked, M_NVLIST);
3301 nvlist_destroy(nvl);
3302 break;
3303 }
3304
3305 case DIOCGETETHRULESET: {
3306 struct epoch_tracker et;
3307 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3308 nvlist_t *nvl = NULL;
3309 void *nvlpacked = NULL;
3310 struct pf_keth_ruleset *ruleset;
3311 struct pf_keth_anchor *anchor;
3312 int nr = 0, req_nr = 0;
3313 bool found = false;
3314
3315 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3316
3317 if (nv->len > pf_ioctl_maxcount)
3318 ERROUT(ENOMEM);
3319
3320 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3321 error = copyin(nv->data, nvlpacked, nv->len);
3322 if (error)
3323 ERROUT(error);
3324
3325 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3326 if (nvl == NULL)
3327 ERROUT(EBADMSG);
3328 if (! nvlist_exists_string(nvl, "path"))
3329 ERROUT(EBADMSG);
3330 if (! nvlist_exists_number(nvl, "nr"))
3331 ERROUT(EBADMSG);
3332
3333 req_nr = nvlist_get_number(nvl, "nr");
3334
3335 NET_EPOCH_ENTER(et);
3336
3337 if ((ruleset = pf_find_keth_ruleset(
3338 nvlist_get_string(nvl, "path"))) == NULL) {
3339 NET_EPOCH_EXIT(et);
3340 ERROUT(ENOENT);
3341 }
3342
3343 nvlist_destroy(nvl);
3344 nvl = NULL;
3345 free(nvlpacked, M_NVLIST);
3346 nvlpacked = NULL;
3347
3348 nvl = nvlist_create(0);
3349 if (nvl == NULL) {
3350 NET_EPOCH_EXIT(et);
3351 ERROUT(ENOMEM);
3352 }
3353
3354 if (ruleset->anchor == NULL) {
3355 RB_FOREACH(anchor, pf_keth_anchor_global,
3356 &V_pf_keth_anchors) {
3357 if (anchor->parent == NULL && nr++ == req_nr) {
3358 found = true;
3359 break;
3360 }
3361 }
3362 } else {
3363 RB_FOREACH(anchor, pf_keth_anchor_node,
3364 &ruleset->anchor->children) {
3365 if (nr++ == req_nr) {
3366 found = true;
3367 break;
3368 }
3369 }
3370 }
3371
3372 NET_EPOCH_EXIT(et);
3373 if (found) {
3374 nvlist_add_number(nvl, "nr", nr);
3375 nvlist_add_string(nvl, "name", anchor->name);
3376 if (ruleset->anchor)
3377 nvlist_add_string(nvl, "path",
3378 ruleset->anchor->path);
3379 else
3380 nvlist_add_string(nvl, "path", "");
3381 } else {
3382 ERROUT(EBUSY);
3383 }
3384
3385 nvlpacked = nvlist_pack(nvl, &nv->len);
3386 if (nvlpacked == NULL)
3387 ERROUT(ENOMEM);
3388
3389 if (nv->size == 0)
3390 ERROUT(0);
3391 else if (nv->size < nv->len)
3392 ERROUT(ENOSPC);
3393
3394 error = copyout(nvlpacked, nv->data, nv->len);
3395
3396 #undef ERROUT
3397 DIOCGETETHRULESET_error:
3398 free(nvlpacked, M_NVLIST);
3399 nvlist_destroy(nvl);
3400 break;
3401 }
3402
3403 case DIOCADDRULENV: {
3404 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3405 nvlist_t *nvl = NULL;
3406 void *nvlpacked = NULL;
3407 struct pf_krule *rule = NULL;
3408 const char *anchor = "", *anchor_call = "";
3409 uint32_t ticket = 0, pool_ticket = 0;
3410
3411 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
3412
3413 if (nv->len > pf_ioctl_maxcount)
3414 ERROUT(ENOMEM);
3415
3416 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3417 error = copyin(nv->data, nvlpacked, nv->len);
3418 if (error)
3419 ERROUT(error);
3420
3421 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3422 if (nvl == NULL)
3423 ERROUT(EBADMSG);
3424
3425 if (! nvlist_exists_number(nvl, "ticket"))
3426 ERROUT(EINVAL);
3427 ticket = nvlist_get_number(nvl, "ticket");
3428
3429 if (! nvlist_exists_number(nvl, "pool_ticket"))
3430 ERROUT(EINVAL);
3431 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3432
3433 if (! nvlist_exists_nvlist(nvl, "rule"))
3434 ERROUT(EINVAL);
3435
3436 rule = pf_krule_alloc();
3437 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3438 rule);
3439 if (error)
3440 ERROUT(error);
3441
3442 if (nvlist_exists_string(nvl, "anchor"))
3443 anchor = nvlist_get_string(nvl, "anchor");
3444 if (nvlist_exists_string(nvl, "anchor_call"))
3445 anchor_call = nvlist_get_string(nvl, "anchor_call");
3446
3447 if ((error = nvlist_error(nvl)))
3448 ERROUT(error);
3449
3450 /* Frees rule on error */
3451 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3452 anchor_call, td->td_ucred->cr_ruid,
3453 td->td_proc ? td->td_proc->p_pid : 0);
3454
3455 nvlist_destroy(nvl);
3456 free(nvlpacked, M_NVLIST);
3457 break;
3458 #undef ERROUT
3459 DIOCADDRULENV_error:
3460 pf_krule_free(rule);
3461 nvlist_destroy(nvl);
3462 free(nvlpacked, M_NVLIST);
3463
3464 break;
3465 }
3466 case DIOCADDRULE: {
3467 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3468 struct pf_krule *rule;
3469
3470 rule = pf_krule_alloc();
3471 error = pf_rule_to_krule(&pr->rule, rule);
3472 if (error != 0) {
3473 pf_krule_free(rule);
3474 break;
3475 }
3476
3477 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3478
3479 /* Frees rule on error */
3480 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3481 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3482 td->td_proc ? td->td_proc->p_pid : 0);
3483 break;
3484 }
3485
3486 case DIOCGETRULES: {
3487 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
3488
3489 pr->anchor[sizeof(pr->anchor) - 1] = 0;
3490
3491 error = pf_ioctl_getrules(pr);
3492
3493 break;
3494 }
3495
3496 case DIOCGETRULENV: {
3497 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3498 nvlist_t *nvrule = NULL;
3499 nvlist_t *nvl = NULL;
3500 struct pf_kruleset *ruleset;
3501 struct pf_krule *rule;
3502 void *nvlpacked = NULL;
3503 int rs_num, nr;
3504 bool clear_counter = false;
3505
3506 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
3507
3508 if (nv->len > pf_ioctl_maxcount)
3509 ERROUT(ENOMEM);
3510
3511 /* Copy the request in */
3512 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3513 error = copyin(nv->data, nvlpacked, nv->len);
3514 if (error)
3515 ERROUT(error);
3516
3517 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3518 if (nvl == NULL)
3519 ERROUT(EBADMSG);
3520
3521 if (! nvlist_exists_string(nvl, "anchor"))
3522 ERROUT(EBADMSG);
3523 if (! nvlist_exists_number(nvl, "ruleset"))
3524 ERROUT(EBADMSG);
3525 if (! nvlist_exists_number(nvl, "ticket"))
3526 ERROUT(EBADMSG);
3527 if (! nvlist_exists_number(nvl, "nr"))
3528 ERROUT(EBADMSG);
3529
3530 if (nvlist_exists_bool(nvl, "clear_counter"))
3531 clear_counter = nvlist_get_bool(nvl, "clear_counter");
3532
3533 if (clear_counter && !(flags & FWRITE))
3534 ERROUT(EACCES);
3535
3536 nr = nvlist_get_number(nvl, "nr");
3537
3538 PF_RULES_WLOCK();
3539 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3540 if (ruleset == NULL) {
3541 PF_RULES_WUNLOCK();
3542 ERROUT(ENOENT);
3543 }
3544
3545 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3546 if (rs_num >= PF_RULESET_MAX) {
3547 PF_RULES_WUNLOCK();
3548 ERROUT(EINVAL);
3549 }
3550
3551 if (nvlist_get_number(nvl, "ticket") !=
3552 ruleset->rules[rs_num].active.ticket) {
3553 PF_RULES_WUNLOCK();
3554 ERROUT(EBUSY);
3555 }
3556
3557 if ((error = nvlist_error(nvl))) {
3558 PF_RULES_WUNLOCK();
3559 ERROUT(error);
3560 }
3561
3562 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3563 while ((rule != NULL) && (rule->nr != nr))
3564 rule = TAILQ_NEXT(rule, entries);
3565 if (rule == NULL) {
3566 PF_RULES_WUNLOCK();
3567 ERROUT(EBUSY);
3568 }
3569
3570 nvrule = pf_krule_to_nvrule(rule);
3571
3572 nvlist_destroy(nvl);
3573 nvl = nvlist_create(0);
3574 if (nvl == NULL) {
3575 PF_RULES_WUNLOCK();
3576 ERROUT(ENOMEM);
3577 }
3578 nvlist_add_number(nvl, "nr", nr);
3579 nvlist_add_nvlist(nvl, "rule", nvrule);
3580 nvlist_destroy(nvrule);
3581 nvrule = NULL;
3582 if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3583 PF_RULES_WUNLOCK();
3584 ERROUT(EBUSY);
3585 }
3586
3587 free(nvlpacked, M_NVLIST);
3588 nvlpacked = nvlist_pack(nvl, &nv->len);
3589 if (nvlpacked == NULL) {
3590 PF_RULES_WUNLOCK();
3591 ERROUT(ENOMEM);
3592 }
3593
3594 if (nv->size == 0) {
3595 PF_RULES_WUNLOCK();
3596 ERROUT(0);
3597 }
3598 else if (nv->size < nv->len) {
3599 PF_RULES_WUNLOCK();
3600 ERROUT(ENOSPC);
3601 }
3602
3603 if (clear_counter)
3604 pf_krule_clear_counters(rule);
3605
3606 PF_RULES_WUNLOCK();
3607
3608 error = copyout(nvlpacked, nv->data, nv->len);
3609
3610 #undef ERROUT
3611 DIOCGETRULENV_error:
3612 free(nvlpacked, M_NVLIST);
3613 nvlist_destroy(nvrule);
3614 nvlist_destroy(nvl);
3615
3616 break;
3617 }
3618
3619 case DIOCCHANGERULE: {
3620 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
3621 struct pf_kruleset *ruleset;
3622 struct pf_krule *oldrule = NULL, *newrule = NULL;
3623 struct pfi_kkif *kif = NULL;
3624 struct pf_kpooladdr *pa;
3625 u_int32_t nr = 0;
3626 int rs_num;
3627
3628 pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3629
3630 if (pcr->action < PF_CHANGE_ADD_HEAD ||
3631 pcr->action > PF_CHANGE_GET_TICKET) {
3632 error = EINVAL;
3633 break;
3634 }
3635 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3636 error = EINVAL;
3637 break;
3638 }
3639
3640 if (pcr->action != PF_CHANGE_REMOVE) {
3641 newrule = pf_krule_alloc();
3642 error = pf_rule_to_krule(&pcr->rule, newrule);
3643 if (error != 0) {
3644 pf_krule_free(newrule);
3645 break;
3646 }
3647
3648 if (newrule->ifname[0])
3649 kif = pf_kkif_create(M_WAITOK);
3650 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3651 for (int i = 0; i < 2; i++) {
3652 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3653 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3654 }
3655 newrule->states_cur = counter_u64_alloc(M_WAITOK);
3656 newrule->states_tot = counter_u64_alloc(M_WAITOK);
3657 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
3658 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
3659 newrule->cuid = td->td_ucred->cr_ruid;
3660 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3661 TAILQ_INIT(&newrule->nat.list);
3662 TAILQ_INIT(&newrule->rdr.list);
3663 TAILQ_INIT(&newrule->route.list);
3664 }
3665 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3666
3667 PF_CONFIG_LOCK();
3668 PF_RULES_WLOCK();
3669 #ifdef PF_WANT_32_TO_64_COUNTER
3670 if (newrule != NULL) {
3671 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3672 newrule->allrulelinked = true;
3673 V_pf_allrulecount++;
3674 }
3675 #endif
3676
3677 if (!(pcr->action == PF_CHANGE_REMOVE ||
3678 pcr->action == PF_CHANGE_GET_TICKET) &&
3679 pcr->pool_ticket != V_ticket_pabuf)
3680 ERROUT(EBUSY);
3681
3682 ruleset = pf_find_kruleset(pcr->anchor);
3683 if (ruleset == NULL)
3684 ERROUT(EINVAL);
3685
3686 rs_num = pf_get_ruleset_number(pcr->rule.action);
3687 if (rs_num >= PF_RULESET_MAX)
3688 ERROUT(EINVAL);
3689
3690 /*
3691 * XXXMJG: there is no guarantee that the ruleset was
3692 * created by the usual route of calling DIOCXBEGIN.
3693 * As a result it is possible the rule tree will not
3694 * be allocated yet. Hack around it by doing it here.
3695 * Note it is fine to let the tree persist in case of
3696 * error as it will be freed down the road on future
3697 * updates (if need be).
3698 */
3699 if (ruleset->rules[rs_num].active.tree == NULL) {
3700 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3701 if (ruleset->rules[rs_num].active.tree == NULL) {
3702 ERROUT(ENOMEM);
3703 }
3704 }
3705
3706 if (pcr->action == PF_CHANGE_GET_TICKET) {
3707 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3708 ERROUT(0);
3709 } else if (pcr->ticket !=
3710 ruleset->rules[rs_num].active.ticket)
3711 ERROUT(EINVAL);
3712
3713 if (pcr->action != PF_CHANGE_REMOVE) {
3714 if (newrule->ifname[0]) {
3715 newrule->kif = pfi_kkif_attach(kif,
3716 newrule->ifname);
3717 kif = NULL;
3718 pfi_kkif_ref(newrule->kif);
3719 } else
3720 newrule->kif = NULL;
3721
3722 if (newrule->rtableid > 0 &&
3723 newrule->rtableid >= rt_numfibs)
3724 error = EBUSY;
3725
3726 #ifdef ALTQ
3727 /* set queue IDs */
3728 if (newrule->qname[0] != 0) {
3729 if ((newrule->qid =
3730 pf_qname2qid(newrule->qname)) == 0)
3731 error = EBUSY;
3732 else if (newrule->pqname[0] != 0) {
3733 if ((newrule->pqid =
3734 pf_qname2qid(newrule->pqname)) == 0)
3735 error = EBUSY;
3736 } else
3737 newrule->pqid = newrule->qid;
3738 }
3739 #endif /* ALTQ */
3740 if (newrule->tagname[0])
3741 if ((newrule->tag =
3742 pf_tagname2tag(newrule->tagname)) == 0)
3743 error = EBUSY;
3744 if (newrule->match_tagname[0])
3745 if ((newrule->match_tag = pf_tagname2tag(
3746 newrule->match_tagname)) == 0)
3747 error = EBUSY;
3748 if (newrule->rt && !newrule->direction)
3749 error = EINVAL;
3750 if (!newrule->log)
3751 newrule->logif = 0;
3752 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3753 error = ENOMEM;
3754 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3755 error = ENOMEM;
3756 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3757 error = EINVAL;
3758 for (int i = 0; i < 3; i++) {
3759 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
3760 if (pa->addr.type == PF_ADDR_TABLE) {
3761 pa->addr.p.tbl =
3762 pfr_attach_table(ruleset,
3763 pa->addr.v.tblname);
3764 if (pa->addr.p.tbl == NULL)
3765 error = ENOMEM;
3766 }
3767 }
3768
3769 newrule->overload_tbl = NULL;
3770 if (newrule->overload_tblname[0]) {
3771 if ((newrule->overload_tbl = pfr_attach_table(
3772 ruleset, newrule->overload_tblname)) ==
3773 NULL)
3774 error = EINVAL;
3775 else
3776 newrule->overload_tbl->pfrkt_flags |=
3777 PFR_TFLAG_ACTIVE;
3778 }
3779
3780 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
3781 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
3782 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list);
3783 if (((((newrule->action == PF_NAT) ||
3784 (newrule->action == PF_RDR) ||
3785 (newrule->action == PF_BINAT) ||
3786 (newrule->rt > PF_NOPFROUTE)) &&
3787 !newrule->anchor)) &&
3788 (TAILQ_FIRST(&newrule->rdr.list) == NULL))
3789 error = EINVAL;
3790
3791 if (error) {
3792 pf_free_rule(newrule);
3793 PF_RULES_WUNLOCK();
3794 PF_CONFIG_UNLOCK();
3795 break;
3796 }
3797
3798 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
3799 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
3800 }
3801 pf_empty_kpool(&V_pf_pabuf[0]);
3802 pf_empty_kpool(&V_pf_pabuf[1]);
3803 pf_empty_kpool(&V_pf_pabuf[2]);
3804
3805 if (pcr->action == PF_CHANGE_ADD_HEAD)
3806 oldrule = TAILQ_FIRST(
3807 ruleset->rules[rs_num].active.ptr);
3808 else if (pcr->action == PF_CHANGE_ADD_TAIL)
3809 oldrule = TAILQ_LAST(
3810 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3811 else {
3812 oldrule = TAILQ_FIRST(
3813 ruleset->rules[rs_num].active.ptr);
3814 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3815 oldrule = TAILQ_NEXT(oldrule, entries);
3816 if (oldrule == NULL) {
3817 if (newrule != NULL)
3818 pf_free_rule(newrule);
3819 PF_RULES_WUNLOCK();
3820 PF_CONFIG_UNLOCK();
3821 error = EINVAL;
3822 break;
3823 }
3824 }
3825
3826 if (pcr->action == PF_CHANGE_REMOVE) {
3827 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3828 oldrule);
3829 RB_REMOVE(pf_krule_global,
3830 ruleset->rules[rs_num].active.tree, oldrule);
3831 ruleset->rules[rs_num].active.rcount--;
3832 } else {
3833 pf_hash_rule(newrule);
3834 if (RB_INSERT(pf_krule_global,
3835 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3836 pf_free_rule(newrule);
3837 PF_RULES_WUNLOCK();
3838 PF_CONFIG_UNLOCK();
3839 error = EEXIST;
3840 break;
3841 }
3842
3843 if (oldrule == NULL)
3844 TAILQ_INSERT_TAIL(
3845 ruleset->rules[rs_num].active.ptr,
3846 newrule, entries);
3847 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3848 pcr->action == PF_CHANGE_ADD_BEFORE)
3849 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3850 else
3851 TAILQ_INSERT_AFTER(
3852 ruleset->rules[rs_num].active.ptr,
3853 oldrule, newrule, entries);
3854 ruleset->rules[rs_num].active.rcount++;
3855 }
3856
3857 nr = 0;
3858 TAILQ_FOREACH(oldrule,
3859 ruleset->rules[rs_num].active.ptr, entries)
3860 oldrule->nr = nr++;
3861
3862 ruleset->rules[rs_num].active.ticket++;
3863
3864 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3865 pf_remove_if_empty_kruleset(ruleset);
3866
3867 PF_RULES_WUNLOCK();
3868 PF_CONFIG_UNLOCK();
3869 break;
3870
3871 #undef ERROUT
3872 DIOCCHANGERULE_error:
3873 PF_RULES_WUNLOCK();
3874 PF_CONFIG_UNLOCK();
3875 pf_krule_free(newrule);
3876 pf_kkif_free(kif);
3877 break;
3878 }
3879
3880 case DIOCCLRSTATESNV: {
3881 error = pf_clearstates_nv((struct pfioc_nv *)addr);
3882 break;
3883 }
3884
3885 case DIOCKILLSTATESNV: {
3886 error = pf_killstates_nv((struct pfioc_nv *)addr);
3887 break;
3888 }
3889
3890 case DIOCADDSTATE: {
3891 struct pfioc_state *ps = (struct pfioc_state *)addr;
3892 struct pfsync_state_1301 *sp = &ps->state;
3893
3894 if (sp->timeout >= PFTM_MAX) {
3895 error = EINVAL;
3896 break;
3897 }
3898 if (V_pfsync_state_import_ptr != NULL) {
3899 PF_RULES_RLOCK();
3900 error = V_pfsync_state_import_ptr(
3901 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3902 PFSYNC_MSG_VERSION_1301);
3903 PF_RULES_RUNLOCK();
3904 } else
3905 error = EOPNOTSUPP;
3906 break;
3907 }
3908
3909 case DIOCGETSTATE: {
3910 struct pfioc_state *ps = (struct pfioc_state *)addr;
3911 struct pf_kstate *s;
3912
3913 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3914 if (s == NULL) {
3915 error = ENOENT;
3916 break;
3917 }
3918
3919 pfsync_state_export((union pfsync_state_union*)&ps->state,
3920 s, PFSYNC_MSG_VERSION_1301);
3921 PF_STATE_UNLOCK(s);
3922 break;
3923 }
3924
3925 case DIOCGETSTATENV: {
3926 error = pf_getstate((struct pfioc_nv *)addr);
3927 break;
3928 }
3929
3930 #ifdef COMPAT_FREEBSD14
3931 case DIOCGETSTATES: {
3932 struct pfioc_states *ps = (struct pfioc_states *)addr;
3933 struct pf_kstate *s;
3934 struct pfsync_state_1301 *pstore, *p;
3935 int i, nr;
3936 size_t slice_count = 16, count;
3937 void *out;
3938
3939 if (ps->ps_len <= 0) {
3940 nr = uma_zone_get_cur(V_pf_state_z);
3941 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3942 break;
3943 }
3944
3945 out = ps->ps_states;
3946 pstore = mallocarray(slice_count,
3947 sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3948 nr = 0;
3949
3950 for (i = 0; i <= V_pf_hashmask; i++) {
3951 struct pf_idhash *ih = &V_pf_idhash[i];
3952
3953 DIOCGETSTATES_retry:
3954 p = pstore;
3955
3956 if (LIST_EMPTY(&ih->states))
3957 continue;
3958
3959 PF_HASHROW_LOCK(ih);
3960 count = 0;
3961 LIST_FOREACH(s, &ih->states, entry) {
3962 if (s->timeout == PFTM_UNLINKED)
3963 continue;
3964 count++;
3965 }
3966
3967 if (count > slice_count) {
3968 PF_HASHROW_UNLOCK(ih);
3969 free(pstore, M_TEMP);
3970 slice_count = count * 2;
3971 pstore = mallocarray(slice_count,
3972 sizeof(struct pfsync_state_1301), M_TEMP,
3973 M_WAITOK | M_ZERO);
3974 goto DIOCGETSTATES_retry;
3975 }
3976
3977 if ((nr+count) * sizeof(*p) > ps->ps_len) {
3978 PF_HASHROW_UNLOCK(ih);
3979 goto DIOCGETSTATES_full;
3980 }
3981
3982 LIST_FOREACH(s, &ih->states, entry) {
3983 if (s->timeout == PFTM_UNLINKED)
3984 continue;
3985
3986 pfsync_state_export((union pfsync_state_union*)p,
3987 s, PFSYNC_MSG_VERSION_1301);
3988 p++;
3989 nr++;
3990 }
3991 PF_HASHROW_UNLOCK(ih);
3992 error = copyout(pstore, out,
3993 sizeof(struct pfsync_state_1301) * count);
3994 if (error)
3995 break;
3996 out = ps->ps_states + nr;
3997 }
3998 DIOCGETSTATES_full:
3999 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4000 free(pstore, M_TEMP);
4001
4002 break;
4003 }
4004
4005 case DIOCGETSTATESV2: {
4006 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
4007 struct pf_kstate *s;
4008 struct pf_state_export *pstore, *p;
4009 int i, nr;
4010 size_t slice_count = 16, count;
4011 void *out;
4012
4013 if (ps->ps_req_version > PF_STATE_VERSION) {
4014 error = ENOTSUP;
4015 break;
4016 }
4017
4018 if (ps->ps_len <= 0) {
4019 nr = uma_zone_get_cur(V_pf_state_z);
4020 ps->ps_len = sizeof(struct pf_state_export) * nr;
4021 break;
4022 }
4023
4024 out = ps->ps_states;
4025 pstore = mallocarray(slice_count,
4026 sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
4027 nr = 0;
4028
4029 for (i = 0; i <= V_pf_hashmask; i++) {
4030 struct pf_idhash *ih = &V_pf_idhash[i];
4031
4032 DIOCGETSTATESV2_retry:
4033 p = pstore;
4034
4035 if (LIST_EMPTY(&ih->states))
4036 continue;
4037
4038 PF_HASHROW_LOCK(ih);
4039 count = 0;
4040 LIST_FOREACH(s, &ih->states, entry) {
4041 if (s->timeout == PFTM_UNLINKED)
4042 continue;
4043 count++;
4044 }
4045
4046 if (count > slice_count) {
4047 PF_HASHROW_UNLOCK(ih);
4048 free(pstore, M_TEMP);
4049 slice_count = count * 2;
4050 pstore = mallocarray(slice_count,
4051 sizeof(struct pf_state_export), M_TEMP,
4052 M_WAITOK | M_ZERO);
4053 goto DIOCGETSTATESV2_retry;
4054 }
4055
4056 if ((nr+count) * sizeof(*p) > ps->ps_len) {
4057 PF_HASHROW_UNLOCK(ih);
4058 goto DIOCGETSTATESV2_full;
4059 }
4060
4061 LIST_FOREACH(s, &ih->states, entry) {
4062 if (s->timeout == PFTM_UNLINKED)
4063 continue;
4064
4065 pf_state_export(p, s);
4066 p++;
4067 nr++;
4068 }
4069 PF_HASHROW_UNLOCK(ih);
4070 error = copyout(pstore, out,
4071 sizeof(struct pf_state_export) * count);
4072 if (error)
4073 break;
4074 out = ps->ps_states + nr;
4075 }
4076 DIOCGETSTATESV2_full:
4077 ps->ps_len = nr * sizeof(struct pf_state_export);
4078 free(pstore, M_TEMP);
4079
4080 break;
4081 }
4082 #endif
4083 case DIOCGETSTATUSNV: {
4084 error = pf_getstatus((struct pfioc_nv *)addr);
4085 break;
4086 }
4087
4088 case DIOCSETSTATUSIF: {
4089 struct pfioc_if *pi = (struct pfioc_if *)addr;
4090
4091 if (pi->ifname[0] == 0) {
4092 bzero(V_pf_status.ifname, IFNAMSIZ);
4093 break;
4094 }
4095 PF_RULES_WLOCK();
4096 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
4097 PF_RULES_WUNLOCK();
4098 break;
4099 }
4100
4101 case DIOCCLRSTATUS: {
4102 pf_ioctl_clear_status();
4103 break;
4104 }
4105
4106 case DIOCNATLOOK: {
4107 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
4108 struct pf_state_key *sk;
4109 struct pf_kstate *state;
4110 struct pf_state_key_cmp key;
4111 int m = 0, direction = pnl->direction;
4112 int sidx, didx;
4113
4114 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
4115 sidx = (direction == PF_IN) ? 1 : 0;
4116 didx = (direction == PF_IN) ? 0 : 1;
4117
4118 if (!pnl->proto ||
4119 PF_AZERO(&pnl->saddr, pnl->af) ||
4120 PF_AZERO(&pnl->daddr, pnl->af) ||
4121 ((pnl->proto == IPPROTO_TCP ||
4122 pnl->proto == IPPROTO_UDP) &&
4123 (!pnl->dport || !pnl->sport)))
4124 error = EINVAL;
4125 else {
4126 bzero(&key, sizeof(key));
4127 key.af = pnl->af;
4128 key.proto = pnl->proto;
4129 PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
4130 key.port[sidx] = pnl->sport;
4131 PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
4132 key.port[didx] = pnl->dport;
4133
4134 state = pf_find_state_all(&key, direction, &m);
4135 if (state == NULL) {
4136 error = ENOENT;
4137 } else {
4138 if (m > 1) {
4139 PF_STATE_UNLOCK(state);
4140 error = E2BIG; /* more than one state */
4141 } else {
4142 sk = state->key[sidx];
4143 PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4144 pnl->rsport = sk->port[sidx];
4145 PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4146 pnl->rdport = sk->port[didx];
4147 PF_STATE_UNLOCK(state);
4148 }
4149 }
4150 }
4151 break;
4152 }
4153
4154 case DIOCSETTIMEOUT: {
4155 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4156
4157 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4158 &pt->seconds);
4159 break;
4160 }
4161
4162 case DIOCGETTIMEOUT: {
4163 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
4164
4165 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4166 break;
4167 }
4168
4169 case DIOCGETLIMIT: {
4170 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4171
4172 error = pf_ioctl_get_limit(pl->index, &pl->limit);
4173 break;
4174 }
4175
4176 case DIOCSETLIMIT: {
4177 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
4178 unsigned int old_limit;
4179
4180 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4181 pl->limit = old_limit;
4182 break;
4183 }
4184
4185 case DIOCSETDEBUG: {
4186 u_int32_t *level = (u_int32_t *)addr;
4187
4188 PF_RULES_WLOCK();
4189 V_pf_status.debug = *level;
4190 PF_RULES_WUNLOCK();
4191 break;
4192 }
4193
4194 case DIOCCLRRULECTRS: {
4195 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4196 struct pf_kruleset *ruleset = &pf_main_ruleset;
4197 struct pf_krule *rule;
4198
4199 PF_RULES_WLOCK();
4200 TAILQ_FOREACH(rule,
4201 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4202 pf_counter_u64_zero(&rule->evaluations);
4203 for (int i = 0; i < 2; i++) {
4204 pf_counter_u64_zero(&rule->packets[i]);
4205 pf_counter_u64_zero(&rule->bytes[i]);
4206 }
4207 }
4208 PF_RULES_WUNLOCK();
4209 break;
4210 }
4211
4212 case DIOCGIFSPEEDV0:
4213 case DIOCGIFSPEEDV1: {
4214 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
4215 struct pf_ifspeed_v1 ps;
4216 struct ifnet *ifp;
4217
4218 if (psp->ifname[0] == '\0') {
4219 error = EINVAL;
4220 break;
4221 }
4222
4223 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4224 if (error != 0)
4225 break;
4226 ifp = ifunit(ps.ifname);
4227 if (ifp != NULL) {
4228 psp->baudrate32 =
4229 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4230 if (cmd == DIOCGIFSPEEDV1)
4231 psp->baudrate = ifp->if_baudrate;
4232 } else {
4233 error = EINVAL;
4234 }
4235 break;
4236 }
4237
4238 #ifdef ALTQ
4239 case DIOCSTARTALTQ: {
4240 struct pf_altq *altq;
4241
4242 PF_RULES_WLOCK();
4243 /* enable all altq interfaces on active list */
4244 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4245 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4246 error = pf_enable_altq(altq);
4247 if (error != 0)
4248 break;
4249 }
4250 }
4251 if (error == 0)
4252 V_pf_altq_running = 1;
4253 PF_RULES_WUNLOCK();
4254 DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4255 break;
4256 }
4257
4258 case DIOCSTOPALTQ: {
4259 struct pf_altq *altq;
4260
4261 PF_RULES_WLOCK();
4262 /* disable all altq interfaces on active list */
4263 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4264 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4265 error = pf_disable_altq(altq);
4266 if (error != 0)
4267 break;
4268 }
4269 }
4270 if (error == 0)
4271 V_pf_altq_running = 0;
4272 PF_RULES_WUNLOCK();
4273 DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4274 break;
4275 }
4276
4277 case DIOCADDALTQV0:
4278 case DIOCADDALTQV1: {
4279 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4280 struct pf_altq *altq, *a;
4281 struct ifnet *ifp;
4282
4283 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4284 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4285 if (error)
4286 break;
4287 altq->local_flags = 0;
4288
4289 PF_RULES_WLOCK();
4290 if (pa->ticket != V_ticket_altqs_inactive) {
4291 PF_RULES_WUNLOCK();
4292 free(altq, M_PFALTQ);
4293 error = EBUSY;
4294 break;
4295 }
4296
4297 /*
4298 * if this is for a queue, find the discipline and
4299 * copy the necessary fields
4300 */
4301 if (altq->qname[0] != 0) {
4302 if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4303 PF_RULES_WUNLOCK();
4304 error = EBUSY;
4305 free(altq, M_PFALTQ);
4306 break;
4307 }
4308 altq->altq_disc = NULL;
4309 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4310 if (strncmp(a->ifname, altq->ifname,
4311 IFNAMSIZ) == 0) {
4312 altq->altq_disc = a->altq_disc;
4313 break;
4314 }
4315 }
4316 }
4317
4318 if ((ifp = ifunit(altq->ifname)) == NULL)
4319 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4320 else
4321 error = altq_add(ifp, altq);
4322
4323 if (error) {
4324 PF_RULES_WUNLOCK();
4325 free(altq, M_PFALTQ);
4326 break;
4327 }
4328
4329 if (altq->qname[0] != 0)
4330 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4331 else
4332 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4333 /* version error check done on import above */
4334 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4335 PF_RULES_WUNLOCK();
4336 break;
4337 }
4338
4339 case DIOCGETALTQSV0:
4340 case DIOCGETALTQSV1: {
4341 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4342 struct pf_altq *altq;
4343
4344 PF_RULES_RLOCK();
4345 pa->nr = 0;
4346 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4347 pa->nr++;
4348 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4349 pa->nr++;
4350 pa->ticket = V_ticket_altqs_active;
4351 PF_RULES_RUNLOCK();
4352 break;
4353 }
4354
4355 case DIOCGETALTQV0:
4356 case DIOCGETALTQV1: {
4357 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
4358 struct pf_altq *altq;
4359
4360 PF_RULES_RLOCK();
4361 if (pa->ticket != V_ticket_altqs_active) {
4362 PF_RULES_RUNLOCK();
4363 error = EBUSY;
4364 break;
4365 }
4366 altq = pf_altq_get_nth_active(pa->nr);
4367 if (altq == NULL) {
4368 PF_RULES_RUNLOCK();
4369 error = EBUSY;
4370 break;
4371 }
4372 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4373 PF_RULES_RUNLOCK();
4374 break;
4375 }
4376
4377 case DIOCCHANGEALTQV0:
4378 case DIOCCHANGEALTQV1:
4379 /* CHANGEALTQ not supported yet! */
4380 error = ENODEV;
4381 break;
4382
4383 case DIOCGETQSTATSV0:
4384 case DIOCGETQSTATSV1: {
4385 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
4386 struct pf_altq *altq;
4387 int nbytes;
4388 u_int32_t version;
4389
4390 PF_RULES_RLOCK();
4391 if (pq->ticket != V_ticket_altqs_active) {
4392 PF_RULES_RUNLOCK();
4393 error = EBUSY;
4394 break;
4395 }
4396 nbytes = pq->nbytes;
4397 altq = pf_altq_get_nth_active(pq->nr);
4398 if (altq == NULL) {
4399 PF_RULES_RUNLOCK();
4400 error = EBUSY;
4401 break;
4402 }
4403
4404 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4405 PF_RULES_RUNLOCK();
4406 error = ENXIO;
4407 break;
4408 }
4409 PF_RULES_RUNLOCK();
4410 if (cmd == DIOCGETQSTATSV0)
4411 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
4412 else
4413 version = pq->version;
4414 error = altq_getqstats(altq, pq->buf, &nbytes, version);
4415 if (error == 0) {
4416 pq->scheduler = altq->scheduler;
4417 pq->nbytes = nbytes;
4418 }
4419 break;
4420 }
4421 #endif /* ALTQ */
4422
4423 case DIOCBEGINADDRS: {
4424 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4425
4426 error = pf_ioctl_begin_addrs(&pp->ticket);
4427 break;
4428 }
4429
4430 case DIOCADDADDR: {
4431 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4432 struct pf_nl_pooladdr npp = {};
4433
4434 npp.which = PF_RDR;
4435 memcpy(&npp, pp, sizeof(*pp));
4436 error = pf_ioctl_add_addr(&npp);
4437 break;
4438 }
4439
4440 case DIOCGETADDRS: {
4441 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4442 struct pf_nl_pooladdr npp = {};
4443
4444 npp.which = PF_RDR;
4445 memcpy(&npp, pp, sizeof(*pp));
4446 error = pf_ioctl_get_addrs(&npp);
4447 memcpy(pp, &npp, sizeof(*pp));
4448
4449 break;
4450 }
4451
4452 case DIOCGETADDR: {
4453 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
4454 struct pf_nl_pooladdr npp = {};
4455
4456 npp.which = PF_RDR;
4457 memcpy(&npp, pp, sizeof(*pp));
4458 error = pf_ioctl_get_addr(&npp);
4459 memcpy(pp, &npp, sizeof(*pp));
4460
4461 break;
4462 }
4463
4464 case DIOCCHANGEADDR: {
4465 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
4466 struct pf_kpool *pool;
4467 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
4468 struct pf_kruleset *ruleset;
4469 struct pfi_kkif *kif = NULL;
4470
4471 pca->anchor[sizeof(pca->anchor) - 1] = 0;
4472
4473 if (pca->action < PF_CHANGE_ADD_HEAD ||
4474 pca->action > PF_CHANGE_REMOVE) {
4475 error = EINVAL;
4476 break;
4477 }
4478 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4479 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4480 pca->addr.addr.type != PF_ADDR_TABLE) {
4481 error = EINVAL;
4482 break;
4483 }
4484 if (pca->addr.addr.p.dyn != NULL) {
4485 error = EINVAL;
4486 break;
4487 }
4488
4489 if (pca->action != PF_CHANGE_REMOVE) {
4490 #ifndef INET
4491 if (pca->af == AF_INET) {
4492 error = EAFNOSUPPORT;
4493 break;
4494 }
4495 #endif /* INET */
4496 #ifndef INET6
4497 if (pca->af == AF_INET6) {
4498 error = EAFNOSUPPORT;
4499 break;
4500 }
4501 #endif /* INET6 */
4502 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4503 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4504 if (newpa->ifname[0])
4505 kif = pf_kkif_create(M_WAITOK);
4506 newpa->kif = NULL;
4507 }
4508 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4509 PF_RULES_WLOCK();
4510 ruleset = pf_find_kruleset(pca->anchor);
4511 if (ruleset == NULL)
4512 ERROUT(EBUSY);
4513
4514 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4515 pca->r_num, pca->r_last, 1, 1, PF_RDR);
4516 if (pool == NULL)
4517 ERROUT(EBUSY);
4518
4519 if (pca->action != PF_CHANGE_REMOVE) {
4520 if (newpa->ifname[0]) {
4521 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4522 pfi_kkif_ref(newpa->kif);
4523 kif = NULL;
4524 }
4525
4526 switch (newpa->addr.type) {
4527 case PF_ADDR_DYNIFTL:
4528 error = pfi_dynaddr_setup(&newpa->addr,
4529 pca->af);
4530 break;
4531 case PF_ADDR_TABLE:
4532 newpa->addr.p.tbl = pfr_attach_table(ruleset,
4533 newpa->addr.v.tblname);
4534 if (newpa->addr.p.tbl == NULL)
4535 error = ENOMEM;
4536 break;
4537 }
4538 if (error)
4539 goto DIOCCHANGEADDR_error;
4540 }
4541
4542 switch (pca->action) {
4543 case PF_CHANGE_ADD_HEAD:
4544 oldpa = TAILQ_FIRST(&pool->list);
4545 break;
4546 case PF_CHANGE_ADD_TAIL:
4547 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4548 break;
4549 default:
4550 oldpa = TAILQ_FIRST(&pool->list);
4551 for (int i = 0; oldpa && i < pca->nr; i++)
4552 oldpa = TAILQ_NEXT(oldpa, entries);
4553
4554 if (oldpa == NULL)
4555 ERROUT(EINVAL);
4556 }
4557
4558 if (pca->action == PF_CHANGE_REMOVE) {
4559 TAILQ_REMOVE(&pool->list, oldpa, entries);
4560 switch (oldpa->addr.type) {
4561 case PF_ADDR_DYNIFTL:
4562 pfi_dynaddr_remove(oldpa->addr.p.dyn);
4563 break;
4564 case PF_ADDR_TABLE:
4565 pfr_detach_table(oldpa->addr.p.tbl);
4566 break;
4567 }
4568 if (oldpa->kif)
4569 pfi_kkif_unref(oldpa->kif);
4570 free(oldpa, M_PFRULE);
4571 } else {
4572 if (oldpa == NULL)
4573 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4574 else if (pca->action == PF_CHANGE_ADD_HEAD ||
4575 pca->action == PF_CHANGE_ADD_BEFORE)
4576 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4577 else
4578 TAILQ_INSERT_AFTER(&pool->list, oldpa,
4579 newpa, entries);
4580 }
4581
4582 pool->cur = TAILQ_FIRST(&pool->list);
4583 PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4584 PF_RULES_WUNLOCK();
4585 break;
4586
4587 #undef ERROUT
4588 DIOCCHANGEADDR_error:
4589 if (newpa != NULL) {
4590 if (newpa->kif)
4591 pfi_kkif_unref(newpa->kif);
4592 free(newpa, M_PFRULE);
4593 }
4594 PF_RULES_WUNLOCK();
4595 pf_kkif_free(kif);
4596 break;
4597 }
4598
4599 case DIOCGETRULESETS: {
4600 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4601
4602 pr->path[sizeof(pr->path) - 1] = 0;
4603
4604 error = pf_ioctl_get_rulesets(pr);
4605 break;
4606 }
4607
4608 case DIOCGETRULESET: {
4609 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
4610
4611 pr->path[sizeof(pr->path) - 1] = 0;
4612
4613 error = pf_ioctl_get_ruleset(pr);
4614 break;
4615 }
4616
4617 case DIOCRCLRTABLES: {
4618 struct pfioc_table *io = (struct pfioc_table *)addr;
4619
4620 if (io->pfrio_esize != 0) {
4621 error = ENODEV;
4622 break;
4623 }
4624 PF_RULES_WLOCK();
4625 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4626 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4627 PF_RULES_WUNLOCK();
4628 break;
4629 }
4630
4631 case DIOCRADDTABLES: {
4632 struct pfioc_table *io = (struct pfioc_table *)addr;
4633 struct pfr_table *pfrts;
4634 size_t totlen;
4635
4636 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4637 error = ENODEV;
4638 break;
4639 }
4640
4641 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4642 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4643 error = ENOMEM;
4644 break;
4645 }
4646
4647 totlen = io->pfrio_size * sizeof(struct pfr_table);
4648 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4649 M_TEMP, M_WAITOK);
4650 error = copyin(io->pfrio_buffer, pfrts, totlen);
4651 if (error) {
4652 free(pfrts, M_TEMP);
4653 break;
4654 }
4655 PF_RULES_WLOCK();
4656 error = pfr_add_tables(pfrts, io->pfrio_size,
4657 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4658 PF_RULES_WUNLOCK();
4659 free(pfrts, M_TEMP);
4660 break;
4661 }
4662
4663 case DIOCRDELTABLES: {
4664 struct pfioc_table *io = (struct pfioc_table *)addr;
4665 struct pfr_table *pfrts;
4666 size_t totlen;
4667
4668 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4669 error = ENODEV;
4670 break;
4671 }
4672
4673 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4674 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4675 error = ENOMEM;
4676 break;
4677 }
4678
4679 totlen = io->pfrio_size * sizeof(struct pfr_table);
4680 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4681 M_TEMP, M_WAITOK);
4682 error = copyin(io->pfrio_buffer, pfrts, totlen);
4683 if (error) {
4684 free(pfrts, M_TEMP);
4685 break;
4686 }
4687 PF_RULES_WLOCK();
4688 error = pfr_del_tables(pfrts, io->pfrio_size,
4689 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4690 PF_RULES_WUNLOCK();
4691 free(pfrts, M_TEMP);
4692 break;
4693 }
4694
4695 case DIOCRGETTABLES: {
4696 struct pfioc_table *io = (struct pfioc_table *)addr;
4697 struct pfr_table *pfrts;
4698 size_t totlen;
4699 int n;
4700
4701 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4702 error = ENODEV;
4703 break;
4704 }
4705 PF_RULES_RLOCK();
4706 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4707 if (n < 0) {
4708 PF_RULES_RUNLOCK();
4709 error = EINVAL;
4710 break;
4711 }
4712 io->pfrio_size = min(io->pfrio_size, n);
4713
4714 totlen = io->pfrio_size * sizeof(struct pfr_table);
4715
4716 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4717 M_TEMP, M_NOWAIT | M_ZERO);
4718 if (pfrts == NULL) {
4719 error = ENOMEM;
4720 PF_RULES_RUNLOCK();
4721 break;
4722 }
4723 error = pfr_get_tables(&io->pfrio_table, pfrts,
4724 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4725 PF_RULES_RUNLOCK();
4726 if (error == 0)
4727 error = copyout(pfrts, io->pfrio_buffer, totlen);
4728 free(pfrts, M_TEMP);
4729 break;
4730 }
4731
4732 case DIOCRGETTSTATS: {
4733 struct pfioc_table *io = (struct pfioc_table *)addr;
4734 struct pfr_tstats *pfrtstats;
4735 size_t totlen;
4736 int n;
4737
4738 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4739 error = ENODEV;
4740 break;
4741 }
4742 PF_TABLE_STATS_LOCK();
4743 PF_RULES_RLOCK();
4744 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4745 if (n < 0) {
4746 PF_RULES_RUNLOCK();
4747 PF_TABLE_STATS_UNLOCK();
4748 error = EINVAL;
4749 break;
4750 }
4751 io->pfrio_size = min(io->pfrio_size, n);
4752
4753 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4754 pfrtstats = mallocarray(io->pfrio_size,
4755 sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4756 if (pfrtstats == NULL) {
4757 error = ENOMEM;
4758 PF_RULES_RUNLOCK();
4759 PF_TABLE_STATS_UNLOCK();
4760 break;
4761 }
4762 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4763 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4764 PF_RULES_RUNLOCK();
4765 PF_TABLE_STATS_UNLOCK();
4766 if (error == 0)
4767 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4768 free(pfrtstats, M_TEMP);
4769 break;
4770 }
4771
4772 case DIOCRCLRTSTATS: {
4773 struct pfioc_table *io = (struct pfioc_table *)addr;
4774 struct pfr_table *pfrts;
4775 size_t totlen;
4776
4777 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4778 error = ENODEV;
4779 break;
4780 }
4781
4782 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4783 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4784 /* We used to count tables and use the minimum required
4785 * size, so we didn't fail on overly large requests.
4786 * Keep doing so. */
4787 io->pfrio_size = pf_ioctl_maxcount;
4788 break;
4789 }
4790
4791 totlen = io->pfrio_size * sizeof(struct pfr_table);
4792 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4793 M_TEMP, M_WAITOK);
4794 error = copyin(io->pfrio_buffer, pfrts, totlen);
4795 if (error) {
4796 free(pfrts, M_TEMP);
4797 break;
4798 }
4799
4800 PF_TABLE_STATS_LOCK();
4801 PF_RULES_RLOCK();
4802 error = pfr_clr_tstats(pfrts, io->pfrio_size,
4803 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4804 PF_RULES_RUNLOCK();
4805 PF_TABLE_STATS_UNLOCK();
4806 free(pfrts, M_TEMP);
4807 break;
4808 }
4809
4810 case DIOCRSETTFLAGS: {
4811 struct pfioc_table *io = (struct pfioc_table *)addr;
4812 struct pfr_table *pfrts;
4813 size_t totlen;
4814 int n;
4815
4816 if (io->pfrio_esize != sizeof(struct pfr_table)) {
4817 error = ENODEV;
4818 break;
4819 }
4820
4821 PF_RULES_RLOCK();
4822 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4823 if (n < 0) {
4824 PF_RULES_RUNLOCK();
4825 error = EINVAL;
4826 break;
4827 }
4828
4829 io->pfrio_size = min(io->pfrio_size, n);
4830 PF_RULES_RUNLOCK();
4831
4832 totlen = io->pfrio_size * sizeof(struct pfr_table);
4833 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4834 M_TEMP, M_WAITOK);
4835 error = copyin(io->pfrio_buffer, pfrts, totlen);
4836 if (error) {
4837 free(pfrts, M_TEMP);
4838 break;
4839 }
4840 PF_RULES_WLOCK();
4841 error = pfr_set_tflags(pfrts, io->pfrio_size,
4842 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4843 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4844 PF_RULES_WUNLOCK();
4845 free(pfrts, M_TEMP);
4846 break;
4847 }
4848
4849 case DIOCRCLRADDRS: {
4850 struct pfioc_table *io = (struct pfioc_table *)addr;
4851
4852 if (io->pfrio_esize != 0) {
4853 error = ENODEV;
4854 break;
4855 }
4856 PF_RULES_WLOCK();
4857 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4858 io->pfrio_flags | PFR_FLAG_USERIOCTL);
4859 PF_RULES_WUNLOCK();
4860 break;
4861 }
4862
4863 case DIOCRADDADDRS: {
4864 struct pfioc_table *io = (struct pfioc_table *)addr;
4865 struct pfr_addr *pfras;
4866 size_t totlen;
4867
4868 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4869 error = ENODEV;
4870 break;
4871 }
4872 if (io->pfrio_size < 0 ||
4873 io->pfrio_size > pf_ioctl_maxcount ||
4874 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4875 error = EINVAL;
4876 break;
4877 }
4878 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4879 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4880 M_TEMP, M_WAITOK);
4881 error = copyin(io->pfrio_buffer, pfras, totlen);
4882 if (error) {
4883 free(pfras, M_TEMP);
4884 break;
4885 }
4886 PF_RULES_WLOCK();
4887 error = pfr_add_addrs(&io->pfrio_table, pfras,
4888 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4889 PFR_FLAG_USERIOCTL);
4890 PF_RULES_WUNLOCK();
4891 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4892 error = copyout(pfras, io->pfrio_buffer, totlen);
4893 free(pfras, M_TEMP);
4894 break;
4895 }
4896
4897 case DIOCRDELADDRS: {
4898 struct pfioc_table *io = (struct pfioc_table *)addr;
4899 struct pfr_addr *pfras;
4900 size_t totlen;
4901
4902 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4903 error = ENODEV;
4904 break;
4905 }
4906 if (io->pfrio_size < 0 ||
4907 io->pfrio_size > pf_ioctl_maxcount ||
4908 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4909 error = EINVAL;
4910 break;
4911 }
4912 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4913 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4914 M_TEMP, M_WAITOK);
4915 error = copyin(io->pfrio_buffer, pfras, totlen);
4916 if (error) {
4917 free(pfras, M_TEMP);
4918 break;
4919 }
4920 PF_RULES_WLOCK();
4921 error = pfr_del_addrs(&io->pfrio_table, pfras,
4922 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4923 PFR_FLAG_USERIOCTL);
4924 PF_RULES_WUNLOCK();
4925 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4926 error = copyout(pfras, io->pfrio_buffer, totlen);
4927 free(pfras, M_TEMP);
4928 break;
4929 }
4930
4931 case DIOCRSETADDRS: {
4932 struct pfioc_table *io = (struct pfioc_table *)addr;
4933 struct pfr_addr *pfras;
4934 size_t totlen, count;
4935
4936 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4937 error = ENODEV;
4938 break;
4939 }
4940 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4941 error = EINVAL;
4942 break;
4943 }
4944 count = max(io->pfrio_size, io->pfrio_size2);
4945 if (count > pf_ioctl_maxcount ||
4946 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4947 error = EINVAL;
4948 break;
4949 }
4950 totlen = count * sizeof(struct pfr_addr);
4951 pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4952 M_WAITOK);
4953 error = copyin(io->pfrio_buffer, pfras, totlen);
4954 if (error) {
4955 free(pfras, M_TEMP);
4956 break;
4957 }
4958 PF_RULES_WLOCK();
4959 error = pfr_set_addrs(&io->pfrio_table, pfras,
4960 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4961 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4962 PFR_FLAG_USERIOCTL, 0);
4963 PF_RULES_WUNLOCK();
4964 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4965 error = copyout(pfras, io->pfrio_buffer, totlen);
4966 free(pfras, M_TEMP);
4967 break;
4968 }
4969
4970 case DIOCRGETADDRS: {
4971 struct pfioc_table *io = (struct pfioc_table *)addr;
4972 struct pfr_addr *pfras;
4973 size_t totlen;
4974
4975 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4976 error = ENODEV;
4977 break;
4978 }
4979 if (io->pfrio_size < 0 ||
4980 io->pfrio_size > pf_ioctl_maxcount ||
4981 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4982 error = EINVAL;
4983 break;
4984 }
4985 totlen = io->pfrio_size * sizeof(struct pfr_addr);
4986 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4987 M_TEMP, M_WAITOK | M_ZERO);
4988 PF_RULES_RLOCK();
4989 error = pfr_get_addrs(&io->pfrio_table, pfras,
4990 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4991 PF_RULES_RUNLOCK();
4992 if (error == 0)
4993 error = copyout(pfras, io->pfrio_buffer, totlen);
4994 free(pfras, M_TEMP);
4995 break;
4996 }
4997
4998 case DIOCRGETASTATS: {
4999 struct pfioc_table *io = (struct pfioc_table *)addr;
5000 struct pfr_astats *pfrastats;
5001 size_t totlen;
5002
5003 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
5004 error = ENODEV;
5005 break;
5006 }
5007 if (io->pfrio_size < 0 ||
5008 io->pfrio_size > pf_ioctl_maxcount ||
5009 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
5010 error = EINVAL;
5011 break;
5012 }
5013 totlen = io->pfrio_size * sizeof(struct pfr_astats);
5014 pfrastats = mallocarray(io->pfrio_size,
5015 sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
5016 PF_RULES_RLOCK();
5017 error = pfr_get_astats(&io->pfrio_table, pfrastats,
5018 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5019 PF_RULES_RUNLOCK();
5020 if (error == 0)
5021 error = copyout(pfrastats, io->pfrio_buffer, totlen);
5022 free(pfrastats, M_TEMP);
5023 break;
5024 }
5025
5026 case DIOCRCLRASTATS: {
5027 struct pfioc_table *io = (struct pfioc_table *)addr;
5028 struct pfr_addr *pfras;
5029 size_t totlen;
5030
5031 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5032 error = ENODEV;
5033 break;
5034 }
5035 if (io->pfrio_size < 0 ||
5036 io->pfrio_size > pf_ioctl_maxcount ||
5037 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5038 error = EINVAL;
5039 break;
5040 }
5041 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5042 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5043 M_TEMP, M_WAITOK);
5044 error = copyin(io->pfrio_buffer, pfras, totlen);
5045 if (error) {
5046 free(pfras, M_TEMP);
5047 break;
5048 }
5049 PF_RULES_WLOCK();
5050 error = pfr_clr_astats(&io->pfrio_table, pfras,
5051 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
5052 PFR_FLAG_USERIOCTL);
5053 PF_RULES_WUNLOCK();
5054 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5055 error = copyout(pfras, io->pfrio_buffer, totlen);
5056 free(pfras, M_TEMP);
5057 break;
5058 }
5059
5060 case DIOCRTSTADDRS: {
5061 struct pfioc_table *io = (struct pfioc_table *)addr;
5062 struct pfr_addr *pfras;
5063 size_t totlen;
5064
5065 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5066 error = ENODEV;
5067 break;
5068 }
5069 if (io->pfrio_size < 0 ||
5070 io->pfrio_size > pf_ioctl_maxcount ||
5071 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5072 error = EINVAL;
5073 break;
5074 }
5075 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5076 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5077 M_TEMP, M_WAITOK);
5078 error = copyin(io->pfrio_buffer, pfras, totlen);
5079 if (error) {
5080 free(pfras, M_TEMP);
5081 break;
5082 }
5083 PF_RULES_RLOCK();
5084 error = pfr_tst_addrs(&io->pfrio_table, pfras,
5085 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5086 PFR_FLAG_USERIOCTL);
5087 PF_RULES_RUNLOCK();
5088 if (error == 0)
5089 error = copyout(pfras, io->pfrio_buffer, totlen);
5090 free(pfras, M_TEMP);
5091 break;
5092 }
5093
5094 case DIOCRINADEFINE: {
5095 struct pfioc_table *io = (struct pfioc_table *)addr;
5096 struct pfr_addr *pfras;
5097 size_t totlen;
5098
5099 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5100 error = ENODEV;
5101 break;
5102 }
5103 if (io->pfrio_size < 0 ||
5104 io->pfrio_size > pf_ioctl_maxcount ||
5105 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5106 error = EINVAL;
5107 break;
5108 }
5109 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5110 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5111 M_TEMP, M_WAITOK);
5112 error = copyin(io->pfrio_buffer, pfras, totlen);
5113 if (error) {
5114 free(pfras, M_TEMP);
5115 break;
5116 }
5117 PF_RULES_WLOCK();
5118 error = pfr_ina_define(&io->pfrio_table, pfras,
5119 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5120 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5121 PF_RULES_WUNLOCK();
5122 free(pfras, M_TEMP);
5123 break;
5124 }
5125
5126 case DIOCOSFPADD: {
5127 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5128 PF_RULES_WLOCK();
5129 error = pf_osfp_add(io);
5130 PF_RULES_WUNLOCK();
5131 break;
5132 }
5133
5134 case DIOCOSFPGET: {
5135 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5136 PF_RULES_RLOCK();
5137 error = pf_osfp_get(io);
5138 PF_RULES_RUNLOCK();
5139 break;
5140 }
5141
5142 case DIOCXBEGIN: {
5143 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5144 struct pfioc_trans_e *ioes, *ioe;
5145 size_t totlen;
5146 int i;
5147
5148 if (io->esize != sizeof(*ioe)) {
5149 error = ENODEV;
5150 break;
5151 }
5152 if (io->size < 0 ||
5153 io->size > pf_ioctl_maxcount ||
5154 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5155 error = EINVAL;
5156 break;
5157 }
5158 totlen = sizeof(struct pfioc_trans_e) * io->size;
5159 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5160 M_TEMP, M_WAITOK);
5161 error = copyin(io->array, ioes, totlen);
5162 if (error) {
5163 free(ioes, M_TEMP);
5164 break;
5165 }
5166 PF_RULES_WLOCK();
5167 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5168 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5169 switch (ioe->rs_num) {
5170 case PF_RULESET_ETH:
5171 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5172 PF_RULES_WUNLOCK();
5173 free(ioes, M_TEMP);
5174 goto fail;
5175 }
5176 break;
5177 #ifdef ALTQ
5178 case PF_RULESET_ALTQ:
5179 if (ioe->anchor[0]) {
5180 PF_RULES_WUNLOCK();
5181 free(ioes, M_TEMP);
5182 error = EINVAL;
5183 goto fail;
5184 }
5185 if ((error = pf_begin_altq(&ioe->ticket))) {
5186 PF_RULES_WUNLOCK();
5187 free(ioes, M_TEMP);
5188 goto fail;
5189 }
5190 break;
5191 #endif /* ALTQ */
5192 case PF_RULESET_TABLE:
5193 {
5194 struct pfr_table table;
5195
5196 bzero(&table, sizeof(table));
5197 strlcpy(table.pfrt_anchor, ioe->anchor,
5198 sizeof(table.pfrt_anchor));
5199 if ((error = pfr_ina_begin(&table,
5200 &ioe->ticket, NULL, 0))) {
5201 PF_RULES_WUNLOCK();
5202 free(ioes, M_TEMP);
5203 goto fail;
5204 }
5205 break;
5206 }
5207 default:
5208 if ((error = pf_begin_rules(&ioe->ticket,
5209 ioe->rs_num, ioe->anchor))) {
5210 PF_RULES_WUNLOCK();
5211 free(ioes, M_TEMP);
5212 goto fail;
5213 }
5214 break;
5215 }
5216 }
5217 PF_RULES_WUNLOCK();
5218 error = copyout(ioes, io->array, totlen);
5219 free(ioes, M_TEMP);
5220 break;
5221 }
5222
5223 case DIOCXROLLBACK: {
5224 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5225 struct pfioc_trans_e *ioe, *ioes;
5226 size_t totlen;
5227 int i;
5228
5229 if (io->esize != sizeof(*ioe)) {
5230 error = ENODEV;
5231 break;
5232 }
5233 if (io->size < 0 ||
5234 io->size > pf_ioctl_maxcount ||
5235 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5236 error = EINVAL;
5237 break;
5238 }
5239 totlen = sizeof(struct pfioc_trans_e) * io->size;
5240 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5241 M_TEMP, M_WAITOK);
5242 error = copyin(io->array, ioes, totlen);
5243 if (error) {
5244 free(ioes, M_TEMP);
5245 break;
5246 }
5247 PF_RULES_WLOCK();
5248 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5249 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5250 switch (ioe->rs_num) {
5251 case PF_RULESET_ETH:
5252 if ((error = pf_rollback_eth(ioe->ticket,
5253 ioe->anchor))) {
5254 PF_RULES_WUNLOCK();
5255 free(ioes, M_TEMP);
5256 goto fail; /* really bad */
5257 }
5258 break;
5259 #ifdef ALTQ
5260 case PF_RULESET_ALTQ:
5261 if (ioe->anchor[0]) {
5262 PF_RULES_WUNLOCK();
5263 free(ioes, M_TEMP);
5264 error = EINVAL;
5265 goto fail;
5266 }
5267 if ((error = pf_rollback_altq(ioe->ticket))) {
5268 PF_RULES_WUNLOCK();
5269 free(ioes, M_TEMP);
5270 goto fail; /* really bad */
5271 }
5272 break;
5273 #endif /* ALTQ */
5274 case PF_RULESET_TABLE:
5275 {
5276 struct pfr_table table;
5277
5278 bzero(&table, sizeof(table));
5279 strlcpy(table.pfrt_anchor, ioe->anchor,
5280 sizeof(table.pfrt_anchor));
5281 if ((error = pfr_ina_rollback(&table,
5282 ioe->ticket, NULL, 0))) {
5283 PF_RULES_WUNLOCK();
5284 free(ioes, M_TEMP);
5285 goto fail; /* really bad */
5286 }
5287 break;
5288 }
5289 default:
5290 if ((error = pf_rollback_rules(ioe->ticket,
5291 ioe->rs_num, ioe->anchor))) {
5292 PF_RULES_WUNLOCK();
5293 free(ioes, M_TEMP);
5294 goto fail; /* really bad */
5295 }
5296 break;
5297 }
5298 }
5299 PF_RULES_WUNLOCK();
5300 free(ioes, M_TEMP);
5301 break;
5302 }
5303
5304 case DIOCXCOMMIT: {
5305 struct pfioc_trans *io = (struct pfioc_trans *)addr;
5306 struct pfioc_trans_e *ioe, *ioes;
5307 struct pf_kruleset *rs;
5308 struct pf_keth_ruleset *ers;
5309 size_t totlen;
5310 int i;
5311
5312 if (io->esize != sizeof(*ioe)) {
5313 error = ENODEV;
5314 break;
5315 }
5316
5317 if (io->size < 0 ||
5318 io->size > pf_ioctl_maxcount ||
5319 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5320 error = EINVAL;
5321 break;
5322 }
5323
5324 totlen = sizeof(struct pfioc_trans_e) * io->size;
5325 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5326 M_TEMP, M_WAITOK);
5327 error = copyin(io->array, ioes, totlen);
5328 if (error) {
5329 free(ioes, M_TEMP);
5330 break;
5331 }
5332 PF_RULES_WLOCK();
5333 /* First makes sure everything will succeed. */
5334 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5335 ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5336 switch (ioe->rs_num) {
5337 case PF_RULESET_ETH:
5338 ers = pf_find_keth_ruleset(ioe->anchor);
5339 if (ers == NULL || ioe->ticket == 0 ||
5340 ioe->ticket != ers->inactive.ticket) {
5341 PF_RULES_WUNLOCK();
5342 free(ioes, M_TEMP);
5343 error = EINVAL;
5344 goto fail;
5345 }
5346 break;
5347 #ifdef ALTQ
5348 case PF_RULESET_ALTQ:
5349 if (ioe->anchor[0]) {
5350 PF_RULES_WUNLOCK();
5351 free(ioes, M_TEMP);
5352 error = EINVAL;
5353 goto fail;
5354 }
5355 if (!V_altqs_inactive_open || ioe->ticket !=
5356 V_ticket_altqs_inactive) {
5357 PF_RULES_WUNLOCK();
5358 free(ioes, M_TEMP);
5359 error = EBUSY;
5360 goto fail;
5361 }
5362 break;
5363 #endif /* ALTQ */
5364 case PF_RULESET_TABLE:
5365 rs = pf_find_kruleset(ioe->anchor);
5366 if (rs == NULL || !rs->topen || ioe->ticket !=
5367 rs->tticket) {
5368 PF_RULES_WUNLOCK();
5369 free(ioes, M_TEMP);
5370 error = EBUSY;
5371 goto fail;
5372 }
5373 break;
5374 default:
5375 if (ioe->rs_num < 0 || ioe->rs_num >=
5376 PF_RULESET_MAX) {
5377 PF_RULES_WUNLOCK();
5378 free(ioes, M_TEMP);
5379 error = EINVAL;
5380 goto fail;
5381 }
5382 rs = pf_find_kruleset(ioe->anchor);
5383 if (rs == NULL ||
5384 !rs->rules[ioe->rs_num].inactive.open ||
5385 rs->rules[ioe->rs_num].inactive.ticket !=
5386 ioe->ticket) {
5387 PF_RULES_WUNLOCK();
5388 free(ioes, M_TEMP);
5389 error = EBUSY;
5390 goto fail;
5391 }
5392 break;
5393 }
5394 }
5395 /* Now do the commit - no errors should happen here. */
5396 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5397 switch (ioe->rs_num) {
5398 case PF_RULESET_ETH:
5399 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5400 PF_RULES_WUNLOCK();
5401 free(ioes, M_TEMP);
5402 goto fail; /* really bad */
5403 }
5404 break;
5405 #ifdef ALTQ
5406 case PF_RULESET_ALTQ:
5407 if ((error = pf_commit_altq(ioe->ticket))) {
5408 PF_RULES_WUNLOCK();
5409 free(ioes, M_TEMP);
5410 goto fail; /* really bad */
5411 }
5412 break;
5413 #endif /* ALTQ */
5414 case PF_RULESET_TABLE:
5415 {
5416 struct pfr_table table;
5417
5418 bzero(&table, sizeof(table));
5419 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
5420 sizeof(table.pfrt_anchor));
5421 if ((error = pfr_ina_commit(&table,
5422 ioe->ticket, NULL, NULL, 0))) {
5423 PF_RULES_WUNLOCK();
5424 free(ioes, M_TEMP);
5425 goto fail; /* really bad */
5426 }
5427 break;
5428 }
5429 default:
5430 if ((error = pf_commit_rules(ioe->ticket,
5431 ioe->rs_num, ioe->anchor))) {
5432 PF_RULES_WUNLOCK();
5433 free(ioes, M_TEMP);
5434 goto fail; /* really bad */
5435 }
5436 break;
5437 }
5438 }
5439 PF_RULES_WUNLOCK();
5440
5441 /* Only hook into EtherNet taffic if we've got rules for it. */
5442 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5443 hook_pf_eth();
5444 else
5445 dehook_pf_eth();
5446
5447 free(ioes, M_TEMP);
5448 break;
5449 }
5450
5451 case DIOCGETSRCNODES: {
5452 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
5453 struct pf_srchash *sh;
5454 struct pf_ksrc_node *n;
5455 struct pf_src_node *p, *pstore;
5456 uint32_t i, nr = 0;
5457
5458 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5459 i++, sh++) {
5460 PF_HASHROW_LOCK(sh);
5461 LIST_FOREACH(n, &sh->nodes, entry)
5462 nr++;
5463 PF_HASHROW_UNLOCK(sh);
5464 }
5465
5466 psn->psn_len = min(psn->psn_len,
5467 sizeof(struct pf_src_node) * nr);
5468
5469 if (psn->psn_len == 0) {
5470 psn->psn_len = sizeof(struct pf_src_node) * nr;
5471 break;
5472 }
5473
5474 nr = 0;
5475
5476 p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5477 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5478 i++, sh++) {
5479 PF_HASHROW_LOCK(sh);
5480 LIST_FOREACH(n, &sh->nodes, entry) {
5481
5482 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5483 break;
5484
5485 pf_src_node_copy(n, p);
5486
5487 p++;
5488 nr++;
5489 }
5490 PF_HASHROW_UNLOCK(sh);
5491 }
5492 error = copyout(pstore, psn->psn_src_nodes,
5493 sizeof(struct pf_src_node) * nr);
5494 if (error) {
5495 free(pstore, M_TEMP);
5496 break;
5497 }
5498 psn->psn_len = sizeof(struct pf_src_node) * nr;
5499 free(pstore, M_TEMP);
5500 break;
5501 }
5502
5503 case DIOCCLRSRCNODES: {
5504 pf_kill_srcnodes(NULL);
5505 break;
5506 }
5507
5508 case DIOCKILLSRCNODES:
5509 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5510 break;
5511
5512 #ifdef COMPAT_FREEBSD13
5513 case DIOCKEEPCOUNTERS_FREEBSD13:
5514 #endif
5515 case DIOCKEEPCOUNTERS:
5516 error = pf_keepcounters((struct pfioc_nv *)addr);
5517 break;
5518
5519 case DIOCGETSYNCOOKIES:
5520 error = pf_get_syncookies((struct pfioc_nv *)addr);
5521 break;
5522
5523 case DIOCSETSYNCOOKIES:
5524 error = pf_set_syncookies((struct pfioc_nv *)addr);
5525 break;
5526
5527 case DIOCSETHOSTID: {
5528 u_int32_t *hostid = (u_int32_t *)addr;
5529
5530 PF_RULES_WLOCK();
5531 if (*hostid == 0)
5532 V_pf_status.hostid = arc4random();
5533 else
5534 V_pf_status.hostid = *hostid;
5535 PF_RULES_WUNLOCK();
5536 break;
5537 }
5538
5539 case DIOCOSFPFLUSH:
5540 PF_RULES_WLOCK();
5541 pf_osfp_flush();
5542 PF_RULES_WUNLOCK();
5543 break;
5544
5545 case DIOCIGETIFACES: {
5546 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5547 struct pfi_kif *ifstore;
5548 size_t bufsiz;
5549
5550 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5551 error = ENODEV;
5552 break;
5553 }
5554
5555 if (io->pfiio_size < 0 ||
5556 io->pfiio_size > pf_ioctl_maxcount ||
5557 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5558 error = EINVAL;
5559 break;
5560 }
5561
5562 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5563
5564 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5565 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5566 M_TEMP, M_WAITOK | M_ZERO);
5567
5568 PF_RULES_RLOCK();
5569 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5570 PF_RULES_RUNLOCK();
5571 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5572 free(ifstore, M_TEMP);
5573 break;
5574 }
5575
5576 case DIOCSETIFFLAG: {
5577 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5578
5579 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5580
5581 PF_RULES_WLOCK();
5582 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5583 PF_RULES_WUNLOCK();
5584 break;
5585 }
5586
5587 case DIOCCLRIFFLAG: {
5588 struct pfioc_iface *io = (struct pfioc_iface *)addr;
5589
5590 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5591
5592 PF_RULES_WLOCK();
5593 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5594 PF_RULES_WUNLOCK();
5595 break;
5596 }
5597
5598 case DIOCSETREASS: {
5599 u_int32_t *reass = (u_int32_t *)addr;
5600
5601 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5602 /* Removal of DF flag without reassembly enabled is not a
5603 * valid combination. Disable reassembly in such case. */
5604 if (!(V_pf_status.reass & PF_REASS_ENABLED))
5605 V_pf_status.reass = 0;
5606 break;
5607 }
5608
5609 default:
5610 error = ENODEV;
5611 break;
5612 }
5613 fail:
5614 CURVNET_RESTORE();
5615
5616 #undef ERROUT_IOCTL
5617
5618 return (error);
5619 }
5620
5621 void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)5622 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5623 {
5624 bzero(sp, sizeof(union pfsync_state_union));
5625
5626 /* copy from state key */
5627 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5628 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5629 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5630 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5631 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5632 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5633 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5634 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5635 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5636 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5637
5638 /* copy from state */
5639 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5640 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5641 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5642 sp->pfs_1301.expire = pf_state_expires(st);
5643 if (sp->pfs_1301.expire <= time_uptime)
5644 sp->pfs_1301.expire = htonl(0);
5645 else
5646 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5647
5648 sp->pfs_1301.direction = st->direction;
5649 sp->pfs_1301.log = st->act.log;
5650 sp->pfs_1301.timeout = st->timeout;
5651
5652 switch (msg_version) {
5653 case PFSYNC_MSG_VERSION_1301:
5654 sp->pfs_1301.state_flags = st->state_flags;
5655 break;
5656 case PFSYNC_MSG_VERSION_1400:
5657 sp->pfs_1400.state_flags = htons(st->state_flags);
5658 sp->pfs_1400.qid = htons(st->act.qid);
5659 sp->pfs_1400.pqid = htons(st->act.pqid);
5660 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5661 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5662 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5663 sp->pfs_1400.min_ttl = st->act.min_ttl;
5664 sp->pfs_1400.set_tos = st->act.set_tos;
5665 sp->pfs_1400.max_mss = htons(st->act.max_mss);
5666 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5667 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5668 sp->pfs_1400.rt = st->act.rt;
5669 if (st->act.rt_kif)
5670 strlcpy(sp->pfs_1400.rt_ifname,
5671 st->act.rt_kif->pfik_name,
5672 sizeof(sp->pfs_1400.rt_ifname));
5673 break;
5674 default:
5675 panic("%s: Unsupported pfsync_msg_version %d",
5676 __func__, msg_version);
5677 }
5678
5679 /*
5680 * XXX Why do we bother pfsyncing source node information if source
5681 * nodes are not synced? Showing users that there is source tracking
5682 * when there is none seems useless.
5683 */
5684 if (st->sns[PF_SN_LIMIT] != NULL)
5685 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5686 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
5687 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5688
5689 sp->pfs_1301.id = st->id;
5690 sp->pfs_1301.creatorid = st->creatorid;
5691 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5692 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5693
5694 if (st->rule == NULL)
5695 sp->pfs_1301.rule = htonl(-1);
5696 else
5697 sp->pfs_1301.rule = htonl(st->rule->nr);
5698 if (st->anchor == NULL)
5699 sp->pfs_1301.anchor = htonl(-1);
5700 else
5701 sp->pfs_1301.anchor = htonl(st->anchor->nr);
5702 if (st->nat_rule == NULL)
5703 sp->pfs_1301.nat_rule = htonl(-1);
5704 else
5705 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
5706
5707 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5708 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5709 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5710 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5711 }
5712
5713 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)5714 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5715 {
5716 bzero(sp, sizeof(*sp));
5717
5718 sp->version = PF_STATE_VERSION;
5719
5720 /* copy from state key */
5721 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5722 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5723 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5724 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5725 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5726 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5727 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5728 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5729 sp->proto = st->key[PF_SK_WIRE]->proto;
5730 sp->af = st->key[PF_SK_WIRE]->af;
5731
5732 /* copy from state */
5733 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5734 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5735 sizeof(sp->orig_ifname));
5736 bcopy(&st->act.rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5737 sp->creation = htonl(time_uptime - (st->creation / 1000));
5738 sp->expire = pf_state_expires(st);
5739 if (sp->expire <= time_uptime)
5740 sp->expire = htonl(0);
5741 else
5742 sp->expire = htonl(sp->expire - time_uptime);
5743
5744 sp->direction = st->direction;
5745 sp->log = st->act.log;
5746 sp->timeout = st->timeout;
5747 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5748 sp->state_flags_compat = st->state_flags;
5749 sp->state_flags = htons(st->state_flags);
5750 if (st->sns[PF_SN_LIMIT] != NULL)
5751 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5752 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL)
5753 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5754 sp->id = st->id;
5755 sp->creatorid = st->creatorid;
5756 pf_state_peer_hton(&st->src, &sp->src);
5757 pf_state_peer_hton(&st->dst, &sp->dst);
5758
5759 if (st->rule == NULL)
5760 sp->rule = htonl(-1);
5761 else
5762 sp->rule = htonl(st->rule->nr);
5763 if (st->anchor == NULL)
5764 sp->anchor = htonl(-1);
5765 else
5766 sp->anchor = htonl(st->anchor->nr);
5767 if (st->nat_rule == NULL)
5768 sp->nat_rule = htonl(-1);
5769 else
5770 sp->nat_rule = htonl(st->nat_rule->nr);
5771
5772 sp->packets[0] = st->packets[0];
5773 sp->packets[1] = st->packets[1];
5774 sp->bytes[0] = st->bytes[0];
5775 sp->bytes[1] = st->bytes[1];
5776
5777 sp->qid = htons(st->act.qid);
5778 sp->pqid = htons(st->act.pqid);
5779 sp->dnpipe = htons(st->act.dnpipe);
5780 sp->dnrpipe = htons(st->act.dnrpipe);
5781 sp->rtableid = htonl(st->act.rtableid);
5782 sp->min_ttl = st->act.min_ttl;
5783 sp->set_tos = st->act.set_tos;
5784 sp->max_mss = htons(st->act.max_mss);
5785 sp->rt = st->act.rt;
5786 if (st->act.rt_kif)
5787 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
5788 sizeof(sp->rt_ifname));
5789 sp->set_prio[0] = st->act.set_prio[0];
5790 sp->set_prio[1] = st->act.set_prio[1];
5791
5792 }
5793
5794 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)5795 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5796 {
5797 struct pfr_ktable *kt;
5798
5799 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5800
5801 kt = aw->p.tbl;
5802 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5803 kt = kt->pfrkt_root;
5804 aw->p.tbl = NULL;
5805 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5806 kt->pfrkt_cnt : -1;
5807 }
5808
5809 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)5810 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5811 size_t number, char **names)
5812 {
5813 nvlist_t *nvc;
5814
5815 nvc = nvlist_create(0);
5816 if (nvc == NULL)
5817 return (ENOMEM);
5818
5819 for (int i = 0; i < number; i++) {
5820 nvlist_append_number_array(nvc, "counters",
5821 counter_u64_fetch(counters[i]));
5822 nvlist_append_string_array(nvc, "names",
5823 names[i]);
5824 nvlist_append_number_array(nvc, "ids",
5825 i);
5826 }
5827 nvlist_add_nvlist(nvl, name, nvc);
5828 nvlist_destroy(nvc);
5829
5830 return (0);
5831 }
5832
5833 static int
pf_getstatus(struct pfioc_nv * nv)5834 pf_getstatus(struct pfioc_nv *nv)
5835 {
5836 nvlist_t *nvl = NULL, *nvc = NULL;
5837 void *nvlpacked = NULL;
5838 int error;
5839 struct pf_status s;
5840 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5841 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5842 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5843 PF_RULES_RLOCK_TRACKER;
5844
5845 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
5846
5847 PF_RULES_RLOCK();
5848
5849 nvl = nvlist_create(0);
5850 if (nvl == NULL)
5851 ERROUT(ENOMEM);
5852
5853 nvlist_add_bool(nvl, "running", V_pf_status.running);
5854 nvlist_add_number(nvl, "since", V_pf_status.since);
5855 nvlist_add_number(nvl, "debug", V_pf_status.debug);
5856 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5857 nvlist_add_number(nvl, "states", V_pf_status.states);
5858 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5859 nvlist_add_number(nvl, "reass", V_pf_status.reass);
5860 nvlist_add_bool(nvl, "syncookies_active",
5861 V_pf_status.syncookies_active);
5862 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5863
5864 /* counters */
5865 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5866 PFRES_MAX, pf_reasons);
5867 if (error != 0)
5868 ERROUT(error);
5869
5870 /* lcounters */
5871 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5872 KLCNT_MAX, pf_lcounter);
5873 if (error != 0)
5874 ERROUT(error);
5875
5876 /* fcounters */
5877 nvc = nvlist_create(0);
5878 if (nvc == NULL)
5879 ERROUT(ENOMEM);
5880
5881 for (int i = 0; i < FCNT_MAX; i++) {
5882 nvlist_append_number_array(nvc, "counters",
5883 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5884 nvlist_append_string_array(nvc, "names",
5885 pf_fcounter[i]);
5886 nvlist_append_number_array(nvc, "ids",
5887 i);
5888 }
5889 nvlist_add_nvlist(nvl, "fcounters", nvc);
5890 nvlist_destroy(nvc);
5891 nvc = NULL;
5892
5893 /* scounters */
5894 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5895 SCNT_MAX, pf_fcounter);
5896 if (error != 0)
5897 ERROUT(error);
5898
5899 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5900 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5901 PF_MD5_DIGEST_LENGTH);
5902
5903 pfi_update_status(V_pf_status.ifname, &s);
5904
5905 /* pcounters / bcounters */
5906 for (int i = 0; i < 2; i++) {
5907 for (int j = 0; j < 2; j++) {
5908 for (int k = 0; k < 2; k++) {
5909 nvlist_append_number_array(nvl, "pcounters",
5910 s.pcounters[i][j][k]);
5911 }
5912 nvlist_append_number_array(nvl, "bcounters",
5913 s.bcounters[i][j]);
5914 }
5915 }
5916
5917 nvlpacked = nvlist_pack(nvl, &nv->len);
5918 if (nvlpacked == NULL)
5919 ERROUT(ENOMEM);
5920
5921 if (nv->size == 0)
5922 ERROUT(0);
5923 else if (nv->size < nv->len)
5924 ERROUT(ENOSPC);
5925
5926 PF_RULES_RUNLOCK();
5927 error = copyout(nvlpacked, nv->data, nv->len);
5928 goto done;
5929
5930 #undef ERROUT
5931 errout:
5932 PF_RULES_RUNLOCK();
5933 done:
5934 free(nvlpacked, M_NVLIST);
5935 nvlist_destroy(nvc);
5936 nvlist_destroy(nvl);
5937
5938 return (error);
5939 }
5940
5941 /*
5942 * XXX - Check for version mismatch!!!
5943 */
5944 static void
pf_clear_all_states(void)5945 pf_clear_all_states(void)
5946 {
5947 struct epoch_tracker et;
5948 struct pf_kstate *s;
5949 u_int i;
5950
5951 NET_EPOCH_ENTER(et);
5952 for (i = 0; i <= V_pf_hashmask; i++) {
5953 struct pf_idhash *ih = &V_pf_idhash[i];
5954 relock:
5955 PF_HASHROW_LOCK(ih);
5956 LIST_FOREACH(s, &ih->states, entry) {
5957 s->timeout = PFTM_PURGE;
5958 /* Don't send out individual delete messages. */
5959 s->state_flags |= PFSTATE_NOSYNC;
5960 pf_unlink_state(s);
5961 goto relock;
5962 }
5963 PF_HASHROW_UNLOCK(ih);
5964 }
5965 NET_EPOCH_EXIT(et);
5966 }
5967
5968 static int
pf_clear_tables(void)5969 pf_clear_tables(void)
5970 {
5971 struct pfioc_table io;
5972 int error;
5973
5974 bzero(&io, sizeof(io));
5975 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5976
5977 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5978 io.pfrio_flags);
5979
5980 return (error);
5981 }
5982
5983 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)5984 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5985 {
5986 struct pf_ksrc_node_list kill;
5987 u_int killed;
5988
5989 LIST_INIT(&kill);
5990 for (int i = 0; i <= V_pf_srchashmask; i++) {
5991 struct pf_srchash *sh = &V_pf_srchash[i];
5992 struct pf_ksrc_node *sn, *tmp;
5993
5994 PF_HASHROW_LOCK(sh);
5995 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5996 if (psnk == NULL ||
5997 (PF_MATCHA(psnk->psnk_src.neg,
5998 &psnk->psnk_src.addr.v.a.addr,
5999 &psnk->psnk_src.addr.v.a.mask,
6000 &sn->addr, sn->af) &&
6001 PF_MATCHA(psnk->psnk_dst.neg,
6002 &psnk->psnk_dst.addr.v.a.addr,
6003 &psnk->psnk_dst.addr.v.a.mask,
6004 &sn->raddr, sn->af))) {
6005 pf_unlink_src_node(sn);
6006 LIST_INSERT_HEAD(&kill, sn, entry);
6007 sn->expire = 1;
6008 }
6009 PF_HASHROW_UNLOCK(sh);
6010 }
6011
6012 for (int i = 0; i <= V_pf_hashmask; i++) {
6013 struct pf_idhash *ih = &V_pf_idhash[i];
6014 struct pf_kstate *s;
6015
6016 PF_HASHROW_LOCK(ih);
6017 LIST_FOREACH(s, &ih->states, entry) {
6018 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX;
6019 sn_type++) {
6020 if (s->sns[sn_type] &&
6021 s->sns[sn_type]->expire == 1) {
6022 s->sns[sn_type] = NULL;
6023 }
6024 }
6025 }
6026 PF_HASHROW_UNLOCK(ih);
6027 }
6028
6029 killed = pf_free_src_nodes(&kill);
6030
6031 if (psnk != NULL)
6032 psnk->psnk_killed = killed;
6033 }
6034
6035 static int
pf_keepcounters(struct pfioc_nv * nv)6036 pf_keepcounters(struct pfioc_nv *nv)
6037 {
6038 nvlist_t *nvl = NULL;
6039 void *nvlpacked = NULL;
6040 int error = 0;
6041
6042 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6043
6044 if (nv->len > pf_ioctl_maxcount)
6045 ERROUT(ENOMEM);
6046
6047 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6048 error = copyin(nv->data, nvlpacked, nv->len);
6049 if (error)
6050 ERROUT(error);
6051
6052 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6053 if (nvl == NULL)
6054 ERROUT(EBADMSG);
6055
6056 if (! nvlist_exists_bool(nvl, "keep_counters"))
6057 ERROUT(EBADMSG);
6058
6059 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6060
6061 on_error:
6062 nvlist_destroy(nvl);
6063 free(nvlpacked, M_NVLIST);
6064 return (error);
6065 }
6066
6067 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)6068 pf_clear_states(const struct pf_kstate_kill *kill)
6069 {
6070 struct pf_state_key_cmp match_key;
6071 struct pf_kstate *s;
6072 struct pfi_kkif *kif;
6073 int idx;
6074 unsigned int killed = 0, dir;
6075
6076 NET_EPOCH_ASSERT();
6077
6078 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6079 struct pf_idhash *ih = &V_pf_idhash[i];
6080
6081 relock_DIOCCLRSTATES:
6082 PF_HASHROW_LOCK(ih);
6083 LIST_FOREACH(s, &ih->states, entry) {
6084 /* For floating states look at the original kif. */
6085 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6086
6087 if (kill->psk_ifname[0] &&
6088 strcmp(kill->psk_ifname,
6089 kif->pfik_name))
6090 continue;
6091
6092 if (kill->psk_kill_match) {
6093 bzero(&match_key, sizeof(match_key));
6094
6095 if (s->direction == PF_OUT) {
6096 dir = PF_IN;
6097 idx = PF_SK_STACK;
6098 } else {
6099 dir = PF_OUT;
6100 idx = PF_SK_WIRE;
6101 }
6102
6103 match_key.af = s->key[idx]->af;
6104 match_key.proto = s->key[idx]->proto;
6105 PF_ACPY(&match_key.addr[0],
6106 &s->key[idx]->addr[1], match_key.af);
6107 match_key.port[0] = s->key[idx]->port[1];
6108 PF_ACPY(&match_key.addr[1],
6109 &s->key[idx]->addr[0], match_key.af);
6110 match_key.port[1] = s->key[idx]->port[0];
6111 }
6112
6113 /*
6114 * Don't send out individual
6115 * delete messages.
6116 */
6117 s->state_flags |= PFSTATE_NOSYNC;
6118 pf_unlink_state(s);
6119 killed++;
6120
6121 if (kill->psk_kill_match)
6122 killed += pf_kill_matching_state(&match_key,
6123 dir);
6124
6125 goto relock_DIOCCLRSTATES;
6126 }
6127 PF_HASHROW_UNLOCK(ih);
6128 }
6129
6130 if (V_pfsync_clear_states_ptr != NULL)
6131 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6132
6133 return (killed);
6134 }
6135
6136 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)6137 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6138 {
6139 struct pf_kstate *s;
6140
6141 NET_EPOCH_ASSERT();
6142 if (kill->psk_pfcmp.id) {
6143 if (kill->psk_pfcmp.creatorid == 0)
6144 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6145 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6146 kill->psk_pfcmp.creatorid))) {
6147 pf_unlink_state(s);
6148 *killed = 1;
6149 }
6150 return;
6151 }
6152
6153 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6154 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6155 }
6156
6157 static int
pf_killstates_nv(struct pfioc_nv * nv)6158 pf_killstates_nv(struct pfioc_nv *nv)
6159 {
6160 struct pf_kstate_kill kill;
6161 struct epoch_tracker et;
6162 nvlist_t *nvl = NULL;
6163 void *nvlpacked = NULL;
6164 int error = 0;
6165 unsigned int killed = 0;
6166
6167 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6168
6169 if (nv->len > pf_ioctl_maxcount)
6170 ERROUT(ENOMEM);
6171
6172 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6173 error = copyin(nv->data, nvlpacked, nv->len);
6174 if (error)
6175 ERROUT(error);
6176
6177 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6178 if (nvl == NULL)
6179 ERROUT(EBADMSG);
6180
6181 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6182 if (error)
6183 ERROUT(error);
6184
6185 NET_EPOCH_ENTER(et);
6186 pf_killstates(&kill, &killed);
6187 NET_EPOCH_EXIT(et);
6188
6189 free(nvlpacked, M_NVLIST);
6190 nvlpacked = NULL;
6191 nvlist_destroy(nvl);
6192 nvl = nvlist_create(0);
6193 if (nvl == NULL)
6194 ERROUT(ENOMEM);
6195
6196 nvlist_add_number(nvl, "killed", killed);
6197
6198 nvlpacked = nvlist_pack(nvl, &nv->len);
6199 if (nvlpacked == NULL)
6200 ERROUT(ENOMEM);
6201
6202 if (nv->size == 0)
6203 ERROUT(0);
6204 else if (nv->size < nv->len)
6205 ERROUT(ENOSPC);
6206
6207 error = copyout(nvlpacked, nv->data, nv->len);
6208
6209 on_error:
6210 nvlist_destroy(nvl);
6211 free(nvlpacked, M_NVLIST);
6212 return (error);
6213 }
6214
6215 static int
pf_clearstates_nv(struct pfioc_nv * nv)6216 pf_clearstates_nv(struct pfioc_nv *nv)
6217 {
6218 struct pf_kstate_kill kill;
6219 struct epoch_tracker et;
6220 nvlist_t *nvl = NULL;
6221 void *nvlpacked = NULL;
6222 int error = 0;
6223 unsigned int killed;
6224
6225 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
6226
6227 if (nv->len > pf_ioctl_maxcount)
6228 ERROUT(ENOMEM);
6229
6230 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6231 error = copyin(nv->data, nvlpacked, nv->len);
6232 if (error)
6233 ERROUT(error);
6234
6235 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6236 if (nvl == NULL)
6237 ERROUT(EBADMSG);
6238
6239 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6240 if (error)
6241 ERROUT(error);
6242
6243 NET_EPOCH_ENTER(et);
6244 killed = pf_clear_states(&kill);
6245 NET_EPOCH_EXIT(et);
6246
6247 free(nvlpacked, M_NVLIST);
6248 nvlpacked = NULL;
6249 nvlist_destroy(nvl);
6250 nvl = nvlist_create(0);
6251 if (nvl == NULL)
6252 ERROUT(ENOMEM);
6253
6254 nvlist_add_number(nvl, "killed", killed);
6255
6256 nvlpacked = nvlist_pack(nvl, &nv->len);
6257 if (nvlpacked == NULL)
6258 ERROUT(ENOMEM);
6259
6260 if (nv->size == 0)
6261 ERROUT(0);
6262 else if (nv->size < nv->len)
6263 ERROUT(ENOSPC);
6264
6265 error = copyout(nvlpacked, nv->data, nv->len);
6266
6267 #undef ERROUT
6268 on_error:
6269 nvlist_destroy(nvl);
6270 free(nvlpacked, M_NVLIST);
6271 return (error);
6272 }
6273
6274 static int
pf_getstate(struct pfioc_nv * nv)6275 pf_getstate(struct pfioc_nv *nv)
6276 {
6277 nvlist_t *nvl = NULL, *nvls;
6278 void *nvlpacked = NULL;
6279 struct pf_kstate *s = NULL;
6280 int error = 0;
6281 uint64_t id, creatorid;
6282
6283 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6284
6285 if (nv->len > pf_ioctl_maxcount)
6286 ERROUT(ENOMEM);
6287
6288 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6289 error = copyin(nv->data, nvlpacked, nv->len);
6290 if (error)
6291 ERROUT(error);
6292
6293 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6294 if (nvl == NULL)
6295 ERROUT(EBADMSG);
6296
6297 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6298 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6299
6300 s = pf_find_state_byid(id, creatorid);
6301 if (s == NULL)
6302 ERROUT(ENOENT);
6303
6304 free(nvlpacked, M_NVLIST);
6305 nvlpacked = NULL;
6306 nvlist_destroy(nvl);
6307 nvl = nvlist_create(0);
6308 if (nvl == NULL)
6309 ERROUT(ENOMEM);
6310
6311 nvls = pf_state_to_nvstate(s);
6312 if (nvls == NULL)
6313 ERROUT(ENOMEM);
6314
6315 nvlist_add_nvlist(nvl, "state", nvls);
6316 nvlist_destroy(nvls);
6317
6318 nvlpacked = nvlist_pack(nvl, &nv->len);
6319 if (nvlpacked == NULL)
6320 ERROUT(ENOMEM);
6321
6322 if (nv->size == 0)
6323 ERROUT(0);
6324 else if (nv->size < nv->len)
6325 ERROUT(ENOSPC);
6326
6327 error = copyout(nvlpacked, nv->data, nv->len);
6328
6329 #undef ERROUT
6330 errout:
6331 if (s != NULL)
6332 PF_STATE_UNLOCK(s);
6333 free(nvlpacked, M_NVLIST);
6334 nvlist_destroy(nvl);
6335 return (error);
6336 }
6337
6338 /*
6339 * XXX - Check for version mismatch!!!
6340 */
6341
6342 /*
6343 * Duplicate pfctl -Fa operation to get rid of as much as we can.
6344 */
6345 static int
shutdown_pf(void)6346 shutdown_pf(void)
6347 {
6348 int error = 0;
6349 u_int32_t t[5];
6350 char nn = '\0';
6351 struct pf_kanchor *anchor;
6352 struct pf_keth_anchor *eth_anchor;
6353 int rs_num;
6354
6355 do {
6356 /* Unlink rules of all user defined anchors */
6357 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6358 /* Wildcard based anchors may not have a respective
6359 * explicit anchor rule or they may be left empty
6360 * without rules. It leads to anchor.refcnt=0, and the
6361 * rest of the logic does not expect it. */
6362 if (anchor->refcnt == 0)
6363 anchor->refcnt = 1;
6364 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6365 if ((error = pf_begin_rules(&t[rs_num], rs_num,
6366 anchor->path)) != 0) {
6367 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6368 "anchor.path=%s rs_num=%d\n",
6369 anchor->path, rs_num));
6370 goto error; /* XXX: rollback? */
6371 }
6372 }
6373 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6374 error = pf_commit_rules(t[rs_num], rs_num,
6375 anchor->path);
6376 MPASS(error == 0);
6377 }
6378 }
6379
6380 /* Unlink rules of all user defined ether anchors */
6381 RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6382 &V_pf_keth_anchors) {
6383 /* Wildcard based anchors may not have a respective
6384 * explicit anchor rule or they may be left empty
6385 * without rules. It leads to anchor.refcnt=0, and the
6386 * rest of the logic does not expect it. */
6387 if (eth_anchor->refcnt == 0)
6388 eth_anchor->refcnt = 1;
6389 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6390 != 0) {
6391 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6392 "anchor.path=%s\n", eth_anchor->path));
6393 goto error;
6394 }
6395 error = pf_commit_eth(t[0], eth_anchor->path);
6396 MPASS(error == 0);
6397 }
6398
6399 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6400 != 0) {
6401 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6402 break;
6403 }
6404 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6405 != 0) {
6406 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6407 break; /* XXX: rollback? */
6408 }
6409 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6410 != 0) {
6411 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6412 break; /* XXX: rollback? */
6413 }
6414 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6415 != 0) {
6416 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6417 break; /* XXX: rollback? */
6418 }
6419 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6420 != 0) {
6421 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6422 break; /* XXX: rollback? */
6423 }
6424
6425 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6426 MPASS(error == 0);
6427 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6428 MPASS(error == 0);
6429 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6430 MPASS(error == 0);
6431 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6432 MPASS(error == 0);
6433 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6434 MPASS(error == 0);
6435
6436 if ((error = pf_clear_tables()) != 0)
6437 break;
6438
6439 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6440 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6441 break;
6442 }
6443 error = pf_commit_eth(t[0], &nn);
6444 MPASS(error == 0);
6445
6446 #ifdef ALTQ
6447 if ((error = pf_begin_altq(&t[0])) != 0) {
6448 DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6449 break;
6450 }
6451 pf_commit_altq(t[0]);
6452 #endif
6453
6454 pf_clear_all_states();
6455
6456 pf_kill_srcnodes(NULL);
6457
6458 /* status does not use malloced mem so no need to cleanup */
6459 /* fingerprints and interfaces have their own cleanup code */
6460 } while(0);
6461
6462 error:
6463 return (error);
6464 }
6465
6466 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)6467 pf_check_return(int chk, struct mbuf **m)
6468 {
6469
6470 switch (chk) {
6471 case PF_PASS:
6472 if (*m == NULL)
6473 return (PFIL_CONSUMED);
6474 else
6475 return (PFIL_PASS);
6476 break;
6477 default:
6478 if (*m != NULL) {
6479 m_freem(*m);
6480 *m = NULL;
6481 }
6482 return (PFIL_DROPPED);
6483 }
6484 }
6485
6486 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6487 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6488 void *ruleset __unused, struct inpcb *inp)
6489 {
6490 int chk;
6491
6492 CURVNET_ASSERT_SET();
6493
6494 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6495
6496 return (pf_check_return(chk, m));
6497 }
6498
6499 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6500 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6501 void *ruleset __unused, struct inpcb *inp)
6502 {
6503 int chk;
6504
6505 CURVNET_ASSERT_SET();
6506
6507 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6508
6509 return (pf_check_return(chk, m));
6510 }
6511
6512 #ifdef INET
6513 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6514 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6515 void *ruleset __unused, struct inpcb *inp)
6516 {
6517 int chk;
6518
6519 CURVNET_ASSERT_SET();
6520
6521 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
6522
6523 return (pf_check_return(chk, m));
6524 }
6525
6526 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6527 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6528 void *ruleset __unused, struct inpcb *inp)
6529 {
6530 int chk;
6531
6532 CURVNET_ASSERT_SET();
6533
6534 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
6535
6536 return (pf_check_return(chk, m));
6537 }
6538 #endif
6539
6540 #ifdef INET6
6541 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6542 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6543 void *ruleset __unused, struct inpcb *inp)
6544 {
6545 int chk;
6546
6547 CURVNET_ASSERT_SET();
6548
6549 /*
6550 * In case of loopback traffic IPv6 uses the real interface in
6551 * order to support scoped addresses. In order to support stateful
6552 * filtering we have change this to lo0 as it is the case in IPv4.
6553 */
6554 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6555 m, inp, NULL);
6556
6557 return (pf_check_return(chk, m));
6558 }
6559
6560 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)6561 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6562 void *ruleset __unused, struct inpcb *inp)
6563 {
6564 int chk;
6565
6566 CURVNET_ASSERT_SET();
6567
6568 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL);
6569
6570 return (pf_check_return(chk, m));
6571 }
6572 #endif /* INET6 */
6573
6574 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6575 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6576 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
6577 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
6578
6579 #ifdef INET
6580 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6581 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6582 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
6583 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
6584 #endif
6585 #ifdef INET6
6586 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6587 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6588 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
6589 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
6590 #endif
6591
6592 static void
hook_pf_eth(void)6593 hook_pf_eth(void)
6594 {
6595 struct pfil_hook_args pha = {
6596 .pa_version = PFIL_VERSION,
6597 .pa_modname = "pf",
6598 .pa_type = PFIL_TYPE_ETHERNET,
6599 };
6600 struct pfil_link_args pla = {
6601 .pa_version = PFIL_VERSION,
6602 };
6603 int ret __diagused;
6604
6605 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6606 return;
6607
6608 pha.pa_mbuf_chk = pf_eth_check_in;
6609 pha.pa_flags = PFIL_IN;
6610 pha.pa_rulname = "eth-in";
6611 V_pf_eth_in_hook = pfil_add_hook(&pha);
6612 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6613 pla.pa_head = V_link_pfil_head;
6614 pla.pa_hook = V_pf_eth_in_hook;
6615 ret = pfil_link(&pla);
6616 MPASS(ret == 0);
6617 pha.pa_mbuf_chk = pf_eth_check_out;
6618 pha.pa_flags = PFIL_OUT;
6619 pha.pa_rulname = "eth-out";
6620 V_pf_eth_out_hook = pfil_add_hook(&pha);
6621 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6622 pla.pa_head = V_link_pfil_head;
6623 pla.pa_hook = V_pf_eth_out_hook;
6624 ret = pfil_link(&pla);
6625 MPASS(ret == 0);
6626
6627 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6628 }
6629
6630 static void
hook_pf(void)6631 hook_pf(void)
6632 {
6633 struct pfil_hook_args pha = {
6634 .pa_version = PFIL_VERSION,
6635 .pa_modname = "pf",
6636 };
6637 struct pfil_link_args pla = {
6638 .pa_version = PFIL_VERSION,
6639 };
6640 int ret __diagused;
6641
6642 if (atomic_load_bool(&V_pf_pfil_hooked))
6643 return;
6644
6645 #ifdef INET
6646 pha.pa_type = PFIL_TYPE_IP4;
6647 pha.pa_mbuf_chk = pf_check_in;
6648 pha.pa_flags = PFIL_IN;
6649 pha.pa_rulname = "default-in";
6650 V_pf_ip4_in_hook = pfil_add_hook(&pha);
6651 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6652 pla.pa_head = V_inet_pfil_head;
6653 pla.pa_hook = V_pf_ip4_in_hook;
6654 ret = pfil_link(&pla);
6655 MPASS(ret == 0);
6656 pha.pa_mbuf_chk = pf_check_out;
6657 pha.pa_flags = PFIL_OUT;
6658 pha.pa_rulname = "default-out";
6659 V_pf_ip4_out_hook = pfil_add_hook(&pha);
6660 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6661 pla.pa_head = V_inet_pfil_head;
6662 pla.pa_hook = V_pf_ip4_out_hook;
6663 ret = pfil_link(&pla);
6664 MPASS(ret == 0);
6665 if (V_pf_filter_local) {
6666 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6667 pla.pa_head = V_inet_local_pfil_head;
6668 pla.pa_hook = V_pf_ip4_out_hook;
6669 ret = pfil_link(&pla);
6670 MPASS(ret == 0);
6671 }
6672 #endif
6673 #ifdef INET6
6674 pha.pa_type = PFIL_TYPE_IP6;
6675 pha.pa_mbuf_chk = pf_check6_in;
6676 pha.pa_flags = PFIL_IN;
6677 pha.pa_rulname = "default-in6";
6678 V_pf_ip6_in_hook = pfil_add_hook(&pha);
6679 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6680 pla.pa_head = V_inet6_pfil_head;
6681 pla.pa_hook = V_pf_ip6_in_hook;
6682 ret = pfil_link(&pla);
6683 MPASS(ret == 0);
6684 pha.pa_mbuf_chk = pf_check6_out;
6685 pha.pa_rulname = "default-out6";
6686 pha.pa_flags = PFIL_OUT;
6687 V_pf_ip6_out_hook = pfil_add_hook(&pha);
6688 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6689 pla.pa_head = V_inet6_pfil_head;
6690 pla.pa_hook = V_pf_ip6_out_hook;
6691 ret = pfil_link(&pla);
6692 MPASS(ret == 0);
6693 if (V_pf_filter_local) {
6694 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6695 pla.pa_head = V_inet6_local_pfil_head;
6696 pla.pa_hook = V_pf_ip6_out_hook;
6697 ret = pfil_link(&pla);
6698 MPASS(ret == 0);
6699 }
6700 #endif
6701
6702 atomic_store_bool(&V_pf_pfil_hooked, true);
6703 }
6704
6705 static void
dehook_pf_eth(void)6706 dehook_pf_eth(void)
6707 {
6708
6709 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6710 return;
6711
6712 pfil_remove_hook(V_pf_eth_in_hook);
6713 pfil_remove_hook(V_pf_eth_out_hook);
6714
6715 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6716 }
6717
6718 static void
dehook_pf(void)6719 dehook_pf(void)
6720 {
6721
6722 if (!atomic_load_bool(&V_pf_pfil_hooked))
6723 return;
6724
6725 #ifdef INET
6726 pfil_remove_hook(V_pf_ip4_in_hook);
6727 pfil_remove_hook(V_pf_ip4_out_hook);
6728 #endif
6729 #ifdef INET6
6730 pfil_remove_hook(V_pf_ip6_in_hook);
6731 pfil_remove_hook(V_pf_ip6_out_hook);
6732 #endif
6733
6734 atomic_store_bool(&V_pf_pfil_hooked, false);
6735 }
6736
6737 static void
pf_load_vnet(void)6738 pf_load_vnet(void)
6739 {
6740 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6741 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6742
6743 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6744 sx_init(&V_pf_ioctl_lock, "pf ioctl");
6745
6746 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6747 PF_RULE_TAG_HASH_SIZE_DEFAULT);
6748 #ifdef ALTQ
6749 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6750 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6751 #endif
6752
6753 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6754
6755 pfattach_vnet();
6756 V_pf_vnet_active = 1;
6757 }
6758
6759 static int
pf_load(void)6760 pf_load(void)
6761 {
6762 int error;
6763
6764 sx_init(&pf_end_lock, "pf end thread");
6765
6766 pf_mtag_initialize();
6767
6768 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6769 if (pf_dev == NULL)
6770 return (ENOMEM);
6771
6772 pf_end_threads = 0;
6773 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6774 if (error != 0)
6775 return (error);
6776
6777 pfi_initialize();
6778
6779 return (0);
6780 }
6781
6782 static void
pf_unload_vnet(void)6783 pf_unload_vnet(void)
6784 {
6785 int ret __diagused;
6786
6787 V_pf_vnet_active = 0;
6788 V_pf_status.running = 0;
6789 dehook_pf();
6790 dehook_pf_eth();
6791
6792 PF_RULES_WLOCK();
6793 pf_syncookies_cleanup();
6794 shutdown_pf();
6795 PF_RULES_WUNLOCK();
6796
6797 ret = swi_remove(V_pf_swi_cookie);
6798 MPASS(ret == 0);
6799 ret = intr_event_destroy(V_pf_swi_ie);
6800 MPASS(ret == 0);
6801
6802 pf_unload_vnet_purge();
6803
6804 pf_normalize_cleanup();
6805 PF_RULES_WLOCK();
6806 pfi_cleanup_vnet();
6807 PF_RULES_WUNLOCK();
6808 pfr_cleanup();
6809 pf_osfp_flush();
6810 pf_cleanup();
6811 if (IS_DEFAULT_VNET(curvnet))
6812 pf_mtag_cleanup();
6813
6814 pf_cleanup_tagset(&V_pf_tags);
6815 #ifdef ALTQ
6816 pf_cleanup_tagset(&V_pf_qids);
6817 #endif
6818 uma_zdestroy(V_pf_tag_z);
6819
6820 #ifdef PF_WANT_32_TO_64_COUNTER
6821 PF_RULES_WLOCK();
6822 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6823
6824 MPASS(LIST_EMPTY(&V_pf_allkiflist));
6825 MPASS(V_pf_allkifcount == 0);
6826
6827 LIST_REMOVE(&V_pf_default_rule, allrulelist);
6828 V_pf_allrulecount--;
6829 LIST_REMOVE(V_pf_rulemarker, allrulelist);
6830
6831 MPASS(LIST_EMPTY(&V_pf_allrulelist));
6832 MPASS(V_pf_allrulecount == 0);
6833
6834 PF_RULES_WUNLOCK();
6835
6836 free(V_pf_kifmarker, PFI_MTYPE);
6837 free(V_pf_rulemarker, M_PFRULE);
6838 #endif
6839
6840 /* Free counters last as we updated them during shutdown. */
6841 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6842 for (int i = 0; i < 2; i++) {
6843 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6844 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6845 }
6846 counter_u64_free(V_pf_default_rule.states_cur);
6847 counter_u64_free(V_pf_default_rule.states_tot);
6848 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
6849 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]);
6850 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6851
6852 for (int i = 0; i < PFRES_MAX; i++)
6853 counter_u64_free(V_pf_status.counters[i]);
6854 for (int i = 0; i < KLCNT_MAX; i++)
6855 counter_u64_free(V_pf_status.lcounters[i]);
6856 for (int i = 0; i < FCNT_MAX; i++)
6857 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6858 for (int i = 0; i < SCNT_MAX; i++)
6859 counter_u64_free(V_pf_status.scounters[i]);
6860
6861 rm_destroy(&V_pf_rules_lock);
6862 sx_destroy(&V_pf_ioctl_lock);
6863 }
6864
6865 static void
pf_unload(void)6866 pf_unload(void)
6867 {
6868
6869 sx_xlock(&pf_end_lock);
6870 pf_end_threads = 1;
6871 while (pf_end_threads < 2) {
6872 wakeup_one(pf_purge_thread);
6873 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6874 }
6875 sx_xunlock(&pf_end_lock);
6876
6877 pf_nl_unregister();
6878
6879 if (pf_dev != NULL)
6880 destroy_dev(pf_dev);
6881
6882 pfi_cleanup();
6883
6884 sx_destroy(&pf_end_lock);
6885 }
6886
6887 static void
vnet_pf_init(void * unused __unused)6888 vnet_pf_init(void *unused __unused)
6889 {
6890
6891 pf_load_vnet();
6892 }
6893 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6894 vnet_pf_init, NULL);
6895
6896 static void
vnet_pf_uninit(const void * unused __unused)6897 vnet_pf_uninit(const void *unused __unused)
6898 {
6899
6900 pf_unload_vnet();
6901 }
6902 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6903 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6904 vnet_pf_uninit, NULL);
6905
6906 static int
pf_modevent(module_t mod,int type,void * data)6907 pf_modevent(module_t mod, int type, void *data)
6908 {
6909 int error = 0;
6910
6911 switch(type) {
6912 case MOD_LOAD:
6913 error = pf_load();
6914 pf_nl_register();
6915 break;
6916 case MOD_UNLOAD:
6917 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6918 * the vnet_pf_uninit()s */
6919 break;
6920 default:
6921 error = EINVAL;
6922 break;
6923 }
6924
6925 return (error);
6926 }
6927
6928 static moduledata_t pf_mod = {
6929 "pf",
6930 pf_modevent,
6931 0
6932 };
6933
6934 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6935 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6936 MODULE_VERSION(pf, PF_MODVER);
6937