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 void pf_qid_unref(uint16_t);
120 #endif /* ALTQ */
121 static int pf_begin_rules(u_int32_t *, int, const char *);
122 static int pf_rollback_rules(u_int32_t, int, char *);
123 static int pf_setup_pfsync_matching(struct pf_kruleset *);
124 static void pf_hash_rule_rolling(MD5_CTX *, struct pf_krule *);
125 static void pf_hash_rule(struct pf_krule *);
126 static void pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
127 static int pf_commit_rules(u_int32_t, int, char *);
128 static int pf_addr_setup(struct pf_kruleset *,
129 struct pf_addr_wrap *, sa_family_t);
130 static void pf_src_node_copy(const struct pf_ksrc_node *,
131 struct pf_src_node *);
132 #ifdef ALTQ
133 static int pf_export_kaltq(struct pf_altq *,
134 struct pfioc_altq_v1 *, size_t);
135 static int pf_import_kaltq(struct pfioc_altq_v1 *,
136 struct pf_altq *, size_t);
137 #endif /* ALTQ */
138
139 static void pf_statelim_commit(void);
140 static void pf_statelim_rollback(void);
141 static int pf_sourcelim_check(void);
142 static void pf_sourcelim_commit(void);
143 static void pf_sourcelim_rollback(void);
144
145 VNET_DEFINE(struct pf_krule, pf_default_rule);
146
147 static __inline int pf_krule_compare(struct pf_krule *,
148 struct pf_krule *);
149
150 RB_GENERATE(pf_krule_global, pf_krule, entry_global, pf_krule_compare);
151
152 #ifdef ALTQ
153 VNET_DEFINE_STATIC(int, pf_altq_running);
154 #define V_pf_altq_running VNET(pf_altq_running)
155 #endif
156
157 #define TAGID_MAX 50000
158 struct pf_tagname {
159 TAILQ_ENTRY(pf_tagname) namehash_entries;
160 TAILQ_ENTRY(pf_tagname) taghash_entries;
161 char name[PF_TAG_NAME_SIZE];
162 uint16_t tag;
163 int ref;
164 };
165
166 struct pf_tagset {
167 TAILQ_HEAD(, pf_tagname) *namehash;
168 TAILQ_HEAD(, pf_tagname) *taghash;
169 unsigned int mask;
170 uint32_t seed;
171 BITSET_DEFINE(, TAGID_MAX) avail;
172 };
173
174 VNET_DEFINE(struct pf_tagset, pf_tags);
175 #define V_pf_tags VNET(pf_tags)
176 static unsigned int pf_rule_tag_hashsize;
177 #define PF_RULE_TAG_HASH_SIZE_DEFAULT 128
178 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
179 &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
180 "Size of pf(4) rule tag hashtable");
181
182 #ifdef ALTQ
183 VNET_DEFINE(struct pf_tagset, pf_qids);
184 #define V_pf_qids VNET(pf_qids)
185 static unsigned int pf_queue_tag_hashsize;
186 #define PF_QUEUE_TAG_HASH_SIZE_DEFAULT 128
187 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
188 &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
189 "Size of pf(4) queue tag hashtable");
190 #endif
191 VNET_DEFINE(uma_zone_t, pf_tag_z);
192 #define V_pf_tag_z VNET(pf_tag_z)
193 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
194 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
195 MALLOC_DEFINE(M_PF, "pf", "pf(4)");
196 MALLOC_DEFINE(M_PF_STATE_LIM, "pf_state_lim", "pf(4) state limiter");
197
198 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
199 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
200 #endif
201
202 VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
203 #define V_pf_filter_local VNET(pf_filter_local)
204 SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
205 &VNET_NAME(pf_filter_local), false,
206 "Enable filtering for packets delivered to local network stack");
207
208 #ifdef PF_DEFAULT_TO_DROP
209 VNET_DEFINE_STATIC(bool, default_to_drop) = true;
210 #else
211 VNET_DEFINE_STATIC(bool, default_to_drop);
212 #endif
213 #define V_default_to_drop VNET(default_to_drop)
214 SYSCTL_BOOL(_net_pf, OID_AUTO, default_to_drop, CTLFLAG_RDTUN | CTLFLAG_VNET,
215 &VNET_NAME(default_to_drop), false,
216 "Make the default rule drop all packets.");
217
218 static void pf_init_tagset(struct pf_tagset *, unsigned int *,
219 unsigned int);
220 static void pf_cleanup_tagset(struct pf_tagset *);
221 static uint16_t tagname2hashindex(const struct pf_tagset *, const char *);
222 static uint16_t tag2hashindex(const struct pf_tagset *, uint16_t);
223 static u_int16_t tagname2tag(struct pf_tagset *, const char *, bool);
224 static void tag_unref(struct pf_tagset *, u_int16_t);
225
226 struct cdev *pf_dev;
227
228 /*
229 * XXX - These are new and need to be checked when moveing to a new version
230 */
231 static void pf_clear_all_states(void);
232 static int pf_killstates_row(struct pf_kstate_kill *,
233 struct pf_idhash *);
234 static int pf_killstates_nv(struct pfioc_nv *);
235 static int pf_clearstates_nv(struct pfioc_nv *);
236 static int pf_getstate(struct pfioc_nv *);
237 static int pf_getstatus(struct pfioc_nv *);
238 static int pf_clear_tables(void);
239 static void pf_kill_srcnodes(struct pfioc_src_node_kill *);
240 static int pf_keepcounters(struct pfioc_nv *);
241 static void pf_tbladdr_copyout(struct pf_addr_wrap *);
242
243 /*
244 * Wrapper functions for pfil(9) hooks
245 */
246 static pfil_return_t pf_eth_check_in(struct mbuf **m, struct ifnet *ifp,
247 int flags, void *ruleset __unused, struct inpcb *inp);
248 static pfil_return_t pf_eth_check_out(struct mbuf **m, struct ifnet *ifp,
249 int flags, void *ruleset __unused, struct inpcb *inp);
250 #ifdef INET
251 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
252 int flags, void *ruleset __unused, struct inpcb *inp);
253 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
254 int flags, void *ruleset __unused, struct inpcb *inp);
255 #endif
256 #ifdef INET6
257 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
258 int flags, void *ruleset __unused, struct inpcb *inp);
259 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
260 int flags, void *ruleset __unused, struct inpcb *inp);
261 #endif
262
263 static void hook_pf_eth(void);
264 static void hook_pf(void);
265 static void dehook_pf_eth(void);
266 static void dehook_pf(void);
267 static int shutdown_pf(void);
268 static int pf_load(void);
269 static void pf_unload(void *);
270
271 static struct cdevsw pf_cdevsw = {
272 .d_ioctl = pfioctl,
273 .d_name = PF_NAME,
274 .d_version = D_VERSION,
275 };
276
277 VNET_DEFINE_STATIC(bool, pf_pfil_hooked);
278 #define V_pf_pfil_hooked VNET(pf_pfil_hooked)
279 VNET_DEFINE_STATIC(bool, pf_pfil_eth_hooked);
280 #define V_pf_pfil_eth_hooked VNET(pf_pfil_eth_hooked)
281
282 /*
283 * We need a flag that is neither hooked nor running to know when
284 * the VNET is "valid". We primarily need this to control (global)
285 * external event, e.g., eventhandlers.
286 */
287 VNET_DEFINE(int, pf_vnet_active);
288 #define V_pf_vnet_active VNET(pf_vnet_active)
289
290 int pf_end_threads;
291 struct proc *pf_purge_proc;
292
293 VNET_DEFINE(struct rmlock, pf_rules_lock);
294 VNET_DEFINE(struct rmlock, pf_tags_lock);
295 VNET_DEFINE_STATIC(struct sx, pf_ioctl_lock);
296 #define V_pf_ioctl_lock VNET(pf_ioctl_lock)
297 struct sx pf_end_lock;
298
299 /* pfsync */
300 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
301 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
302 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
303 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
304 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
305 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
306 VNET_DEFINE(pflow_export_state_t *, pflow_export_state_ptr);
307 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
308
309 /* pflog */
310 pflog_packet_t *pflog_packet_ptr = NULL;
311
312 /*
313 * Copy a user-provided string, returning an error if truncation would occur.
314 * Avoid scanning past "sz" bytes in the source string since there's no
315 * guarantee that it's nul-terminated.
316 */
317 static int
pf_user_strcpy(char * dst,const char * src,size_t sz)318 pf_user_strcpy(char *dst, const char *src, size_t sz)
319 {
320 if (strnlen(src, sz) == sz)
321 return (EINVAL);
322 (void)strlcpy(dst, src, sz);
323 return (0);
324 }
325
326 static void
pfattach_vnet(void)327 pfattach_vnet(void)
328 {
329 u_int32_t *my_timeout = V_pf_default_rule.timeout;
330
331 bzero(&V_pf_status, sizeof(V_pf_status));
332
333 pf_initialize();
334 pfr_initialize();
335 pfi_initialize_vnet();
336 pf_normalize_init();
337 pf_syncookies_init();
338
339 V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
340 V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
341 V_pf_limits[PF_LIMIT_ANCHORS].limit = PF_ANCHOR_HIWAT;
342 V_pf_limits[PF_LIMIT_ETH_ANCHORS].limit = PF_ANCHOR_HIWAT;
343
344 RB_INIT(&V_pf_anchors);
345 pf_init_kruleset(&pf_main_ruleset);
346
347 pf_init_keth(V_pf_keth);
348
349 /* default rule should never be garbage collected */
350 V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
351 V_pf_default_rule.action = V_default_to_drop ? PF_DROP : PF_PASS;
352 V_pf_default_rule.nr = (uint32_t)-1;
353 V_pf_default_rule.rtableid = -1;
354
355 pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
356 for (int i = 0; i < 2; i++) {
357 pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
358 pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
359 }
360 V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
361 V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
362 for (pf_sn_types_t sn_type = 0; sn_type<PF_SN_MAX; sn_type++)
363 V_pf_default_rule.src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
364
365 V_pf_default_rule.timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
366 M_WAITOK | M_ZERO);
367
368 #ifdef PF_WANT_32_TO_64_COUNTER
369 V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
370 V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
371 PF_RULES_WLOCK();
372 LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
373 LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
374 V_pf_allrulecount++;
375 LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
376 PF_RULES_WUNLOCK();
377 #endif
378
379 /* initialize default timeouts */
380 my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
381 my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
382 my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
383 my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
384 my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
385 my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
386 my_timeout[PFTM_SCTP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
387 my_timeout[PFTM_SCTP_OPENING] = PFTM_TCP_OPENING_VAL;
388 my_timeout[PFTM_SCTP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
389 my_timeout[PFTM_SCTP_CLOSING] = PFTM_TCP_CLOSING_VAL;
390 my_timeout[PFTM_SCTP_CLOSED] = PFTM_TCP_CLOSED_VAL;
391 my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
392 my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
393 my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
394 my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
395 my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
396 my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
397 my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
398 my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
399 my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
400 my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
401 my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
402 my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
403 my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
404 my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
405
406 V_pf_status.debug = PF_DEBUG_URGENT;
407 /*
408 * XXX This is different than in OpenBSD where reassembly is enabled by
409 * defult. In FreeBSD we expect people to still use scrub rules and
410 * switch to the new syntax later. Only when they switch they must
411 * explicitly enable reassemle. We could change the default once the
412 * scrub rule functionality is hopefully removed some day in future.
413 */
414 V_pf_status.reass = 0;
415
416 V_pf_pfil_hooked = false;
417 V_pf_pfil_eth_hooked = false;
418
419 /* XXX do our best to avoid a conflict */
420 V_pf_status.hostid = arc4random();
421
422 for (int i = 0; i < PFRES_MAX; i++)
423 V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
424 for (int i = 0; i < KLCNT_MAX; i++)
425 V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
426 for (int i = 0; i < FCNT_MAX; i++)
427 pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
428 for (int i = 0; i < SCNT_MAX; i++)
429 V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
430 for (int i = 0; i < NCNT_MAX; i++)
431 V_pf_status.ncounters[i] = counter_u64_alloc(M_WAITOK);
432
433 if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
434 INTR_MPSAFE, &V_pf_swi_cookie) != 0)
435 /* XXXGL: leaked all above. */
436 return;
437 }
438
439 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)440 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
441 u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
442 u_int8_t check_ticket, int which)
443 {
444 struct pf_kruleset *ruleset;
445 struct pf_krule *rule;
446 int rs_num;
447
448 MPASS(which == PF_RDR || which == PF_NAT || which == PF_RT);
449
450 ruleset = pf_find_kruleset(anchor);
451 if (ruleset == NULL)
452 return (NULL);
453 rs_num = pf_get_ruleset_number(rule_action);
454 if (rs_num >= PF_RULESET_MAX)
455 return (NULL);
456 if (active) {
457 if (check_ticket && ticket !=
458 ruleset->rules[rs_num].active.ticket)
459 return (NULL);
460 if (r_last)
461 rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
462 pf_krulequeue);
463 else
464 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
465 } else {
466 if (check_ticket && ticket !=
467 ruleset->rules[rs_num].inactive.ticket)
468 return (NULL);
469 if (r_last)
470 rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
471 pf_krulequeue);
472 else
473 rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
474 }
475 if (!r_last) {
476 while ((rule != NULL) && (rule->nr != rule_number))
477 rule = TAILQ_NEXT(rule, entries);
478 }
479 if (rule == NULL)
480 return (NULL);
481
482 switch (which) {
483 case PF_RDR:
484 return (&rule->rdr);
485 case PF_NAT:
486 return (&rule->nat);
487 case PF_RT:
488 return (&rule->route);
489 default:
490 panic("Unknow pool type %d", which);
491 }
492 }
493
494 static void
pf_mv_kpool(struct pf_kpalist * poola,struct pf_kpalist * poolb)495 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
496 {
497 struct pf_kpooladdr *mv_pool_pa;
498
499 while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
500 TAILQ_REMOVE(poola, mv_pool_pa, entries);
501 TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
502 }
503 }
504
505 static void
pf_empty_kpool(struct pf_kpalist * poola)506 pf_empty_kpool(struct pf_kpalist *poola)
507 {
508 struct pf_kpooladdr *pa;
509
510 while ((pa = TAILQ_FIRST(poola)) != NULL) {
511 switch (pa->addr.type) {
512 case PF_ADDR_DYNIFTL:
513 pfi_dynaddr_remove(pa->addr.p.dyn);
514 break;
515 case PF_ADDR_TABLE:
516 /* XXX: this could be unfinished pooladdr on pabuf */
517 if (pa->addr.p.tbl != NULL)
518 pfr_detach_table(pa->addr.p.tbl);
519 break;
520 }
521 if (pa->kif)
522 pfi_kkif_unref(pa->kif);
523 TAILQ_REMOVE(poola, pa, entries);
524 free(pa, M_PFRULE);
525 }
526 }
527
528 static void
pf_unlink_rule_locked(struct pf_krulequeue * rulequeue,struct pf_krule * rule)529 pf_unlink_rule_locked(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
530 {
531
532 PF_RULES_WASSERT();
533 PF_UNLNKDRULES_ASSERT();
534
535 TAILQ_REMOVE(rulequeue, rule, entries);
536
537 rule->rule_ref |= PFRULE_REFS;
538 TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
539 }
540
541 static void
pf_unlink_rule(struct pf_krulequeue * rulequeue,struct pf_krule * rule)542 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
543 {
544
545 PF_RULES_WASSERT();
546
547 PF_UNLNKDRULES_LOCK();
548 pf_unlink_rule_locked(rulequeue, rule);
549 PF_UNLNKDRULES_UNLOCK();
550 }
551
552 static void
pf_free_eth_rule(struct pf_keth_rule * rule)553 pf_free_eth_rule(struct pf_keth_rule *rule)
554 {
555 PF_RULES_WASSERT();
556
557 if (rule == NULL)
558 return;
559
560 if (rule->tag)
561 tag_unref(&V_pf_tags, rule->tag);
562 if (rule->match_tag)
563 tag_unref(&V_pf_tags, rule->match_tag);
564 #ifdef ALTQ
565 pf_qid_unref(rule->qid);
566 #endif
567
568 if (rule->bridge_to)
569 pfi_kkif_unref(rule->bridge_to);
570 if (rule->kif)
571 pfi_kkif_unref(rule->kif);
572
573 if (rule->ipsrc.addr.type == PF_ADDR_TABLE)
574 pfr_detach_table(rule->ipsrc.addr.p.tbl);
575 if (rule->ipdst.addr.type == PF_ADDR_TABLE)
576 pfr_detach_table(rule->ipdst.addr.p.tbl);
577
578 counter_u64_free(rule->evaluations);
579 for (int i = 0; i < 2; i++) {
580 counter_u64_free(rule->packets[i]);
581 counter_u64_free(rule->bytes[i]);
582 }
583 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
584 pf_keth_anchor_remove(rule);
585
586 free(rule, M_PFRULE);
587 }
588
589 void
pf_free_rule(struct pf_krule * rule)590 pf_free_rule(struct pf_krule *rule)
591 {
592
593 PF_RULES_WASSERT();
594 PF_CONFIG_ASSERT();
595
596 if (rule->tag)
597 tag_unref(&V_pf_tags, rule->tag);
598 if (rule->match_tag)
599 tag_unref(&V_pf_tags, rule->match_tag);
600 #ifdef ALTQ
601 if (rule->pqid != rule->qid)
602 pf_qid_unref(rule->pqid);
603 pf_qid_unref(rule->qid);
604 #endif
605 switch (rule->src.addr.type) {
606 case PF_ADDR_DYNIFTL:
607 pfi_dynaddr_remove(rule->src.addr.p.dyn);
608 break;
609 case PF_ADDR_TABLE:
610 pfr_detach_table(rule->src.addr.p.tbl);
611 break;
612 }
613 switch (rule->dst.addr.type) {
614 case PF_ADDR_DYNIFTL:
615 pfi_dynaddr_remove(rule->dst.addr.p.dyn);
616 break;
617 case PF_ADDR_TABLE:
618 pfr_detach_table(rule->dst.addr.p.tbl);
619 break;
620 }
621 if (rule->overload_tbl)
622 pfr_detach_table(rule->overload_tbl);
623 if (rule->kif)
624 pfi_kkif_unref(rule->kif);
625 if (rule->rcv_kif)
626 pfi_kkif_unref(rule->rcv_kif);
627 pf_remove_kanchor(rule);
628 pf_empty_kpool(&rule->rdr.list);
629 pf_empty_kpool(&rule->nat.list);
630 pf_empty_kpool(&rule->route.list);
631
632 pf_krule_free(rule);
633 }
634
635 static void
pf_init_tagset(struct pf_tagset * ts,unsigned int * tunable_size,unsigned int default_size)636 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
637 unsigned int default_size)
638 {
639 unsigned int i;
640 unsigned int hashsize;
641
642 if (*tunable_size == 0 || !powerof2(*tunable_size))
643 *tunable_size = default_size;
644
645 hashsize = *tunable_size;
646 ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
647 M_WAITOK);
648 ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
649 M_WAITOK);
650 ts->mask = hashsize - 1;
651 ts->seed = arc4random();
652 for (i = 0; i < hashsize; i++) {
653 TAILQ_INIT(&ts->namehash[i]);
654 TAILQ_INIT(&ts->taghash[i]);
655 }
656 BIT_FILL(TAGID_MAX, &ts->avail);
657 }
658
659 static void
pf_cleanup_tagset(struct pf_tagset * ts)660 pf_cleanup_tagset(struct pf_tagset *ts)
661 {
662 unsigned int i;
663 unsigned int hashsize;
664 struct pf_tagname *t, *tmp;
665
666 /*
667 * Only need to clean up one of the hashes as each tag is hashed
668 * into each table.
669 */
670 hashsize = ts->mask + 1;
671 for (i = 0; i < hashsize; i++)
672 TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
673 uma_zfree(V_pf_tag_z, t);
674
675 free(ts->namehash, M_PFHASH);
676 free(ts->taghash, M_PFHASH);
677 }
678
679 static uint16_t
tagname2hashindex(const struct pf_tagset * ts,const char * tagname)680 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
681 {
682 size_t len;
683
684 len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
685 return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
686 }
687
688 static uint16_t
tag2hashindex(const struct pf_tagset * ts,uint16_t tag)689 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
690 {
691
692 return (tag & ts->mask);
693 }
694
695 static u_int16_t
tagname2tag(struct pf_tagset * ts,const char * tagname,bool add_new)696 tagname2tag(struct pf_tagset *ts, const char *tagname, bool add_new)
697 {
698 struct pf_tagname *tag;
699 u_int32_t index;
700 u_int16_t new_tagid;
701
702 PF_TAGS_RLOCK_TRACKER;
703
704 PF_TAGS_RLOCK();
705
706 index = tagname2hashindex(ts, tagname);
707 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
708 if (strcmp(tagname, tag->name) == 0) {
709 tag->ref++;
710 new_tagid = tag->tag;
711 PF_TAGS_RUNLOCK();
712 return (new_tagid);
713 }
714
715 /*
716 * When used for pfsync with queues we must not create new entries.
717 * Pf tags can be created just fine by this function, but queues
718 * require additional configuration. If they are missing on the target
719 * system we just ignore them
720 */
721 if (add_new == false) {
722 printf("%s: Not creating a new tag\n", __func__);
723 PF_TAGS_RUNLOCK();
724 return (0);
725 }
726
727 /*
728 * If a new entry must be created do it under a write lock.
729 * But first search again, somebody could have created the tag
730 * between unlocking the read lock and locking the write lock.
731 */
732 PF_TAGS_RUNLOCK();
733 PF_TAGS_WLOCK();
734 TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
735 if (strcmp(tagname, tag->name) == 0) {
736 tag->ref++;
737 new_tagid = tag->tag;
738 PF_TAGS_WUNLOCK();
739 return (new_tagid);
740 }
741
742 /*
743 * new entry
744 *
745 * to avoid fragmentation, we do a linear search from the beginning
746 * and take the first free slot we find.
747 */
748 new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
749 /*
750 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
751 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
752 * set. It may also return a bit number greater than TAGID_MAX due
753 * to rounding of the number of bits in the vector up to a multiple
754 * of the vector word size at declaration/allocation time.
755 */
756 if ((new_tagid == 0) || (new_tagid > TAGID_MAX)) {
757 PF_TAGS_WUNLOCK();
758 return (0);
759 }
760
761 /* Mark the tag as in use. Bits are 0-based for BIT_CLR() */
762 BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
763
764 /* allocate and fill new struct pf_tagname */
765 tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
766 if (tag == NULL) {
767 PF_TAGS_WUNLOCK();
768 return (0);
769 }
770 strlcpy(tag->name, tagname, sizeof(tag->name));
771 tag->tag = new_tagid;
772 tag->ref = 1;
773
774 /* Insert into namehash */
775 TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
776
777 /* Insert into taghash */
778 index = tag2hashindex(ts, new_tagid);
779 TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
780
781 PF_TAGS_WUNLOCK();
782 return (new_tagid);
783 }
784
785 static char *
tag2tagname(struct pf_tagset * ts,u_int16_t tag)786 tag2tagname(struct pf_tagset *ts, u_int16_t tag)
787 {
788 struct pf_tagname *t;
789 uint16_t index;
790
791 PF_TAGS_RLOCK_TRACKER;
792
793 PF_TAGS_RLOCK();
794
795 index = tag2hashindex(ts, tag);
796 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
797 if (tag == t->tag) {
798 PF_TAGS_RUNLOCK();
799 return (t->name);
800 }
801
802 PF_TAGS_RUNLOCK();
803 return (NULL);
804 }
805
806 static void
tag_unref(struct pf_tagset * ts,u_int16_t tag)807 tag_unref(struct pf_tagset *ts, u_int16_t tag)
808 {
809 struct pf_tagname *t;
810 uint16_t index;
811
812 PF_TAGS_WLOCK();
813
814 index = tag2hashindex(ts, tag);
815 TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
816 if (tag == t->tag) {
817 if (--t->ref == 0) {
818 TAILQ_REMOVE(&ts->taghash[index], t,
819 taghash_entries);
820 index = tagname2hashindex(ts, t->name);
821 TAILQ_REMOVE(&ts->namehash[index], t,
822 namehash_entries);
823 /* Bits are 0-based for BIT_SET() */
824 BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
825 uma_zfree(V_pf_tag_z, t);
826 }
827 break;
828 }
829
830 PF_TAGS_WUNLOCK();
831 }
832
833 uint16_t
pf_tagname2tag(const char * tagname)834 pf_tagname2tag(const char *tagname)
835 {
836 return (tagname2tag(&V_pf_tags, tagname, true));
837 }
838
839 static const char *
pf_tag2tagname(uint16_t tag)840 pf_tag2tagname(uint16_t tag)
841 {
842 return (tag2tagname(&V_pf_tags, tag));
843 }
844
845 static int
pf_begin_eth(uint32_t * ticket,const char * anchor)846 pf_begin_eth(uint32_t *ticket, const char *anchor)
847 {
848 struct pf_keth_rule *rule, *tmp;
849 struct pf_keth_ruleset *rs;
850
851 PF_RULES_WASSERT();
852
853 rs = pf_find_or_create_keth_ruleset(anchor);
854 if (rs == NULL)
855 return (EINVAL);
856
857 /* Purge old inactive rules. */
858 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
859 tmp) {
860 TAILQ_REMOVE(rs->inactive.rules, rule,
861 entries);
862 pf_free_eth_rule(rule);
863 }
864
865 *ticket = ++rs->inactive.ticket;
866 rs->inactive.open = 1;
867
868 return (0);
869 }
870
871 static int
pf_rollback_eth(uint32_t ticket,const char * anchor)872 pf_rollback_eth(uint32_t ticket, const char *anchor)
873 {
874 struct pf_keth_rule *rule, *tmp;
875 struct pf_keth_ruleset *rs;
876
877 PF_RULES_WASSERT();
878
879 rs = pf_find_keth_ruleset(anchor);
880 if (rs == NULL)
881 return (EINVAL);
882
883 if (!rs->inactive.open ||
884 ticket != rs->inactive.ticket)
885 return (0);
886
887 /* Purge old inactive rules. */
888 TAILQ_FOREACH_SAFE(rule, rs->inactive.rules, entries,
889 tmp) {
890 TAILQ_REMOVE(rs->inactive.rules, rule, entries);
891 pf_free_eth_rule(rule);
892 }
893
894 rs->inactive.open = 0;
895
896 pf_remove_if_empty_keth_ruleset(rs);
897
898 return (0);
899 }
900
901 #define PF_SET_SKIP_STEPS(i) \
902 do { \
903 while (head[i] != cur) { \
904 head[i]->skip[i].ptr = cur; \
905 head[i] = TAILQ_NEXT(head[i], entries); \
906 } \
907 } while (0)
908
909 static void
pf_eth_calc_skip_steps(struct pf_keth_ruleq * rules)910 pf_eth_calc_skip_steps(struct pf_keth_ruleq *rules)
911 {
912 struct pf_keth_rule *cur, *prev, *head[PFE_SKIP_COUNT];
913 int i;
914
915 cur = TAILQ_FIRST(rules);
916 prev = cur;
917 for (i = 0; i < PFE_SKIP_COUNT; ++i)
918 head[i] = cur;
919 while (cur != NULL) {
920 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
921 PF_SET_SKIP_STEPS(PFE_SKIP_IFP);
922 if (cur->direction != prev->direction)
923 PF_SET_SKIP_STEPS(PFE_SKIP_DIR);
924 if (cur->proto != prev->proto)
925 PF_SET_SKIP_STEPS(PFE_SKIP_PROTO);
926 if (memcmp(&cur->src, &prev->src, sizeof(cur->src)) != 0)
927 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_ADDR);
928 if (memcmp(&cur->dst, &prev->dst, sizeof(cur->dst)) != 0)
929 PF_SET_SKIP_STEPS(PFE_SKIP_DST_ADDR);
930 if (cur->ipsrc.neg != prev->ipsrc.neg ||
931 pf_addr_wrap_neq(&cur->ipsrc.addr, &prev->ipsrc.addr))
932 PF_SET_SKIP_STEPS(PFE_SKIP_SRC_IP_ADDR);
933 if (cur->ipdst.neg != prev->ipdst.neg ||
934 pf_addr_wrap_neq(&cur->ipdst.addr, &prev->ipdst.addr))
935 PF_SET_SKIP_STEPS(PFE_SKIP_DST_IP_ADDR);
936
937 prev = cur;
938 cur = TAILQ_NEXT(cur, entries);
939 }
940 for (i = 0; i < PFE_SKIP_COUNT; ++i)
941 PF_SET_SKIP_STEPS(i);
942 }
943
944 static int
pf_commit_eth(uint32_t ticket,const char * anchor)945 pf_commit_eth(uint32_t ticket, const char *anchor)
946 {
947 struct pf_keth_ruleq *rules;
948 struct pf_keth_ruleset *rs;
949
950 rs = pf_find_keth_ruleset(anchor);
951 if (rs == NULL) {
952 return (EINVAL);
953 }
954
955 if (!rs->inactive.open ||
956 ticket != rs->inactive.ticket)
957 return (EBUSY);
958
959 PF_RULES_WASSERT();
960
961 pf_eth_calc_skip_steps(rs->inactive.rules);
962
963 rules = rs->active.rules;
964 atomic_store_ptr(&rs->active.rules, rs->inactive.rules);
965 rs->inactive.rules = rules;
966 rs->inactive.ticket = rs->active.ticket;
967
968 return (pf_rollback_eth(rs->inactive.ticket,
969 rs->anchor ? rs->anchor->path : ""));
970 }
971
972 #ifdef ALTQ
973 uint16_t
pf_qname2qid(const char * qname,bool add_new)974 pf_qname2qid(const char *qname, bool add_new)
975 {
976 return (tagname2tag(&V_pf_qids, qname, add_new));
977 }
978
979 static void
pf_qid_unref(uint16_t qid)980 pf_qid_unref(uint16_t qid)
981 {
982 tag_unref(&V_pf_qids, qid);
983 }
984
985 static int
pf_begin_altq(u_int32_t * ticket)986 pf_begin_altq(u_int32_t *ticket)
987 {
988 struct pf_altq *altq, *tmp;
989 int error = 0;
990
991 PF_RULES_WASSERT();
992
993 /* Purge the old altq lists */
994 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
995 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
996 /* detach and destroy the discipline */
997 error = altq_remove(altq);
998 }
999 free(altq, M_PFALTQ);
1000 }
1001 TAILQ_INIT(V_pf_altq_ifs_inactive);
1002 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1003 pf_qid_unref(altq->qid);
1004 free(altq, M_PFALTQ);
1005 }
1006 TAILQ_INIT(V_pf_altqs_inactive);
1007 if (error)
1008 return (error);
1009 *ticket = ++V_ticket_altqs_inactive;
1010 V_altqs_inactive_open = 1;
1011 return (0);
1012 }
1013
1014 static int
pf_rollback_altq(u_int32_t ticket)1015 pf_rollback_altq(u_int32_t ticket)
1016 {
1017 struct pf_altq *altq, *tmp;
1018 int error = 0;
1019
1020 PF_RULES_WASSERT();
1021
1022 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
1023 return (0);
1024 /* Purge the old altq lists */
1025 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1026 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1027 /* detach and destroy the discipline */
1028 error = altq_remove(altq);
1029 }
1030 free(altq, M_PFALTQ);
1031 }
1032 TAILQ_INIT(V_pf_altq_ifs_inactive);
1033 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1034 pf_qid_unref(altq->qid);
1035 free(altq, M_PFALTQ);
1036 }
1037 TAILQ_INIT(V_pf_altqs_inactive);
1038 V_altqs_inactive_open = 0;
1039 return (error);
1040 }
1041
1042 static int
pf_commit_altq(u_int32_t ticket)1043 pf_commit_altq(u_int32_t ticket)
1044 {
1045 struct pf_altqqueue *old_altqs, *old_altq_ifs;
1046 struct pf_altq *altq, *tmp;
1047 int err, error = 0;
1048
1049 PF_RULES_WASSERT();
1050
1051 if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
1052 return (EBUSY);
1053
1054 /* swap altqs, keep the old. */
1055 old_altqs = V_pf_altqs_active;
1056 old_altq_ifs = V_pf_altq_ifs_active;
1057 V_pf_altqs_active = V_pf_altqs_inactive;
1058 V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
1059 V_pf_altqs_inactive = old_altqs;
1060 V_pf_altq_ifs_inactive = old_altq_ifs;
1061 V_ticket_altqs_active = V_ticket_altqs_inactive;
1062
1063 /* Attach new disciplines */
1064 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1065 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1066 /* attach the discipline */
1067 error = altq_pfattach(altq);
1068 if (error == 0 && V_pf_altq_running)
1069 error = pf_enable_altq(altq);
1070 if (error != 0)
1071 return (error);
1072 }
1073 }
1074
1075 /* Purge the old altq lists */
1076 TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
1077 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
1078 /* detach and destroy the discipline */
1079 if (V_pf_altq_running)
1080 error = pf_disable_altq(altq);
1081 err = altq_pfdetach(altq);
1082 if (err != 0 && error == 0)
1083 error = err;
1084 err = altq_remove(altq);
1085 if (err != 0 && error == 0)
1086 error = err;
1087 }
1088 free(altq, M_PFALTQ);
1089 }
1090 TAILQ_INIT(V_pf_altq_ifs_inactive);
1091 TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
1092 pf_qid_unref(altq->qid);
1093 free(altq, M_PFALTQ);
1094 }
1095 TAILQ_INIT(V_pf_altqs_inactive);
1096
1097 V_altqs_inactive_open = 0;
1098 return (error);
1099 }
1100
1101 static int
pf_enable_altq(struct pf_altq * altq)1102 pf_enable_altq(struct pf_altq *altq)
1103 {
1104 struct ifnet *ifp;
1105 struct tb_profile tb;
1106 int error = 0;
1107
1108 if ((ifp = ifunit(altq->ifname)) == NULL)
1109 return (EINVAL);
1110
1111 if (ifp->if_snd.altq_type != ALTQT_NONE)
1112 error = altq_enable(&ifp->if_snd);
1113
1114 /* set tokenbucket regulator */
1115 if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1116 tb.rate = altq->ifbandwidth;
1117 tb.depth = altq->tbrsize;
1118 error = tbr_set(&ifp->if_snd, &tb);
1119 }
1120
1121 return (error);
1122 }
1123
1124 static int
pf_disable_altq(struct pf_altq * altq)1125 pf_disable_altq(struct pf_altq *altq)
1126 {
1127 struct ifnet *ifp;
1128 struct tb_profile tb;
1129 int error;
1130
1131 if ((ifp = ifunit(altq->ifname)) == NULL)
1132 return (EINVAL);
1133
1134 /*
1135 * when the discipline is no longer referenced, it was overridden
1136 * by a new one. if so, just return.
1137 */
1138 if (altq->altq_disc != ifp->if_snd.altq_disc)
1139 return (0);
1140
1141 error = altq_disable(&ifp->if_snd);
1142
1143 if (error == 0) {
1144 /* clear tokenbucket regulator */
1145 tb.rate = 0;
1146 error = tbr_set(&ifp->if_snd, &tb);
1147 }
1148
1149 return (error);
1150 }
1151
1152 static int
pf_altq_ifnet_event_add(struct ifnet * ifp,int remove,u_int32_t ticket,struct pf_altq * altq)1153 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
1154 struct pf_altq *altq)
1155 {
1156 struct ifnet *ifp1;
1157 int error = 0;
1158
1159 /* Deactivate the interface in question */
1160 altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
1161 if ((ifp1 = ifunit(altq->ifname)) == NULL ||
1162 (remove && ifp1 == ifp)) {
1163 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
1164 } else {
1165 error = altq_add(ifp1, altq);
1166
1167 if (ticket != V_ticket_altqs_inactive)
1168 error = EBUSY;
1169
1170 if (error)
1171 free(altq, M_PFALTQ);
1172 }
1173
1174 return (error);
1175 }
1176
1177 void
pf_altq_ifnet_event(struct ifnet * ifp,int remove)1178 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
1179 {
1180 struct pf_altq *a1, *a2, *a3;
1181 u_int32_t ticket;
1182 int error = 0;
1183
1184 /*
1185 * No need to re-evaluate the configuration for events on interfaces
1186 * that do not support ALTQ, as it's not possible for such
1187 * interfaces to be part of the configuration.
1188 */
1189 if (!ALTQ_IS_READY(&ifp->if_snd))
1190 return;
1191
1192 /* Interrupt userland queue modifications */
1193 if (V_altqs_inactive_open)
1194 pf_rollback_altq(V_ticket_altqs_inactive);
1195
1196 /* Start new altq ruleset */
1197 if (pf_begin_altq(&ticket))
1198 return;
1199
1200 /* Copy the current active set */
1201 TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
1202 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1203 if (a2 == NULL) {
1204 error = ENOMEM;
1205 break;
1206 }
1207 bcopy(a1, a2, sizeof(struct pf_altq));
1208
1209 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1210 if (error)
1211 break;
1212
1213 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
1214 }
1215 if (error)
1216 goto out;
1217 TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
1218 a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
1219 if (a2 == NULL) {
1220 error = ENOMEM;
1221 break;
1222 }
1223 bcopy(a1, a2, sizeof(struct pf_altq));
1224
1225 if ((a2->qid = pf_qname2qid(a2->qname, true)) == 0) {
1226 error = EBUSY;
1227 free(a2, M_PFALTQ);
1228 break;
1229 }
1230 a2->altq_disc = NULL;
1231 TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
1232 if (strncmp(a3->ifname, a2->ifname,
1233 IFNAMSIZ) == 0) {
1234 a2->altq_disc = a3->altq_disc;
1235 break;
1236 }
1237 }
1238 error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
1239 if (error)
1240 break;
1241
1242 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
1243 }
1244
1245 out:
1246 if (error != 0)
1247 pf_rollback_altq(ticket);
1248 else
1249 pf_commit_altq(ticket);
1250 }
1251 #endif /* ALTQ */
1252
1253 static struct pf_krule_global *
pf_rule_tree_alloc(int flags)1254 pf_rule_tree_alloc(int flags)
1255 {
1256 struct pf_krule_global *tree;
1257
1258 tree = malloc(sizeof(struct pf_krule_global), M_PF, flags);
1259 if (tree == NULL)
1260 return (NULL);
1261 RB_INIT(tree);
1262 return (tree);
1263 }
1264
1265 void
pf_rule_tree_free(struct pf_krule_global * tree)1266 pf_rule_tree_free(struct pf_krule_global *tree)
1267 {
1268
1269 free(tree, M_PF);
1270 }
1271
1272 static int
pf_begin_rules(u_int32_t * ticket,int rs_num,const char * anchor)1273 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1274 {
1275 struct pf_krule_global *tree;
1276 struct pf_kruleset *rs;
1277 struct pf_krule *rule;
1278
1279 PF_RULES_WASSERT();
1280
1281 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1282 return (EINVAL);
1283 tree = pf_rule_tree_alloc(M_NOWAIT);
1284 if (tree == NULL)
1285 return (ENOMEM);
1286 rs = pf_find_or_create_kruleset(anchor);
1287 if (rs == NULL) {
1288 pf_rule_tree_free(tree);
1289 return (EINVAL);
1290 }
1291 pf_rule_tree_free(rs->rules[rs_num].inactive.tree);
1292 rs->rules[rs_num].inactive.tree = tree;
1293
1294 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1295 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1296 rs->rules[rs_num].inactive.rcount--;
1297 }
1298 *ticket = ++rs->rules[rs_num].inactive.ticket;
1299 rs->rules[rs_num].inactive.open = 1;
1300 return (0);
1301 }
1302
1303 static int
pf_rollback_rules(u_int32_t ticket,int rs_num,char * anchor)1304 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1305 {
1306 struct pf_kruleset *rs;
1307 struct pf_krule *rule;
1308
1309 PF_RULES_WASSERT();
1310
1311 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1312 return (EINVAL);
1313 rs = pf_find_kruleset(anchor);
1314 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1315 rs->rules[rs_num].inactive.ticket != ticket)
1316 return (0);
1317 while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
1318 pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
1319 rs->rules[rs_num].inactive.rcount--;
1320 }
1321 rs->rules[rs_num].inactive.open = 0;
1322
1323 if (anchor[0])
1324 return (0);
1325
1326 pf_statelim_rollback();
1327 pf_sourcelim_rollback();
1328 return (0);
1329 }
1330
1331 #define PF_MD5_UPD(st, elm) \
1332 MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
1333
1334 #define PF_MD5_UPD_STR(st, elm) \
1335 MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
1336
1337 #define PF_MD5_UPD_HTONL(st, elm, stor) do { \
1338 (stor) = htonl((st)->elm); \
1339 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
1340 } while (0)
1341
1342 #define PF_MD5_UPD_HTONS(st, elm, stor) do { \
1343 (stor) = htons((st)->elm); \
1344 MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
1345 } while (0)
1346
1347 static void
pf_hash_rule_addr(MD5_CTX * ctx,struct pf_rule_addr * pfr)1348 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1349 {
1350 PF_MD5_UPD(pfr, addr.type);
1351 switch (pfr->addr.type) {
1352 case PF_ADDR_DYNIFTL:
1353 PF_MD5_UPD(pfr, addr.v.ifname);
1354 PF_MD5_UPD(pfr, addr.iflags);
1355 break;
1356 case PF_ADDR_TABLE:
1357 if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX,
1358 strlen(PF_OPTIMIZER_TABLE_PFX)))
1359 PF_MD5_UPD(pfr, addr.v.tblname);
1360 break;
1361 case PF_ADDR_ADDRMASK:
1362 case PF_ADDR_RANGE:
1363 /* XXX ignore af? */
1364 PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1365 PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1366 break;
1367 case PF_ADDR_NONE:
1368 case PF_ADDR_NOROUTE:
1369 case PF_ADDR_URPFFAILED:
1370 /* These do not use any address data. */
1371 break;
1372 default:
1373 panic("Unknown address type %d", pfr->addr.type);
1374 }
1375
1376 PF_MD5_UPD(pfr, port[0]);
1377 PF_MD5_UPD(pfr, port[1]);
1378 PF_MD5_UPD(pfr, neg);
1379 PF_MD5_UPD(pfr, port_op);
1380 }
1381
1382 static void
pf_hash_pool(MD5_CTX * ctx,struct pf_kpool * pool)1383 pf_hash_pool(MD5_CTX *ctx, struct pf_kpool *pool)
1384 {
1385 uint16_t x;
1386 int y;
1387
1388 if (pool->cur) {
1389 PF_MD5_UPD(pool, cur->addr);
1390 PF_MD5_UPD_STR(pool, cur->ifname);
1391 PF_MD5_UPD(pool, cur->af);
1392 }
1393 PF_MD5_UPD(pool, key);
1394 PF_MD5_UPD(pool, counter);
1395
1396 PF_MD5_UPD(pool, mape.offset);
1397 PF_MD5_UPD(pool, mape.psidlen);
1398 PF_MD5_UPD_HTONS(pool, mape.psid, x);
1399 PF_MD5_UPD_HTONL(pool, tblidx, y);
1400 PF_MD5_UPD_HTONS(pool, proxy_port[0], x);
1401 PF_MD5_UPD_HTONS(pool, proxy_port[1], x);
1402 PF_MD5_UPD(pool, opts);
1403 PF_MD5_UPD(pool, ipv6_nexthop_af);
1404 }
1405
1406 static void
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)1407 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1408 {
1409 u_int16_t x;
1410 u_int32_t y;
1411
1412 pf_hash_rule_addr(ctx, &rule->src);
1413 pf_hash_rule_addr(ctx, &rule->dst);
1414 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1415 PF_MD5_UPD_STR(rule, label[i]);
1416 PF_MD5_UPD_HTONL(rule, ridentifier, y);
1417 PF_MD5_UPD_STR(rule, ifname);
1418 PF_MD5_UPD_STR(rule, rcv_ifname);
1419 PF_MD5_UPD_STR(rule, qname);
1420 PF_MD5_UPD_STR(rule, pqname);
1421 PF_MD5_UPD_STR(rule, tagname);
1422 PF_MD5_UPD_STR(rule, match_tagname);
1423
1424 PF_MD5_UPD_STR(rule, overload_tblname);
1425
1426 pf_hash_pool(ctx, &rule->nat);
1427 pf_hash_pool(ctx, &rule->rdr);
1428 pf_hash_pool(ctx, &rule->route);
1429 PF_MD5_UPD_HTONL(rule, pktrate.limit, y);
1430 PF_MD5_UPD_HTONL(rule, pktrate.seconds, y);
1431
1432 PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1433
1434 PF_MD5_UPD_HTONL(rule, rtableid, y);
1435 for (int i = 0; i < PFTM_MAX; i++)
1436 PF_MD5_UPD_HTONL(rule, timeout[i], y);
1437 PF_MD5_UPD_HTONL(rule, max_states, y);
1438 PF_MD5_UPD_HTONL(rule, max_src_nodes, y);
1439 PF_MD5_UPD_HTONL(rule, max_src_states, y);
1440 PF_MD5_UPD_HTONL(rule, max_src_conn, y);
1441 PF_MD5_UPD_HTONL(rule, max_src_conn_rate.limit, y);
1442 PF_MD5_UPD_HTONL(rule, max_src_conn_rate.seconds, y);
1443 PF_MD5_UPD_HTONS(rule, max_pkt_size, y);
1444 PF_MD5_UPD_HTONS(rule, qid, x);
1445 PF_MD5_UPD_HTONS(rule, pqid, x);
1446 PF_MD5_UPD_HTONS(rule, dnpipe, x);
1447 PF_MD5_UPD_HTONS(rule, dnrpipe, x);
1448 PF_MD5_UPD_HTONL(rule, free_flags, y);
1449 PF_MD5_UPD_HTONL(rule, prob, y);
1450
1451 PF_MD5_UPD_HTONS(rule, return_icmp, x);
1452 PF_MD5_UPD_HTONS(rule, return_icmp6, x);
1453 PF_MD5_UPD_HTONS(rule, max_mss, x);
1454 PF_MD5_UPD_HTONS(rule, tag, x); /* dup? */
1455 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1456 PF_MD5_UPD_HTONS(rule, scrub_flags, x);
1457
1458 PF_MD5_UPD(rule, uid.op);
1459 PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1460 PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1461 PF_MD5_UPD(rule, gid.op);
1462 PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1463 PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1464
1465 PF_MD5_UPD_HTONL(rule, rule_flag, y);
1466 PF_MD5_UPD_HTONL(rule, rule_ref, y);
1467 PF_MD5_UPD(rule, action);
1468 PF_MD5_UPD(rule, direction);
1469 PF_MD5_UPD(rule, log);
1470 PF_MD5_UPD(rule, logif);
1471 PF_MD5_UPD(rule, quick);
1472 PF_MD5_UPD(rule, ifnot);
1473 PF_MD5_UPD(rule, match_tag_not);
1474 PF_MD5_UPD(rule, natpass);
1475
1476 PF_MD5_UPD(rule, keep_state);
1477 PF_MD5_UPD(rule, af);
1478 PF_MD5_UPD(rule, proto);
1479 PF_MD5_UPD_HTONS(rule, type, x);
1480 PF_MD5_UPD_HTONS(rule, code, x);
1481 PF_MD5_UPD(rule, flags);
1482 PF_MD5_UPD(rule, flagset);
1483 PF_MD5_UPD(rule, min_ttl);
1484 PF_MD5_UPD(rule, allow_opts);
1485 PF_MD5_UPD(rule, rt);
1486 PF_MD5_UPD(rule, return_ttl);
1487 PF_MD5_UPD(rule, tos);
1488 PF_MD5_UPD(rule, set_tos);
1489 PF_MD5_UPD(rule, anchor_relative);
1490 PF_MD5_UPD(rule, anchor_wildcard);
1491
1492 PF_MD5_UPD(rule, flush);
1493 PF_MD5_UPD(rule, prio);
1494 PF_MD5_UPD(rule, set_prio[0]);
1495 PF_MD5_UPD(rule, set_prio[1]);
1496 PF_MD5_UPD(rule, naf);
1497 PF_MD5_UPD(rule, rcvifnot);
1498 PF_MD5_UPD(rule, statelim.id);
1499 PF_MD5_UPD_HTONL(rule, statelim.limiter_action, y);
1500 PF_MD5_UPD(rule, sourcelim.id);
1501 PF_MD5_UPD_HTONL(rule, sourcelim.limiter_action, y);
1502
1503 PF_MD5_UPD(rule, divert.addr);
1504 PF_MD5_UPD_HTONS(rule, divert.port, x);
1505
1506 if (rule->anchor != NULL)
1507 PF_MD5_UPD_STR(rule, anchor->path);
1508 }
1509
1510 static void
pf_hash_rule(struct pf_krule * rule)1511 pf_hash_rule(struct pf_krule *rule)
1512 {
1513 MD5_CTX ctx;
1514
1515 MD5Init(&ctx);
1516 pf_hash_rule_rolling(&ctx, rule);
1517 MD5Final(rule->md5sum, &ctx);
1518 }
1519
1520 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1521 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1522 {
1523
1524 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1525 }
1526
1527 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1528 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1529 {
1530 struct pf_kruleset *rs;
1531 struct pf_krule *rule, *old_rule;
1532 struct pf_krulequeue *old_rules;
1533 struct pf_krule_global *old_tree;
1534 int error;
1535 u_int32_t old_rcount;
1536 bool is_main_ruleset = anchor[0] == '\0';
1537
1538 PF_RULES_WASSERT();
1539
1540 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1541 return (EINVAL);
1542 rs = pf_find_kruleset(anchor);
1543 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1544 ticket != rs->rules[rs_num].inactive.ticket)
1545 return (EBUSY);
1546
1547 /* Calculate checksum for the main ruleset */
1548 if (rs == &pf_main_ruleset) {
1549 error = pf_sourcelim_check();
1550 if (error != 0)
1551 return (error);
1552 error = pf_setup_pfsync_matching(rs);
1553 if (error != 0)
1554 return (error);
1555 }
1556
1557 /* Swap rules, keep the old. */
1558 old_rules = rs->rules[rs_num].active.ptr;
1559 old_rcount = rs->rules[rs_num].active.rcount;
1560 old_tree = rs->rules[rs_num].active.tree;
1561
1562 rs->rules[rs_num].active.ptr =
1563 rs->rules[rs_num].inactive.ptr;
1564 rs->rules[rs_num].active.tree =
1565 rs->rules[rs_num].inactive.tree;
1566 rs->rules[rs_num].active.rcount =
1567 rs->rules[rs_num].inactive.rcount;
1568
1569 /* Attempt to preserve counter information. */
1570 if (V_pf_status.keep_counters && old_tree != NULL) {
1571 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1572 entries) {
1573 old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1574 if (old_rule == NULL) {
1575 continue;
1576 }
1577 pf_counter_u64_critical_enter();
1578 pf_counter_u64_rollup_protected(&rule->evaluations,
1579 pf_counter_u64_fetch(&old_rule->evaluations));
1580 pf_counter_u64_rollup_protected(&rule->packets[0],
1581 pf_counter_u64_fetch(&old_rule->packets[0]));
1582 pf_counter_u64_rollup_protected(&rule->packets[1],
1583 pf_counter_u64_fetch(&old_rule->packets[1]));
1584 pf_counter_u64_rollup_protected(&rule->bytes[0],
1585 pf_counter_u64_fetch(&old_rule->bytes[0]));
1586 pf_counter_u64_rollup_protected(&rule->bytes[1],
1587 pf_counter_u64_fetch(&old_rule->bytes[1]));
1588 pf_counter_u64_critical_exit();
1589 }
1590 }
1591
1592 rs->rules[rs_num].inactive.ptr = old_rules;
1593 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1594 rs->rules[rs_num].inactive.rcount = old_rcount;
1595
1596 rs->rules[rs_num].active.ticket =
1597 rs->rules[rs_num].inactive.ticket;
1598 pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1599
1600 /* Purge the old rule list. */
1601 PF_UNLNKDRULES_LOCK();
1602 while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1603 pf_unlink_rule_locked(old_rules, rule);
1604 PF_UNLNKDRULES_UNLOCK();
1605 rs->rules[rs_num].inactive.rcount = 0;
1606 rs->rules[rs_num].inactive.open = 0;
1607 pf_remove_if_empty_kruleset(rs);
1608 pf_rule_tree_free(old_tree);
1609
1610 /* statelim/sourcelim/queue defs only in the main ruleset */
1611 if (! is_main_ruleset || rs_num != PF_RULESET_FILTER)
1612 return (0);
1613
1614 pf_statelim_commit();
1615 pf_sourcelim_commit();
1616
1617 return (0);
1618 }
1619
1620 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1621 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1622 {
1623 MD5_CTX ctx;
1624 struct pf_krule *rule;
1625 int rs_cnt;
1626 u_int8_t digest[PF_MD5_DIGEST_LENGTH];
1627
1628 MD5Init(&ctx);
1629 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1630 /* XXX PF_RULESET_SCRUB as well? */
1631 if (rs_cnt == PF_RULESET_SCRUB)
1632 continue;
1633
1634 if (rs->rules[rs_cnt].inactive.rcount) {
1635 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1636 entries) {
1637 pf_hash_rule_rolling(&ctx, rule);
1638 }
1639 }
1640 }
1641
1642 MD5Final(digest, &ctx);
1643 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1644 return (0);
1645 }
1646
1647 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1648 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1649 {
1650 int error = 0;
1651
1652 switch (addr->type) {
1653 case PF_ADDR_TABLE:
1654 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1655 if (addr->p.tbl == NULL)
1656 error = ENOMEM;
1657 break;
1658 default:
1659 error = EINVAL;
1660 }
1661
1662 return (error);
1663 }
1664
1665 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1666 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1667 sa_family_t af)
1668 {
1669 int error = 0;
1670
1671 switch (addr->type) {
1672 case PF_ADDR_TABLE:
1673 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1674 if (addr->p.tbl == NULL)
1675 error = ENOMEM;
1676 break;
1677 case PF_ADDR_DYNIFTL:
1678 error = pfi_dynaddr_setup(addr, af);
1679 break;
1680 }
1681
1682 return (error);
1683 }
1684
1685 void
pf_addr_copyout(struct pf_addr_wrap * addr)1686 pf_addr_copyout(struct pf_addr_wrap *addr)
1687 {
1688
1689 switch (addr->type) {
1690 case PF_ADDR_DYNIFTL:
1691 pfi_dynaddr_copyout(addr);
1692 break;
1693 case PF_ADDR_TABLE:
1694 pf_tbladdr_copyout(addr);
1695 break;
1696 }
1697 }
1698
1699 int
pf_statelim_add(const struct pfioc_statelim * ioc)1700 pf_statelim_add(const struct pfioc_statelim *ioc)
1701 {
1702 struct pf_statelim *pfstlim;
1703 int error;
1704 size_t namelen;
1705
1706 if (ioc->id < PF_STATELIM_ID_MIN ||
1707 ioc->id > PF_STATELIM_ID_MAX)
1708 return (EINVAL);
1709
1710 if (ioc->limit < PF_STATELIM_LIMIT_MIN ||
1711 ioc->limit > PF_STATELIM_LIMIT_MAX)
1712 return (EINVAL);
1713
1714 if ((ioc->rate.limit == 0) != (ioc->rate.seconds == 0))
1715 return (EINVAL);
1716
1717 namelen = strnlen(ioc->name, sizeof(ioc->name));
1718 /* is the name from userland nul terminated? */
1719 if (namelen == sizeof(ioc->name))
1720 return (EINVAL);
1721
1722 pfstlim = malloc(sizeof(*pfstlim), M_PF_STATE_LIM, M_WAITOK | M_ZERO);
1723 if (pfstlim == NULL)
1724 return (ENOMEM);
1725
1726 pfstlim->pfstlim_id = ioc->id;
1727 if (strlcpy(pfstlim->pfstlim_nm, ioc->name,
1728 sizeof(pfstlim->pfstlim_nm)) >= sizeof(pfstlim->pfstlim_nm)) {
1729 error = EINVAL;
1730 goto free;
1731 }
1732 pfstlim->pfstlim_limit = ioc->limit;
1733 pfstlim->pfstlim_rate.limit = ioc->rate.limit;
1734 pfstlim->pfstlim_rate.seconds = ioc->rate.seconds;
1735
1736 if (pfstlim->pfstlim_rate.limit) {
1737 uint64_t bucket = SEC_TO_NSEC(pfstlim->pfstlim_rate.seconds);
1738 struct timespec ts;
1739
1740 getnanouptime(&ts);
1741
1742 pfstlim->pfstlim_rate_ts = SEC_TO_NSEC(ts.tv_sec) + ts.tv_nsec -
1743 bucket;
1744 pfstlim->pfstlim_rate_token = bucket /
1745 pfstlim->pfstlim_rate.limit;
1746 pfstlim->pfstlim_rate_bucket = bucket;
1747 }
1748
1749 TAILQ_INIT(&pfstlim->pfstlim_states);
1750 mtx_init(&pfstlim->pfstlim_lock, "pf state limit", NULL, MTX_DEF);
1751
1752 PF_RULES_WLOCK();
1753 if (ioc->ticket != pf_main_ruleset.rules[PF_RULESET_FILTER].inactive.ticket) {
1754 error = EBUSY;
1755 goto unlock;
1756 }
1757
1758 if (RB_INSERT(pf_statelim_id_tree, &V_pf_statelim_id_tree_inactive,
1759 pfstlim) != NULL) {
1760 error = EBUSY;
1761 goto unlock;
1762 }
1763
1764 if (RB_INSERT(pf_statelim_nm_tree, &V_pf_statelim_nm_tree_inactive,
1765 pfstlim) != NULL) {
1766 RB_REMOVE(pf_statelim_id_tree, &V_pf_statelim_id_tree_inactive,
1767 pfstlim);
1768 error = EBUSY;
1769 goto unlock;
1770 }
1771
1772 TAILQ_INSERT_HEAD(&V_pf_statelim_list_inactive, pfstlim, pfstlim_list);
1773
1774 PF_RULES_WUNLOCK();
1775
1776 return (0);
1777
1778 unlock:
1779 PF_RULES_WUNLOCK();
1780
1781 free:
1782 free(pfstlim, M_PF_STATE_LIM);
1783
1784 return (error);
1785 }
1786
1787 static void
pf_statelim_unlink(struct pf_statelim * pfstlim,struct pf_state_link_list * garbage)1788 pf_statelim_unlink(struct pf_statelim *pfstlim,
1789 struct pf_state_link_list *garbage)
1790 {
1791 struct pf_state_link *pfl;
1792
1793
1794 /* unwire the links */
1795 TAILQ_FOREACH(pfl, &pfstlim->pfstlim_states, pfl_link) {
1796 struct pf_kstate *s = pfl->pfl_state;
1797
1798 /* if !rmst */
1799 PF_STATE_LOCK(s);
1800 s->statelim = 0;
1801 SLIST_REMOVE(&s->linkage, pfl, pf_state_link, pfl_linkage);
1802 PF_STATE_UNLOCK(s);
1803 }
1804
1805 /* take the list away */
1806 TAILQ_CONCAT(garbage, &pfstlim->pfstlim_states, pfl_link);
1807 pfstlim->pfstlim_inuse = 0;
1808 }
1809
1810 void
pf_statelim_commit(void)1811 pf_statelim_commit(void)
1812 {
1813 struct pf_statelim *pfstlim, *npfstlim, *opfstlim;
1814 struct pf_statelim_list l = TAILQ_HEAD_INITIALIZER(l);
1815 struct pf_state_link_list garbage = TAILQ_HEAD_INITIALIZER(garbage);
1816 struct pf_state_link *pfl, *npfl;
1817
1818 PF_RULES_WASSERT();
1819
1820 /* merge the new statelims into the current set */
1821
1822 /* start with an empty active list */
1823 TAILQ_CONCAT(&l, &V_pf_statelim_list_active, pfstlim_list);
1824
1825 /* beware, the inactive bits gets messed up here */
1826
1827 /* try putting pending statelims into the active tree */
1828 TAILQ_FOREACH_SAFE(pfstlim, &V_pf_statelim_list_inactive, pfstlim_list,
1829 npfstlim) {
1830 opfstlim = RB_INSERT(pf_statelim_id_tree,
1831 &V_pf_statelim_id_tree_active, pfstlim);
1832 if (opfstlim != NULL) {
1833 /* this statelim already exists, merge */
1834 opfstlim->pfstlim_limit = pfstlim->pfstlim_limit;
1835 opfstlim->pfstlim_rate.limit =
1836 pfstlim->pfstlim_rate.limit;
1837 opfstlim->pfstlim_rate.seconds =
1838 pfstlim->pfstlim_rate.seconds;
1839
1840 opfstlim->pfstlim_rate_ts = pfstlim->pfstlim_rate_ts;
1841 opfstlim->pfstlim_rate_token =
1842 pfstlim->pfstlim_rate_token;
1843 opfstlim->pfstlim_rate_bucket =
1844 pfstlim->pfstlim_rate_bucket;
1845
1846 memcpy(opfstlim->pfstlim_nm, pfstlim->pfstlim_nm,
1847 sizeof(opfstlim->pfstlim_nm));
1848
1849 /* use the existing statelim instead */
1850 free(pfstlim, M_PF_STATE_LIM);
1851 TAILQ_REMOVE(&l, opfstlim, pfstlim_list);
1852 pfstlim = opfstlim;
1853 }
1854
1855 TAILQ_INSERT_TAIL(&V_pf_statelim_list_active, pfstlim,
1856 pfstlim_list);
1857 }
1858
1859 /* clean up the now unused statelims from the old set */
1860 TAILQ_FOREACH_SAFE(pfstlim, &l, pfstlim_list, npfstlim) {
1861 pf_statelim_unlink(pfstlim, &garbage);
1862
1863 RB_REMOVE(pf_statelim_id_tree, &V_pf_statelim_id_tree_active,
1864 pfstlim);
1865
1866 free(pfstlim, M_PF_STATE_LIM);
1867 }
1868
1869 /* fix up the inactive tree */
1870 RB_INIT(&V_pf_statelim_id_tree_inactive);
1871 RB_INIT(&V_pf_statelim_nm_tree_inactive);
1872 TAILQ_INIT(&V_pf_statelim_list_inactive);
1873
1874 TAILQ_FOREACH_SAFE(pfl, &garbage, pfl_link, npfl)
1875 free(pfl, M_PF_STATE_LINK);
1876 }
1877
1878 static void
pf_sourcelim_unlink(struct pf_sourcelim * pfsrlim,struct pf_state_link_list * garbage)1879 pf_sourcelim_unlink(struct pf_sourcelim *pfsrlim,
1880 struct pf_state_link_list *garbage)
1881 {
1882 extern struct pf_source_list pf_source_gc;
1883 struct pf_source *pfsr;
1884 struct pf_state_link *pfl;
1885
1886 PF_RULES_WASSERT();
1887
1888 while ((pfsr = RB_ROOT(&pfsrlim->pfsrlim_sources)) != NULL) {
1889 RB_REMOVE(pf_source_tree, &pfsrlim->pfsrlim_sources, pfsr);
1890 RB_REMOVE(pf_source_ioc_tree, &pfsrlim->pfsrlim_ioc_sources,
1891 pfsr);
1892 if (pfsr->pfsr_inuse == 0)
1893 TAILQ_REMOVE(&pf_source_gc, pfsr, pfsr_empty_gc);
1894
1895 /* unwire the links */
1896 TAILQ_FOREACH(pfl, &pfsr->pfsr_states, pfl_link) {
1897 struct pf_kstate *s = pfl->pfl_state;
1898
1899 PF_STATE_LOCK(s);
1900 /* if !rmst */
1901 s->sourcelim = 0;
1902 SLIST_REMOVE(&s->linkage, pfl, pf_state_link,
1903 pfl_linkage);
1904 PF_STATE_UNLOCK(s);
1905 }
1906
1907 /* take the list away */
1908 TAILQ_CONCAT(garbage, &pfsr->pfsr_states, pfl_link);
1909
1910 free(pfsr, M_PF_SOURCE_LIM);
1911 }
1912 }
1913
1914 int
pf_sourcelim_check(void)1915 pf_sourcelim_check(void)
1916 {
1917 struct pf_sourcelim *pfsrlim, *npfsrlim;
1918
1919 PF_RULES_WASSERT();
1920
1921 /* check if we can merge */
1922
1923 TAILQ_FOREACH(pfsrlim, &V_pf_sourcelim_list_inactive, pfsrlim_list) {
1924 npfsrlim = RB_FIND(pf_sourcelim_id_tree,
1925 &V_pf_sourcelim_id_tree_active, pfsrlim);
1926
1927 /* new config, no conflict */
1928 if (npfsrlim == NULL)
1929 continue;
1930
1931 /* nothing is tracked at the moment, no conflict */
1932 if (RB_EMPTY(&npfsrlim->pfsrlim_sources))
1933 continue;
1934
1935 if (strcmp(npfsrlim->pfsrlim_overload.name,
1936 pfsrlim->pfsrlim_overload.name) != 0)
1937 return (EBUSY);
1938
1939 /*
1940 * we should allow the prefixlens to get shorter
1941 * and merge pf_source entries.
1942 */
1943
1944 if ((npfsrlim->pfsrlim_ipv4_prefix !=
1945 pfsrlim->pfsrlim_ipv4_prefix) ||
1946 (npfsrlim->pfsrlim_ipv6_prefix !=
1947 pfsrlim->pfsrlim_ipv6_prefix))
1948 return (EBUSY);
1949 }
1950
1951 return (0);
1952 }
1953
1954 void
pf_sourcelim_commit(void)1955 pf_sourcelim_commit(void)
1956 {
1957 struct pf_sourcelim *pfsrlim, *npfsrlim, *opfsrlim;
1958 struct pf_sourcelim_list l = TAILQ_HEAD_INITIALIZER(l);
1959 struct pf_state_link_list garbage = TAILQ_HEAD_INITIALIZER(garbage);
1960 struct pf_state_link *pfl, *npfl;
1961
1962 PF_RULES_WASSERT();
1963
1964 /* merge the new sourcelims into the current set */
1965
1966 /* start with an empty active list */
1967 TAILQ_CONCAT(&l, &V_pf_sourcelim_list_active, pfsrlim_list);
1968
1969 /* beware, the inactive bits gets messed up here */
1970
1971 /* try putting pending sourcelims into the active tree */
1972 TAILQ_FOREACH_SAFE(pfsrlim, &V_pf_sourcelim_list_inactive, pfsrlim_list,
1973 npfsrlim) {
1974 opfsrlim = RB_INSERT(pf_sourcelim_id_tree,
1975 &V_pf_sourcelim_id_tree_active, pfsrlim);
1976 if (opfsrlim != NULL) {
1977 /* this sourcelim already exists, merge */
1978 opfsrlim->pfsrlim_entries = pfsrlim->pfsrlim_entries;
1979 opfsrlim->pfsrlim_limit = pfsrlim->pfsrlim_limit;
1980 opfsrlim->pfsrlim_ipv4_prefix =
1981 pfsrlim->pfsrlim_ipv4_prefix;
1982 opfsrlim->pfsrlim_ipv6_prefix =
1983 pfsrlim->pfsrlim_ipv6_prefix;
1984 opfsrlim->pfsrlim_rate.limit =
1985 pfsrlim->pfsrlim_rate.limit;
1986 opfsrlim->pfsrlim_rate.seconds =
1987 pfsrlim->pfsrlim_rate.seconds;
1988
1989 opfsrlim->pfsrlim_ipv4_mask =
1990 pfsrlim->pfsrlim_ipv4_mask;
1991 opfsrlim->pfsrlim_ipv6_mask =
1992 pfsrlim->pfsrlim_ipv6_mask;
1993
1994 /* keep the existing pfstlim_rate_ts */
1995
1996 opfsrlim->pfsrlim_rate_token =
1997 pfsrlim->pfsrlim_rate_token;
1998 opfsrlim->pfsrlim_rate_bucket =
1999 pfsrlim->pfsrlim_rate_bucket;
2000
2001 if (opfsrlim->pfsrlim_overload.table != NULL) {
2002 pfr_detach_table(
2003 opfsrlim->pfsrlim_overload.table);
2004 }
2005
2006 strlcpy(opfsrlim->pfsrlim_overload.name,
2007 pfsrlim->pfsrlim_overload.name,
2008 sizeof(opfsrlim->pfsrlim_overload.name));
2009 opfsrlim->pfsrlim_overload.hwm =
2010 pfsrlim->pfsrlim_overload.hwm;
2011 opfsrlim->pfsrlim_overload.lwm =
2012 pfsrlim->pfsrlim_overload.lwm;
2013 opfsrlim->pfsrlim_overload.table =
2014 pfsrlim->pfsrlim_overload.table,
2015
2016 memcpy(opfsrlim->pfsrlim_nm, pfsrlim->pfsrlim_nm,
2017 sizeof(opfsrlim->pfsrlim_nm));
2018
2019 /* use the existing sourcelim instead */
2020 free(pfsrlim, M_PF_SOURCE_LIM);
2021 TAILQ_REMOVE(&l, opfsrlim, pfsrlim_list);
2022 pfsrlim = opfsrlim;
2023 }
2024
2025 TAILQ_INSERT_TAIL(&V_pf_sourcelim_list_active, pfsrlim,
2026 pfsrlim_list);
2027 }
2028
2029 /* clean up the now unused sourcelims from the old set */
2030 TAILQ_FOREACH_SAFE(pfsrlim, &l, pfsrlim_list, npfsrlim) {
2031 pf_sourcelim_unlink(pfsrlim, &garbage);
2032
2033 RB_REMOVE(pf_sourcelim_id_tree, &V_pf_sourcelim_id_tree_active,
2034 pfsrlim);
2035
2036 if (pfsrlim->pfsrlim_overload.table != NULL)
2037 pfr_detach_table(pfsrlim->pfsrlim_overload.table);
2038
2039 free(pfsrlim, M_PF_SOURCE_LIM);
2040 }
2041
2042 /* fix up the inactive tree */
2043 RB_INIT(&V_pf_sourcelim_id_tree_inactive);
2044 RB_INIT(&V_pf_sourcelim_nm_tree_inactive);
2045 TAILQ_INIT(&V_pf_sourcelim_list_inactive);
2046
2047 TAILQ_FOREACH_SAFE(pfl, &garbage, pfl_link, npfl)
2048 free(pfl, M_PF_STATE_LINK);
2049 }
2050
2051 void
pf_statelim_rollback(void)2052 pf_statelim_rollback(void)
2053 {
2054 struct pf_statelim *pfstlim, *npfstlim;
2055
2056 PF_RULES_WASSERT();
2057
2058 TAILQ_FOREACH_SAFE(pfstlim, &V_pf_statelim_list_inactive, pfstlim_list,
2059 npfstlim)
2060 free(pfstlim, M_PF_STATE_LIM);
2061
2062 TAILQ_INIT(&V_pf_statelim_list_inactive);
2063 RB_INIT(&V_pf_statelim_id_tree_inactive);
2064 RB_INIT(&V_pf_statelim_nm_tree_inactive);
2065 }
2066
2067 struct pf_statelim *
pf_statelim_rb_find(struct pf_statelim_id_tree * tree,struct pf_statelim * key)2068 pf_statelim_rb_find(struct pf_statelim_id_tree *tree, struct pf_statelim *key)
2069 {
2070 PF_RULES_ASSERT();
2071
2072 return (RB_FIND(pf_statelim_id_tree, tree, key));
2073 }
2074
2075 struct pf_statelim *
pf_statelim_rb_nfind(struct pf_statelim_id_tree * tree,struct pf_statelim * key)2076 pf_statelim_rb_nfind(struct pf_statelim_id_tree *tree, struct pf_statelim *key)
2077 {
2078 PF_RULES_ASSERT();
2079
2080 return (RB_NFIND(pf_statelim_id_tree, tree, key));
2081 }
2082
2083 int
pf_statelim_get(struct pfioc_statelim * ioc,struct pf_statelim * (* rbt_op)(struct pf_statelim_id_tree *,struct pf_statelim *))2084 pf_statelim_get(struct pfioc_statelim *ioc,
2085 struct pf_statelim *(*rbt_op)(struct pf_statelim_id_tree *,
2086 struct pf_statelim *))
2087 {
2088 struct pf_statelim key = { .pfstlim_id = ioc->id };
2089 struct pf_statelim *pfstlim;
2090 int error = 0;
2091 PF_RULES_RLOCK_TRACKER;
2092
2093 PF_RULES_RLOCK();
2094
2095 pfstlim = (*rbt_op)(&V_pf_statelim_id_tree_active, &key);
2096 if (pfstlim == NULL) {
2097 error = ENOENT;
2098 goto unlock;
2099 }
2100
2101 ioc->id = pfstlim->pfstlim_id;
2102 ioc->limit = pfstlim->pfstlim_limit;
2103 ioc->rate.limit = pfstlim->pfstlim_rate.limit;
2104 ioc->rate.seconds = pfstlim->pfstlim_rate.seconds;
2105 CTASSERT(sizeof(ioc->name) == sizeof(pfstlim->pfstlim_nm));
2106 memcpy(ioc->name, pfstlim->pfstlim_nm, sizeof(ioc->name));
2107
2108 ioc->inuse = pfstlim->pfstlim_inuse;
2109 ioc->admitted = pfstlim->pfstlim_counters.admitted;
2110 ioc->hardlimited = pfstlim->pfstlim_counters.hardlimited;
2111 ioc->ratelimited = pfstlim->pfstlim_counters.ratelimited;
2112
2113 unlock:
2114 PF_RULES_RUNLOCK();
2115
2116 return (error);
2117 }
2118
2119 int
pf_sourcelim_add(const struct pfioc_sourcelim * ioc)2120 pf_sourcelim_add(const struct pfioc_sourcelim *ioc)
2121 {
2122 struct pf_sourcelim *pfsrlim;
2123 int error;
2124 size_t namelen, tablelen;
2125 unsigned int prefix;
2126 size_t i;
2127
2128 if (ioc->id < PF_SOURCELIM_ID_MIN ||
2129 ioc->id > PF_SOURCELIM_ID_MAX)
2130 return (EINVAL);
2131
2132 if (ioc->entries < 1)
2133 return (EINVAL);
2134
2135 if (ioc->limit < 1)
2136 return (EINVAL);
2137
2138 if ((ioc->rate.limit == 0) != (ioc->rate.seconds == 0))
2139 return (EINVAL);
2140
2141 if (ioc->inet_prefix > 32)
2142 return (EINVAL);
2143 if (ioc->inet6_prefix > 128)
2144 return (EINVAL);
2145
2146 namelen = strnlen(ioc->name, sizeof(ioc->name));
2147 /* is the name from userland nul terminated? */
2148 if (namelen == sizeof(ioc->name))
2149 return (EINVAL);
2150
2151 tablelen = strnlen(ioc->overload_tblname,
2152 sizeof(ioc->overload_tblname));
2153 /* is the name from userland nul terminated? */
2154 if (tablelen == sizeof(ioc->overload_tblname))
2155 return (EINVAL);
2156 if (tablelen != 0) {
2157 if (ioc->overload_hwm == 0)
2158 return (EINVAL);
2159
2160 if (ioc->overload_hwm < ioc->overload_lwm)
2161 return (EINVAL);
2162 }
2163
2164 pfsrlim = malloc(sizeof(*pfsrlim), M_PF_SOURCE_LIM, M_WAITOK | M_ZERO);
2165 if (pfsrlim == NULL)
2166 return (ENOMEM);
2167
2168 pfsrlim->pfsrlim_id = ioc->id;
2169 pfsrlim->pfsrlim_entries = ioc->entries;
2170 pfsrlim->pfsrlim_limit = ioc->limit;
2171 pfsrlim->pfsrlim_ipv4_prefix = ioc->inet_prefix;
2172 pfsrlim->pfsrlim_ipv6_prefix = ioc->inet6_prefix;
2173 pfsrlim->pfsrlim_rate.limit = ioc->rate.limit;
2174 pfsrlim->pfsrlim_rate.seconds = ioc->rate.seconds;
2175 if (strlcpy(pfsrlim->pfsrlim_overload.name, ioc->overload_tblname,
2176 sizeof(pfsrlim->pfsrlim_overload.name)) >=
2177 sizeof(pfsrlim->pfsrlim_overload.name)) {
2178 error = EINVAL;
2179 goto free;
2180 }
2181 pfsrlim->pfsrlim_overload.hwm = ioc->overload_hwm;
2182 pfsrlim->pfsrlim_overload.lwm = ioc->overload_lwm;
2183 if (strlcpy(pfsrlim->pfsrlim_nm, ioc->name,
2184 sizeof(pfsrlim->pfsrlim_nm)) >= sizeof(pfsrlim->pfsrlim_nm)) {
2185 error = EINVAL;
2186 goto free;
2187 }
2188
2189 if (pfsrlim->pfsrlim_rate.limit) {
2190 uint64_t bucket = pfsrlim->pfsrlim_rate.seconds * 1000000000ULL;
2191
2192 pfsrlim->pfsrlim_rate_token = bucket /
2193 pfsrlim->pfsrlim_rate.limit;
2194 pfsrlim->pfsrlim_rate_bucket = bucket;
2195 }
2196
2197 pfsrlim->pfsrlim_ipv4_mask.v4.s_addr = htonl(
2198 0xffffffff << (32 - pfsrlim->pfsrlim_ipv4_prefix));
2199
2200 prefix = pfsrlim->pfsrlim_ipv6_prefix;
2201 for (i = 0; i < nitems(pfsrlim->pfsrlim_ipv6_mask.addr32); i++) {
2202 if (prefix == 0) {
2203 /* the memory is already zeroed */
2204 break;
2205 }
2206 if (prefix < 32) {
2207 pfsrlim->pfsrlim_ipv6_mask.addr32[i] = htonl(
2208 0xffffffff << (32 - prefix));
2209 break;
2210 }
2211
2212 pfsrlim->pfsrlim_ipv6_mask.addr32[i] = htonl(0xffffffff);
2213 prefix -= 32;
2214 }
2215
2216 RB_INIT(&pfsrlim->pfsrlim_sources);
2217 mtx_init(&pfsrlim->pfsrlim_lock, "pf source limit", NULL, MTX_DEF);
2218
2219 PF_RULES_WLOCK();
2220 if (ioc->ticket != pf_main_ruleset.rules[PF_RULESET_FILTER].inactive.ticket) {
2221 error = EBUSY;
2222 goto unlock;
2223 }
2224
2225 if (pfsrlim->pfsrlim_overload.name[0] != '\0') {
2226 pfsrlim->pfsrlim_overload.table = pfr_attach_table(
2227 &pf_main_ruleset, pfsrlim->pfsrlim_overload.name);
2228 if (pfsrlim->pfsrlim_overload.table == NULL) {
2229 error = EINVAL;
2230 goto unlock;
2231 }
2232 }
2233
2234 if (RB_INSERT(pf_sourcelim_id_tree, &V_pf_sourcelim_id_tree_inactive,
2235 pfsrlim) != NULL) {
2236 error = EBUSY;
2237 goto unlock;
2238 }
2239
2240 if (RB_INSERT(pf_sourcelim_nm_tree, &V_pf_sourcelim_nm_tree_inactive,
2241 pfsrlim) != NULL) {
2242 RB_INSERT(pf_sourcelim_nm_tree, &V_pf_sourcelim_nm_tree_inactive,
2243 pfsrlim);
2244 error = EBUSY;
2245 goto unlock;
2246 }
2247
2248 TAILQ_INSERT_HEAD(&V_pf_sourcelim_list_inactive, pfsrlim, pfsrlim_list);
2249
2250 PF_RULES_WUNLOCK();
2251
2252 return (0);
2253
2254 unlock:
2255 PF_RULES_WUNLOCK();
2256
2257 free:
2258 free(pfsrlim, M_PF_SOURCE_LIM);
2259
2260 return (error);
2261 }
2262
2263 void
pf_sourcelim_rollback(void)2264 pf_sourcelim_rollback(void)
2265 {
2266 struct pf_sourcelim *pfsrlim, *npfsrlim;
2267
2268 PF_RULES_WASSERT();
2269
2270 TAILQ_FOREACH_SAFE(pfsrlim, &V_pf_sourcelim_list_inactive, pfsrlim_list,
2271 npfsrlim) {
2272 if (pfsrlim->pfsrlim_overload.table != NULL)
2273 pfr_detach_table(pfsrlim->pfsrlim_overload.table);
2274
2275 free(pfsrlim, M_PF_SOURCE_LIM);
2276 }
2277
2278 TAILQ_INIT(&V_pf_sourcelim_list_inactive);
2279 RB_INIT(&V_pf_sourcelim_id_tree_inactive);
2280 RB_INIT(&V_pf_sourcelim_nm_tree_inactive);
2281 }
2282
2283 struct pf_sourcelim *
pf_sourcelim_rb_find(struct pf_sourcelim_id_tree * tree,struct pf_sourcelim * key)2284 pf_sourcelim_rb_find(struct pf_sourcelim_id_tree *tree,
2285 struct pf_sourcelim *key)
2286 {
2287 PF_RULES_ASSERT();
2288 return (RB_FIND(pf_sourcelim_id_tree, tree, key));
2289 }
2290
2291 struct pf_sourcelim *
pf_sourcelim_rb_nfind(struct pf_sourcelim_id_tree * tree,struct pf_sourcelim * key)2292 pf_sourcelim_rb_nfind(struct pf_sourcelim_id_tree *tree,
2293 struct pf_sourcelim *key)
2294 {
2295 PF_RULES_ASSERT();
2296 return (RB_NFIND(pf_sourcelim_id_tree, tree, key));
2297 }
2298
2299 int
pf_sourcelim_get(struct pfioc_sourcelim * ioc,struct pf_sourcelim * (* rbt_op)(struct pf_sourcelim_id_tree *,struct pf_sourcelim *))2300 pf_sourcelim_get(struct pfioc_sourcelim *ioc,
2301 struct pf_sourcelim *(*rbt_op)(struct pf_sourcelim_id_tree *,
2302 struct pf_sourcelim *))
2303 {
2304 struct pf_sourcelim key = { .pfsrlim_id = ioc->id };
2305 struct pf_sourcelim *pfsrlim;
2306 int error = 0;
2307 PF_RULES_RLOCK_TRACKER;
2308
2309 PF_RULES_RLOCK();
2310
2311 pfsrlim = (*rbt_op)(&V_pf_sourcelim_id_tree_active, &key);
2312 if (pfsrlim == NULL) {
2313 error = ESRCH;
2314 goto unlock;
2315 }
2316
2317 ioc->id = pfsrlim->pfsrlim_id;
2318 ioc->entries = pfsrlim->pfsrlim_entries;
2319 ioc->limit = pfsrlim->pfsrlim_limit;
2320 ioc->inet_prefix = pfsrlim->pfsrlim_ipv4_prefix;
2321 ioc->inet6_prefix = pfsrlim->pfsrlim_ipv6_prefix;
2322 ioc->rate.limit = pfsrlim->pfsrlim_rate.limit;
2323 ioc->rate.seconds = pfsrlim->pfsrlim_rate.seconds;
2324
2325 CTASSERT(sizeof(ioc->overload_tblname) ==
2326 sizeof(pfsrlim->pfsrlim_overload.name));
2327 memcpy(ioc->overload_tblname, pfsrlim->pfsrlim_overload.name,
2328 sizeof(pfsrlim->pfsrlim_overload.name));
2329 ioc->overload_hwm = pfsrlim->pfsrlim_overload.hwm;
2330 ioc->overload_lwm = pfsrlim->pfsrlim_overload.lwm;
2331
2332 CTASSERT(sizeof(ioc->name) == sizeof(pfsrlim->pfsrlim_nm));
2333 memcpy(ioc->name, pfsrlim->pfsrlim_nm, sizeof(ioc->name));
2334 /* XXX overload table thing */
2335
2336 ioc->nentries = pfsrlim->pfsrlim_nsources;
2337
2338 ioc->inuse = pfsrlim->pfsrlim_counters.inuse;
2339 ioc->addrallocs = pfsrlim->pfsrlim_counters.addrallocs;
2340 ioc->addrnomem = pfsrlim->pfsrlim_counters.addrnomem;
2341 ioc->admitted = pfsrlim->pfsrlim_counters.admitted;
2342 ioc->addrlimited = pfsrlim->pfsrlim_counters.addrlimited;
2343 ioc->hardlimited = pfsrlim->pfsrlim_counters.hardlimited;
2344 ioc->ratelimited = pfsrlim->pfsrlim_counters.ratelimited;
2345
2346 unlock:
2347 PF_RULES_RUNLOCK();
2348
2349 return (error);
2350 }
2351
2352 struct pf_source *
pf_source_rb_find(struct pf_source_ioc_tree * tree,struct pf_source * key)2353 pf_source_rb_find(struct pf_source_ioc_tree *tree,
2354 struct pf_source *key)
2355 {
2356 PF_RULES_ASSERT();
2357
2358 return (RB_FIND(pf_source_ioc_tree, tree, key));
2359 }
2360
2361 struct pf_source *
pf_source_rb_nfind(struct pf_source_ioc_tree * tree,struct pf_source * key)2362 pf_source_rb_nfind(struct pf_source_ioc_tree *tree,
2363 struct pf_source *key)
2364 {
2365 PF_RULES_ASSERT();
2366
2367 return (RB_NFIND(pf_source_ioc_tree, tree, key));
2368 }
2369
2370 int
pf_source_clr(struct pfioc_source_kill * ioc)2371 pf_source_clr(struct pfioc_source_kill *ioc)
2372 {
2373 extern struct pf_source_list pf_source_gc;
2374 struct pf_sourcelim plkey = {
2375 .pfsrlim_id = ioc->id,
2376 };
2377 struct pf_source skey = {
2378 .pfsr_af = ioc->af,
2379 .pfsr_rdomain = ioc->rdomain,
2380 .pfsr_addr = ioc->addr,
2381 };
2382 struct pf_sourcelim *pfsrlim;
2383 struct pf_source *pfsr;
2384 struct pf_state_link *pfl, *npfl;
2385 int error = 0;
2386 unsigned int gen;
2387
2388 if (ioc->rmstates) {
2389 /* XXX userland wants the states removed too */
2390 return (EOPNOTSUPP);
2391 }
2392
2393 PF_RULES_WLOCK();
2394
2395 pfsrlim = pf_sourcelim_rb_find(&V_pf_sourcelim_id_tree_active, &plkey);
2396 if (pfsrlim == NULL) {
2397 error = ESRCH;
2398 goto unlock;
2399 }
2400
2401 pfsr = pf_source_rb_find(&pfsrlim->pfsrlim_ioc_sources, &skey);
2402 if (pfsr == NULL) {
2403 error = ENOENT;
2404 goto unlock;
2405 }
2406
2407 RB_REMOVE(pf_source_tree, &pfsrlim->pfsrlim_sources, pfsr);
2408 RB_REMOVE(pf_source_ioc_tree, &pfsrlim->pfsrlim_ioc_sources, pfsr);
2409 if (pfsr->pfsr_inuse == 0)
2410 TAILQ_REMOVE(&pf_source_gc, pfsr, pfsr_empty_gc);
2411
2412 gen = pf_sourcelim_enter(pfsrlim);
2413 pfsrlim->pfsrlim_nsources--;
2414 pfsrlim->pfsrlim_counters.inuse -= pfsr->pfsr_inuse;
2415 pf_sourcelim_leave(pfsrlim, gen);
2416
2417 /* unwire the links */
2418 TAILQ_FOREACH(pfl, &pfsr->pfsr_states, pfl_link) {
2419 struct pf_kstate *st = pfl->pfl_state;
2420
2421 /* if !rmst */
2422 st->sourcelim = 0;
2423 SLIST_REMOVE(&st->linkage, pfl, pf_state_link, pfl_linkage);
2424 }
2425
2426 PF_RULES_WUNLOCK();
2427
2428 TAILQ_FOREACH_SAFE(pfl, &pfsr->pfsr_states, pfl_link, npfl)
2429 free(pfl, M_PF_STATE_LINK);
2430
2431 free(pfsr, M_PF_SOURCE_LIM);
2432
2433 return (0);
2434
2435 unlock:
2436 PF_RULES_WUNLOCK();
2437
2438 return (error);
2439 }
2440
2441 static void
pf_src_node_copy(const struct pf_ksrc_node * in,struct pf_src_node * out)2442 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
2443 {
2444 int secs = time_uptime;
2445
2446 bzero(out, sizeof(struct pf_src_node));
2447
2448 bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
2449 bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
2450
2451 if (in->rule != NULL)
2452 out->rule.nr = in->rule->nr;
2453
2454 for (int i = 0; i < 2; i++) {
2455 out->bytes[i] = counter_u64_fetch(in->bytes[i]);
2456 out->packets[i] = counter_u64_fetch(in->packets[i]);
2457 }
2458
2459 out->states = in->states;
2460 out->conn = in->conn;
2461 out->af = in->af;
2462 out->ruletype = in->ruletype;
2463
2464 out->creation = secs - in->creation;
2465 if (out->expire > secs)
2466 out->expire -= secs;
2467 else
2468 out->expire = 0;
2469
2470 /* Adjust the connection rate estimate. */
2471 out->conn_rate.limit = in->conn_rate.limit;
2472 out->conn_rate.seconds = in->conn_rate.seconds;
2473 /* If there's no limit there's no counter_rate. */
2474 if (in->conn_rate.cr != NULL)
2475 out->conn_rate.count = counter_rate_get(in->conn_rate.cr);
2476 }
2477
2478 #ifdef ALTQ
2479 /*
2480 * Handle export of struct pf_kaltq to user binaries that may be using any
2481 * version of struct pf_altq.
2482 */
2483 static int
pf_export_kaltq(struct pf_altq * q,struct pfioc_altq_v1 * pa,size_t ioc_size)2484 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
2485 {
2486 u_int32_t version;
2487
2488 if (ioc_size == sizeof(struct pfioc_altq_v0))
2489 version = 0;
2490 else
2491 version = pa->version;
2492
2493 if (version > PFIOC_ALTQ_VERSION)
2494 return (EINVAL);
2495
2496 #define ASSIGN(x) exported_q->x = q->x
2497 #define COPY(x) \
2498 bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
2499 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
2500 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
2501
2502 switch (version) {
2503 case 0: {
2504 struct pf_altq_v0 *exported_q =
2505 &((struct pfioc_altq_v0 *)pa)->altq;
2506
2507 COPY(ifname);
2508
2509 ASSIGN(scheduler);
2510 ASSIGN(tbrsize);
2511 exported_q->tbrsize = SATU16(q->tbrsize);
2512 exported_q->ifbandwidth = SATU32(q->ifbandwidth);
2513
2514 COPY(qname);
2515 COPY(parent);
2516 ASSIGN(parent_qid);
2517 exported_q->bandwidth = SATU32(q->bandwidth);
2518 ASSIGN(priority);
2519 ASSIGN(local_flags);
2520
2521 ASSIGN(qlimit);
2522 ASSIGN(flags);
2523
2524 if (q->scheduler == ALTQT_HFSC) {
2525 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
2526 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
2527 SATU32(q->pq_u.hfsc_opts.x)
2528
2529 ASSIGN_OPT_SATU32(rtsc_m1);
2530 ASSIGN_OPT(rtsc_d);
2531 ASSIGN_OPT_SATU32(rtsc_m2);
2532
2533 ASSIGN_OPT_SATU32(lssc_m1);
2534 ASSIGN_OPT(lssc_d);
2535 ASSIGN_OPT_SATU32(lssc_m2);
2536
2537 ASSIGN_OPT_SATU32(ulsc_m1);
2538 ASSIGN_OPT(ulsc_d);
2539 ASSIGN_OPT_SATU32(ulsc_m2);
2540
2541 ASSIGN_OPT(flags);
2542
2543 #undef ASSIGN_OPT
2544 #undef ASSIGN_OPT_SATU32
2545 } else
2546 COPY(pq_u);
2547
2548 ASSIGN(qid);
2549 break;
2550 }
2551 case 1: {
2552 struct pf_altq_v1 *exported_q =
2553 &((struct pfioc_altq_v1 *)pa)->altq;
2554
2555 COPY(ifname);
2556
2557 ASSIGN(scheduler);
2558 ASSIGN(tbrsize);
2559 ASSIGN(ifbandwidth);
2560
2561 COPY(qname);
2562 COPY(parent);
2563 ASSIGN(parent_qid);
2564 ASSIGN(bandwidth);
2565 ASSIGN(priority);
2566 ASSIGN(local_flags);
2567
2568 ASSIGN(qlimit);
2569 ASSIGN(flags);
2570 COPY(pq_u);
2571
2572 ASSIGN(qid);
2573 break;
2574 }
2575 default:
2576 panic("%s: unhandled struct pfioc_altq version", __func__);
2577 break;
2578 }
2579
2580 #undef ASSIGN
2581 #undef COPY
2582 #undef SATU16
2583 #undef SATU32
2584
2585 return (0);
2586 }
2587
2588 /*
2589 * Handle import to struct pf_kaltq of struct pf_altq from user binaries
2590 * that may be using any version of it.
2591 */
2592 static int
pf_import_kaltq(struct pfioc_altq_v1 * pa,struct pf_altq * q,size_t ioc_size)2593 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
2594 {
2595 u_int32_t version;
2596
2597 if (ioc_size == sizeof(struct pfioc_altq_v0))
2598 version = 0;
2599 else
2600 version = pa->version;
2601
2602 if (version > PFIOC_ALTQ_VERSION)
2603 return (EINVAL);
2604
2605 #define ASSIGN(x) q->x = imported_q->x
2606 #define COPY(x) \
2607 bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
2608
2609 switch (version) {
2610 case 0: {
2611 struct pf_altq_v0 *imported_q =
2612 &((struct pfioc_altq_v0 *)pa)->altq;
2613
2614 COPY(ifname);
2615
2616 ASSIGN(scheduler);
2617 ASSIGN(tbrsize); /* 16-bit -> 32-bit */
2618 ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
2619
2620 COPY(qname);
2621 COPY(parent);
2622 ASSIGN(parent_qid);
2623 ASSIGN(bandwidth); /* 32-bit -> 64-bit */
2624 ASSIGN(priority);
2625 ASSIGN(local_flags);
2626
2627 ASSIGN(qlimit);
2628 ASSIGN(flags);
2629
2630 if (imported_q->scheduler == ALTQT_HFSC) {
2631 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
2632
2633 /*
2634 * The m1 and m2 parameters are being copied from
2635 * 32-bit to 64-bit.
2636 */
2637 ASSIGN_OPT(rtsc_m1);
2638 ASSIGN_OPT(rtsc_d);
2639 ASSIGN_OPT(rtsc_m2);
2640
2641 ASSIGN_OPT(lssc_m1);
2642 ASSIGN_OPT(lssc_d);
2643 ASSIGN_OPT(lssc_m2);
2644
2645 ASSIGN_OPT(ulsc_m1);
2646 ASSIGN_OPT(ulsc_d);
2647 ASSIGN_OPT(ulsc_m2);
2648
2649 ASSIGN_OPT(flags);
2650
2651 #undef ASSIGN_OPT
2652 } else
2653 COPY(pq_u);
2654
2655 ASSIGN(qid);
2656 break;
2657 }
2658 case 1: {
2659 struct pf_altq_v1 *imported_q =
2660 &((struct pfioc_altq_v1 *)pa)->altq;
2661
2662 COPY(ifname);
2663
2664 ASSIGN(scheduler);
2665 ASSIGN(tbrsize);
2666 ASSIGN(ifbandwidth);
2667
2668 COPY(qname);
2669 COPY(parent);
2670 ASSIGN(parent_qid);
2671 ASSIGN(bandwidth);
2672 ASSIGN(priority);
2673 ASSIGN(local_flags);
2674
2675 ASSIGN(qlimit);
2676 ASSIGN(flags);
2677 COPY(pq_u);
2678
2679 ASSIGN(qid);
2680 break;
2681 }
2682 default:
2683 panic("%s: unhandled struct pfioc_altq version", __func__);
2684 break;
2685 }
2686
2687 #undef ASSIGN
2688 #undef COPY
2689
2690 return (0);
2691 }
2692
2693 static struct pf_altq *
pf_altq_get_nth_active(u_int32_t n)2694 pf_altq_get_nth_active(u_int32_t n)
2695 {
2696 struct pf_altq *altq;
2697 u_int32_t nr;
2698
2699 nr = 0;
2700 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
2701 if (nr == n)
2702 return (altq);
2703 nr++;
2704 }
2705
2706 TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
2707 if (nr == n)
2708 return (altq);
2709 nr++;
2710 }
2711
2712 return (NULL);
2713 }
2714 #endif /* ALTQ */
2715
2716 struct pf_krule *
pf_krule_alloc(void)2717 pf_krule_alloc(void)
2718 {
2719 struct pf_krule *rule;
2720
2721 rule = malloc(sizeof(struct pf_krule), M_PFRULE, M_WAITOK | M_ZERO);
2722 mtx_init(&rule->nat.mtx, "pf_krule_nat_pool", NULL, MTX_DEF);
2723 mtx_init(&rule->rdr.mtx, "pf_krule_rdr_pool", NULL, MTX_DEF);
2724 mtx_init(&rule->route.mtx, "pf_krule_route_pool", NULL, MTX_DEF);
2725 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
2726 M_WAITOK | M_ZERO);
2727 return (rule);
2728 }
2729
2730 void
pf_krule_free(struct pf_krule * rule)2731 pf_krule_free(struct pf_krule *rule)
2732 {
2733 #ifdef PF_WANT_32_TO_64_COUNTER
2734 bool wowned;
2735 #endif
2736
2737 if (rule == NULL)
2738 return;
2739
2740 #ifdef PF_WANT_32_TO_64_COUNTER
2741 if (rule->allrulelinked) {
2742 wowned = PF_RULES_WOWNED();
2743 if (!wowned)
2744 PF_RULES_WLOCK();
2745 LIST_REMOVE(rule, allrulelist);
2746 V_pf_allrulecount--;
2747 if (!wowned)
2748 PF_RULES_WUNLOCK();
2749 }
2750 #endif
2751
2752 pf_counter_u64_deinit(&rule->evaluations);
2753 for (int i = 0; i < 2; i++) {
2754 pf_counter_u64_deinit(&rule->packets[i]);
2755 pf_counter_u64_deinit(&rule->bytes[i]);
2756 }
2757 counter_u64_free(rule->states_cur);
2758 counter_u64_free(rule->states_tot);
2759 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
2760 counter_u64_free(rule->src_nodes[sn_type]);
2761 uma_zfree_pcpu(pf_timestamp_pcpu_zone, rule->timestamp);
2762
2763 mtx_destroy(&rule->nat.mtx);
2764 mtx_destroy(&rule->rdr.mtx);
2765 mtx_destroy(&rule->route.mtx);
2766 free(rule, M_PFRULE);
2767 }
2768
2769 void
pf_krule_clear_counters(struct pf_krule * rule)2770 pf_krule_clear_counters(struct pf_krule *rule)
2771 {
2772 pf_counter_u64_zero(&rule->evaluations);
2773 for (int i = 0; i < 2; i++) {
2774 pf_counter_u64_zero(&rule->packets[i]);
2775 pf_counter_u64_zero(&rule->bytes[i]);
2776 }
2777 counter_u64_zero(rule->states_tot);
2778 }
2779
2780 static void
pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr * kpool,struct pf_pooladdr * pool)2781 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
2782 struct pf_pooladdr *pool)
2783 {
2784
2785 bzero(pool, sizeof(*pool));
2786 bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
2787 strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
2788 }
2789
2790 static int
pf_pooladdr_to_kpooladdr(const struct pf_pooladdr * pool,struct pf_kpooladdr * kpool)2791 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
2792 struct pf_kpooladdr *kpool)
2793 {
2794 int ret;
2795
2796 bzero(kpool, sizeof(*kpool));
2797 bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
2798 ret = pf_user_strcpy(kpool->ifname, pool->ifname,
2799 sizeof(kpool->ifname));
2800 return (ret);
2801 }
2802
2803 static void
pf_pool_to_kpool(const struct pf_pool * pool,struct pf_kpool * kpool)2804 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
2805 {
2806 _Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
2807 _Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
2808
2809 bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
2810 bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
2811
2812 kpool->tblidx = pool->tblidx;
2813 kpool->proxy_port[0] = pool->proxy_port[0];
2814 kpool->proxy_port[1] = pool->proxy_port[1];
2815 kpool->opts = pool->opts;
2816 }
2817
2818 static int
pf_rule_to_krule(const struct pf_rule * rule,struct pf_krule * krule)2819 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
2820 {
2821 int ret;
2822
2823 #ifndef INET
2824 if (rule->af == AF_INET) {
2825 return (EAFNOSUPPORT);
2826 }
2827 #endif /* INET */
2828 #ifndef INET6
2829 if (rule->af == AF_INET6) {
2830 return (EAFNOSUPPORT);
2831 }
2832 #endif /* INET6 */
2833
2834 ret = pf_check_rule_addr(&rule->src);
2835 if (ret != 0)
2836 return (ret);
2837 ret = pf_check_rule_addr(&rule->dst);
2838 if (ret != 0)
2839 return (ret);
2840
2841 bcopy(&rule->src, &krule->src, sizeof(rule->src));
2842 bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
2843
2844 ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
2845 if (ret != 0)
2846 return (ret);
2847 ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
2848 if (ret != 0)
2849 return (ret);
2850 ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
2851 if (ret != 0)
2852 return (ret);
2853 ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
2854 if (ret != 0)
2855 return (ret);
2856 ret = pf_user_strcpy(krule->tagname, rule->tagname,
2857 sizeof(rule->tagname));
2858 if (ret != 0)
2859 return (ret);
2860 ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
2861 sizeof(rule->match_tagname));
2862 if (ret != 0)
2863 return (ret);
2864 ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
2865 sizeof(rule->overload_tblname));
2866 if (ret != 0)
2867 return (ret);
2868
2869 pf_pool_to_kpool(&rule->rpool, &krule->rdr);
2870
2871 /* Don't allow userspace to set evaluations, packets or bytes. */
2872 /* kif, anchor, overload_tbl are not copied over. */
2873
2874 krule->os_fingerprint = rule->os_fingerprint;
2875
2876 krule->rtableid = rule->rtableid;
2877 /* pf_rule->timeout is smaller than pf_krule->timeout */
2878 bcopy(rule->timeout, krule->timeout, sizeof(rule->timeout));
2879 krule->max_states = rule->max_states;
2880 krule->max_src_nodes = rule->max_src_nodes;
2881 krule->max_src_states = rule->max_src_states;
2882 krule->max_src_conn = rule->max_src_conn;
2883 krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
2884 krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
2885 krule->qid = rule->qid;
2886 krule->pqid = rule->pqid;
2887 krule->nr = rule->nr;
2888 krule->prob = rule->prob;
2889 krule->cuid = rule->cuid;
2890 krule->cpid = rule->cpid;
2891
2892 krule->return_icmp = rule->return_icmp;
2893 krule->return_icmp6 = rule->return_icmp6;
2894 krule->max_mss = rule->max_mss;
2895 krule->tag = rule->tag;
2896 krule->match_tag = rule->match_tag;
2897 krule->scrub_flags = rule->scrub_flags;
2898
2899 bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
2900 bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
2901
2902 krule->rule_flag = rule->rule_flag;
2903 krule->action = rule->action;
2904 krule->direction = rule->direction;
2905 krule->log = rule->log;
2906 krule->logif = rule->logif;
2907 krule->quick = rule->quick;
2908 krule->ifnot = rule->ifnot;
2909 krule->match_tag_not = rule->match_tag_not;
2910 krule->natpass = rule->natpass;
2911
2912 krule->keep_state = rule->keep_state;
2913 krule->af = rule->af;
2914 krule->proto = rule->proto;
2915 krule->type = rule->type;
2916 krule->code = rule->code;
2917 krule->flags = rule->flags;
2918 krule->flagset = rule->flagset;
2919 krule->min_ttl = rule->min_ttl;
2920 krule->allow_opts = rule->allow_opts;
2921 krule->rt = rule->rt;
2922 krule->return_ttl = rule->return_ttl;
2923 krule->tos = rule->tos;
2924 krule->set_tos = rule->set_tos;
2925
2926 krule->flush = rule->flush;
2927 krule->prio = rule->prio;
2928 krule->set_prio[0] = rule->set_prio[0];
2929 krule->set_prio[1] = rule->set_prio[1];
2930
2931 bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
2932
2933 return (0);
2934 }
2935
2936 int
pf_ioctl_getrules(struct pfioc_rule * pr)2937 pf_ioctl_getrules(struct pfioc_rule *pr)
2938 {
2939 PF_RULES_RLOCK_TRACKER;
2940 struct pf_kruleset *ruleset;
2941 struct pf_krule *tail;
2942 int rs_num;
2943
2944 PF_RULES_RLOCK();
2945 ruleset = pf_find_kruleset(pr->anchor);
2946 if (ruleset == NULL) {
2947 PF_RULES_RUNLOCK();
2948 return (EINVAL);
2949 }
2950 rs_num = pf_get_ruleset_number(pr->rule.action);
2951 if (rs_num >= PF_RULESET_MAX) {
2952 PF_RULES_RUNLOCK();
2953 return (EINVAL);
2954 }
2955 tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2956 pf_krulequeue);
2957 if (tail)
2958 pr->nr = tail->nr + 1;
2959 else
2960 pr->nr = 0;
2961 pr->ticket = ruleset->rules[rs_num].active.ticket;
2962 PF_RULES_RUNLOCK();
2963
2964 return (0);
2965 }
2966
2967 static int
pf_rule_checkaf(struct pf_krule * r)2968 pf_rule_checkaf(struct pf_krule *r)
2969 {
2970 switch (r->af) {
2971 case 0:
2972 if (r->rule_flag & PFRULE_AFTO)
2973 return (EPFNOSUPPORT);
2974 break;
2975 case AF_INET:
2976 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET6)
2977 return (EPFNOSUPPORT);
2978 break;
2979 #ifdef INET6
2980 case AF_INET6:
2981 if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET)
2982 return (EPFNOSUPPORT);
2983 break;
2984 #endif /* INET6 */
2985 default:
2986 return (EPFNOSUPPORT);
2987 }
2988
2989 if ((r->rule_flag & PFRULE_AFTO) == 0 && r->naf != 0)
2990 return (EPFNOSUPPORT);
2991
2992 return (0);
2993 }
2994
2995 static int
pf_validate_range(uint8_t op,uint16_t port[2])2996 pf_validate_range(uint8_t op, uint16_t port[2])
2997 {
2998 uint16_t a = ntohs(port[0]);
2999 uint16_t b = ntohs(port[1]);
3000
3001 if ((op == PF_OP_RRG && a > b) || /* 34:12, i.e. none */
3002 (op == PF_OP_IRG && a >= b) || /* 34><12, i.e. none */
3003 (op == PF_OP_XRG && a > b)) /* 34<>22, i.e. all */
3004 return 1;
3005 return 0;
3006 }
3007
3008 static int
pf_chk_limiter_action(int limiter_action)3009 pf_chk_limiter_action(int limiter_action)
3010 {
3011 int rv;
3012
3013 switch (limiter_action) {
3014 case PF_LIMITER_NOMATCH:
3015 case PF_LIMITER_BLOCK:
3016 rv = 0;
3017 break;
3018 default:
3019 rv = 1;
3020 }
3021
3022 return (rv);
3023 }
3024
3025 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)3026 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
3027 uint32_t pool_ticket, const char *anchor, const char *anchor_call,
3028 uid_t uid, pid_t pid)
3029 {
3030 struct pf_kruleset *ruleset;
3031 struct pf_krule *tail;
3032 struct pf_kpooladdr *pa;
3033 struct pfi_kkif *kif = NULL, *rcv_kif = NULL;
3034 int rs_num;
3035 int error = 0;
3036
3037 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
3038 #define ERROUT_UNLOCKED(x) ERROUT_FUNCTION(errout_unlocked, x)
3039
3040 if ((rule->return_icmp >> 8) > ICMP_MAXTYPE)
3041 ERROUT_UNLOCKED(EINVAL);
3042
3043 if ((error = pf_rule_checkaf(rule)))
3044 ERROUT_UNLOCKED(error);
3045 if (pf_validate_range(rule->src.port_op, rule->src.port))
3046 ERROUT_UNLOCKED(EINVAL);
3047 if (pf_validate_range(rule->dst.port_op, rule->dst.port))
3048 ERROUT_UNLOCKED(EINVAL);
3049 if (pf_chk_limiter_action(rule->statelim.limiter_action) ||
3050 pf_chk_limiter_action(rule->sourcelim.limiter_action))
3051 ERROUT_UNLOCKED(EINVAL);
3052
3053 if (rule->ifname[0])
3054 kif = pf_kkif_create(M_WAITOK);
3055 if (rule->rcv_ifname[0])
3056 rcv_kif = pf_kkif_create(M_WAITOK);
3057 pf_counter_u64_init(&rule->evaluations, M_WAITOK);
3058 for (int i = 0; i < 2; i++) {
3059 pf_counter_u64_init(&rule->packets[i], M_WAITOK);
3060 pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
3061 }
3062 rule->states_cur = counter_u64_alloc(M_WAITOK);
3063 rule->states_tot = counter_u64_alloc(M_WAITOK);
3064 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
3065 rule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
3066 rule->cuid = uid;
3067 rule->cpid = pid;
3068 TAILQ_INIT(&rule->rdr.list);
3069 TAILQ_INIT(&rule->nat.list);
3070 TAILQ_INIT(&rule->route.list);
3071
3072 PF_CONFIG_LOCK();
3073 PF_RULES_WLOCK();
3074 #ifdef PF_WANT_32_TO_64_COUNTER
3075 LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
3076 MPASS(!rule->allrulelinked);
3077 rule->allrulelinked = true;
3078 V_pf_allrulecount++;
3079 #endif
3080 ruleset = pf_find_kruleset(anchor);
3081 if (ruleset == NULL)
3082 ERROUT(EINVAL);
3083 rs_num = pf_get_ruleset_number(rule->action);
3084 if (rs_num >= PF_RULESET_MAX)
3085 ERROUT(EINVAL);
3086 if (ticket != ruleset->rules[rs_num].inactive.ticket) {
3087 DPFPRINTF(PF_DEBUG_MISC,
3088 "ticket: %d != [%d]%d", ticket, rs_num,
3089 ruleset->rules[rs_num].inactive.ticket);
3090 ERROUT(EBUSY);
3091 }
3092 if (pool_ticket != V_ticket_pabuf) {
3093 DPFPRINTF(PF_DEBUG_MISC,
3094 "pool_ticket: %d != %d", pool_ticket,
3095 V_ticket_pabuf);
3096 ERROUT(EBUSY);
3097 }
3098 /*
3099 * XXXMJG hack: there is no mechanism to ensure they started the
3100 * transaction. Ticket checked above may happen to match by accident,
3101 * even if nobody called DIOCXBEGIN, let alone this process.
3102 * Partially work around it by checking if the RB tree got allocated,
3103 * see pf_begin_rules.
3104 */
3105 if (ruleset->rules[rs_num].inactive.tree == NULL) {
3106 ERROUT(EINVAL);
3107 }
3108
3109 tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
3110 pf_krulequeue);
3111 if (tail)
3112 rule->nr = tail->nr + 1;
3113 else
3114 rule->nr = 0;
3115 if (rule->ifname[0]) {
3116 rule->kif = pfi_kkif_attach(kif, rule->ifname);
3117 kif = NULL;
3118 pfi_kkif_ref(rule->kif);
3119 } else
3120 rule->kif = NULL;
3121
3122 if (rule->rcv_ifname[0]) {
3123 rule->rcv_kif = pfi_kkif_attach(rcv_kif, rule->rcv_ifname);
3124 rcv_kif = NULL;
3125 pfi_kkif_ref(rule->rcv_kif);
3126 } else
3127 rule->rcv_kif = NULL;
3128
3129 if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
3130 ERROUT(EBUSY);
3131 #ifdef ALTQ
3132 /* set queue IDs */
3133 if (rule->qname[0] != 0) {
3134 if ((rule->qid = pf_qname2qid(rule->qname, true)) == 0)
3135 ERROUT(EBUSY);
3136 else if (rule->pqname[0] != 0) {
3137 if ((rule->pqid =
3138 pf_qname2qid(rule->pqname, true)) == 0)
3139 ERROUT(EBUSY);
3140 } else
3141 rule->pqid = rule->qid;
3142 }
3143 #endif
3144 if (rule->tagname[0])
3145 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3146 ERROUT(EBUSY);
3147 if (rule->match_tagname[0])
3148 if ((rule->match_tag =
3149 pf_tagname2tag(rule->match_tagname)) == 0)
3150 ERROUT(EBUSY);
3151 if (rule->rt && !rule->direction)
3152 ERROUT(EINVAL);
3153 if (!rule->log)
3154 rule->logif = 0;
3155 if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit,
3156 rule->pktrate.seconds))
3157 ERROUT(ENOMEM);
3158 if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
3159 ERROUT(ENOMEM);
3160 if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
3161 ERROUT(ENOMEM);
3162 if (pf_kanchor_setup(rule, ruleset, anchor_call))
3163 ERROUT(EINVAL);
3164 if (rule->scrub_flags & PFSTATE_SETPRIO &&
3165 (rule->set_prio[0] > PF_PRIO_MAX ||
3166 rule->set_prio[1] > PF_PRIO_MAX))
3167 ERROUT(EINVAL);
3168 for (int i = 0; i < 3; i++) {
3169 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
3170 if (pa->addr.type == PF_ADDR_TABLE) {
3171 pa->addr.p.tbl = pfr_attach_table(ruleset,
3172 pa->addr.v.tblname);
3173 if (pa->addr.p.tbl == NULL)
3174 ERROUT(ENOMEM);
3175 }
3176 }
3177
3178 rule->overload_tbl = NULL;
3179 if (rule->overload_tblname[0]) {
3180 if ((rule->overload_tbl = pfr_attach_table(ruleset,
3181 rule->overload_tblname)) == NULL)
3182 ERROUT(EINVAL);
3183 else
3184 rule->overload_tbl->pfrkt_flags |=
3185 PFR_TFLAG_ACTIVE;
3186 }
3187
3188 pf_mv_kpool(&V_pf_pabuf[0], &rule->nat.list);
3189
3190 /*
3191 * Old version of pfctl provide route redirection pools in single
3192 * common redirection pool rdr. New versions use rdr only for
3193 * rdr-to rules.
3194 */
3195 if (rule->rt > PF_NOPFROUTE && TAILQ_EMPTY(&V_pf_pabuf[2])) {
3196 pf_mv_kpool(&V_pf_pabuf[1], &rule->route.list);
3197 } else {
3198 pf_mv_kpool(&V_pf_pabuf[1], &rule->rdr.list);
3199 pf_mv_kpool(&V_pf_pabuf[2], &rule->route.list);
3200 }
3201
3202 if (((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
3203 (rule->action == PF_BINAT)) && rule->anchor == NULL &&
3204 TAILQ_FIRST(&rule->rdr.list) == NULL) {
3205 ERROUT(EINVAL);
3206 }
3207
3208 if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) {
3209 ERROUT(EINVAL);
3210 }
3211
3212 if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR ||
3213 rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) {
3214 ERROUT(EINVAL);
3215 }
3216
3217 MPASS(error == 0);
3218
3219 rule->nat.cur = TAILQ_FIRST(&rule->nat.list);
3220 rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list);
3221 rule->route.cur = TAILQ_FIRST(&rule->route.list);
3222 rule->route.ipv6_nexthop_af = AF_INET6;
3223 TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
3224 rule, entries);
3225 ruleset->rules[rs_num].inactive.rcount++;
3226
3227 PF_RULES_WUNLOCK();
3228 pf_hash_rule(rule);
3229 if (RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule) != NULL) {
3230 PF_RULES_WLOCK();
3231 TAILQ_REMOVE(ruleset->rules[rs_num].inactive.ptr, rule, entries);
3232 ruleset->rules[rs_num].inactive.rcount--;
3233 pf_free_rule(rule);
3234 rule = NULL;
3235 ERROUT(EEXIST);
3236 }
3237 PF_CONFIG_UNLOCK();
3238
3239 return (0);
3240
3241 #undef ERROUT
3242 #undef ERROUT_UNLOCKED
3243 errout:
3244 PF_RULES_WUNLOCK();
3245 PF_CONFIG_UNLOCK();
3246 errout_unlocked:
3247 pf_kkif_free(rcv_kif);
3248 pf_kkif_free(kif);
3249 pf_krule_free(rule);
3250 return (error);
3251 }
3252
3253 static bool
pf_label_match(const struct pf_krule * rule,const char * label)3254 pf_label_match(const struct pf_krule *rule, const char *label)
3255 {
3256 int i = 0;
3257
3258 while (*rule->label[i]) {
3259 if (strcmp(rule->label[i], label) == 0)
3260 return (true);
3261 i++;
3262 }
3263
3264 return (false);
3265 }
3266
3267 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)3268 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
3269 {
3270 struct pf_kstate *s;
3271 int more = 0;
3272
3273 s = pf_find_state_all(key, dir, &more);
3274 if (s == NULL)
3275 return (0);
3276
3277 if (more) {
3278 PF_STATE_UNLOCK(s);
3279 return (0);
3280 }
3281
3282 pf_remove_state(s);
3283 return (1);
3284 }
3285
3286 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)3287 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
3288 {
3289 struct pf_kstate *s;
3290 struct pf_state_key *sk;
3291 struct pf_addr *srcaddr, *dstaddr;
3292 struct pf_state_key_cmp match_key;
3293 int idx, killed = 0;
3294 unsigned int dir;
3295 u_int16_t srcport, dstport;
3296 struct pfi_kkif *kif;
3297
3298 relock_DIOCKILLSTATES:
3299 PF_HASHROW_LOCK(ih);
3300 LIST_FOREACH(s, &ih->states, entry) {
3301 /* For floating states look at the original kif. */
3302 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
3303
3304 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
3305 if (s->direction == PF_OUT) {
3306 srcaddr = &sk->addr[1];
3307 dstaddr = &sk->addr[0];
3308 srcport = sk->port[1];
3309 dstport = sk->port[0];
3310 } else {
3311 srcaddr = &sk->addr[0];
3312 dstaddr = &sk->addr[1];
3313 srcport = sk->port[0];
3314 dstport = sk->port[1];
3315 }
3316
3317 if (psk->psk_af && sk->af != psk->psk_af)
3318 continue;
3319
3320 if (psk->psk_proto && psk->psk_proto != sk->proto)
3321 continue;
3322
3323 if (! pf_match_addr(psk->psk_src.neg,
3324 &psk->psk_src.addr.v.a.addr,
3325 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
3326 continue;
3327
3328 if (! pf_match_addr(psk->psk_dst.neg,
3329 &psk->psk_dst.addr.v.a.addr,
3330 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
3331 continue;
3332
3333 if (! pf_match_addr(psk->psk_rt_addr.neg,
3334 &psk->psk_rt_addr.addr.v.a.addr,
3335 &psk->psk_rt_addr.addr.v.a.mask,
3336 &s->act.rt_addr, sk->af))
3337 continue;
3338
3339 if (psk->psk_src.port_op != 0 &&
3340 ! pf_match_port(psk->psk_src.port_op,
3341 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
3342 continue;
3343
3344 if (psk->psk_dst.port_op != 0 &&
3345 ! pf_match_port(psk->psk_dst.port_op,
3346 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
3347 continue;
3348
3349 if (psk->psk_label[0] &&
3350 ! pf_label_match(s->rule, psk->psk_label))
3351 continue;
3352
3353 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
3354 kif->pfik_name))
3355 continue;
3356
3357 if (psk->psk_kill_match) {
3358 /* Create the key to find matching states, with lock
3359 * held. */
3360
3361 bzero(&match_key, sizeof(match_key));
3362
3363 if (s->direction == PF_OUT) {
3364 dir = PF_IN;
3365 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
3366 } else {
3367 dir = PF_OUT;
3368 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
3369 }
3370
3371 match_key.af = s->key[idx]->af;
3372 match_key.proto = s->key[idx]->proto;
3373 pf_addrcpy(&match_key.addr[0],
3374 &s->key[idx]->addr[1], match_key.af);
3375 match_key.port[0] = s->key[idx]->port[1];
3376 pf_addrcpy(&match_key.addr[1],
3377 &s->key[idx]->addr[0], match_key.af);
3378 match_key.port[1] = s->key[idx]->port[0];
3379 }
3380
3381 pf_remove_state(s);
3382 killed++;
3383
3384 if (psk->psk_kill_match)
3385 killed += pf_kill_matching_state(&match_key, dir);
3386
3387 goto relock_DIOCKILLSTATES;
3388 }
3389 PF_HASHROW_UNLOCK(ih);
3390
3391 return (killed);
3392 }
3393
3394 int
pf_start(void)3395 pf_start(void)
3396 {
3397 int error = 0;
3398
3399 sx_xlock(&V_pf_ioctl_lock);
3400 if (V_pf_status.running)
3401 error = EEXIST;
3402 else {
3403 hook_pf();
3404 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
3405 hook_pf_eth();
3406 V_pf_status.running = 1;
3407 V_pf_status.since = time_uptime;
3408 new_unrhdr64(&V_pf_stateid, time_second);
3409
3410 DPFPRINTF(PF_DEBUG_MISC, "pf: started");
3411 }
3412 sx_xunlock(&V_pf_ioctl_lock);
3413
3414 return (error);
3415 }
3416
3417 int
pf_stop(void)3418 pf_stop(void)
3419 {
3420 int error = 0;
3421
3422 sx_xlock(&V_pf_ioctl_lock);
3423 if (!V_pf_status.running)
3424 error = ENOENT;
3425 else {
3426 V_pf_status.running = 0;
3427 dehook_pf();
3428 dehook_pf_eth();
3429 V_pf_status.since = time_uptime;
3430 DPFPRINTF(PF_DEBUG_MISC, "pf: stopped");
3431 }
3432 sx_xunlock(&V_pf_ioctl_lock);
3433
3434 return (error);
3435 }
3436
3437 void
pf_ioctl_clear_status(void)3438 pf_ioctl_clear_status(void)
3439 {
3440 PF_RULES_WLOCK();
3441 for (int i = 0; i < PFRES_MAX; i++)
3442 counter_u64_zero(V_pf_status.counters[i]);
3443 for (int i = 0; i < FCNT_MAX; i++)
3444 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3445 for (int i = 0; i < SCNT_MAX; i++)
3446 counter_u64_zero(V_pf_status.scounters[i]);
3447 for (int i = 0; i < NCNT_MAX; i++)
3448 counter_u64_zero(V_pf_status.ncounters[i]);
3449 for (int i = 0; i < KLCNT_MAX; i++)
3450 counter_u64_zero(V_pf_status.lcounters[i]);
3451 V_pf_status.since = time_uptime;
3452 if (*V_pf_status.ifname)
3453 pfi_update_status(V_pf_status.ifname, NULL);
3454 PF_RULES_WUNLOCK();
3455 }
3456
3457 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)3458 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
3459 {
3460 uint32_t old;
3461
3462 if (timeout < 0 || timeout >= PFTM_MAX ||
3463 seconds < 0)
3464 return (EINVAL);
3465
3466 PF_RULES_WLOCK();
3467 old = V_pf_default_rule.timeout[timeout];
3468 if (timeout == PFTM_INTERVAL && seconds == 0)
3469 seconds = 1;
3470 V_pf_default_rule.timeout[timeout] = seconds;
3471 if (timeout == PFTM_INTERVAL && seconds < old)
3472 wakeup(pf_purge_thread);
3473
3474 if (prev_seconds != NULL)
3475 *prev_seconds = old;
3476
3477 PF_RULES_WUNLOCK();
3478
3479 return (0);
3480 }
3481
3482 int
pf_ioctl_get_timeout(int timeout,int * seconds)3483 pf_ioctl_get_timeout(int timeout, int *seconds)
3484 {
3485 PF_RULES_RLOCK_TRACKER;
3486
3487 if (timeout < 0 || timeout >= PFTM_MAX)
3488 return (EINVAL);
3489
3490 PF_RULES_RLOCK();
3491 *seconds = V_pf_default_rule.timeout[timeout];
3492 PF_RULES_RUNLOCK();
3493
3494 return (0);
3495 }
3496
3497 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)3498 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
3499 {
3500
3501 PF_RULES_WLOCK();
3502 if (index < 0 || index >= PF_LIMIT_MAX ||
3503 V_pf_limits[index].zone == NULL) {
3504 PF_RULES_WUNLOCK();
3505 return (EINVAL);
3506 }
3507 uma_zone_set_max(V_pf_limits[index].zone,
3508 limit == 0 ? INT_MAX : limit);
3509 if (old_limit != NULL)
3510 *old_limit = V_pf_limits[index].limit;
3511 V_pf_limits[index].limit = limit;
3512 PF_RULES_WUNLOCK();
3513
3514 return (0);
3515 }
3516
3517 int
pf_ioctl_get_limit(int index,unsigned int * limit)3518 pf_ioctl_get_limit(int index, unsigned int *limit)
3519 {
3520 PF_RULES_RLOCK_TRACKER;
3521
3522 if (index < 0 || index >= PF_LIMIT_MAX)
3523 return (EINVAL);
3524
3525 PF_RULES_RLOCK();
3526 *limit = V_pf_limits[index].limit;
3527 PF_RULES_RUNLOCK();
3528
3529 return (0);
3530 }
3531
3532 int
pf_ioctl_begin_addrs(uint32_t * ticket)3533 pf_ioctl_begin_addrs(uint32_t *ticket)
3534 {
3535 PF_RULES_WLOCK();
3536 pf_empty_kpool(&V_pf_pabuf[0]);
3537 pf_empty_kpool(&V_pf_pabuf[1]);
3538 pf_empty_kpool(&V_pf_pabuf[2]);
3539 *ticket = ++V_ticket_pabuf;
3540 PF_RULES_WUNLOCK();
3541
3542 return (0);
3543 }
3544
3545 int
pf_ioctl_add_addr(struct pf_nl_pooladdr * pp)3546 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
3547 {
3548 struct pf_kpooladdr *pa = NULL;
3549 struct pfi_kkif *kif = NULL;
3550 int error;
3551
3552 if (pp->which != PF_RDR && pp->which != PF_NAT &&
3553 pp->which != PF_RT)
3554 return (EINVAL);
3555
3556 switch (pp->af) {
3557 #ifdef INET
3558 case AF_INET:
3559 /* FALLTHROUGH */
3560 #endif /* INET */
3561 #ifdef INET6
3562 case AF_INET6:
3563 /* FALLTHROUGH */
3564 #endif /* INET6 */
3565 case AF_UNSPEC:
3566 break;
3567 default:
3568 return (EAFNOSUPPORT);
3569 }
3570
3571 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
3572 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
3573 pp->addr.addr.type != PF_ADDR_TABLE)
3574 return (EINVAL);
3575
3576 if (pp->addr.addr.p.dyn != NULL)
3577 return (EINVAL);
3578
3579 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
3580 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
3581 if (error != 0)
3582 goto out;
3583 if (pa->ifname[0])
3584 kif = pf_kkif_create(M_WAITOK);
3585 PF_RULES_WLOCK();
3586 if (pp->ticket != V_ticket_pabuf) {
3587 PF_RULES_WUNLOCK();
3588 if (pa->ifname[0])
3589 pf_kkif_free(kif);
3590 error = EBUSY;
3591 goto out;
3592 }
3593 if (pa->ifname[0]) {
3594 pa->kif = pfi_kkif_attach(kif, pa->ifname);
3595 kif = NULL;
3596 pfi_kkif_ref(pa->kif);
3597 } else
3598 pa->kif = NULL;
3599 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
3600 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
3601 if (pa->ifname[0])
3602 pfi_kkif_unref(pa->kif);
3603 PF_RULES_WUNLOCK();
3604 goto out;
3605 }
3606 pa->af = pp->af;
3607 switch (pp->which) {
3608 case PF_NAT:
3609 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries);
3610 break;
3611 case PF_RDR:
3612 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries);
3613 break;
3614 case PF_RT:
3615 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries);
3616 break;
3617 }
3618 PF_RULES_WUNLOCK();
3619
3620 return (0);
3621
3622 out:
3623 free(pa, M_PFRULE);
3624 return (error);
3625 }
3626
3627 int
pf_ioctl_get_addrs(struct pf_nl_pooladdr * pp)3628 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
3629 {
3630 struct pf_kpool *pool;
3631 struct pf_kpooladdr *pa;
3632
3633 PF_RULES_RLOCK_TRACKER;
3634
3635 if (pp->which != PF_RDR && pp->which != PF_NAT &&
3636 pp->which != PF_RT)
3637 return (EINVAL);
3638
3639 pp->anchor[sizeof(pp->anchor) - 1] = 0;
3640 pp->nr = 0;
3641
3642 PF_RULES_RLOCK();
3643 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3644 pp->r_num, 0, 1, 0, pp->which);
3645 if (pool == NULL) {
3646 PF_RULES_RUNLOCK();
3647 return (EBUSY);
3648 }
3649 TAILQ_FOREACH(pa, &pool->list, entries)
3650 pp->nr++;
3651 PF_RULES_RUNLOCK();
3652
3653 return (0);
3654 }
3655
3656 int
pf_ioctl_get_addr(struct pf_nl_pooladdr * pp)3657 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
3658 {
3659 struct pf_kpool *pool;
3660 struct pf_kpooladdr *pa;
3661 u_int32_t nr = 0;
3662
3663 if (pp->which != PF_RDR && pp->which != PF_NAT &&
3664 pp->which != PF_RT)
3665 return (EINVAL);
3666
3667 PF_RULES_RLOCK_TRACKER;
3668
3669 pp->anchor[sizeof(pp->anchor) - 1] = '\0';
3670
3671 PF_RULES_RLOCK();
3672 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3673 pp->r_num, 0, 1, 1, pp->which);
3674 if (pool == NULL) {
3675 PF_RULES_RUNLOCK();
3676 return (EBUSY);
3677 }
3678 pa = TAILQ_FIRST(&pool->list);
3679 while ((pa != NULL) && (nr < pp->nr)) {
3680 pa = TAILQ_NEXT(pa, entries);
3681 nr++;
3682 }
3683 if (pa == NULL) {
3684 PF_RULES_RUNLOCK();
3685 return (EBUSY);
3686 }
3687 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3688 pp->af = pa->af;
3689 pf_addr_copyout(&pp->addr.addr);
3690 PF_RULES_RUNLOCK();
3691
3692 return (0);
3693 }
3694
3695 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)3696 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
3697 {
3698 struct pf_kruleset *ruleset;
3699 struct pf_kanchor *anchor;
3700
3701 PF_RULES_RLOCK_TRACKER;
3702
3703 pr->path[sizeof(pr->path) - 1] = '\0';
3704
3705 PF_RULES_RLOCK();
3706 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3707 PF_RULES_RUNLOCK();
3708 return (ENOENT);
3709 }
3710 pr->nr = 0;
3711 if (ruleset == &pf_main_ruleset) {
3712 /* XXX kludge for pf_main_ruleset */
3713 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3714 if (anchor->parent == NULL)
3715 pr->nr++;
3716 } else {
3717 RB_FOREACH(anchor, pf_kanchor_node,
3718 &ruleset->anchor->children)
3719 pr->nr++;
3720 }
3721 PF_RULES_RUNLOCK();
3722
3723 return (0);
3724 }
3725
3726 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)3727 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
3728 {
3729 struct pf_kruleset *ruleset;
3730 struct pf_kanchor *anchor;
3731 u_int32_t nr = 0;
3732 int error = 0;
3733
3734 PF_RULES_RLOCK_TRACKER;
3735
3736 PF_RULES_RLOCK();
3737 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3738 PF_RULES_RUNLOCK();
3739 return (ENOENT);
3740 }
3741
3742 pr->name[0] = '\0';
3743 if (ruleset == &pf_main_ruleset) {
3744 /* XXX kludge for pf_main_ruleset */
3745 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3746 if (anchor->parent == NULL && nr++ == pr->nr) {
3747 strlcpy(pr->name, anchor->name,
3748 sizeof(pr->name));
3749 break;
3750 }
3751 } else {
3752 RB_FOREACH(anchor, pf_kanchor_node,
3753 &ruleset->anchor->children)
3754 if (nr++ == pr->nr) {
3755 strlcpy(pr->name, anchor->name,
3756 sizeof(pr->name));
3757 break;
3758 }
3759 }
3760 if (!pr->name[0])
3761 error = EBUSY;
3762 PF_RULES_RUNLOCK();
3763
3764 return (error);
3765 }
3766
3767 int
pf_ioctl_natlook(struct pfioc_natlook * pnl)3768 pf_ioctl_natlook(struct pfioc_natlook *pnl)
3769 {
3770 struct pf_state_key *sk;
3771 struct pf_kstate *state;
3772 struct pf_state_key_cmp key;
3773 int m = 0, direction = pnl->direction;
3774 int sidx, didx;
3775
3776 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
3777 sidx = (direction == PF_IN) ? 1 : 0;
3778 didx = (direction == PF_IN) ? 0 : 1;
3779
3780 if (!pnl->proto ||
3781 PF_AZERO(&pnl->saddr, pnl->af) ||
3782 PF_AZERO(&pnl->daddr, pnl->af) ||
3783 ((pnl->proto == IPPROTO_TCP ||
3784 pnl->proto == IPPROTO_UDP) &&
3785 (!pnl->dport || !pnl->sport)))
3786 return (EINVAL);
3787
3788 switch (pnl->direction) {
3789 case PF_IN:
3790 case PF_OUT:
3791 case PF_INOUT:
3792 break;
3793 default:
3794 return (EINVAL);
3795 }
3796
3797 switch (pnl->af) {
3798 #ifdef INET
3799 case AF_INET:
3800 break;
3801 #endif /* INET */
3802 #ifdef INET6
3803 case AF_INET6:
3804 break;
3805 #endif /* INET6 */
3806 default:
3807 return (EAFNOSUPPORT);
3808 }
3809
3810 bzero(&key, sizeof(key));
3811 key.af = pnl->af;
3812 key.proto = pnl->proto;
3813 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af);
3814 key.port[sidx] = pnl->sport;
3815 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af);
3816 key.port[didx] = pnl->dport;
3817
3818 state = pf_find_state_all(&key, direction, &m);
3819 if (state == NULL)
3820 return (ENOENT);
3821
3822 if (m > 1) {
3823 PF_STATE_UNLOCK(state);
3824 return (E2BIG); /* more than one state */
3825 }
3826
3827 sk = state->key[sidx];
3828 pf_addrcpy(&pnl->rsaddr,
3829 &sk->addr[sidx], sk->af);
3830 pnl->rsport = sk->port[sidx];
3831 pf_addrcpy(&pnl->rdaddr,
3832 &sk->addr[didx], sk->af);
3833 pnl->rdport = sk->port[didx];
3834 PF_STATE_UNLOCK(state);
3835
3836 return (0);
3837 }
3838
3839 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)3840 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
3841 {
3842 int error = 0;
3843 PF_RULES_RLOCK_TRACKER;
3844
3845 #define ERROUT_IOCTL(target, x) \
3846 do { \
3847 error = (x); \
3848 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
3849 goto target; \
3850 } while (0)
3851
3852
3853 /* XXX keep in sync with switch() below */
3854 if (securelevel_gt(td->td_ucred, 2))
3855 switch (cmd) {
3856 case DIOCGETRULES:
3857 case DIOCGETRULENV:
3858 case DIOCGETADDRS:
3859 case DIOCGETADDR:
3860 case DIOCGETSTATE:
3861 case DIOCGETSTATENV:
3862 case DIOCSETSTATUSIF:
3863 case DIOCGETSTATUSNV:
3864 case DIOCCLRSTATUS:
3865 case DIOCNATLOOK:
3866 case DIOCSETDEBUG:
3867 #ifdef COMPAT_FREEBSD14
3868 case DIOCGETSTATES:
3869 case DIOCGETSTATESV2:
3870 #endif
3871 case DIOCGETTIMEOUT:
3872 case DIOCCLRRULECTRS:
3873 case DIOCGETLIMIT:
3874 case DIOCGETALTQSV0:
3875 case DIOCGETALTQSV1:
3876 case DIOCGETALTQV0:
3877 case DIOCGETALTQV1:
3878 case DIOCGETQSTATSV0:
3879 case DIOCGETQSTATSV1:
3880 case DIOCGETRULESETS:
3881 case DIOCGETRULESET:
3882 case DIOCRGETTABLES:
3883 case DIOCRGETTSTATS:
3884 case DIOCRCLRTSTATS:
3885 case DIOCRCLRADDRS:
3886 case DIOCRADDADDRS:
3887 case DIOCRDELADDRS:
3888 case DIOCRSETADDRS:
3889 case DIOCRGETADDRS:
3890 case DIOCRGETASTATS:
3891 case DIOCRCLRASTATS:
3892 case DIOCRTSTADDRS:
3893 case DIOCOSFPGET:
3894 case DIOCGETSRCNODES:
3895 case DIOCCLRSRCNODES:
3896 case DIOCGETSYNCOOKIES:
3897 case DIOCIGETIFACES:
3898 case DIOCGIFSPEEDV0:
3899 case DIOCGIFSPEEDV1:
3900 case DIOCSETIFFLAG:
3901 case DIOCCLRIFFLAG:
3902 case DIOCGETETHRULES:
3903 case DIOCGETETHRULE:
3904 case DIOCGETETHRULESETS:
3905 case DIOCGETETHRULESET:
3906 break;
3907 case DIOCRCLRTABLES:
3908 case DIOCRADDTABLES:
3909 case DIOCRDELTABLES:
3910 case DIOCRSETTFLAGS:
3911 if (((struct pfioc_table *)addr)->pfrio_flags &
3912 PFR_FLAG_DUMMY)
3913 break; /* dummy operation ok */
3914 return (EPERM);
3915 default:
3916 return (EPERM);
3917 }
3918
3919 if (!(flags & FWRITE))
3920 switch (cmd) {
3921 case DIOCGETRULES:
3922 case DIOCGETADDRS:
3923 case DIOCGETADDR:
3924 case DIOCGETSTATE:
3925 case DIOCGETSTATENV:
3926 case DIOCGETSTATUSNV:
3927 #ifdef COMPAT_FREEBSD14
3928 case DIOCGETSTATES:
3929 case DIOCGETSTATESV2:
3930 #endif
3931 case DIOCGETTIMEOUT:
3932 case DIOCGETLIMIT:
3933 case DIOCGETALTQSV0:
3934 case DIOCGETALTQSV1:
3935 case DIOCGETALTQV0:
3936 case DIOCGETALTQV1:
3937 case DIOCGETQSTATSV0:
3938 case DIOCGETQSTATSV1:
3939 case DIOCGETRULESETS:
3940 case DIOCGETRULESET:
3941 case DIOCNATLOOK:
3942 case DIOCRGETTABLES:
3943 case DIOCRGETTSTATS:
3944 case DIOCRGETADDRS:
3945 case DIOCRGETASTATS:
3946 case DIOCRTSTADDRS:
3947 case DIOCOSFPGET:
3948 case DIOCGETSRCNODES:
3949 case DIOCGETSYNCOOKIES:
3950 case DIOCIGETIFACES:
3951 case DIOCGIFSPEEDV1:
3952 case DIOCGIFSPEEDV0:
3953 case DIOCGETRULENV:
3954 case DIOCGETETHRULES:
3955 case DIOCGETETHRULE:
3956 case DIOCGETETHRULESETS:
3957 case DIOCGETETHRULESET:
3958 break;
3959 case DIOCRCLRTABLES:
3960 case DIOCRADDTABLES:
3961 case DIOCRDELTABLES:
3962 case DIOCRCLRTSTATS:
3963 case DIOCRCLRADDRS:
3964 case DIOCRADDADDRS:
3965 case DIOCRDELADDRS:
3966 case DIOCRSETADDRS:
3967 case DIOCRSETTFLAGS:
3968 if (((struct pfioc_table *)addr)->pfrio_flags &
3969 PFR_FLAG_DUMMY) {
3970 flags |= FWRITE; /* need write lock for dummy */
3971 break; /* dummy operation ok */
3972 }
3973 return (EACCES);
3974 default:
3975 return (EACCES);
3976 }
3977
3978 CURVNET_SET(TD_TO_VNET(td));
3979
3980 switch (cmd) {
3981 #ifdef COMPAT_FREEBSD14
3982 case DIOCSTART:
3983 error = pf_start();
3984 break;
3985
3986 case DIOCSTOP:
3987 error = pf_stop();
3988 break;
3989 #endif
3990
3991 case DIOCGETETHRULES: {
3992 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3993 nvlist_t *nvl;
3994 void *packed;
3995 struct pf_keth_rule *tail;
3996 struct pf_keth_ruleset *rs;
3997 u_int32_t ticket, nr;
3998 const char *anchor = "";
3999
4000 nvl = NULL;
4001 packed = NULL;
4002
4003 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
4004
4005 if (nv->len > pf_ioctl_maxcount)
4006 ERROUT(ENOMEM);
4007
4008 /* Copy the request in */
4009 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
4010 error = copyin(nv->data, packed, nv->len);
4011 if (error)
4012 ERROUT(error);
4013
4014 nvl = nvlist_unpack(packed, nv->len, 0);
4015 if (nvl == NULL)
4016 ERROUT(EBADMSG);
4017
4018 if (! nvlist_exists_string(nvl, "anchor"))
4019 ERROUT(EBADMSG);
4020
4021 anchor = nvlist_get_string(nvl, "anchor");
4022
4023 rs = pf_find_keth_ruleset(anchor);
4024
4025 nvlist_destroy(nvl);
4026 nvl = NULL;
4027 free(packed, M_NVLIST);
4028 packed = NULL;
4029
4030 if (rs == NULL)
4031 ERROUT(ENOENT);
4032
4033 /* Reply */
4034 nvl = nvlist_create(0);
4035 if (nvl == NULL)
4036 ERROUT(ENOMEM);
4037
4038 PF_RULES_RLOCK();
4039
4040 ticket = rs->active.ticket;
4041 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
4042 if (tail)
4043 nr = tail->nr + 1;
4044 else
4045 nr = 0;
4046
4047 PF_RULES_RUNLOCK();
4048
4049 nvlist_add_number(nvl, "ticket", ticket);
4050 nvlist_add_number(nvl, "nr", nr);
4051
4052 packed = nvlist_pack(nvl, &nv->len);
4053 if (packed == NULL)
4054 ERROUT(ENOMEM);
4055
4056 if (nv->size == 0)
4057 ERROUT(0);
4058 else if (nv->size < nv->len)
4059 ERROUT(ENOSPC);
4060
4061 error = copyout(packed, nv->data, nv->len);
4062
4063 #undef ERROUT
4064 DIOCGETETHRULES_error:
4065 free(packed, M_NVLIST);
4066 nvlist_destroy(nvl);
4067 break;
4068 }
4069
4070 case DIOCGETETHRULE: {
4071 struct epoch_tracker et;
4072 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4073 nvlist_t *nvl = NULL;
4074 void *nvlpacked = NULL;
4075 struct pf_keth_rule *rule = NULL;
4076 struct pf_keth_ruleset *rs;
4077 u_int32_t ticket, nr;
4078 bool clear = false;
4079 const char *anchor;
4080
4081 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
4082
4083 if (nv->len > pf_ioctl_maxcount)
4084 ERROUT(ENOMEM);
4085
4086 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4087 error = copyin(nv->data, nvlpacked, nv->len);
4088 if (error)
4089 ERROUT(error);
4090
4091 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4092 if (nvl == NULL)
4093 ERROUT(EBADMSG);
4094 if (! nvlist_exists_number(nvl, "ticket"))
4095 ERROUT(EBADMSG);
4096 ticket = nvlist_get_number(nvl, "ticket");
4097 if (! nvlist_exists_string(nvl, "anchor"))
4098 ERROUT(EBADMSG);
4099 anchor = nvlist_get_string(nvl, "anchor");
4100
4101 if (nvlist_exists_bool(nvl, "clear"))
4102 clear = nvlist_get_bool(nvl, "clear");
4103
4104 if (clear && !(flags & FWRITE))
4105 ERROUT(EACCES);
4106
4107 if (! nvlist_exists_number(nvl, "nr"))
4108 ERROUT(EBADMSG);
4109 nr = nvlist_get_number(nvl, "nr");
4110
4111 PF_RULES_RLOCK();
4112 rs = pf_find_keth_ruleset(anchor);
4113 if (rs == NULL) {
4114 PF_RULES_RUNLOCK();
4115 ERROUT(ENOENT);
4116 }
4117 if (ticket != rs->active.ticket) {
4118 PF_RULES_RUNLOCK();
4119 ERROUT(EBUSY);
4120 }
4121
4122 nvlist_destroy(nvl);
4123 nvl = NULL;
4124 free(nvlpacked, M_NVLIST);
4125 nvlpacked = NULL;
4126
4127 rule = TAILQ_FIRST(rs->active.rules);
4128 while ((rule != NULL) && (rule->nr != nr))
4129 rule = TAILQ_NEXT(rule, entries);
4130 if (rule == NULL) {
4131 PF_RULES_RUNLOCK();
4132 ERROUT(ENOENT);
4133 }
4134 /* Make sure rule can't go away. */
4135 NET_EPOCH_ENTER(et);
4136 PF_RULES_RUNLOCK();
4137 nvl = pf_keth_rule_to_nveth_rule(rule);
4138 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
4139 NET_EPOCH_EXIT(et);
4140 ERROUT(EBUSY);
4141 }
4142 NET_EPOCH_EXIT(et);
4143 if (nvl == NULL)
4144 ERROUT(ENOMEM);
4145
4146 nvlpacked = nvlist_pack(nvl, &nv->len);
4147 if (nvlpacked == NULL)
4148 ERROUT(ENOMEM);
4149
4150 if (nv->size == 0)
4151 ERROUT(0);
4152 else if (nv->size < nv->len)
4153 ERROUT(ENOSPC);
4154
4155 error = copyout(nvlpacked, nv->data, nv->len);
4156 if (error == 0 && clear) {
4157 counter_u64_zero(rule->evaluations);
4158 for (int i = 0; i < 2; i++) {
4159 counter_u64_zero(rule->packets[i]);
4160 counter_u64_zero(rule->bytes[i]);
4161 }
4162 }
4163
4164 #undef ERROUT
4165 DIOCGETETHRULE_error:
4166 free(nvlpacked, M_NVLIST);
4167 nvlist_destroy(nvl);
4168 break;
4169 }
4170
4171 case DIOCADDETHRULE: {
4172 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4173 nvlist_t *nvl = NULL;
4174 void *nvlpacked = NULL;
4175 struct pf_keth_rule *rule = NULL, *tail = NULL;
4176 struct pf_keth_ruleset *ruleset = NULL;
4177 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
4178 const char *anchor = "", *anchor_call = "";
4179
4180 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
4181
4182 if (nv->len > pf_ioctl_maxcount)
4183 ERROUT(ENOMEM);
4184
4185 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4186 error = copyin(nv->data, nvlpacked, nv->len);
4187 if (error)
4188 ERROUT(error);
4189
4190 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4191 if (nvl == NULL)
4192 ERROUT(EBADMSG);
4193
4194 if (! nvlist_exists_number(nvl, "ticket"))
4195 ERROUT(EBADMSG);
4196
4197 if (nvlist_exists_string(nvl, "anchor"))
4198 anchor = nvlist_get_string(nvl, "anchor");
4199 if (nvlist_exists_string(nvl, "anchor_call"))
4200 anchor_call = nvlist_get_string(nvl, "anchor_call");
4201
4202 ruleset = pf_find_keth_ruleset(anchor);
4203 if (ruleset == NULL)
4204 ERROUT(EINVAL);
4205
4206 if (nvlist_get_number(nvl, "ticket") !=
4207 ruleset->inactive.ticket) {
4208 DPFPRINTF(PF_DEBUG_MISC,
4209 "ticket: %d != %d",
4210 (u_int32_t)nvlist_get_number(nvl, "ticket"),
4211 ruleset->inactive.ticket);
4212 ERROUT(EBUSY);
4213 }
4214
4215 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
4216 rule->timestamp = NULL;
4217
4218 error = pf_nveth_rule_to_keth_rule(nvl, rule);
4219 if (error != 0)
4220 ERROUT(error);
4221
4222 if (rule->ifname[0])
4223 kif = pf_kkif_create(M_WAITOK);
4224 if (rule->bridge_to_name[0])
4225 bridge_to_kif = pf_kkif_create(M_WAITOK);
4226 rule->evaluations = counter_u64_alloc(M_WAITOK);
4227 for (int i = 0; i < 2; i++) {
4228 rule->packets[i] = counter_u64_alloc(M_WAITOK);
4229 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
4230 }
4231 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
4232 M_WAITOK | M_ZERO);
4233
4234 PF_RULES_WLOCK();
4235
4236 if (rule->ifname[0]) {
4237 rule->kif = pfi_kkif_attach(kif, rule->ifname);
4238 pfi_kkif_ref(rule->kif);
4239 } else
4240 rule->kif = NULL;
4241 if (rule->bridge_to_name[0]) {
4242 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
4243 rule->bridge_to_name);
4244 pfi_kkif_ref(rule->bridge_to);
4245 } else
4246 rule->bridge_to = NULL;
4247
4248 #ifdef ALTQ
4249 /* set queue IDs */
4250 if (rule->qname[0] != 0) {
4251 if ((rule->qid = pf_qname2qid(rule->qname, true)) == 0)
4252 error = EBUSY;
4253 else
4254 rule->qid = rule->qid;
4255 }
4256 #endif
4257 if (rule->tagname[0])
4258 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
4259 error = EBUSY;
4260 if (rule->match_tagname[0])
4261 if ((rule->match_tag = pf_tagname2tag(
4262 rule->match_tagname)) == 0)
4263 error = EBUSY;
4264
4265 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
4266 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
4267 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
4268 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
4269
4270 if (error) {
4271 pf_free_eth_rule(rule);
4272 PF_RULES_WUNLOCK();
4273 ERROUT(error);
4274 }
4275
4276 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
4277 pf_free_eth_rule(rule);
4278 PF_RULES_WUNLOCK();
4279 ERROUT(EINVAL);
4280 }
4281
4282 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
4283 if (tail)
4284 rule->nr = tail->nr + 1;
4285 else
4286 rule->nr = 0;
4287
4288 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
4289
4290 PF_RULES_WUNLOCK();
4291
4292 #undef ERROUT
4293 DIOCADDETHRULE_error:
4294 nvlist_destroy(nvl);
4295 free(nvlpacked, M_NVLIST);
4296 break;
4297 }
4298
4299 case DIOCGETETHRULESETS: {
4300 struct epoch_tracker et;
4301 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4302 nvlist_t *nvl = NULL;
4303 void *nvlpacked = NULL;
4304 struct pf_keth_ruleset *ruleset;
4305 struct pf_keth_anchor *anchor;
4306 int nr = 0;
4307
4308 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
4309
4310 if (nv->len > pf_ioctl_maxcount)
4311 ERROUT(ENOMEM);
4312
4313 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4314 error = copyin(nv->data, nvlpacked, nv->len);
4315 if (error)
4316 ERROUT(error);
4317
4318 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4319 if (nvl == NULL)
4320 ERROUT(EBADMSG);
4321 if (! nvlist_exists_string(nvl, "path"))
4322 ERROUT(EBADMSG);
4323
4324 NET_EPOCH_ENTER(et);
4325
4326 if ((ruleset = pf_find_keth_ruleset(
4327 nvlist_get_string(nvl, "path"))) == NULL) {
4328 NET_EPOCH_EXIT(et);
4329 ERROUT(ENOENT);
4330 }
4331
4332 if (ruleset->anchor == NULL) {
4333 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
4334 if (anchor->parent == NULL)
4335 nr++;
4336 } else {
4337 RB_FOREACH(anchor, pf_keth_anchor_node,
4338 &ruleset->anchor->children)
4339 nr++;
4340 }
4341
4342 NET_EPOCH_EXIT(et);
4343
4344 nvlist_destroy(nvl);
4345 nvl = NULL;
4346 free(nvlpacked, M_NVLIST);
4347 nvlpacked = NULL;
4348
4349 nvl = nvlist_create(0);
4350 if (nvl == NULL)
4351 ERROUT(ENOMEM);
4352
4353 nvlist_add_number(nvl, "nr", nr);
4354
4355 nvlpacked = nvlist_pack(nvl, &nv->len);
4356 if (nvlpacked == NULL)
4357 ERROUT(ENOMEM);
4358
4359 if (nv->size == 0)
4360 ERROUT(0);
4361 else if (nv->size < nv->len)
4362 ERROUT(ENOSPC);
4363
4364 error = copyout(nvlpacked, nv->data, nv->len);
4365
4366 #undef ERROUT
4367 DIOCGETETHRULESETS_error:
4368 free(nvlpacked, M_NVLIST);
4369 nvlist_destroy(nvl);
4370 break;
4371 }
4372
4373 case DIOCGETETHRULESET: {
4374 struct epoch_tracker et;
4375 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4376 nvlist_t *nvl = NULL;
4377 void *nvlpacked = NULL;
4378 struct pf_keth_ruleset *ruleset;
4379 struct pf_keth_anchor *anchor;
4380 int nr = 0, req_nr = 0;
4381 bool found = false;
4382
4383 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
4384
4385 if (nv->len > pf_ioctl_maxcount)
4386 ERROUT(ENOMEM);
4387
4388 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4389 error = copyin(nv->data, nvlpacked, nv->len);
4390 if (error)
4391 ERROUT(error);
4392
4393 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4394 if (nvl == NULL)
4395 ERROUT(EBADMSG);
4396 if (! nvlist_exists_string(nvl, "path"))
4397 ERROUT(EBADMSG);
4398 if (! nvlist_exists_number(nvl, "nr"))
4399 ERROUT(EBADMSG);
4400
4401 req_nr = nvlist_get_number(nvl, "nr");
4402
4403 NET_EPOCH_ENTER(et);
4404
4405 if ((ruleset = pf_find_keth_ruleset(
4406 nvlist_get_string(nvl, "path"))) == NULL) {
4407 NET_EPOCH_EXIT(et);
4408 ERROUT(ENOENT);
4409 }
4410
4411 nvlist_destroy(nvl);
4412 nvl = NULL;
4413 free(nvlpacked, M_NVLIST);
4414 nvlpacked = NULL;
4415
4416 nvl = nvlist_create(0);
4417 if (nvl == NULL) {
4418 NET_EPOCH_EXIT(et);
4419 ERROUT(ENOMEM);
4420 }
4421
4422 if (ruleset->anchor == NULL) {
4423 RB_FOREACH(anchor, pf_keth_anchor_global,
4424 &V_pf_keth_anchors) {
4425 if (anchor->parent == NULL && nr++ == req_nr) {
4426 found = true;
4427 break;
4428 }
4429 }
4430 } else {
4431 RB_FOREACH(anchor, pf_keth_anchor_node,
4432 &ruleset->anchor->children) {
4433 if (nr++ == req_nr) {
4434 found = true;
4435 break;
4436 }
4437 }
4438 }
4439
4440 NET_EPOCH_EXIT(et);
4441 if (found) {
4442 nvlist_add_number(nvl, "nr", nr);
4443 nvlist_add_string(nvl, "name", anchor->name);
4444 if (ruleset->anchor)
4445 nvlist_add_string(nvl, "path",
4446 ruleset->anchor->path);
4447 else
4448 nvlist_add_string(nvl, "path", "");
4449 } else {
4450 ERROUT(EBUSY);
4451 }
4452
4453 nvlpacked = nvlist_pack(nvl, &nv->len);
4454 if (nvlpacked == NULL)
4455 ERROUT(ENOMEM);
4456
4457 if (nv->size == 0)
4458 ERROUT(0);
4459 else if (nv->size < nv->len)
4460 ERROUT(ENOSPC);
4461
4462 error = copyout(nvlpacked, nv->data, nv->len);
4463
4464 #undef ERROUT
4465 DIOCGETETHRULESET_error:
4466 free(nvlpacked, M_NVLIST);
4467 nvlist_destroy(nvl);
4468 break;
4469 }
4470
4471 case DIOCADDRULENV: {
4472 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4473 nvlist_t *nvl = NULL;
4474 void *nvlpacked = NULL;
4475 struct pf_krule *rule = NULL;
4476 const char *anchor = "", *anchor_call = "";
4477 uint32_t ticket = 0, pool_ticket = 0;
4478
4479 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
4480
4481 if (nv->len > pf_ioctl_maxcount)
4482 ERROUT(ENOMEM);
4483
4484 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4485 error = copyin(nv->data, nvlpacked, nv->len);
4486 if (error)
4487 ERROUT(error);
4488
4489 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4490 if (nvl == NULL)
4491 ERROUT(EBADMSG);
4492
4493 if (! nvlist_exists_number(nvl, "ticket"))
4494 ERROUT(EINVAL);
4495 ticket = nvlist_get_number(nvl, "ticket");
4496
4497 if (! nvlist_exists_number(nvl, "pool_ticket"))
4498 ERROUT(EINVAL);
4499 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
4500
4501 if (! nvlist_exists_nvlist(nvl, "rule"))
4502 ERROUT(EINVAL);
4503
4504 rule = pf_krule_alloc();
4505 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
4506 rule);
4507 if (error)
4508 ERROUT(error);
4509
4510 if (nvlist_exists_string(nvl, "anchor"))
4511 anchor = nvlist_get_string(nvl, "anchor");
4512 if (nvlist_exists_string(nvl, "anchor_call"))
4513 anchor_call = nvlist_get_string(nvl, "anchor_call");
4514
4515 if ((error = nvlist_error(nvl)))
4516 ERROUT(error);
4517
4518 /* Frees rule on error */
4519 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
4520 anchor_call, td->td_ucred->cr_ruid,
4521 td->td_proc ? td->td_proc->p_pid : 0);
4522
4523 nvlist_destroy(nvl);
4524 free(nvlpacked, M_NVLIST);
4525 break;
4526 #undef ERROUT
4527 DIOCADDRULENV_error:
4528 pf_krule_free(rule);
4529 nvlist_destroy(nvl);
4530 free(nvlpacked, M_NVLIST);
4531
4532 break;
4533 }
4534 case DIOCADDRULE: {
4535 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
4536 struct pf_krule *rule;
4537
4538 rule = pf_krule_alloc();
4539 error = pf_rule_to_krule(&pr->rule, rule);
4540 if (error != 0) {
4541 pf_krule_free(rule);
4542 goto fail;
4543 }
4544
4545 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
4546
4547 /* Frees rule on error */
4548 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
4549 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
4550 td->td_proc ? td->td_proc->p_pid : 0);
4551 break;
4552 }
4553
4554 case DIOCGETRULES: {
4555 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
4556
4557 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
4558
4559 error = pf_ioctl_getrules(pr);
4560
4561 break;
4562 }
4563
4564 case DIOCGETRULENV: {
4565 PF_RULES_RLOCK_TRACKER;
4566 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4567 nvlist_t *nvrule = NULL;
4568 nvlist_t *nvl = NULL;
4569 struct pf_kruleset *ruleset;
4570 struct pf_krule *rule;
4571 void *nvlpacked = NULL;
4572 int rs_num, nr;
4573 bool clear_counter = false;
4574
4575 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
4576 #define ERROUT_LOCKED(x) do { \
4577 if (clear_counter) \
4578 PF_RULES_WUNLOCK(); \
4579 else \
4580 PF_RULES_RUNLOCK(); \
4581 ERROUT(x); \
4582 } while (0)
4583
4584 if (nv->len > pf_ioctl_maxcount)
4585 ERROUT(ENOMEM);
4586
4587 /* Copy the request in */
4588 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4589 error = copyin(nv->data, nvlpacked, nv->len);
4590 if (error)
4591 ERROUT(error);
4592
4593 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4594 if (nvl == NULL)
4595 ERROUT(EBADMSG);
4596
4597 if (! nvlist_exists_string(nvl, "anchor"))
4598 ERROUT(EBADMSG);
4599 if (! nvlist_exists_number(nvl, "ruleset"))
4600 ERROUT(EBADMSG);
4601 if (! nvlist_exists_number(nvl, "ticket"))
4602 ERROUT(EBADMSG);
4603 if (! nvlist_exists_number(nvl, "nr"))
4604 ERROUT(EBADMSG);
4605
4606 if (nvlist_exists_bool(nvl, "clear_counter"))
4607 clear_counter = nvlist_get_bool(nvl, "clear_counter");
4608
4609 if (clear_counter && !(flags & FWRITE))
4610 ERROUT(EACCES);
4611
4612 nr = nvlist_get_number(nvl, "nr");
4613
4614 if (clear_counter)
4615 PF_RULES_WLOCK();
4616 else
4617 PF_RULES_RLOCK();
4618 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
4619 if (ruleset == NULL)
4620 ERROUT_LOCKED(ENOENT);
4621
4622 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
4623 if (rs_num >= PF_RULESET_MAX)
4624 ERROUT_LOCKED(EINVAL);
4625
4626 if (nvlist_get_number(nvl, "ticket") !=
4627 ruleset->rules[rs_num].active.ticket)
4628 ERROUT_LOCKED(EBUSY);
4629
4630 if ((error = nvlist_error(nvl)))
4631 ERROUT_LOCKED(error);
4632
4633 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
4634 while ((rule != NULL) && (rule->nr != nr))
4635 rule = TAILQ_NEXT(rule, entries);
4636 if (rule == NULL)
4637 ERROUT_LOCKED(EBUSY);
4638
4639 nvrule = pf_krule_to_nvrule(rule);
4640
4641 nvlist_destroy(nvl);
4642 nvl = nvlist_create(0);
4643 if (nvl == NULL)
4644 ERROUT_LOCKED(ENOMEM);
4645 nvlist_add_number(nvl, "nr", nr);
4646 nvlist_add_nvlist(nvl, "rule", nvrule);
4647 nvlist_destroy(nvrule);
4648 nvrule = NULL;
4649 if (pf_kanchor_nvcopyout(ruleset, rule, nvl))
4650 ERROUT_LOCKED(EBUSY);
4651
4652 free(nvlpacked, M_NVLIST);
4653 nvlpacked = nvlist_pack(nvl, &nv->len);
4654 if (nvlpacked == NULL)
4655 ERROUT_LOCKED(ENOMEM);
4656
4657 if (nv->size == 0)
4658 ERROUT_LOCKED(0);
4659 else if (nv->size < nv->len)
4660 ERROUT_LOCKED(ENOSPC);
4661
4662 if (clear_counter) {
4663 pf_krule_clear_counters(rule);
4664 PF_RULES_WUNLOCK();
4665 } else {
4666 PF_RULES_RUNLOCK();
4667 }
4668
4669 error = copyout(nvlpacked, nv->data, nv->len);
4670
4671 #undef ERROUT_LOCKED
4672 #undef ERROUT
4673 DIOCGETRULENV_error:
4674 free(nvlpacked, M_NVLIST);
4675 nvlist_destroy(nvrule);
4676 nvlist_destroy(nvl);
4677
4678 break;
4679 }
4680
4681 case DIOCCHANGERULE: {
4682 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
4683 struct pf_kruleset *ruleset;
4684 struct pf_krule *oldrule = NULL, *newrule = NULL;
4685 struct pfi_kkif *kif = NULL;
4686 struct pf_kpooladdr *pa;
4687 u_int32_t nr = 0;
4688 int rs_num;
4689
4690 pcr->anchor[sizeof(pcr->anchor) - 1] = '\0';
4691
4692 if (pcr->action < PF_CHANGE_ADD_HEAD ||
4693 pcr->action > PF_CHANGE_GET_TICKET) {
4694 error = EINVAL;
4695 goto fail;
4696 }
4697 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
4698 error = EINVAL;
4699 goto fail;
4700 }
4701
4702 if (pcr->action != PF_CHANGE_REMOVE) {
4703 newrule = pf_krule_alloc();
4704 error = pf_rule_to_krule(&pcr->rule, newrule);
4705 if (error != 0) {
4706 pf_krule_free(newrule);
4707 goto fail;
4708 }
4709
4710 if ((error = pf_rule_checkaf(newrule))) {
4711 pf_krule_free(newrule);
4712 goto fail;
4713 }
4714 if (newrule->ifname[0])
4715 kif = pf_kkif_create(M_WAITOK);
4716 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
4717 for (int i = 0; i < 2; i++) {
4718 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
4719 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
4720 }
4721 newrule->states_cur = counter_u64_alloc(M_WAITOK);
4722 newrule->states_tot = counter_u64_alloc(M_WAITOK);
4723 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
4724 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
4725 newrule->cuid = td->td_ucred->cr_ruid;
4726 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
4727 TAILQ_INIT(&newrule->nat.list);
4728 TAILQ_INIT(&newrule->rdr.list);
4729 TAILQ_INIT(&newrule->route.list);
4730 }
4731 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
4732
4733 PF_CONFIG_LOCK();
4734 PF_RULES_WLOCK();
4735 #ifdef PF_WANT_32_TO_64_COUNTER
4736 if (newrule != NULL) {
4737 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
4738 newrule->allrulelinked = true;
4739 V_pf_allrulecount++;
4740 }
4741 #endif
4742
4743 if (!(pcr->action == PF_CHANGE_REMOVE ||
4744 pcr->action == PF_CHANGE_GET_TICKET) &&
4745 pcr->pool_ticket != V_ticket_pabuf)
4746 ERROUT(EBUSY);
4747
4748 ruleset = pf_find_kruleset(pcr->anchor);
4749 if (ruleset == NULL)
4750 ERROUT(EINVAL);
4751
4752 rs_num = pf_get_ruleset_number(pcr->rule.action);
4753 if (rs_num >= PF_RULESET_MAX)
4754 ERROUT(EINVAL);
4755
4756 /*
4757 * XXXMJG: there is no guarantee that the ruleset was
4758 * created by the usual route of calling DIOCXBEGIN.
4759 * As a result it is possible the rule tree will not
4760 * be allocated yet. Hack around it by doing it here.
4761 * Note it is fine to let the tree persist in case of
4762 * error as it will be freed down the road on future
4763 * updates (if need be).
4764 */
4765 if (ruleset->rules[rs_num].active.tree == NULL) {
4766 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
4767 if (ruleset->rules[rs_num].active.tree == NULL) {
4768 ERROUT(ENOMEM);
4769 }
4770 }
4771
4772 if (pcr->action == PF_CHANGE_GET_TICKET) {
4773 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
4774 ERROUT(0);
4775 } else if (pcr->ticket !=
4776 ruleset->rules[rs_num].active.ticket)
4777 ERROUT(EINVAL);
4778
4779 if (pcr->action != PF_CHANGE_REMOVE) {
4780 if (newrule->ifname[0]) {
4781 newrule->kif = pfi_kkif_attach(kif,
4782 newrule->ifname);
4783 kif = NULL;
4784 pfi_kkif_ref(newrule->kif);
4785 } else
4786 newrule->kif = NULL;
4787
4788 if (newrule->rtableid > 0 &&
4789 newrule->rtableid >= rt_numfibs)
4790 error = EBUSY;
4791
4792 #ifdef ALTQ
4793 /* set queue IDs */
4794 if (newrule->qname[0] != 0) {
4795 if ((newrule->qid =
4796 pf_qname2qid(newrule->qname, true)) == 0)
4797 error = EBUSY;
4798 else if (newrule->pqname[0] != 0) {
4799 if ((newrule->pqid =
4800 pf_qname2qid(newrule->pqname, true)) == 0)
4801 error = EBUSY;
4802 } else
4803 newrule->pqid = newrule->qid;
4804 }
4805 #endif /* ALTQ */
4806 if (newrule->tagname[0])
4807 if ((newrule->tag =
4808 pf_tagname2tag(newrule->tagname)) == 0)
4809 error = EBUSY;
4810 if (newrule->match_tagname[0])
4811 if ((newrule->match_tag = pf_tagname2tag(
4812 newrule->match_tagname)) == 0)
4813 error = EBUSY;
4814 if (newrule->rt && !newrule->direction)
4815 error = EINVAL;
4816 if (!newrule->log)
4817 newrule->logif = 0;
4818 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
4819 error = ENOMEM;
4820 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
4821 error = ENOMEM;
4822 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
4823 error = EINVAL;
4824 for (int i = 0; i < 3; i++) {
4825 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
4826 if (pa->addr.type == PF_ADDR_TABLE) {
4827 pa->addr.p.tbl =
4828 pfr_attach_table(ruleset,
4829 pa->addr.v.tblname);
4830 if (pa->addr.p.tbl == NULL)
4831 error = ENOMEM;
4832 }
4833 }
4834
4835 newrule->overload_tbl = NULL;
4836 if (newrule->overload_tblname[0]) {
4837 if ((newrule->overload_tbl = pfr_attach_table(
4838 ruleset, newrule->overload_tblname)) ==
4839 NULL)
4840 error = EINVAL;
4841 else
4842 newrule->overload_tbl->pfrkt_flags |=
4843 PFR_TFLAG_ACTIVE;
4844 }
4845
4846 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
4847 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
4848 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list);
4849 if (((((newrule->action == PF_NAT) ||
4850 (newrule->action == PF_RDR) ||
4851 (newrule->action == PF_BINAT) ||
4852 (newrule->rt > PF_NOPFROUTE)) &&
4853 !newrule->anchor)) &&
4854 (TAILQ_FIRST(&newrule->rdr.list) == NULL))
4855 error = EINVAL;
4856
4857 if (error) {
4858 pf_free_rule(newrule);
4859 PF_RULES_WUNLOCK();
4860 PF_CONFIG_UNLOCK();
4861 goto fail;
4862 }
4863
4864 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
4865 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
4866 }
4867 pf_empty_kpool(&V_pf_pabuf[0]);
4868 pf_empty_kpool(&V_pf_pabuf[1]);
4869 pf_empty_kpool(&V_pf_pabuf[2]);
4870
4871 if (pcr->action == PF_CHANGE_ADD_HEAD)
4872 oldrule = TAILQ_FIRST(
4873 ruleset->rules[rs_num].active.ptr);
4874 else if (pcr->action == PF_CHANGE_ADD_TAIL)
4875 oldrule = TAILQ_LAST(
4876 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
4877 else {
4878 oldrule = TAILQ_FIRST(
4879 ruleset->rules[rs_num].active.ptr);
4880 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
4881 oldrule = TAILQ_NEXT(oldrule, entries);
4882 if (oldrule == NULL) {
4883 if (newrule != NULL)
4884 pf_free_rule(newrule);
4885 PF_RULES_WUNLOCK();
4886 PF_CONFIG_UNLOCK();
4887 error = EINVAL;
4888 goto fail;
4889 }
4890 }
4891
4892 if (pcr->action == PF_CHANGE_REMOVE) {
4893 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
4894 oldrule);
4895 RB_REMOVE(pf_krule_global,
4896 ruleset->rules[rs_num].active.tree, oldrule);
4897 ruleset->rules[rs_num].active.rcount--;
4898 } else {
4899 pf_hash_rule(newrule);
4900 if (RB_INSERT(pf_krule_global,
4901 ruleset->rules[rs_num].active.tree, newrule) != NULL) {
4902 pf_free_rule(newrule);
4903 PF_RULES_WUNLOCK();
4904 PF_CONFIG_UNLOCK();
4905 error = EEXIST;
4906 goto fail;
4907 }
4908
4909 if (oldrule == NULL)
4910 TAILQ_INSERT_TAIL(
4911 ruleset->rules[rs_num].active.ptr,
4912 newrule, entries);
4913 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
4914 pcr->action == PF_CHANGE_ADD_BEFORE)
4915 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
4916 else
4917 TAILQ_INSERT_AFTER(
4918 ruleset->rules[rs_num].active.ptr,
4919 oldrule, newrule, entries);
4920 ruleset->rules[rs_num].active.rcount++;
4921 }
4922
4923 nr = 0;
4924 TAILQ_FOREACH(oldrule,
4925 ruleset->rules[rs_num].active.ptr, entries)
4926 oldrule->nr = nr++;
4927
4928 ruleset->rules[rs_num].active.ticket++;
4929
4930 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
4931 pf_remove_if_empty_kruleset(ruleset);
4932
4933 PF_RULES_WUNLOCK();
4934 PF_CONFIG_UNLOCK();
4935 break;
4936
4937 #undef ERROUT
4938 DIOCCHANGERULE_error:
4939 PF_RULES_WUNLOCK();
4940 PF_CONFIG_UNLOCK();
4941 pf_krule_free(newrule);
4942 pf_kkif_free(kif);
4943 break;
4944 }
4945
4946 case DIOCCLRSTATESNV: {
4947 error = pf_clearstates_nv((struct pfioc_nv *)addr);
4948 break;
4949 }
4950
4951 case DIOCKILLSTATESNV: {
4952 error = pf_killstates_nv((struct pfioc_nv *)addr);
4953 break;
4954 }
4955
4956 case DIOCADDSTATE: {
4957 struct pfioc_state *ps = (struct pfioc_state *)addr;
4958 struct pfsync_state_1301 *sp = &ps->state;
4959
4960 if (sp->timeout >= PFTM_MAX) {
4961 error = EINVAL;
4962 goto fail;
4963 }
4964 if (V_pfsync_state_import_ptr != NULL) {
4965 PF_RULES_RLOCK();
4966 error = V_pfsync_state_import_ptr(
4967 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
4968 PFSYNC_MSG_VERSION_1301);
4969 PF_RULES_RUNLOCK();
4970 } else
4971 error = EOPNOTSUPP;
4972 break;
4973 }
4974
4975 case DIOCGETSTATE: {
4976 struct pfioc_state *ps = (struct pfioc_state *)addr;
4977 struct pf_kstate *s;
4978
4979 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
4980 if (s == NULL) {
4981 error = ENOENT;
4982 goto fail;
4983 }
4984
4985 pfsync_state_export_1301(&ps->state, s);
4986 PF_STATE_UNLOCK(s);
4987 break;
4988 }
4989
4990 case DIOCGETSTATENV: {
4991 error = pf_getstate((struct pfioc_nv *)addr);
4992 break;
4993 }
4994
4995 #ifdef COMPAT_FREEBSD14
4996 case DIOCGETSTATES: {
4997 struct pfioc_states *ps = (struct pfioc_states *)addr;
4998 struct pf_kstate *s;
4999 struct pfsync_state_1301 *pstore, *p;
5000 int i, nr;
5001 size_t slice_count = 16, count;
5002 void *out;
5003
5004 if (ps->ps_len <= 0) {
5005 nr = uma_zone_get_cur(V_pf_state_z);
5006 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
5007 break;
5008 }
5009
5010 out = ps->ps_states;
5011 pstore = mallocarray(slice_count,
5012 sizeof(struct pfsync_state_1301), M_PF, M_WAITOK | M_ZERO);
5013 nr = 0;
5014
5015 for (i = 0; i <= V_pf_hashmask; i++) {
5016 struct pf_idhash *ih = &V_pf_idhash[i];
5017
5018 DIOCGETSTATES_retry:
5019 p = pstore;
5020
5021 if (LIST_EMPTY(&ih->states))
5022 continue;
5023
5024 PF_HASHROW_LOCK(ih);
5025 count = 0;
5026 LIST_FOREACH(s, &ih->states, entry) {
5027 if (s->timeout == PFTM_UNLINKED)
5028 continue;
5029 count++;
5030 }
5031
5032 if (count > slice_count) {
5033 PF_HASHROW_UNLOCK(ih);
5034 free(pstore, M_PF);
5035 slice_count = count * 2;
5036 pstore = mallocarray(slice_count,
5037 sizeof(struct pfsync_state_1301), M_PF,
5038 M_WAITOK | M_ZERO);
5039 goto DIOCGETSTATES_retry;
5040 }
5041
5042 if ((nr+count) * sizeof(*p) > ps->ps_len) {
5043 PF_HASHROW_UNLOCK(ih);
5044 goto DIOCGETSTATES_full;
5045 }
5046
5047 LIST_FOREACH(s, &ih->states, entry) {
5048 if (s->timeout == PFTM_UNLINKED)
5049 continue;
5050
5051 pfsync_state_export_1301(p, s);
5052 p++;
5053 nr++;
5054 }
5055 PF_HASHROW_UNLOCK(ih);
5056 error = copyout(pstore, out,
5057 sizeof(struct pfsync_state_1301) * count);
5058 if (error) {
5059 free(pstore, M_PF);
5060 goto fail;
5061 }
5062 out = ps->ps_states + nr;
5063 }
5064 DIOCGETSTATES_full:
5065 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
5066 free(pstore, M_PF);
5067
5068 break;
5069 }
5070
5071 case DIOCGETSTATESV2: {
5072 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
5073 struct pf_kstate *s;
5074 struct pf_state_export *pstore, *p;
5075 int i, nr;
5076 size_t slice_count = 16, count;
5077 void *out;
5078
5079 if (ps->ps_req_version > PF_STATE_VERSION) {
5080 error = ENOTSUP;
5081 goto fail;
5082 }
5083
5084 if (ps->ps_len <= 0) {
5085 nr = uma_zone_get_cur(V_pf_state_z);
5086 ps->ps_len = sizeof(struct pf_state_export) * nr;
5087 break;
5088 }
5089
5090 out = ps->ps_states;
5091 pstore = mallocarray(slice_count,
5092 sizeof(struct pf_state_export), M_PF, M_WAITOK | M_ZERO);
5093 nr = 0;
5094
5095 for (i = 0; i <= V_pf_hashmask; i++) {
5096 struct pf_idhash *ih = &V_pf_idhash[i];
5097
5098 DIOCGETSTATESV2_retry:
5099 p = pstore;
5100
5101 if (LIST_EMPTY(&ih->states))
5102 continue;
5103
5104 PF_HASHROW_LOCK(ih);
5105 count = 0;
5106 LIST_FOREACH(s, &ih->states, entry) {
5107 if (s->timeout == PFTM_UNLINKED)
5108 continue;
5109 count++;
5110 }
5111
5112 if (count > slice_count) {
5113 PF_HASHROW_UNLOCK(ih);
5114 free(pstore, M_PF);
5115 slice_count = count * 2;
5116 pstore = mallocarray(slice_count,
5117 sizeof(struct pf_state_export), M_PF,
5118 M_WAITOK | M_ZERO);
5119 goto DIOCGETSTATESV2_retry;
5120 }
5121
5122 if ((nr+count) * sizeof(*p) > ps->ps_len) {
5123 PF_HASHROW_UNLOCK(ih);
5124 goto DIOCGETSTATESV2_full;
5125 }
5126
5127 LIST_FOREACH(s, &ih->states, entry) {
5128 if (s->timeout == PFTM_UNLINKED)
5129 continue;
5130
5131 pf_state_export(p, s);
5132 p++;
5133 nr++;
5134 }
5135 PF_HASHROW_UNLOCK(ih);
5136 error = copyout(pstore, out,
5137 sizeof(struct pf_state_export) * count);
5138 if (error) {
5139 free(pstore, M_PF);
5140 goto fail;
5141 }
5142 out = ps->ps_states + nr;
5143 }
5144 DIOCGETSTATESV2_full:
5145 ps->ps_len = nr * sizeof(struct pf_state_export);
5146 free(pstore, M_PF);
5147
5148 break;
5149 }
5150 #endif
5151 case DIOCGETSTATUSNV: {
5152 error = pf_getstatus((struct pfioc_nv *)addr);
5153 break;
5154 }
5155
5156 case DIOCSETSTATUSIF: {
5157 struct pfioc_if *pi = (struct pfioc_if *)addr;
5158
5159 if (pi->ifname[0] == 0) {
5160 bzero(V_pf_status.ifname, IFNAMSIZ);
5161 break;
5162 }
5163 PF_RULES_WLOCK();
5164 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
5165 PF_RULES_WUNLOCK();
5166 break;
5167 }
5168
5169 case DIOCCLRSTATUS: {
5170 pf_ioctl_clear_status();
5171 break;
5172 }
5173
5174 case DIOCNATLOOK: {
5175 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
5176
5177 error = pf_ioctl_natlook(pnl);
5178 break;
5179 }
5180
5181 case DIOCSETTIMEOUT: {
5182 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
5183
5184 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
5185 &pt->seconds);
5186 break;
5187 }
5188
5189 case DIOCGETTIMEOUT: {
5190 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
5191
5192 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
5193 break;
5194 }
5195
5196 case DIOCGETLIMIT: {
5197 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
5198
5199 error = pf_ioctl_get_limit(pl->index, &pl->limit);
5200 break;
5201 }
5202
5203 case DIOCSETLIMIT: {
5204 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
5205 unsigned int old_limit;
5206
5207 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
5208 pl->limit = old_limit;
5209 break;
5210 }
5211
5212 case DIOCSETDEBUG: {
5213 u_int32_t *level = (u_int32_t *)addr;
5214
5215 PF_RULES_WLOCK();
5216 V_pf_status.debug = *level;
5217 PF_RULES_WUNLOCK();
5218 break;
5219 }
5220
5221 case DIOCCLRRULECTRS: {
5222 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
5223 struct pf_kruleset *ruleset = &pf_main_ruleset;
5224 struct pf_krule *rule;
5225
5226 PF_RULES_WLOCK();
5227 TAILQ_FOREACH(rule,
5228 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
5229 pf_counter_u64_zero(&rule->evaluations);
5230 for (int i = 0; i < 2; i++) {
5231 pf_counter_u64_zero(&rule->packets[i]);
5232 pf_counter_u64_zero(&rule->bytes[i]);
5233 }
5234 }
5235 PF_RULES_WUNLOCK();
5236 break;
5237 }
5238
5239 case DIOCGIFSPEEDV0:
5240 case DIOCGIFSPEEDV1: {
5241 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
5242 struct pf_ifspeed_v1 ps;
5243 struct ifnet *ifp;
5244
5245 if (psp->ifname[0] == '\0') {
5246 error = EINVAL;
5247 goto fail;
5248 }
5249
5250 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
5251 if (error != 0)
5252 goto fail;
5253 ifp = ifunit(ps.ifname);
5254 if (ifp != NULL) {
5255 psp->baudrate32 =
5256 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
5257 if (cmd == DIOCGIFSPEEDV1)
5258 psp->baudrate = ifp->if_baudrate;
5259 } else {
5260 error = EINVAL;
5261 }
5262 break;
5263 }
5264
5265 #ifdef ALTQ
5266 case DIOCSTARTALTQ: {
5267 struct pf_altq *altq;
5268
5269 PF_RULES_WLOCK();
5270 /* enable all altq interfaces on active list */
5271 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
5272 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
5273 error = pf_enable_altq(altq);
5274 if (error != 0)
5275 break;
5276 }
5277 }
5278 if (error == 0)
5279 V_pf_altq_running = 1;
5280 PF_RULES_WUNLOCK();
5281 DPFPRINTF(PF_DEBUG_MISC, "altq: started");
5282 break;
5283 }
5284
5285 case DIOCSTOPALTQ: {
5286 struct pf_altq *altq;
5287
5288 PF_RULES_WLOCK();
5289 /* disable all altq interfaces on active list */
5290 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
5291 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
5292 error = pf_disable_altq(altq);
5293 if (error != 0)
5294 break;
5295 }
5296 }
5297 if (error == 0)
5298 V_pf_altq_running = 0;
5299 PF_RULES_WUNLOCK();
5300 DPFPRINTF(PF_DEBUG_MISC, "altq: stopped");
5301 break;
5302 }
5303
5304 case DIOCADDALTQV0:
5305 case DIOCADDALTQV1: {
5306 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
5307 struct pf_altq *altq, *a;
5308 struct ifnet *ifp;
5309
5310 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
5311 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
5312 if (error)
5313 goto fail;
5314 altq->local_flags = 0;
5315
5316 PF_RULES_WLOCK();
5317 if (pa->ticket != V_ticket_altqs_inactive) {
5318 PF_RULES_WUNLOCK();
5319 free(altq, M_PFALTQ);
5320 error = EBUSY;
5321 goto fail;
5322 }
5323
5324 /*
5325 * if this is for a queue, find the discipline and
5326 * copy the necessary fields
5327 */
5328 if (altq->qname[0] != 0) {
5329 if ((altq->qid = pf_qname2qid(altq->qname, true)) == 0) {
5330 PF_RULES_WUNLOCK();
5331 error = EBUSY;
5332 free(altq, M_PFALTQ);
5333 goto fail;
5334 }
5335 altq->altq_disc = NULL;
5336 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
5337 if (strncmp(a->ifname, altq->ifname,
5338 IFNAMSIZ) == 0) {
5339 altq->altq_disc = a->altq_disc;
5340 break;
5341 }
5342 }
5343 }
5344
5345 if ((ifp = ifunit(altq->ifname)) == NULL)
5346 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
5347 else
5348 error = altq_add(ifp, altq);
5349
5350 if (error) {
5351 PF_RULES_WUNLOCK();
5352 free(altq, M_PFALTQ);
5353 goto fail;
5354 }
5355
5356 if (altq->qname[0] != 0)
5357 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
5358 else
5359 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
5360 /* version error check done on import above */
5361 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
5362 PF_RULES_WUNLOCK();
5363 break;
5364 }
5365
5366 case DIOCGETALTQSV0:
5367 case DIOCGETALTQSV1: {
5368 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
5369 struct pf_altq *altq;
5370
5371 PF_RULES_RLOCK();
5372 pa->nr = 0;
5373 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
5374 pa->nr++;
5375 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
5376 pa->nr++;
5377 pa->ticket = V_ticket_altqs_active;
5378 PF_RULES_RUNLOCK();
5379 break;
5380 }
5381
5382 case DIOCGETALTQV0:
5383 case DIOCGETALTQV1: {
5384 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
5385 struct pf_altq *altq;
5386
5387 PF_RULES_RLOCK();
5388 if (pa->ticket != V_ticket_altqs_active) {
5389 PF_RULES_RUNLOCK();
5390 error = EBUSY;
5391 goto fail;
5392 }
5393 altq = pf_altq_get_nth_active(pa->nr);
5394 if (altq == NULL) {
5395 PF_RULES_RUNLOCK();
5396 error = EBUSY;
5397 goto fail;
5398 }
5399 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
5400 PF_RULES_RUNLOCK();
5401 break;
5402 }
5403
5404 case DIOCCHANGEALTQV0:
5405 case DIOCCHANGEALTQV1:
5406 /* CHANGEALTQ not supported yet! */
5407 error = ENODEV;
5408 break;
5409
5410 case DIOCGETQSTATSV0:
5411 case DIOCGETQSTATSV1: {
5412 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
5413 struct pf_altq *altq;
5414 int nbytes;
5415 u_int32_t version;
5416
5417 PF_RULES_RLOCK();
5418 if (pq->ticket != V_ticket_altqs_active) {
5419 PF_RULES_RUNLOCK();
5420 error = EBUSY;
5421 goto fail;
5422 }
5423 nbytes = pq->nbytes;
5424 altq = pf_altq_get_nth_active(pq->nr);
5425 if (altq == NULL) {
5426 PF_RULES_RUNLOCK();
5427 error = EBUSY;
5428 goto fail;
5429 }
5430
5431 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
5432 PF_RULES_RUNLOCK();
5433 error = ENXIO;
5434 goto fail;
5435 }
5436 PF_RULES_RUNLOCK();
5437 if (cmd == DIOCGETQSTATSV0)
5438 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
5439 else
5440 version = pq->version;
5441 error = altq_getqstats(altq, pq->buf, &nbytes, version);
5442 if (error == 0) {
5443 pq->scheduler = altq->scheduler;
5444 pq->nbytes = nbytes;
5445 }
5446 break;
5447 }
5448 #endif /* ALTQ */
5449
5450 case DIOCBEGINADDRS: {
5451 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
5452
5453 error = pf_ioctl_begin_addrs(&pp->ticket);
5454 break;
5455 }
5456
5457 case DIOCADDADDR: {
5458 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
5459 struct pf_nl_pooladdr npp = {};
5460
5461 npp.which = PF_RDR;
5462 memcpy(&npp, pp, sizeof(*pp));
5463 error = pf_ioctl_add_addr(&npp);
5464 break;
5465 }
5466
5467 case DIOCGETADDRS: {
5468 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
5469 struct pf_nl_pooladdr npp = {};
5470
5471 npp.which = PF_RDR;
5472 memcpy(&npp, pp, sizeof(*pp));
5473 error = pf_ioctl_get_addrs(&npp);
5474 memcpy(pp, &npp, sizeof(*pp));
5475
5476 break;
5477 }
5478
5479 case DIOCGETADDR: {
5480 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
5481 struct pf_nl_pooladdr npp = {};
5482
5483 npp.which = PF_RDR;
5484 memcpy(&npp, pp, sizeof(*pp));
5485 error = pf_ioctl_get_addr(&npp);
5486 memcpy(pp, &npp, sizeof(*pp));
5487
5488 break;
5489 }
5490
5491 case DIOCCHANGEADDR: {
5492 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
5493 struct pf_kpool *pool;
5494 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
5495 struct pf_kruleset *ruleset;
5496 struct pfi_kkif *kif = NULL;
5497
5498 pca->anchor[sizeof(pca->anchor) - 1] = '\0';
5499
5500 if (pca->action < PF_CHANGE_ADD_HEAD ||
5501 pca->action > PF_CHANGE_REMOVE) {
5502 error = EINVAL;
5503 goto fail;
5504 }
5505 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
5506 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
5507 pca->addr.addr.type != PF_ADDR_TABLE) {
5508 error = EINVAL;
5509 goto fail;
5510 }
5511 if (pca->addr.addr.p.dyn != NULL) {
5512 error = EINVAL;
5513 goto fail;
5514 }
5515
5516 if (pca->action != PF_CHANGE_REMOVE) {
5517 #ifndef INET
5518 if (pca->af == AF_INET) {
5519 error = EAFNOSUPPORT;
5520 goto fail;
5521 }
5522 #endif /* INET */
5523 #ifndef INET6
5524 if (pca->af == AF_INET6) {
5525 error = EAFNOSUPPORT;
5526 goto fail;
5527 }
5528 #endif /* INET6 */
5529 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
5530 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
5531 if (newpa->ifname[0])
5532 kif = pf_kkif_create(M_WAITOK);
5533 newpa->kif = NULL;
5534 }
5535 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
5536 PF_RULES_WLOCK();
5537 ruleset = pf_find_kruleset(pca->anchor);
5538 if (ruleset == NULL)
5539 ERROUT(EBUSY);
5540
5541 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
5542 pca->r_num, pca->r_last, 1, 1, PF_RDR);
5543 if (pool == NULL)
5544 ERROUT(EBUSY);
5545
5546 if (pca->action != PF_CHANGE_REMOVE) {
5547 if (newpa->ifname[0]) {
5548 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
5549 pfi_kkif_ref(newpa->kif);
5550 kif = NULL;
5551 }
5552
5553 switch (newpa->addr.type) {
5554 case PF_ADDR_DYNIFTL:
5555 error = pfi_dynaddr_setup(&newpa->addr,
5556 pca->af);
5557 break;
5558 case PF_ADDR_TABLE:
5559 newpa->addr.p.tbl = pfr_attach_table(ruleset,
5560 newpa->addr.v.tblname);
5561 if (newpa->addr.p.tbl == NULL)
5562 error = ENOMEM;
5563 break;
5564 }
5565 if (error)
5566 goto DIOCCHANGEADDR_error;
5567 }
5568
5569 switch (pca->action) {
5570 case PF_CHANGE_ADD_HEAD:
5571 oldpa = TAILQ_FIRST(&pool->list);
5572 break;
5573 case PF_CHANGE_ADD_TAIL:
5574 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
5575 break;
5576 default:
5577 oldpa = TAILQ_FIRST(&pool->list);
5578 for (int i = 0; oldpa && i < pca->nr; i++)
5579 oldpa = TAILQ_NEXT(oldpa, entries);
5580
5581 if (oldpa == NULL)
5582 ERROUT(EINVAL);
5583 }
5584
5585 if (pca->action == PF_CHANGE_REMOVE) {
5586 TAILQ_REMOVE(&pool->list, oldpa, entries);
5587 switch (oldpa->addr.type) {
5588 case PF_ADDR_DYNIFTL:
5589 pfi_dynaddr_remove(oldpa->addr.p.dyn);
5590 break;
5591 case PF_ADDR_TABLE:
5592 pfr_detach_table(oldpa->addr.p.tbl);
5593 break;
5594 }
5595 if (oldpa->kif)
5596 pfi_kkif_unref(oldpa->kif);
5597 free(oldpa, M_PFRULE);
5598 } else {
5599 if (oldpa == NULL)
5600 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
5601 else if (pca->action == PF_CHANGE_ADD_HEAD ||
5602 pca->action == PF_CHANGE_ADD_BEFORE)
5603 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
5604 else
5605 TAILQ_INSERT_AFTER(&pool->list, oldpa,
5606 newpa, entries);
5607 }
5608
5609 pool->cur = TAILQ_FIRST(&pool->list);
5610 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
5611 PF_RULES_WUNLOCK();
5612 break;
5613
5614 #undef ERROUT
5615 DIOCCHANGEADDR_error:
5616 if (newpa != NULL) {
5617 if (newpa->kif)
5618 pfi_kkif_unref(newpa->kif);
5619 free(newpa, M_PFRULE);
5620 }
5621 PF_RULES_WUNLOCK();
5622 pf_kkif_free(kif);
5623 break;
5624 }
5625
5626 case DIOCGETRULESETS: {
5627 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
5628
5629 pr->path[sizeof(pr->path) - 1] = '\0';
5630
5631 error = pf_ioctl_get_rulesets(pr);
5632 break;
5633 }
5634
5635 case DIOCGETRULESET: {
5636 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
5637
5638 pr->path[sizeof(pr->path) - 1] = '\0';
5639
5640 error = pf_ioctl_get_ruleset(pr);
5641 break;
5642 }
5643
5644 case DIOCRCLRTABLES: {
5645 struct pfioc_table *io = (struct pfioc_table *)addr;
5646
5647 if (io->pfrio_esize != 0) {
5648 error = ENODEV;
5649 goto fail;
5650 }
5651 if (strnlen(io->pfrio_table.pfrt_anchor, MAXPATHLEN)
5652 == MAXPATHLEN) {
5653 error = EINVAL;
5654 goto fail;
5655 }
5656 if (strnlen(io->pfrio_table.pfrt_name, PF_TABLE_NAME_SIZE)
5657 == PF_TABLE_NAME_SIZE) {
5658 error = EINVAL;
5659 goto fail;
5660 }
5661
5662 PF_RULES_WLOCK();
5663 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
5664 io->pfrio_flags | PFR_FLAG_USERIOCTL);
5665 PF_RULES_WUNLOCK();
5666 break;
5667 }
5668
5669 case DIOCRADDTABLES: {
5670 struct pfioc_table *io = (struct pfioc_table *)addr;
5671 struct pfr_table *pfrts;
5672 size_t totlen;
5673
5674 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5675 error = ENODEV;
5676 goto fail;
5677 }
5678
5679 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
5680 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
5681 error = ENOMEM;
5682 goto fail;
5683 }
5684
5685 totlen = io->pfrio_size * sizeof(struct pfr_table);
5686 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5687 M_PF, M_WAITOK);
5688 error = copyin(io->pfrio_buffer, pfrts, totlen);
5689 if (error) {
5690 free(pfrts, M_PF);
5691 goto fail;
5692 }
5693 PF_RULES_WLOCK();
5694 error = pfr_add_tables(pfrts, io->pfrio_size,
5695 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5696 PF_RULES_WUNLOCK();
5697 free(pfrts, M_PF);
5698 break;
5699 }
5700
5701 case DIOCRDELTABLES: {
5702 struct pfioc_table *io = (struct pfioc_table *)addr;
5703 struct pfr_table *pfrts;
5704 size_t totlen;
5705
5706 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5707 error = ENODEV;
5708 goto fail;
5709 }
5710
5711 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
5712 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
5713 error = ENOMEM;
5714 goto fail;
5715 }
5716
5717 totlen = io->pfrio_size * sizeof(struct pfr_table);
5718 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5719 M_PF, M_WAITOK);
5720 error = copyin(io->pfrio_buffer, pfrts, totlen);
5721 if (error) {
5722 free(pfrts, M_PF);
5723 goto fail;
5724 }
5725 PF_RULES_WLOCK();
5726 error = pfr_del_tables(pfrts, io->pfrio_size,
5727 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5728 PF_RULES_WUNLOCK();
5729 free(pfrts, M_PF);
5730 break;
5731 }
5732
5733 case DIOCRGETTABLES: {
5734 struct pfioc_table *io = (struct pfioc_table *)addr;
5735 struct pfr_table *pfrts;
5736 size_t totlen;
5737 int n;
5738
5739 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5740 error = ENODEV;
5741 goto fail;
5742 }
5743 PF_RULES_RLOCK();
5744 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
5745 if (n < 0) {
5746 PF_RULES_RUNLOCK();
5747 error = EINVAL;
5748 goto fail;
5749 }
5750 io->pfrio_size = min(io->pfrio_size, n);
5751
5752 totlen = io->pfrio_size * sizeof(struct pfr_table);
5753
5754 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5755 M_PF, M_NOWAIT | M_ZERO);
5756 if (pfrts == NULL) {
5757 error = ENOMEM;
5758 PF_RULES_RUNLOCK();
5759 goto fail;
5760 }
5761 error = pfr_get_tables(&io->pfrio_table, pfrts,
5762 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5763 PF_RULES_RUNLOCK();
5764 if (error == 0)
5765 error = copyout(pfrts, io->pfrio_buffer, totlen);
5766 free(pfrts, M_PF);
5767 break;
5768 }
5769
5770 case DIOCRGETTSTATS: {
5771 struct pfioc_table *io = (struct pfioc_table *)addr;
5772 struct pfr_tstats *pfrtstats;
5773 size_t totlen;
5774 int n;
5775
5776 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
5777 error = ENODEV;
5778 goto fail;
5779 }
5780 PF_TABLE_STATS_LOCK();
5781 PF_RULES_RLOCK();
5782 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
5783 if (n < 0) {
5784 PF_RULES_RUNLOCK();
5785 PF_TABLE_STATS_UNLOCK();
5786 error = EINVAL;
5787 goto fail;
5788 }
5789 io->pfrio_size = min(io->pfrio_size, n);
5790
5791 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
5792 pfrtstats = mallocarray(io->pfrio_size,
5793 sizeof(struct pfr_tstats), M_PF, M_NOWAIT | M_ZERO);
5794 if (pfrtstats == NULL) {
5795 error = ENOMEM;
5796 PF_RULES_RUNLOCK();
5797 PF_TABLE_STATS_UNLOCK();
5798 goto fail;
5799 }
5800 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
5801 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5802 PF_RULES_RUNLOCK();
5803 PF_TABLE_STATS_UNLOCK();
5804 if (error == 0)
5805 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
5806 free(pfrtstats, M_PF);
5807 break;
5808 }
5809
5810 case DIOCRCLRTSTATS: {
5811 struct pfioc_table *io = (struct pfioc_table *)addr;
5812 struct pfr_table *pfrts;
5813 size_t totlen;
5814
5815 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5816 error = ENODEV;
5817 goto fail;
5818 }
5819
5820 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
5821 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
5822 /* We used to count tables and use the minimum required
5823 * size, so we didn't fail on overly large requests.
5824 * Keep doing so. */
5825 io->pfrio_size = pf_ioctl_maxcount;
5826 goto fail;
5827 }
5828
5829 totlen = io->pfrio_size * sizeof(struct pfr_table);
5830 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5831 M_PF, M_WAITOK);
5832 error = copyin(io->pfrio_buffer, pfrts, totlen);
5833 if (error) {
5834 free(pfrts, M_PF);
5835 goto fail;
5836 }
5837
5838 PF_TABLE_STATS_LOCK();
5839 PF_RULES_RLOCK();
5840 error = pfr_clr_tstats(pfrts, io->pfrio_size,
5841 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5842 PF_RULES_RUNLOCK();
5843 PF_TABLE_STATS_UNLOCK();
5844 free(pfrts, M_PF);
5845 break;
5846 }
5847
5848 case DIOCRSETTFLAGS: {
5849 struct pfioc_table *io = (struct pfioc_table *)addr;
5850 struct pfr_table *pfrts;
5851 size_t totlen;
5852 int n;
5853
5854 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5855 error = ENODEV;
5856 goto fail;
5857 }
5858
5859 PF_RULES_RLOCK();
5860 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
5861 if (n < 0) {
5862 PF_RULES_RUNLOCK();
5863 error = EINVAL;
5864 goto fail;
5865 }
5866
5867 io->pfrio_size = min(io->pfrio_size, n);
5868 PF_RULES_RUNLOCK();
5869
5870 totlen = io->pfrio_size * sizeof(struct pfr_table);
5871 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5872 M_PF, M_WAITOK);
5873 error = copyin(io->pfrio_buffer, pfrts, totlen);
5874 if (error) {
5875 free(pfrts, M_PF);
5876 goto fail;
5877 }
5878 PF_RULES_WLOCK();
5879 error = pfr_set_tflags(pfrts, io->pfrio_size,
5880 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
5881 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5882 PF_RULES_WUNLOCK();
5883 free(pfrts, M_PF);
5884 break;
5885 }
5886
5887 case DIOCRCLRADDRS: {
5888 struct pfioc_table *io = (struct pfioc_table *)addr;
5889
5890 if (io->pfrio_esize != 0) {
5891 error = ENODEV;
5892 goto fail;
5893 }
5894 PF_RULES_WLOCK();
5895 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
5896 io->pfrio_flags | PFR_FLAG_USERIOCTL);
5897 PF_RULES_WUNLOCK();
5898 break;
5899 }
5900
5901 case DIOCRADDADDRS: {
5902 struct pfioc_table *io = (struct pfioc_table *)addr;
5903 struct pfr_addr *pfras;
5904 size_t totlen;
5905
5906 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5907 error = ENODEV;
5908 goto fail;
5909 }
5910 if (io->pfrio_size < 0 ||
5911 io->pfrio_size > pf_ioctl_maxcount ||
5912 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5913 error = EINVAL;
5914 goto fail;
5915 }
5916 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5917 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5918 M_PF, M_WAITOK);
5919 error = copyin(io->pfrio_buffer, pfras, totlen);
5920 if (error) {
5921 free(pfras, M_PF);
5922 goto fail;
5923 }
5924 PF_RULES_WLOCK();
5925 io->pfrio_nadd = 0;
5926 error = pfr_add_addrs(&io->pfrio_table, pfras,
5927 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
5928 PFR_FLAG_USERIOCTL);
5929 PF_RULES_WUNLOCK();
5930 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5931 error = copyout(pfras, io->pfrio_buffer, totlen);
5932 free(pfras, M_PF);
5933 break;
5934 }
5935
5936 case DIOCRDELADDRS: {
5937 struct pfioc_table *io = (struct pfioc_table *)addr;
5938 struct pfr_addr *pfras;
5939 size_t totlen;
5940
5941 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5942 error = ENODEV;
5943 goto fail;
5944 }
5945 if (io->pfrio_size < 0 ||
5946 io->pfrio_size > pf_ioctl_maxcount ||
5947 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5948 error = EINVAL;
5949 goto fail;
5950 }
5951 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5952 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5953 M_PF, M_WAITOK);
5954 error = copyin(io->pfrio_buffer, pfras, totlen);
5955 if (error) {
5956 free(pfras, M_PF);
5957 goto fail;
5958 }
5959 PF_RULES_WLOCK();
5960 error = pfr_del_addrs(&io->pfrio_table, pfras,
5961 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
5962 PFR_FLAG_USERIOCTL);
5963 PF_RULES_WUNLOCK();
5964 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5965 error = copyout(pfras, io->pfrio_buffer, totlen);
5966 free(pfras, M_PF);
5967 break;
5968 }
5969
5970 case DIOCRSETADDRS: {
5971 struct pfioc_table *io = (struct pfioc_table *)addr;
5972 struct pfr_addr *pfras;
5973 size_t totlen, count;
5974
5975 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5976 error = ENODEV;
5977 goto fail;
5978 }
5979 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
5980 error = EINVAL;
5981 goto fail;
5982 }
5983 count = max(io->pfrio_size, io->pfrio_size2);
5984 if (count > pf_ioctl_maxcount ||
5985 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
5986 error = EINVAL;
5987 goto fail;
5988 }
5989 totlen = count * sizeof(struct pfr_addr);
5990 pfras = mallocarray(count, sizeof(struct pfr_addr), M_PF,
5991 M_WAITOK);
5992 error = copyin(io->pfrio_buffer, pfras, totlen);
5993 if (error) {
5994 free(pfras, M_PF);
5995 goto fail;
5996 }
5997 PF_RULES_WLOCK();
5998 error = pfr_set_addrs(&io->pfrio_table, pfras,
5999 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
6000 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
6001 PFR_FLAG_START | PFR_FLAG_DONE | PFR_FLAG_USERIOCTL, 0);
6002 PF_RULES_WUNLOCK();
6003 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
6004 error = copyout(pfras, io->pfrio_buffer, totlen);
6005 free(pfras, M_PF);
6006 break;
6007 }
6008
6009 case DIOCRGETADDRS: {
6010 struct pfioc_table *io = (struct pfioc_table *)addr;
6011 struct pfr_addr *pfras;
6012 size_t totlen;
6013
6014 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
6015 error = ENODEV;
6016 goto fail;
6017 }
6018 if (io->pfrio_size < 0 ||
6019 io->pfrio_size > pf_ioctl_maxcount ||
6020 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
6021 error = EINVAL;
6022 goto fail;
6023 }
6024 totlen = io->pfrio_size * sizeof(struct pfr_addr);
6025 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
6026 M_PF, M_WAITOK | M_ZERO);
6027 PF_RULES_RLOCK();
6028 error = pfr_get_addrs(&io->pfrio_table, pfras,
6029 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
6030 PF_RULES_RUNLOCK();
6031 if (error == 0)
6032 error = copyout(pfras, io->pfrio_buffer, totlen);
6033 free(pfras, M_PF);
6034 break;
6035 }
6036
6037 case DIOCRGETASTATS: {
6038 struct pfioc_table *io = (struct pfioc_table *)addr;
6039 struct pfr_astats *pfrastats;
6040 size_t totlen;
6041
6042 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
6043 error = ENODEV;
6044 goto fail;
6045 }
6046 if (io->pfrio_size < 0 ||
6047 io->pfrio_size > pf_ioctl_maxcount ||
6048 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
6049 error = EINVAL;
6050 goto fail;
6051 }
6052 totlen = io->pfrio_size * sizeof(struct pfr_astats);
6053 pfrastats = mallocarray(io->pfrio_size,
6054 sizeof(struct pfr_astats), M_PF, M_WAITOK | M_ZERO);
6055 PF_RULES_RLOCK();
6056 error = pfr_get_astats(&io->pfrio_table, pfrastats,
6057 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
6058 PF_RULES_RUNLOCK();
6059 if (error == 0)
6060 error = copyout(pfrastats, io->pfrio_buffer, totlen);
6061 free(pfrastats, M_PF);
6062 break;
6063 }
6064
6065 case DIOCRCLRASTATS: {
6066 struct pfioc_table *io = (struct pfioc_table *)addr;
6067 struct pfr_addr *pfras;
6068 size_t totlen;
6069
6070 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
6071 error = ENODEV;
6072 goto fail;
6073 }
6074 if (io->pfrio_size < 0 ||
6075 io->pfrio_size > pf_ioctl_maxcount ||
6076 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
6077 error = EINVAL;
6078 goto fail;
6079 }
6080 totlen = io->pfrio_size * sizeof(struct pfr_addr);
6081 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
6082 M_PF, M_WAITOK);
6083 error = copyin(io->pfrio_buffer, pfras, totlen);
6084 if (error) {
6085 free(pfras, M_PF);
6086 goto fail;
6087 }
6088 PF_RULES_WLOCK();
6089 error = pfr_clr_astats(&io->pfrio_table, pfras,
6090 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
6091 PFR_FLAG_USERIOCTL);
6092 PF_RULES_WUNLOCK();
6093 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
6094 error = copyout(pfras, io->pfrio_buffer, totlen);
6095 free(pfras, M_PF);
6096 break;
6097 }
6098
6099 case DIOCRTSTADDRS: {
6100 struct pfioc_table *io = (struct pfioc_table *)addr;
6101 struct pfr_addr *pfras;
6102 size_t totlen;
6103
6104 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
6105 error = ENODEV;
6106 goto fail;
6107 }
6108 if (io->pfrio_size < 0 ||
6109 io->pfrio_size > pf_ioctl_maxcount ||
6110 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
6111 error = EINVAL;
6112 goto fail;
6113 }
6114 totlen = io->pfrio_size * sizeof(struct pfr_addr);
6115 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
6116 M_PF, M_WAITOK);
6117 error = copyin(io->pfrio_buffer, pfras, totlen);
6118 if (error) {
6119 free(pfras, M_PF);
6120 goto fail;
6121 }
6122 PF_RULES_RLOCK();
6123 error = pfr_tst_addrs(&io->pfrio_table, pfras,
6124 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
6125 PFR_FLAG_USERIOCTL);
6126 PF_RULES_RUNLOCK();
6127 if (error == 0)
6128 error = copyout(pfras, io->pfrio_buffer, totlen);
6129 free(pfras, M_PF);
6130 break;
6131 }
6132
6133 case DIOCRINADEFINE: {
6134 struct pfioc_table *io = (struct pfioc_table *)addr;
6135 struct pfr_addr *pfras;
6136 size_t totlen;
6137
6138 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
6139 error = ENODEV;
6140 goto fail;
6141 }
6142 if (io->pfrio_size < 0 ||
6143 io->pfrio_size > pf_ioctl_maxcount ||
6144 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
6145 error = EINVAL;
6146 goto fail;
6147 }
6148 totlen = io->pfrio_size * sizeof(struct pfr_addr);
6149 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
6150 M_PF, M_WAITOK);
6151 error = copyin(io->pfrio_buffer, pfras, totlen);
6152 if (error) {
6153 free(pfras, M_PF);
6154 goto fail;
6155 }
6156 PF_RULES_WLOCK();
6157 error = pfr_ina_define(&io->pfrio_table, pfras,
6158 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
6159 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
6160 PF_RULES_WUNLOCK();
6161 free(pfras, M_PF);
6162 break;
6163 }
6164
6165 case DIOCOSFPADD: {
6166 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
6167 PF_RULES_WLOCK();
6168 error = pf_osfp_add(io);
6169 PF_RULES_WUNLOCK();
6170 break;
6171 }
6172
6173 case DIOCOSFPGET: {
6174 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
6175 PF_RULES_RLOCK();
6176 error = pf_osfp_get(io);
6177 PF_RULES_RUNLOCK();
6178 break;
6179 }
6180
6181 case DIOCXBEGIN: {
6182 struct pfioc_trans *io = (struct pfioc_trans *)addr;
6183 struct pfioc_trans_e *ioes, *ioe;
6184 size_t totlen;
6185 int i;
6186
6187 if (io->esize != sizeof(*ioe)) {
6188 error = ENODEV;
6189 goto fail;
6190 }
6191 if (io->size < 0 ||
6192 io->size > pf_ioctl_maxcount ||
6193 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
6194 error = EINVAL;
6195 goto fail;
6196 }
6197 totlen = sizeof(struct pfioc_trans_e) * io->size;
6198 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
6199 M_PF, M_WAITOK);
6200 error = copyin(io->array, ioes, totlen);
6201 if (error) {
6202 free(ioes, M_PF);
6203 goto fail;
6204 }
6205 PF_RULES_WLOCK();
6206 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
6207 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
6208 switch (ioe->rs_num) {
6209 case PF_RULESET_ETH:
6210 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
6211 PF_RULES_WUNLOCK();
6212 free(ioes, M_PF);
6213 goto fail;
6214 }
6215 break;
6216 #ifdef ALTQ
6217 case PF_RULESET_ALTQ:
6218 if (ioe->anchor[0]) {
6219 PF_RULES_WUNLOCK();
6220 free(ioes, M_PF);
6221 error = EINVAL;
6222 goto fail;
6223 }
6224 if ((error = pf_begin_altq(&ioe->ticket))) {
6225 PF_RULES_WUNLOCK();
6226 free(ioes, M_PF);
6227 goto fail;
6228 }
6229 break;
6230 #endif /* ALTQ */
6231 case PF_RULESET_TABLE:
6232 {
6233 struct pfr_table table;
6234
6235 bzero(&table, sizeof(table));
6236 strlcpy(table.pfrt_anchor, ioe->anchor,
6237 sizeof(table.pfrt_anchor));
6238 if ((error = pfr_ina_begin(&table,
6239 &ioe->ticket, NULL, 0))) {
6240 PF_RULES_WUNLOCK();
6241 free(ioes, M_PF);
6242 goto fail;
6243 }
6244 break;
6245 }
6246 default:
6247 if ((error = pf_begin_rules(&ioe->ticket,
6248 ioe->rs_num, ioe->anchor))) {
6249 PF_RULES_WUNLOCK();
6250 free(ioes, M_PF);
6251 goto fail;
6252 }
6253 break;
6254 }
6255 }
6256 PF_RULES_WUNLOCK();
6257 error = copyout(ioes, io->array, totlen);
6258 free(ioes, M_PF);
6259 break;
6260 }
6261
6262 case DIOCXROLLBACK: {
6263 struct pfioc_trans *io = (struct pfioc_trans *)addr;
6264 struct pfioc_trans_e *ioe, *ioes;
6265 size_t totlen;
6266 int i;
6267
6268 if (io->esize != sizeof(*ioe)) {
6269 error = ENODEV;
6270 goto fail;
6271 }
6272 if (io->size < 0 ||
6273 io->size > pf_ioctl_maxcount ||
6274 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
6275 error = EINVAL;
6276 goto fail;
6277 }
6278 totlen = sizeof(struct pfioc_trans_e) * io->size;
6279 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
6280 M_PF, M_WAITOK);
6281 error = copyin(io->array, ioes, totlen);
6282 if (error) {
6283 free(ioes, M_PF);
6284 goto fail;
6285 }
6286 PF_RULES_WLOCK();
6287 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
6288 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
6289 switch (ioe->rs_num) {
6290 case PF_RULESET_ETH:
6291 if ((error = pf_rollback_eth(ioe->ticket,
6292 ioe->anchor))) {
6293 PF_RULES_WUNLOCK();
6294 free(ioes, M_PF);
6295 goto fail; /* really bad */
6296 }
6297 break;
6298 #ifdef ALTQ
6299 case PF_RULESET_ALTQ:
6300 if (ioe->anchor[0]) {
6301 PF_RULES_WUNLOCK();
6302 free(ioes, M_PF);
6303 error = EINVAL;
6304 goto fail;
6305 }
6306 if ((error = pf_rollback_altq(ioe->ticket))) {
6307 PF_RULES_WUNLOCK();
6308 free(ioes, M_PF);
6309 goto fail; /* really bad */
6310 }
6311 break;
6312 #endif /* ALTQ */
6313 case PF_RULESET_TABLE:
6314 {
6315 struct pfr_table table;
6316
6317 bzero(&table, sizeof(table));
6318 strlcpy(table.pfrt_anchor, ioe->anchor,
6319 sizeof(table.pfrt_anchor));
6320 if ((error = pfr_ina_rollback(&table,
6321 ioe->ticket, NULL, 0))) {
6322 PF_RULES_WUNLOCK();
6323 free(ioes, M_PF);
6324 goto fail; /* really bad */
6325 }
6326 break;
6327 }
6328 default:
6329 if ((error = pf_rollback_rules(ioe->ticket,
6330 ioe->rs_num, ioe->anchor))) {
6331 PF_RULES_WUNLOCK();
6332 free(ioes, M_PF);
6333 goto fail; /* really bad */
6334 }
6335 break;
6336 }
6337 }
6338 PF_RULES_WUNLOCK();
6339 free(ioes, M_PF);
6340 break;
6341 }
6342
6343 case DIOCXCOMMIT: {
6344 struct pfioc_trans *io = (struct pfioc_trans *)addr;
6345 struct pfioc_trans_e *ioe, *ioes;
6346 struct pf_kruleset *rs;
6347 struct pf_keth_ruleset *ers;
6348 size_t totlen;
6349 int i;
6350
6351 if (io->esize != sizeof(*ioe)) {
6352 error = ENODEV;
6353 goto fail;
6354 }
6355
6356 if (io->size < 0 ||
6357 io->size > pf_ioctl_maxcount ||
6358 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
6359 error = EINVAL;
6360 goto fail;
6361 }
6362
6363 totlen = sizeof(struct pfioc_trans_e) * io->size;
6364 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
6365 M_PF, M_WAITOK);
6366 error = copyin(io->array, ioes, totlen);
6367 if (error) {
6368 free(ioes, M_PF);
6369 goto fail;
6370 }
6371 PF_RULES_WLOCK();
6372 /* First makes sure everything will succeed. */
6373 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
6374 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
6375 switch (ioe->rs_num) {
6376 case PF_RULESET_ETH:
6377 ers = pf_find_keth_ruleset(ioe->anchor);
6378 if (ers == NULL || ioe->ticket == 0 ||
6379 ioe->ticket != ers->inactive.ticket) {
6380 PF_RULES_WUNLOCK();
6381 free(ioes, M_PF);
6382 error = EINVAL;
6383 goto fail;
6384 }
6385 break;
6386 #ifdef ALTQ
6387 case PF_RULESET_ALTQ:
6388 if (ioe->anchor[0]) {
6389 PF_RULES_WUNLOCK();
6390 free(ioes, M_PF);
6391 error = EINVAL;
6392 goto fail;
6393 }
6394 if (!V_altqs_inactive_open || ioe->ticket !=
6395 V_ticket_altqs_inactive) {
6396 PF_RULES_WUNLOCK();
6397 free(ioes, M_PF);
6398 error = EBUSY;
6399 goto fail;
6400 }
6401 break;
6402 #endif /* ALTQ */
6403 case PF_RULESET_TABLE:
6404 rs = pf_find_kruleset(ioe->anchor);
6405 if (rs == NULL || !rs->topen || ioe->ticket !=
6406 rs->tticket) {
6407 PF_RULES_WUNLOCK();
6408 free(ioes, M_PF);
6409 error = EBUSY;
6410 goto fail;
6411 }
6412 break;
6413 default:
6414 if (ioe->rs_num < 0 || ioe->rs_num >=
6415 PF_RULESET_MAX) {
6416 PF_RULES_WUNLOCK();
6417 free(ioes, M_PF);
6418 error = EINVAL;
6419 goto fail;
6420 }
6421 rs = pf_find_kruleset(ioe->anchor);
6422 if (rs == NULL ||
6423 !rs->rules[ioe->rs_num].inactive.open ||
6424 rs->rules[ioe->rs_num].inactive.ticket !=
6425 ioe->ticket) {
6426 PF_RULES_WUNLOCK();
6427 free(ioes, M_PF);
6428 error = EBUSY;
6429 goto fail;
6430 }
6431 break;
6432 }
6433 }
6434 /* Now do the commit - no errors should happen here. */
6435 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
6436 switch (ioe->rs_num) {
6437 case PF_RULESET_ETH:
6438 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
6439 PF_RULES_WUNLOCK();
6440 free(ioes, M_PF);
6441 goto fail; /* really bad */
6442 }
6443 break;
6444 #ifdef ALTQ
6445 case PF_RULESET_ALTQ:
6446 if ((error = pf_commit_altq(ioe->ticket))) {
6447 PF_RULES_WUNLOCK();
6448 free(ioes, M_PF);
6449 goto fail; /* really bad */
6450 }
6451 break;
6452 #endif /* ALTQ */
6453 case PF_RULESET_TABLE:
6454 {
6455 struct pfr_table table;
6456
6457 bzero(&table, sizeof(table));
6458 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
6459 sizeof(table.pfrt_anchor));
6460 if ((error = pfr_ina_commit(&table,
6461 ioe->ticket, NULL, NULL, 0))) {
6462 PF_RULES_WUNLOCK();
6463 free(ioes, M_PF);
6464 goto fail; /* really bad */
6465 }
6466 break;
6467 }
6468 default:
6469 if ((error = pf_commit_rules(ioe->ticket,
6470 ioe->rs_num, ioe->anchor))) {
6471 PF_RULES_WUNLOCK();
6472 free(ioes, M_PF);
6473 goto fail; /* really bad */
6474 }
6475 break;
6476 }
6477 }
6478 PF_RULES_WUNLOCK();
6479
6480 /* Only hook into EtherNet taffic if we've got rules for it. */
6481 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
6482 hook_pf_eth();
6483 else
6484 dehook_pf_eth();
6485
6486 free(ioes, M_PF);
6487 break;
6488 }
6489
6490 case DIOCGETSRCNODES: {
6491 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
6492 struct pf_srchash *sh;
6493 struct pf_ksrc_node *n;
6494 struct pf_src_node *p, *pstore;
6495 uint32_t i, nr = 0;
6496
6497 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
6498 i++, sh++) {
6499 PF_HASHROW_LOCK(sh);
6500 LIST_FOREACH(n, &sh->nodes, entry)
6501 nr++;
6502 PF_HASHROW_UNLOCK(sh);
6503 }
6504
6505 psn->psn_len = min(psn->psn_len,
6506 sizeof(struct pf_src_node) * nr);
6507
6508 if (psn->psn_len == 0) {
6509 psn->psn_len = sizeof(struct pf_src_node) * nr;
6510 goto fail;
6511 }
6512
6513 nr = 0;
6514
6515 p = pstore = malloc(psn->psn_len, M_PF, M_WAITOK | M_ZERO);
6516 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
6517 i++, sh++) {
6518 PF_HASHROW_LOCK(sh);
6519 LIST_FOREACH(n, &sh->nodes, entry) {
6520
6521 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
6522 break;
6523
6524 pf_src_node_copy(n, p);
6525
6526 p++;
6527 nr++;
6528 }
6529 PF_HASHROW_UNLOCK(sh);
6530 }
6531 error = copyout(pstore, psn->psn_src_nodes,
6532 sizeof(struct pf_src_node) * nr);
6533 if (error) {
6534 free(pstore, M_PF);
6535 goto fail;
6536 }
6537 psn->psn_len = sizeof(struct pf_src_node) * nr;
6538 free(pstore, M_PF);
6539 break;
6540 }
6541
6542 case DIOCCLRSRCNODES: {
6543 pf_kill_srcnodes(NULL);
6544 break;
6545 }
6546
6547 case DIOCKILLSRCNODES:
6548 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
6549 break;
6550
6551 #ifdef COMPAT_FREEBSD13
6552 case DIOCKEEPCOUNTERS_FREEBSD13:
6553 #endif
6554 case DIOCKEEPCOUNTERS:
6555 error = pf_keepcounters((struct pfioc_nv *)addr);
6556 break;
6557
6558 case DIOCGETSYNCOOKIES:
6559 error = pf_get_syncookies((struct pfioc_nv *)addr);
6560 break;
6561
6562 case DIOCSETSYNCOOKIES:
6563 error = pf_set_syncookies((struct pfioc_nv *)addr);
6564 break;
6565
6566 case DIOCSETHOSTID: {
6567 u_int32_t *hostid = (u_int32_t *)addr;
6568
6569 PF_RULES_WLOCK();
6570 if (*hostid == 0)
6571 V_pf_status.hostid = arc4random();
6572 else
6573 V_pf_status.hostid = *hostid;
6574 PF_RULES_WUNLOCK();
6575 break;
6576 }
6577
6578 case DIOCOSFPFLUSH:
6579 PF_RULES_WLOCK();
6580 pf_osfp_flush();
6581 PF_RULES_WUNLOCK();
6582 break;
6583
6584 case DIOCIGETIFACES: {
6585 struct pfioc_iface *io = (struct pfioc_iface *)addr;
6586 struct pfi_kif *ifstore;
6587 size_t bufsiz;
6588
6589 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
6590 error = ENODEV;
6591 goto fail;
6592 }
6593
6594 if (io->pfiio_size < 0 ||
6595 io->pfiio_size > pf_ioctl_maxcount ||
6596 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
6597 error = EINVAL;
6598 goto fail;
6599 }
6600
6601 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
6602
6603 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
6604 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
6605 M_PF, M_WAITOK | M_ZERO);
6606
6607 PF_RULES_RLOCK();
6608 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
6609 PF_RULES_RUNLOCK();
6610 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
6611 free(ifstore, M_PF);
6612 break;
6613 }
6614
6615 case DIOCSETIFFLAG: {
6616 struct pfioc_iface *io = (struct pfioc_iface *)addr;
6617
6618 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
6619
6620 PF_RULES_WLOCK();
6621 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
6622 PF_RULES_WUNLOCK();
6623 break;
6624 }
6625
6626 case DIOCCLRIFFLAG: {
6627 struct pfioc_iface *io = (struct pfioc_iface *)addr;
6628
6629 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
6630
6631 PF_RULES_WLOCK();
6632 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
6633 PF_RULES_WUNLOCK();
6634 break;
6635 }
6636
6637 case DIOCSETREASS: {
6638 u_int32_t *reass = (u_int32_t *)addr;
6639
6640 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
6641 /* Removal of DF flag without reassembly enabled is not a
6642 * valid combination. Disable reassembly in such case. */
6643 if (!(V_pf_status.reass & PF_REASS_ENABLED))
6644 V_pf_status.reass = 0;
6645 break;
6646 }
6647
6648 default:
6649 error = ENODEV;
6650 break;
6651 }
6652 fail:
6653 CURVNET_RESTORE();
6654
6655 #undef ERROUT_IOCTL
6656
6657 return (error);
6658 }
6659
6660 static void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)6661 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
6662 {
6663 const char *tagname;
6664
6665 /* copy from state key */
6666 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
6667 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
6668 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
6669 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
6670 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
6671 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
6672 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
6673 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
6674
6675 /* copy from state */
6676 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
6677 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
6678 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
6679 sp->pfs_1301.expire = pf_state_expires(st);
6680 if (sp->pfs_1301.expire <= time_uptime)
6681 sp->pfs_1301.expire = htonl(0);
6682 else
6683 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
6684
6685 switch (msg_version) {
6686 case PFSYNC_MSG_VERSION_1301:
6687 sp->pfs_1301.state_flags = st->state_flags;
6688 sp->pfs_1301.direction = st->direction;
6689 sp->pfs_1301.log = st->act.log;
6690 sp->pfs_1301.timeout = st->timeout;
6691 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
6692 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
6693 /*
6694 * XXX Why do we bother pfsyncing source node information if source
6695 * nodes are not synced? Showing users that there is source tracking
6696 * when there is none seems useless.
6697 */
6698 if (st->sns[PF_SN_LIMIT] != NULL)
6699 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
6700 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
6701 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
6702 break;
6703 case PFSYNC_MSG_VERSION_1400:
6704 sp->pfs_1400.state_flags = htons(st->state_flags);
6705 sp->pfs_1400.direction = st->direction;
6706 sp->pfs_1400.log = st->act.log;
6707 sp->pfs_1400.timeout = st->timeout;
6708 sp->pfs_1400.proto = st->key[PF_SK_WIRE]->proto;
6709 sp->pfs_1400.af = st->key[PF_SK_WIRE]->af;
6710 sp->pfs_1400.qid = htons(st->act.qid);
6711 sp->pfs_1400.pqid = htons(st->act.pqid);
6712 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
6713 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
6714 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
6715 sp->pfs_1400.min_ttl = st->act.min_ttl;
6716 sp->pfs_1400.set_tos = st->act.set_tos;
6717 sp->pfs_1400.max_mss = htons(st->act.max_mss);
6718 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
6719 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
6720 sp->pfs_1400.rt = st->act.rt;
6721 if (st->act.rt_kif)
6722 strlcpy(sp->pfs_1400.rt_ifname,
6723 st->act.rt_kif->pfik_name,
6724 sizeof(sp->pfs_1400.rt_ifname));
6725 /*
6726 * XXX Why do we bother pfsyncing source node information if source
6727 * nodes are not synced? Showing users that there is source tracking
6728 * when there is none seems useless.
6729 */
6730 if (st->sns[PF_SN_LIMIT] != NULL)
6731 sp->pfs_1400.sync_flags |= PFSYNC_FLAG_SRCNODE;
6732 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
6733 sp->pfs_1400.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
6734 break;
6735 case PFSYNC_MSG_VERSION_1500:
6736 sp->pfs_1500.state_flags = htons(st->state_flags);
6737 sp->pfs_1500.direction = st->direction;
6738 sp->pfs_1500.log = st->act.log;
6739 sp->pfs_1500.timeout = st->timeout;
6740 sp->pfs_1500.wire_proto = st->key[PF_SK_WIRE]->proto;
6741 sp->pfs_1500.wire_af = st->key[PF_SK_WIRE]->af;
6742 sp->pfs_1500.stack_proto = st->key[PF_SK_STACK]->proto;
6743 sp->pfs_1500.stack_af = st->key[PF_SK_STACK]->af;
6744 sp->pfs_1500.qid = htons(st->act.qid);
6745 sp->pfs_1500.pqid = htons(st->act.pqid);
6746 sp->pfs_1500.dnpipe = htons(st->act.dnpipe);
6747 sp->pfs_1500.dnrpipe = htons(st->act.dnrpipe);
6748 sp->pfs_1500.rtableid = htonl(st->act.rtableid);
6749 sp->pfs_1500.min_ttl = st->act.min_ttl;
6750 sp->pfs_1500.set_tos = st->act.set_tos;
6751 sp->pfs_1500.max_mss = htons(st->act.max_mss);
6752 sp->pfs_1500.set_prio[0] = st->act.set_prio[0];
6753 sp->pfs_1500.set_prio[1] = st->act.set_prio[1];
6754 sp->pfs_1500.rt = st->act.rt;
6755 sp->pfs_1500.rt_af = st->act.rt_af;
6756 if (st->act.rt_kif)
6757 strlcpy(sp->pfs_1500.rt_ifname,
6758 st->act.rt_kif->pfik_name,
6759 sizeof(sp->pfs_1500.rt_ifname));
6760 strlcpy(sp->pfs_1500.orig_ifname,
6761 st->orig_kif->pfik_name,
6762 sizeof(sp->pfs_1500.orig_ifname));
6763 if ((tagname = pf_tag2tagname(st->tag)) != NULL)
6764 strlcpy(sp->pfs_1500.tagname, tagname,
6765 sizeof(sp->pfs_1500.tagname));
6766 break;
6767 default:
6768 panic("%s: Unsupported pfsync_msg_version %d",
6769 __func__, msg_version);
6770 }
6771
6772 sp->pfs_1301.id = st->id;
6773 sp->pfs_1301.creatorid = st->creatorid;
6774 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
6775 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
6776
6777 if (st->rule == NULL)
6778 sp->pfs_1301.rule = htonl(-1);
6779 else
6780 sp->pfs_1301.rule = htonl(st->rule->nr);
6781 if (st->anchor == NULL)
6782 sp->pfs_1301.anchor = htonl(-1);
6783 else
6784 sp->pfs_1301.anchor = htonl(st->anchor->nr);
6785 if (st->nat_rule == NULL)
6786 sp->pfs_1301.nat_rule = htonl(-1);
6787 else
6788 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
6789
6790 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
6791 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
6792 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
6793 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
6794 }
6795
6796 void
pfsync_state_export_1301(struct pfsync_state_1301 * sp,struct pf_kstate * st)6797 pfsync_state_export_1301(struct pfsync_state_1301 *sp, struct pf_kstate *st)
6798 {
6799 bzero(sp, sizeof(*sp));
6800 pfsync_state_export((union pfsync_state_union *)sp, st,
6801 PFSYNC_MSG_VERSION_1301);
6802 }
6803
6804 void
pfsync_state_export_1400(struct pfsync_state_1400 * sp,struct pf_kstate * st)6805 pfsync_state_export_1400(struct pfsync_state_1400 *sp, struct pf_kstate *st)
6806 {
6807 bzero(sp, sizeof(*sp));
6808 pfsync_state_export((union pfsync_state_union *)sp, st,
6809 PFSYNC_MSG_VERSION_1400);
6810 }
6811
6812 void
pfsync_state_export_1500(struct pfsync_state_1500 * sp,struct pf_kstate * st)6813 pfsync_state_export_1500(struct pfsync_state_1500 *sp, struct pf_kstate *st)
6814 {
6815 bzero(sp, sizeof(*sp));
6816 pfsync_state_export((union pfsync_state_union *)sp, st,
6817 PFSYNC_MSG_VERSION_1500);
6818 }
6819
6820 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)6821 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
6822 {
6823 bzero(sp, sizeof(*sp));
6824
6825 sp->version = PF_STATE_VERSION;
6826
6827 /* copy from state key */
6828 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
6829 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
6830 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
6831 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
6832 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
6833 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
6834 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
6835 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
6836 sp->proto = st->key[PF_SK_WIRE]->proto;
6837 sp->af = st->key[PF_SK_WIRE]->af;
6838
6839 /* copy from state */
6840 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
6841 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
6842 sizeof(sp->orig_ifname));
6843 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr));
6844 sp->creation = htonl(time_uptime - (st->creation / 1000));
6845 sp->expire = pf_state_expires(st);
6846 if (sp->expire <= time_uptime)
6847 sp->expire = htonl(0);
6848 else
6849 sp->expire = htonl(sp->expire - time_uptime);
6850
6851 sp->direction = st->direction;
6852 sp->log = st->act.log;
6853 sp->timeout = st->timeout;
6854 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
6855 sp->state_flags_compat = st->state_flags;
6856 sp->state_flags = htons(st->state_flags);
6857 if (st->sns[PF_SN_LIMIT] != NULL)
6858 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
6859 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL)
6860 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
6861 sp->id = st->id;
6862 sp->creatorid = st->creatorid;
6863 pf_state_peer_hton(&st->src, &sp->src);
6864 pf_state_peer_hton(&st->dst, &sp->dst);
6865
6866 if (st->rule == NULL)
6867 sp->rule = htonl(-1);
6868 else
6869 sp->rule = htonl(st->rule->nr);
6870 if (st->anchor == NULL)
6871 sp->anchor = htonl(-1);
6872 else
6873 sp->anchor = htonl(st->anchor->nr);
6874 if (st->nat_rule == NULL)
6875 sp->nat_rule = htonl(-1);
6876 else
6877 sp->nat_rule = htonl(st->nat_rule->nr);
6878
6879 sp->packets[0] = st->packets[0];
6880 sp->packets[1] = st->packets[1];
6881 sp->bytes[0] = st->bytes[0];
6882 sp->bytes[1] = st->bytes[1];
6883
6884 sp->qid = htons(st->act.qid);
6885 sp->pqid = htons(st->act.pqid);
6886 sp->dnpipe = htons(st->act.dnpipe);
6887 sp->dnrpipe = htons(st->act.dnrpipe);
6888 sp->rtableid = htonl(st->act.rtableid);
6889 sp->min_ttl = st->act.min_ttl;
6890 sp->set_tos = st->act.set_tos;
6891 sp->max_mss = htons(st->act.max_mss);
6892 sp->rt = st->act.rt;
6893 if (st->act.rt_kif)
6894 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
6895 sizeof(sp->rt_ifname));
6896 sp->set_prio[0] = st->act.set_prio[0];
6897 sp->set_prio[1] = st->act.set_prio[1];
6898
6899 }
6900
6901 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)6902 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
6903 {
6904 struct pfr_ktable *kt;
6905
6906 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
6907
6908 kt = aw->p.tbl;
6909 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
6910 kt = kt->pfrkt_root;
6911 aw->p.tbl = NULL;
6912 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
6913 kt->pfrkt_cnt : -1;
6914 }
6915
6916 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)6917 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
6918 size_t number, char **names)
6919 {
6920 nvlist_t *nvc;
6921
6922 nvc = nvlist_create(0);
6923 if (nvc == NULL)
6924 return (ENOMEM);
6925
6926 for (int i = 0; i < number; i++) {
6927 nvlist_append_number_array(nvc, "counters",
6928 counter_u64_fetch(counters[i]));
6929 nvlist_append_string_array(nvc, "names",
6930 names[i]);
6931 nvlist_append_number_array(nvc, "ids",
6932 i);
6933 }
6934 nvlist_add_nvlist(nvl, name, nvc);
6935 nvlist_destroy(nvc);
6936
6937 return (0);
6938 }
6939
6940 static int
pf_getstatus(struct pfioc_nv * nv)6941 pf_getstatus(struct pfioc_nv *nv)
6942 {
6943 nvlist_t *nvl = NULL, *nvc = NULL;
6944 void *nvlpacked = NULL;
6945 int error;
6946 struct pf_status s;
6947 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
6948 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
6949 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
6950 time_t since;
6951
6952 PF_RULES_RLOCK_TRACKER;
6953
6954 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6955
6956 PF_RULES_RLOCK();
6957
6958 nvl = nvlist_create(0);
6959 if (nvl == NULL)
6960 ERROUT(ENOMEM);
6961
6962 since = time_second - (time_uptime - V_pf_status.since);
6963
6964 nvlist_add_bool(nvl, "running", V_pf_status.running);
6965 nvlist_add_number(nvl, "since", since);
6966 nvlist_add_number(nvl, "debug", V_pf_status.debug);
6967 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
6968 nvlist_add_number(nvl, "states", V_pf_status.states);
6969 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
6970 nvlist_add_number(nvl, "reass", V_pf_status.reass);
6971 nvlist_add_bool(nvl, "syncookies_active",
6972 V_pf_status.syncookies_active);
6973 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
6974
6975 /* counters */
6976 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
6977 PFRES_MAX, pf_reasons);
6978 if (error != 0)
6979 ERROUT(error);
6980
6981 /* lcounters */
6982 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
6983 KLCNT_MAX, pf_lcounter);
6984 if (error != 0)
6985 ERROUT(error);
6986
6987 /* fcounters */
6988 nvc = nvlist_create(0);
6989 if (nvc == NULL)
6990 ERROUT(ENOMEM);
6991
6992 for (int i = 0; i < FCNT_MAX; i++) {
6993 nvlist_append_number_array(nvc, "counters",
6994 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
6995 nvlist_append_string_array(nvc, "names",
6996 pf_fcounter[i]);
6997 nvlist_append_number_array(nvc, "ids",
6998 i);
6999 }
7000 nvlist_add_nvlist(nvl, "fcounters", nvc);
7001 nvlist_destroy(nvc);
7002 nvc = NULL;
7003
7004 /* scounters */
7005 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
7006 SCNT_MAX, pf_fcounter);
7007 if (error != 0)
7008 ERROUT(error);
7009
7010 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
7011 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
7012 PF_MD5_DIGEST_LENGTH);
7013
7014 pfi_update_status(V_pf_status.ifname, &s);
7015
7016 /* pcounters / bcounters */
7017 for (int i = 0; i < 2; i++) {
7018 for (int j = 0; j < 2; j++) {
7019 for (int k = 0; k < 2; k++) {
7020 nvlist_append_number_array(nvl, "pcounters",
7021 s.pcounters[i][j][k]);
7022 }
7023 nvlist_append_number_array(nvl, "bcounters",
7024 s.bcounters[i][j]);
7025 }
7026 }
7027
7028 nvlpacked = nvlist_pack(nvl, &nv->len);
7029 if (nvlpacked == NULL)
7030 ERROUT(ENOMEM);
7031
7032 if (nv->size == 0)
7033 ERROUT(0);
7034 else if (nv->size < nv->len)
7035 ERROUT(ENOSPC);
7036
7037 PF_RULES_RUNLOCK();
7038 error = copyout(nvlpacked, nv->data, nv->len);
7039 goto done;
7040
7041 #undef ERROUT
7042 errout:
7043 PF_RULES_RUNLOCK();
7044 done:
7045 free(nvlpacked, M_NVLIST);
7046 nvlist_destroy(nvc);
7047 nvlist_destroy(nvl);
7048
7049 return (error);
7050 }
7051
7052 /*
7053 * XXX - Check for version mismatch!!!
7054 */
7055 static void
pf_clear_all_states(void)7056 pf_clear_all_states(void)
7057 {
7058 struct epoch_tracker et;
7059 struct pf_kstate *s;
7060 u_int i;
7061
7062 NET_EPOCH_ENTER(et);
7063 for (i = 0; i <= V_pf_hashmask; i++) {
7064 struct pf_idhash *ih = &V_pf_idhash[i];
7065 relock:
7066 PF_HASHROW_LOCK(ih);
7067 LIST_FOREACH(s, &ih->states, entry) {
7068 s->timeout = PFTM_PURGE;
7069 /* Don't send out individual delete messages. */
7070 s->state_flags |= PFSTATE_NOSYNC;
7071 pf_remove_state(s);
7072 goto relock;
7073 }
7074 PF_HASHROW_UNLOCK(ih);
7075 }
7076 NET_EPOCH_EXIT(et);
7077 }
7078
7079 static int
pf_clear_tables(void)7080 pf_clear_tables(void)
7081 {
7082 struct pfioc_table io;
7083 int error;
7084
7085 bzero(&io, sizeof(io));
7086 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
7087
7088 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
7089 io.pfrio_flags);
7090
7091 return (error);
7092 }
7093
7094 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)7095 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
7096 {
7097 struct pf_ksrc_node_list kill;
7098 u_int killed;
7099
7100 LIST_INIT(&kill);
7101 for (int i = 0; i <= V_pf_srchashmask; i++) {
7102 struct pf_srchash *sh = &V_pf_srchash[i];
7103 struct pf_ksrc_node *sn, *tmp;
7104
7105 PF_HASHROW_LOCK(sh);
7106 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
7107 if (psnk == NULL ||
7108 (pf_match_addr(psnk->psnk_src.neg,
7109 &psnk->psnk_src.addr.v.a.addr,
7110 &psnk->psnk_src.addr.v.a.mask,
7111 &sn->addr, sn->af) &&
7112 pf_match_addr(psnk->psnk_dst.neg,
7113 &psnk->psnk_dst.addr.v.a.addr,
7114 &psnk->psnk_dst.addr.v.a.mask,
7115 &sn->raddr, sn->af))) {
7116 pf_unlink_src_node(sn);
7117 LIST_INSERT_HEAD(&kill, sn, entry);
7118 sn->expire = 1;
7119 }
7120 PF_HASHROW_UNLOCK(sh);
7121 }
7122
7123 for (int i = 0; i <= V_pf_hashmask; i++) {
7124 struct pf_idhash *ih = &V_pf_idhash[i];
7125 struct pf_kstate *s;
7126
7127 PF_HASHROW_LOCK(ih);
7128 LIST_FOREACH(s, &ih->states, entry) {
7129 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX;
7130 sn_type++) {
7131 if (s->sns[sn_type] &&
7132 s->sns[sn_type]->expire == 1) {
7133 s->sns[sn_type] = NULL;
7134 }
7135 }
7136 }
7137 PF_HASHROW_UNLOCK(ih);
7138 }
7139
7140 killed = pf_free_src_nodes(&kill);
7141
7142 if (psnk != NULL)
7143 psnk->psnk_killed = killed;
7144 }
7145
7146 static int
pf_keepcounters(struct pfioc_nv * nv)7147 pf_keepcounters(struct pfioc_nv *nv)
7148 {
7149 nvlist_t *nvl = NULL;
7150 void *nvlpacked = NULL;
7151 int error = 0;
7152
7153 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
7154
7155 if (nv->len > pf_ioctl_maxcount)
7156 ERROUT(ENOMEM);
7157
7158 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
7159 error = copyin(nv->data, nvlpacked, nv->len);
7160 if (error)
7161 ERROUT(error);
7162
7163 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
7164 if (nvl == NULL)
7165 ERROUT(EBADMSG);
7166
7167 if (! nvlist_exists_bool(nvl, "keep_counters"))
7168 ERROUT(EBADMSG);
7169
7170 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
7171
7172 on_error:
7173 nvlist_destroy(nvl);
7174 free(nvlpacked, M_NVLIST);
7175 return (error);
7176 }
7177
7178 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)7179 pf_clear_states(const struct pf_kstate_kill *kill)
7180 {
7181 struct pf_state_key_cmp match_key;
7182 struct pf_kstate *s;
7183 struct pfi_kkif *kif;
7184 int idx;
7185 unsigned int killed = 0, dir;
7186
7187 NET_EPOCH_ASSERT();
7188
7189 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
7190 struct pf_idhash *ih = &V_pf_idhash[i];
7191
7192 relock_DIOCCLRSTATES:
7193 PF_HASHROW_LOCK(ih);
7194 LIST_FOREACH(s, &ih->states, entry) {
7195 /* For floating states look at the original kif. */
7196 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
7197
7198 if (kill->psk_ifname[0] &&
7199 strcmp(kill->psk_ifname,
7200 kif->pfik_name))
7201 continue;
7202
7203 if (kill->psk_kill_match) {
7204 bzero(&match_key, sizeof(match_key));
7205
7206 if (s->direction == PF_OUT) {
7207 dir = PF_IN;
7208 idx = PF_SK_STACK;
7209 } else {
7210 dir = PF_OUT;
7211 idx = PF_SK_WIRE;
7212 }
7213
7214 match_key.af = s->key[idx]->af;
7215 match_key.proto = s->key[idx]->proto;
7216 pf_addrcpy(&match_key.addr[0],
7217 &s->key[idx]->addr[1], match_key.af);
7218 match_key.port[0] = s->key[idx]->port[1];
7219 pf_addrcpy(&match_key.addr[1],
7220 &s->key[idx]->addr[0], match_key.af);
7221 match_key.port[1] = s->key[idx]->port[0];
7222 }
7223
7224 /*
7225 * Don't send out individual
7226 * delete messages.
7227 */
7228 s->state_flags |= PFSTATE_NOSYNC;
7229 pf_remove_state(s);
7230 killed++;
7231
7232 if (kill->psk_kill_match)
7233 killed += pf_kill_matching_state(&match_key,
7234 dir);
7235
7236 goto relock_DIOCCLRSTATES;
7237 }
7238 PF_HASHROW_UNLOCK(ih);
7239 }
7240
7241 if (V_pfsync_clear_states_ptr != NULL)
7242 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
7243
7244 return (killed);
7245 }
7246
7247 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)7248 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
7249 {
7250 struct pf_kstate *s;
7251
7252 NET_EPOCH_ASSERT();
7253 if (kill->psk_pfcmp.id) {
7254 if (kill->psk_pfcmp.creatorid == 0)
7255 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
7256 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
7257 kill->psk_pfcmp.creatorid))) {
7258 pf_remove_state(s);
7259 *killed = 1;
7260 }
7261 return;
7262 }
7263
7264 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
7265 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
7266 }
7267
7268 static int
pf_killstates_nv(struct pfioc_nv * nv)7269 pf_killstates_nv(struct pfioc_nv *nv)
7270 {
7271 struct pf_kstate_kill kill;
7272 struct epoch_tracker et;
7273 nvlist_t *nvl = NULL;
7274 void *nvlpacked = NULL;
7275 int error = 0;
7276 unsigned int killed = 0;
7277
7278 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
7279
7280 if (nv->len > pf_ioctl_maxcount)
7281 ERROUT(ENOMEM);
7282
7283 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
7284 error = copyin(nv->data, nvlpacked, nv->len);
7285 if (error)
7286 ERROUT(error);
7287
7288 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
7289 if (nvl == NULL)
7290 ERROUT(EBADMSG);
7291
7292 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
7293 if (error)
7294 ERROUT(error);
7295
7296 NET_EPOCH_ENTER(et);
7297 pf_killstates(&kill, &killed);
7298 NET_EPOCH_EXIT(et);
7299
7300 free(nvlpacked, M_NVLIST);
7301 nvlpacked = NULL;
7302 nvlist_destroy(nvl);
7303 nvl = nvlist_create(0);
7304 if (nvl == NULL)
7305 ERROUT(ENOMEM);
7306
7307 nvlist_add_number(nvl, "killed", killed);
7308
7309 nvlpacked = nvlist_pack(nvl, &nv->len);
7310 if (nvlpacked == NULL)
7311 ERROUT(ENOMEM);
7312
7313 if (nv->size == 0)
7314 ERROUT(0);
7315 else if (nv->size < nv->len)
7316 ERROUT(ENOSPC);
7317
7318 error = copyout(nvlpacked, nv->data, nv->len);
7319
7320 on_error:
7321 nvlist_destroy(nvl);
7322 free(nvlpacked, M_NVLIST);
7323 return (error);
7324 }
7325
7326 static int
pf_clearstates_nv(struct pfioc_nv * nv)7327 pf_clearstates_nv(struct pfioc_nv *nv)
7328 {
7329 struct pf_kstate_kill kill;
7330 struct epoch_tracker et;
7331 nvlist_t *nvl = NULL;
7332 void *nvlpacked = NULL;
7333 int error = 0;
7334 unsigned int killed;
7335
7336 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
7337
7338 if (nv->len > pf_ioctl_maxcount)
7339 ERROUT(ENOMEM);
7340
7341 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
7342 error = copyin(nv->data, nvlpacked, nv->len);
7343 if (error)
7344 ERROUT(error);
7345
7346 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
7347 if (nvl == NULL)
7348 ERROUT(EBADMSG);
7349
7350 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
7351 if (error)
7352 ERROUT(error);
7353
7354 NET_EPOCH_ENTER(et);
7355 killed = pf_clear_states(&kill);
7356 NET_EPOCH_EXIT(et);
7357
7358 free(nvlpacked, M_NVLIST);
7359 nvlpacked = NULL;
7360 nvlist_destroy(nvl);
7361 nvl = nvlist_create(0);
7362 if (nvl == NULL)
7363 ERROUT(ENOMEM);
7364
7365 nvlist_add_number(nvl, "killed", killed);
7366
7367 nvlpacked = nvlist_pack(nvl, &nv->len);
7368 if (nvlpacked == NULL)
7369 ERROUT(ENOMEM);
7370
7371 if (nv->size == 0)
7372 ERROUT(0);
7373 else if (nv->size < nv->len)
7374 ERROUT(ENOSPC);
7375
7376 error = copyout(nvlpacked, nv->data, nv->len);
7377
7378 #undef ERROUT
7379 on_error:
7380 nvlist_destroy(nvl);
7381 free(nvlpacked, M_NVLIST);
7382 return (error);
7383 }
7384
7385 static int
pf_getstate(struct pfioc_nv * nv)7386 pf_getstate(struct pfioc_nv *nv)
7387 {
7388 nvlist_t *nvl = NULL, *nvls;
7389 void *nvlpacked = NULL;
7390 struct pf_kstate *s = NULL;
7391 int error = 0;
7392 uint64_t id, creatorid;
7393
7394 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
7395
7396 if (nv->len > pf_ioctl_maxcount)
7397 ERROUT(ENOMEM);
7398
7399 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
7400 error = copyin(nv->data, nvlpacked, nv->len);
7401 if (error)
7402 ERROUT(error);
7403
7404 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
7405 if (nvl == NULL)
7406 ERROUT(EBADMSG);
7407
7408 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
7409 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
7410
7411 s = pf_find_state_byid(id, creatorid);
7412 if (s == NULL)
7413 ERROUT(ENOENT);
7414
7415 free(nvlpacked, M_NVLIST);
7416 nvlpacked = NULL;
7417 nvlist_destroy(nvl);
7418 nvl = nvlist_create(0);
7419 if (nvl == NULL)
7420 ERROUT(ENOMEM);
7421
7422 nvls = pf_state_to_nvstate(s);
7423 if (nvls == NULL)
7424 ERROUT(ENOMEM);
7425
7426 nvlist_add_nvlist(nvl, "state", nvls);
7427 nvlist_destroy(nvls);
7428
7429 nvlpacked = nvlist_pack(nvl, &nv->len);
7430 if (nvlpacked == NULL)
7431 ERROUT(ENOMEM);
7432
7433 if (nv->size == 0)
7434 ERROUT(0);
7435 else if (nv->size < nv->len)
7436 ERROUT(ENOSPC);
7437
7438 error = copyout(nvlpacked, nv->data, nv->len);
7439
7440 #undef ERROUT
7441 errout:
7442 if (s != NULL)
7443 PF_STATE_UNLOCK(s);
7444 free(nvlpacked, M_NVLIST);
7445 nvlist_destroy(nvl);
7446 return (error);
7447 }
7448
7449 /*
7450 * XXX - Check for version mismatch!!!
7451 */
7452
7453 /*
7454 * Duplicate pfctl -Fa operation to get rid of as much as we can.
7455 */
7456 static int
shutdown_pf(void)7457 shutdown_pf(void)
7458 {
7459 int error = 0;
7460 u_int32_t t[5];
7461 char nn = '\0';
7462 struct pf_kanchor *anchor, *tmp_anchor;
7463 struct pf_keth_anchor *eth_anchor, *tmp_eth_anchor;
7464 int rs_num;
7465
7466 do {
7467 /* Unlink rules of all user defined anchors */
7468 RB_FOREACH_SAFE(anchor, pf_kanchor_global, &V_pf_anchors,
7469 tmp_anchor) {
7470 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
7471 if ((error = pf_begin_rules(&t[rs_num], rs_num,
7472 anchor->path)) != 0) {
7473 DPFPRINTF(PF_DEBUG_MISC, "%s: "
7474 "anchor.path=%s rs_num=%d",
7475 __func__, anchor->path, rs_num);
7476 goto error; /* XXX: rollback? */
7477 }
7478 }
7479 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
7480 error = pf_commit_rules(t[rs_num], rs_num,
7481 anchor->path);
7482 MPASS(error == 0);
7483 }
7484 }
7485
7486 /* Unlink rules of all user defined ether anchors */
7487 RB_FOREACH_SAFE(eth_anchor, pf_keth_anchor_global,
7488 &V_pf_keth_anchors, tmp_eth_anchor) {
7489 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
7490 != 0) {
7491 DPFPRINTF(PF_DEBUG_MISC, "%s: eth "
7492 "anchor.path=%s", __func__,
7493 eth_anchor->path);
7494 goto error;
7495 }
7496 error = pf_commit_eth(t[0], eth_anchor->path);
7497 MPASS(error == 0);
7498 }
7499
7500 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
7501 != 0) {
7502 DPFPRINTF(PF_DEBUG_MISC, "%s: SCRUB", __func__);
7503 break;
7504 }
7505 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
7506 != 0) {
7507 DPFPRINTF(PF_DEBUG_MISC, "%s: FILTER", __func__);
7508 break; /* XXX: rollback? */
7509 }
7510 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
7511 != 0) {
7512 DPFPRINTF(PF_DEBUG_MISC, "%s: NAT", __func__);
7513 break; /* XXX: rollback? */
7514 }
7515 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
7516 != 0) {
7517 DPFPRINTF(PF_DEBUG_MISC, "%s: BINAT", __func__);
7518 break; /* XXX: rollback? */
7519 }
7520 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
7521 != 0) {
7522 DPFPRINTF(PF_DEBUG_MISC, "%s: RDR", __func__);
7523 break; /* XXX: rollback? */
7524 }
7525
7526 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
7527 MPASS(error == 0);
7528 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
7529 MPASS(error == 0);
7530 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
7531 MPASS(error == 0);
7532 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
7533 MPASS(error == 0);
7534 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
7535 MPASS(error == 0);
7536
7537 if ((error = pf_clear_tables()) != 0)
7538 break;
7539
7540 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
7541 DPFPRINTF(PF_DEBUG_MISC, "%s: eth", __func__);
7542 break;
7543 }
7544 error = pf_commit_eth(t[0], &nn);
7545 MPASS(error == 0);
7546
7547 #ifdef ALTQ
7548 if ((error = pf_begin_altq(&t[0])) != 0) {
7549 DPFPRINTF(PF_DEBUG_MISC, "%s: ALTQ", __func__);
7550 break;
7551 }
7552 pf_commit_altq(t[0]);
7553 #endif
7554
7555 pf_clear_all_states();
7556
7557 pf_kill_srcnodes(NULL);
7558
7559 for (int i = 0; i < PF_RULESET_MAX; i++) {
7560 pf_rule_tree_free(pf_main_ruleset.rules[i].active.tree);
7561 pf_rule_tree_free(pf_main_ruleset.rules[i].inactive.tree);
7562 }
7563
7564 /* status does not use malloced mem so no need to cleanup */
7565 /* fingerprints and interfaces have their own cleanup code */
7566 } while(0);
7567
7568 error:
7569 return (error);
7570 }
7571
7572 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)7573 pf_check_return(int chk, struct mbuf **m)
7574 {
7575
7576 switch (chk) {
7577 case PF_PASS:
7578 if (*m == NULL)
7579 return (PFIL_CONSUMED);
7580 else
7581 return (PFIL_PASS);
7582 break;
7583 default:
7584 if (*m != NULL) {
7585 m_freem(*m);
7586 *m = NULL;
7587 }
7588 return (PFIL_DROPPED);
7589 }
7590 }
7591
7592 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7593 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
7594 void *ruleset __unused, struct inpcb *inp)
7595 {
7596 int chk;
7597
7598 CURVNET_ASSERT_SET();
7599
7600 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
7601
7602 return (pf_check_return(chk, m));
7603 }
7604
7605 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7606 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
7607 void *ruleset __unused, struct inpcb *inp)
7608 {
7609 int chk;
7610
7611 CURVNET_ASSERT_SET();
7612
7613 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
7614
7615 return (pf_check_return(chk, m));
7616 }
7617
7618 #ifdef INET
7619 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7620 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
7621 void *ruleset __unused, struct inpcb *inp)
7622 {
7623 int chk;
7624
7625 CURVNET_ASSERT_SET();
7626
7627 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
7628
7629 return (pf_check_return(chk, m));
7630 }
7631
7632 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7633 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
7634 void *ruleset __unused, struct inpcb *inp)
7635 {
7636 int chk;
7637
7638 CURVNET_ASSERT_SET();
7639
7640 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
7641
7642 return (pf_check_return(chk, m));
7643 }
7644 #endif
7645
7646 #ifdef INET6
7647 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7648 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
7649 void *ruleset __unused, struct inpcb *inp)
7650 {
7651 int chk;
7652
7653 CURVNET_ASSERT_SET();
7654
7655 /*
7656 * In case of loopback traffic IPv6 uses the real interface in
7657 * order to support scoped addresses. In order to support stateful
7658 * filtering we have change this to lo0 as it is the case in IPv4.
7659 */
7660 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
7661 m, inp, NULL);
7662
7663 return (pf_check_return(chk, m));
7664 }
7665
7666 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7667 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
7668 void *ruleset __unused, struct inpcb *inp)
7669 {
7670 int chk;
7671
7672 CURVNET_ASSERT_SET();
7673
7674 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL);
7675
7676 return (pf_check_return(chk, m));
7677 }
7678 #endif /* INET6 */
7679
7680 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
7681 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
7682 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
7683 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
7684
7685 #ifdef INET
7686 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
7687 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
7688 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
7689 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
7690 #endif
7691 #ifdef INET6
7692 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
7693 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
7694 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
7695 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
7696 #endif
7697
7698 static void
hook_pf_eth(void)7699 hook_pf_eth(void)
7700 {
7701 struct pfil_hook_args pha = {
7702 .pa_version = PFIL_VERSION,
7703 .pa_modname = "pf",
7704 .pa_type = PFIL_TYPE_ETHERNET,
7705 };
7706 struct pfil_link_args pla = {
7707 .pa_version = PFIL_VERSION,
7708 };
7709 int ret __diagused;
7710
7711 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
7712 return;
7713
7714 pha.pa_mbuf_chk = pf_eth_check_in;
7715 pha.pa_flags = PFIL_IN;
7716 pha.pa_rulname = "eth-in";
7717 V_pf_eth_in_hook = pfil_add_hook(&pha);
7718 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
7719 pla.pa_head = V_link_pfil_head;
7720 pla.pa_hook = V_pf_eth_in_hook;
7721 ret = pfil_link(&pla);
7722 MPASS(ret == 0);
7723 pha.pa_mbuf_chk = pf_eth_check_out;
7724 pha.pa_flags = PFIL_OUT;
7725 pha.pa_rulname = "eth-out";
7726 V_pf_eth_out_hook = pfil_add_hook(&pha);
7727 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7728 pla.pa_head = V_link_pfil_head;
7729 pla.pa_hook = V_pf_eth_out_hook;
7730 ret = pfil_link(&pla);
7731 MPASS(ret == 0);
7732
7733 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
7734 }
7735
7736 static void
hook_pf(void)7737 hook_pf(void)
7738 {
7739 struct pfil_hook_args pha = {
7740 .pa_version = PFIL_VERSION,
7741 .pa_modname = "pf",
7742 };
7743 struct pfil_link_args pla = {
7744 .pa_version = PFIL_VERSION,
7745 };
7746 int ret __diagused;
7747
7748 if (atomic_load_bool(&V_pf_pfil_hooked))
7749 return;
7750
7751 #ifdef INET
7752 pha.pa_type = PFIL_TYPE_IP4;
7753 pha.pa_mbuf_chk = pf_check_in;
7754 pha.pa_flags = PFIL_IN;
7755 pha.pa_rulname = "default-in";
7756 V_pf_ip4_in_hook = pfil_add_hook(&pha);
7757 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
7758 pla.pa_head = V_inet_pfil_head;
7759 pla.pa_hook = V_pf_ip4_in_hook;
7760 ret = pfil_link(&pla);
7761 MPASS(ret == 0);
7762 pha.pa_mbuf_chk = pf_check_out;
7763 pha.pa_flags = PFIL_OUT;
7764 pha.pa_rulname = "default-out";
7765 V_pf_ip4_out_hook = pfil_add_hook(&pha);
7766 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7767 pla.pa_head = V_inet_pfil_head;
7768 pla.pa_hook = V_pf_ip4_out_hook;
7769 ret = pfil_link(&pla);
7770 MPASS(ret == 0);
7771 if (V_pf_filter_local) {
7772 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7773 pla.pa_head = V_inet_local_pfil_head;
7774 pla.pa_hook = V_pf_ip4_out_hook;
7775 ret = pfil_link(&pla);
7776 MPASS(ret == 0);
7777 }
7778 #endif
7779 #ifdef INET6
7780 pha.pa_type = PFIL_TYPE_IP6;
7781 pha.pa_mbuf_chk = pf_check6_in;
7782 pha.pa_flags = PFIL_IN;
7783 pha.pa_rulname = "default-in6";
7784 V_pf_ip6_in_hook = pfil_add_hook(&pha);
7785 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
7786 pla.pa_head = V_inet6_pfil_head;
7787 pla.pa_hook = V_pf_ip6_in_hook;
7788 ret = pfil_link(&pla);
7789 MPASS(ret == 0);
7790 pha.pa_mbuf_chk = pf_check6_out;
7791 pha.pa_rulname = "default-out6";
7792 pha.pa_flags = PFIL_OUT;
7793 V_pf_ip6_out_hook = pfil_add_hook(&pha);
7794 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7795 pla.pa_head = V_inet6_pfil_head;
7796 pla.pa_hook = V_pf_ip6_out_hook;
7797 ret = pfil_link(&pla);
7798 MPASS(ret == 0);
7799 if (V_pf_filter_local) {
7800 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7801 pla.pa_head = V_inet6_local_pfil_head;
7802 pla.pa_hook = V_pf_ip6_out_hook;
7803 ret = pfil_link(&pla);
7804 MPASS(ret == 0);
7805 }
7806 #endif
7807
7808 atomic_store_bool(&V_pf_pfil_hooked, true);
7809 }
7810
7811 static void
dehook_pf_eth(void)7812 dehook_pf_eth(void)
7813 {
7814
7815 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
7816 return;
7817
7818 pfil_remove_hook(V_pf_eth_in_hook);
7819 pfil_remove_hook(V_pf_eth_out_hook);
7820
7821 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
7822 }
7823
7824 static void
dehook_pf(void)7825 dehook_pf(void)
7826 {
7827
7828 if (!atomic_load_bool(&V_pf_pfil_hooked))
7829 return;
7830
7831 #ifdef INET
7832 pfil_remove_hook(V_pf_ip4_in_hook);
7833 pfil_remove_hook(V_pf_ip4_out_hook);
7834 #endif
7835 #ifdef INET6
7836 pfil_remove_hook(V_pf_ip6_in_hook);
7837 pfil_remove_hook(V_pf_ip6_out_hook);
7838 #endif
7839
7840 atomic_store_bool(&V_pf_pfil_hooked, false);
7841 }
7842
7843 static void
pf_load_vnet(void)7844 pf_load_vnet(void)
7845 {
7846 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
7847 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
7848
7849 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
7850 rm_init_flags(&V_pf_tags_lock, "pf tags and queues", RM_RECURSE);
7851 sx_init(&V_pf_ioctl_lock, "pf ioctl");
7852
7853 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
7854 PF_RULE_TAG_HASH_SIZE_DEFAULT);
7855 #ifdef ALTQ
7856 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
7857 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
7858 #endif
7859
7860 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
7861
7862 pfattach_vnet();
7863 V_pf_vnet_active = 1;
7864 }
7865
7866 static int
pf_load(void)7867 pf_load(void)
7868 {
7869 int error;
7870
7871 sx_init(&pf_end_lock, "pf end thread");
7872
7873 pf_mtag_initialize();
7874
7875 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
7876 if (pf_dev == NULL)
7877 return (ENOMEM);
7878
7879 pf_end_threads = 0;
7880 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
7881 if (error != 0)
7882 return (error);
7883
7884 pfi_initialize();
7885
7886 return (0);
7887 }
7888
7889 static void
pf_unload_vnet(void)7890 pf_unload_vnet(void)
7891 {
7892 int ret __diagused;
7893
7894 V_pf_vnet_active = 0;
7895 V_pf_status.running = 0;
7896 dehook_pf();
7897 dehook_pf_eth();
7898
7899 PF_RULES_WLOCK();
7900 pf_syncookies_cleanup();
7901 shutdown_pf();
7902 PF_RULES_WUNLOCK();
7903
7904 ret = swi_remove(V_pf_swi_cookie);
7905 MPASS(ret == 0);
7906 ret = intr_event_destroy(V_pf_swi_ie);
7907 MPASS(ret == 0);
7908
7909 pf_unload_vnet_purge();
7910
7911 pf_normalize_cleanup();
7912 PF_RULES_WLOCK();
7913 pfi_cleanup_vnet();
7914 PF_RULES_WUNLOCK();
7915 pfr_cleanup();
7916 pf_osfp_flush();
7917 pf_cleanup();
7918 if (IS_DEFAULT_VNET(curvnet))
7919 pf_mtag_cleanup();
7920
7921 pf_cleanup_tagset(&V_pf_tags);
7922 #ifdef ALTQ
7923 pf_cleanup_tagset(&V_pf_qids);
7924 #endif
7925 uma_zdestroy(V_pf_tag_z);
7926
7927 #ifdef PF_WANT_32_TO_64_COUNTER
7928 PF_RULES_WLOCK();
7929 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
7930
7931 MPASS(LIST_EMPTY(&V_pf_allkiflist));
7932 MPASS(V_pf_allkifcount == 0);
7933
7934 LIST_REMOVE(&V_pf_default_rule, allrulelist);
7935 V_pf_allrulecount--;
7936 LIST_REMOVE(V_pf_rulemarker, allrulelist);
7937
7938 MPASS(LIST_EMPTY(&V_pf_allrulelist));
7939 MPASS(V_pf_allrulecount == 0);
7940
7941 PF_RULES_WUNLOCK();
7942
7943 free(V_pf_kifmarker, PFI_MTYPE);
7944 free(V_pf_rulemarker, M_PFRULE);
7945 #endif
7946
7947 /* Free counters last as we updated them during shutdown. */
7948 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
7949 for (int i = 0; i < 2; i++) {
7950 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
7951 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
7952 }
7953 counter_u64_free(V_pf_default_rule.states_cur);
7954 counter_u64_free(V_pf_default_rule.states_tot);
7955 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
7956 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]);
7957 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
7958
7959 for (int i = 0; i < PFRES_MAX; i++)
7960 counter_u64_free(V_pf_status.counters[i]);
7961 for (int i = 0; i < KLCNT_MAX; i++)
7962 counter_u64_free(V_pf_status.lcounters[i]);
7963 for (int i = 0; i < FCNT_MAX; i++)
7964 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
7965 for (int i = 0; i < SCNT_MAX; i++)
7966 counter_u64_free(V_pf_status.scounters[i]);
7967 for (int i = 0; i < NCNT_MAX; i++)
7968 counter_u64_free(V_pf_status.ncounters[i]);
7969
7970 rm_destroy(&V_pf_rules_lock);
7971 sx_destroy(&V_pf_ioctl_lock);
7972 }
7973
7974 static void
pf_unload(void * dummy __unused)7975 pf_unload(void *dummy __unused)
7976 {
7977
7978 sx_xlock(&pf_end_lock);
7979 pf_end_threads = 1;
7980 while (pf_end_threads < 2) {
7981 wakeup_one(pf_purge_thread);
7982 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
7983 }
7984 sx_xunlock(&pf_end_lock);
7985
7986 pf_nl_unregister();
7987
7988 if (pf_dev != NULL)
7989 destroy_dev(pf_dev);
7990
7991 pfi_cleanup();
7992
7993 sx_destroy(&pf_end_lock);
7994 }
7995
7996 static void
vnet_pf_init(void * unused __unused)7997 vnet_pf_init(void *unused __unused)
7998 {
7999
8000 pf_load_vnet();
8001 }
8002 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
8003 vnet_pf_init, NULL);
8004
8005 static void
vnet_pf_uninit(const void * unused __unused)8006 vnet_pf_uninit(const void *unused __unused)
8007 {
8008
8009 pf_unload_vnet();
8010 }
8011 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
8012 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
8013 vnet_pf_uninit, NULL);
8014
8015 static int
pf_modevent(module_t mod,int type,void * data)8016 pf_modevent(module_t mod, int type, void *data)
8017 {
8018 int error = 0;
8019
8020 switch(type) {
8021 case MOD_LOAD:
8022 error = pf_load();
8023 pf_nl_register();
8024 break;
8025 case MOD_UNLOAD:
8026 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
8027 * the vnet_pf_uninit()s */
8028 break;
8029 default:
8030 error = EINVAL;
8031 break;
8032 }
8033
8034 return (error);
8035 }
8036
8037 static moduledata_t pf_mod = {
8038 "pf",
8039 pf_modevent,
8040 0
8041 };
8042
8043 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
8044 MODULE_DEPEND(pf, netlink, 1, 1, 1);
8045 MODULE_DEPEND(pf, crypto, 1, 1, 1);
8046 MODULE_VERSION(pf, PF_MODVER);
8047