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 PF_MD5_UPD(pfr, addr.v.tblname);
1358 break;
1359 case PF_ADDR_ADDRMASK:
1360 case PF_ADDR_RANGE:
1361 /* XXX ignore af? */
1362 PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1363 PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1364 break;
1365 case PF_ADDR_NONE:
1366 case PF_ADDR_NOROUTE:
1367 case PF_ADDR_URPFFAILED:
1368 /* These do not use any address data. */
1369 break;
1370 default:
1371 panic("Unknown address type %d", pfr->addr.type);
1372 }
1373
1374 PF_MD5_UPD(pfr, port[0]);
1375 PF_MD5_UPD(pfr, port[1]);
1376 PF_MD5_UPD(pfr, neg);
1377 PF_MD5_UPD(pfr, port_op);
1378 }
1379
1380 static void
pf_hash_pool(MD5_CTX * ctx,struct pf_kpool * pool)1381 pf_hash_pool(MD5_CTX *ctx, struct pf_kpool *pool)
1382 {
1383 uint16_t x;
1384 int y;
1385
1386 if (pool->cur) {
1387 PF_MD5_UPD(pool, cur->addr);
1388 PF_MD5_UPD_STR(pool, cur->ifname);
1389 PF_MD5_UPD(pool, cur->af);
1390 }
1391 PF_MD5_UPD(pool, key);
1392 PF_MD5_UPD(pool, counter);
1393
1394 PF_MD5_UPD(pool, mape.offset);
1395 PF_MD5_UPD(pool, mape.psidlen);
1396 PF_MD5_UPD_HTONS(pool, mape.psid, x);
1397 PF_MD5_UPD_HTONL(pool, tblidx, y);
1398 PF_MD5_UPD_HTONS(pool, proxy_port[0], x);
1399 PF_MD5_UPD_HTONS(pool, proxy_port[1], x);
1400 PF_MD5_UPD(pool, opts);
1401 PF_MD5_UPD(pool, ipv6_nexthop_af);
1402 }
1403
1404 static void
pf_hash_rule_rolling(MD5_CTX * ctx,struct pf_krule * rule)1405 pf_hash_rule_rolling(MD5_CTX *ctx, struct pf_krule *rule)
1406 {
1407 u_int16_t x;
1408 u_int32_t y;
1409
1410 pf_hash_rule_addr(ctx, &rule->src);
1411 pf_hash_rule_addr(ctx, &rule->dst);
1412 for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1413 PF_MD5_UPD_STR(rule, label[i]);
1414 PF_MD5_UPD_HTONL(rule, ridentifier, y);
1415 PF_MD5_UPD_STR(rule, ifname);
1416 PF_MD5_UPD_STR(rule, rcv_ifname);
1417 PF_MD5_UPD_STR(rule, qname);
1418 PF_MD5_UPD_STR(rule, pqname);
1419 PF_MD5_UPD_STR(rule, tagname);
1420 PF_MD5_UPD_STR(rule, match_tagname);
1421
1422 PF_MD5_UPD_STR(rule, overload_tblname);
1423
1424 pf_hash_pool(ctx, &rule->nat);
1425 pf_hash_pool(ctx, &rule->rdr);
1426 pf_hash_pool(ctx, &rule->route);
1427 PF_MD5_UPD_HTONL(rule, pktrate.limit, y);
1428 PF_MD5_UPD_HTONL(rule, pktrate.seconds, y);
1429
1430 PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1431
1432 PF_MD5_UPD_HTONL(rule, rtableid, y);
1433 for (int i = 0; i < PFTM_MAX; i++)
1434 PF_MD5_UPD_HTONL(rule, timeout[i], y);
1435 PF_MD5_UPD_HTONL(rule, max_states, y);
1436 PF_MD5_UPD_HTONL(rule, max_src_nodes, y);
1437 PF_MD5_UPD_HTONL(rule, max_src_states, y);
1438 PF_MD5_UPD_HTONL(rule, max_src_conn, y);
1439 PF_MD5_UPD_HTONL(rule, max_src_conn_rate.limit, y);
1440 PF_MD5_UPD_HTONL(rule, max_src_conn_rate.seconds, y);
1441 PF_MD5_UPD_HTONS(rule, max_pkt_size, y);
1442 PF_MD5_UPD_HTONS(rule, qid, x);
1443 PF_MD5_UPD_HTONS(rule, pqid, x);
1444 PF_MD5_UPD_HTONS(rule, dnpipe, x);
1445 PF_MD5_UPD_HTONS(rule, dnrpipe, x);
1446 PF_MD5_UPD_HTONL(rule, free_flags, y);
1447 PF_MD5_UPD_HTONL(rule, prob, y);
1448
1449 PF_MD5_UPD_HTONS(rule, return_icmp, x);
1450 PF_MD5_UPD_HTONS(rule, return_icmp6, x);
1451 PF_MD5_UPD_HTONS(rule, max_mss, x);
1452 PF_MD5_UPD_HTONS(rule, tag, x); /* dup? */
1453 PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1454 PF_MD5_UPD_HTONS(rule, scrub_flags, x);
1455
1456 PF_MD5_UPD(rule, uid.op);
1457 PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1458 PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1459 PF_MD5_UPD(rule, gid.op);
1460 PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1461 PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1462
1463 PF_MD5_UPD_HTONL(rule, rule_flag, y);
1464 PF_MD5_UPD_HTONL(rule, rule_ref, y);
1465 PF_MD5_UPD(rule, action);
1466 PF_MD5_UPD(rule, direction);
1467 PF_MD5_UPD(rule, log);
1468 PF_MD5_UPD(rule, logif);
1469 PF_MD5_UPD(rule, quick);
1470 PF_MD5_UPD(rule, ifnot);
1471 PF_MD5_UPD(rule, match_tag_not);
1472 PF_MD5_UPD(rule, natpass);
1473
1474 PF_MD5_UPD(rule, keep_state);
1475 PF_MD5_UPD(rule, af);
1476 PF_MD5_UPD(rule, proto);
1477 PF_MD5_UPD_HTONS(rule, type, x);
1478 PF_MD5_UPD_HTONS(rule, code, x);
1479 PF_MD5_UPD(rule, flags);
1480 PF_MD5_UPD(rule, flagset);
1481 PF_MD5_UPD(rule, min_ttl);
1482 PF_MD5_UPD(rule, allow_opts);
1483 PF_MD5_UPD(rule, rt);
1484 PF_MD5_UPD(rule, return_ttl);
1485 PF_MD5_UPD(rule, tos);
1486 PF_MD5_UPD(rule, set_tos);
1487 PF_MD5_UPD(rule, anchor_relative);
1488 PF_MD5_UPD(rule, anchor_wildcard);
1489
1490 PF_MD5_UPD(rule, flush);
1491 PF_MD5_UPD(rule, prio);
1492 PF_MD5_UPD(rule, set_prio[0]);
1493 PF_MD5_UPD(rule, set_prio[1]);
1494 PF_MD5_UPD(rule, naf);
1495 PF_MD5_UPD(rule, rcvifnot);
1496 PF_MD5_UPD(rule, statelim.id);
1497 PF_MD5_UPD_HTONL(rule, statelim.limiter_action, y);
1498 PF_MD5_UPD(rule, sourcelim.id);
1499 PF_MD5_UPD_HTONL(rule, sourcelim.limiter_action, y);
1500
1501 PF_MD5_UPD(rule, divert.addr);
1502 PF_MD5_UPD_HTONS(rule, divert.port, x);
1503
1504 if (rule->anchor != NULL)
1505 PF_MD5_UPD_STR(rule, anchor->path);
1506 }
1507
1508 static void
pf_hash_rule(struct pf_krule * rule)1509 pf_hash_rule(struct pf_krule *rule)
1510 {
1511 MD5_CTX ctx;
1512
1513 MD5Init(&ctx);
1514 pf_hash_rule_rolling(&ctx, rule);
1515 MD5Final(rule->md5sum, &ctx);
1516 }
1517
1518 static int
pf_krule_compare(struct pf_krule * a,struct pf_krule * b)1519 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1520 {
1521
1522 return (memcmp(a->md5sum, b->md5sum, PF_MD5_DIGEST_LENGTH));
1523 }
1524
1525 static int
pf_commit_rules(u_int32_t ticket,int rs_num,char * anchor)1526 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1527 {
1528 struct pf_kruleset *rs;
1529 struct pf_krule *rule, *old_rule;
1530 struct pf_krulequeue *old_rules;
1531 struct pf_krule_global *old_tree;
1532 int error;
1533 u_int32_t old_rcount;
1534 bool is_main_ruleset = anchor[0] == '\0';
1535
1536 PF_RULES_WASSERT();
1537
1538 if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1539 return (EINVAL);
1540 rs = pf_find_kruleset(anchor);
1541 if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1542 ticket != rs->rules[rs_num].inactive.ticket)
1543 return (EBUSY);
1544
1545 /* Calculate checksum for the main ruleset */
1546 if (rs == &pf_main_ruleset) {
1547 error = pf_sourcelim_check();
1548 if (error != 0)
1549 return (error);
1550 error = pf_setup_pfsync_matching(rs);
1551 if (error != 0)
1552 return (error);
1553 }
1554
1555 /* Swap rules, keep the old. */
1556 old_rules = rs->rules[rs_num].active.ptr;
1557 old_rcount = rs->rules[rs_num].active.rcount;
1558 old_tree = rs->rules[rs_num].active.tree;
1559
1560 rs->rules[rs_num].active.ptr =
1561 rs->rules[rs_num].inactive.ptr;
1562 rs->rules[rs_num].active.tree =
1563 rs->rules[rs_num].inactive.tree;
1564 rs->rules[rs_num].active.rcount =
1565 rs->rules[rs_num].inactive.rcount;
1566
1567 /* Attempt to preserve counter information. */
1568 if (V_pf_status.keep_counters && old_tree != NULL) {
1569 TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1570 entries) {
1571 old_rule = RB_FIND(pf_krule_global, old_tree, rule);
1572 if (old_rule == NULL) {
1573 continue;
1574 }
1575 pf_counter_u64_critical_enter();
1576 pf_counter_u64_rollup_protected(&rule->evaluations,
1577 pf_counter_u64_fetch(&old_rule->evaluations));
1578 pf_counter_u64_rollup_protected(&rule->packets[0],
1579 pf_counter_u64_fetch(&old_rule->packets[0]));
1580 pf_counter_u64_rollup_protected(&rule->packets[1],
1581 pf_counter_u64_fetch(&old_rule->packets[1]));
1582 pf_counter_u64_rollup_protected(&rule->bytes[0],
1583 pf_counter_u64_fetch(&old_rule->bytes[0]));
1584 pf_counter_u64_rollup_protected(&rule->bytes[1],
1585 pf_counter_u64_fetch(&old_rule->bytes[1]));
1586 pf_counter_u64_critical_exit();
1587 }
1588 }
1589
1590 rs->rules[rs_num].inactive.ptr = old_rules;
1591 rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */
1592 rs->rules[rs_num].inactive.rcount = old_rcount;
1593
1594 rs->rules[rs_num].active.ticket =
1595 rs->rules[rs_num].inactive.ticket;
1596 pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1597
1598 /* Purge the old rule list. */
1599 PF_UNLNKDRULES_LOCK();
1600 while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1601 pf_unlink_rule_locked(old_rules, rule);
1602 PF_UNLNKDRULES_UNLOCK();
1603 rs->rules[rs_num].inactive.rcount = 0;
1604 rs->rules[rs_num].inactive.open = 0;
1605 pf_remove_if_empty_kruleset(rs);
1606 pf_rule_tree_free(old_tree);
1607
1608 /* statelim/sourcelim/queue defs only in the main ruleset */
1609 if (! is_main_ruleset || rs_num != PF_RULESET_FILTER)
1610 return (0);
1611
1612 pf_statelim_commit();
1613 pf_sourcelim_commit();
1614
1615 return (0);
1616 }
1617
1618 static int
pf_setup_pfsync_matching(struct pf_kruleset * rs)1619 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1620 {
1621 MD5_CTX ctx;
1622 struct pf_krule *rule;
1623 int rs_cnt;
1624 u_int8_t digest[PF_MD5_DIGEST_LENGTH];
1625
1626 MD5Init(&ctx);
1627 for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1628 /* XXX PF_RULESET_SCRUB as well? */
1629 if (rs_cnt == PF_RULESET_SCRUB)
1630 continue;
1631
1632 if (rs->rules[rs_cnt].inactive.rcount) {
1633 TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1634 entries) {
1635 pf_hash_rule_rolling(&ctx, rule);
1636 }
1637 }
1638 }
1639
1640 MD5Final(digest, &ctx);
1641 memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1642 return (0);
1643 }
1644
1645 static int
pf_eth_addr_setup(struct pf_keth_ruleset * ruleset,struct pf_addr_wrap * addr)1646 pf_eth_addr_setup(struct pf_keth_ruleset *ruleset, struct pf_addr_wrap *addr)
1647 {
1648 int error = 0;
1649
1650 switch (addr->type) {
1651 case PF_ADDR_TABLE:
1652 addr->p.tbl = pfr_eth_attach_table(ruleset, addr->v.tblname);
1653 if (addr->p.tbl == NULL)
1654 error = ENOMEM;
1655 break;
1656 default:
1657 error = EINVAL;
1658 }
1659
1660 return (error);
1661 }
1662
1663 static int
pf_addr_setup(struct pf_kruleset * ruleset,struct pf_addr_wrap * addr,sa_family_t af)1664 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1665 sa_family_t af)
1666 {
1667 int error = 0;
1668
1669 switch (addr->type) {
1670 case PF_ADDR_TABLE:
1671 addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1672 if (addr->p.tbl == NULL)
1673 error = ENOMEM;
1674 break;
1675 case PF_ADDR_DYNIFTL:
1676 error = pfi_dynaddr_setup(addr, af);
1677 break;
1678 }
1679
1680 return (error);
1681 }
1682
1683 void
pf_addr_copyout(struct pf_addr_wrap * addr)1684 pf_addr_copyout(struct pf_addr_wrap *addr)
1685 {
1686
1687 switch (addr->type) {
1688 case PF_ADDR_DYNIFTL:
1689 pfi_dynaddr_copyout(addr);
1690 break;
1691 case PF_ADDR_TABLE:
1692 pf_tbladdr_copyout(addr);
1693 break;
1694 }
1695 }
1696
1697 int
pf_statelim_add(const struct pfioc_statelim * ioc)1698 pf_statelim_add(const struct pfioc_statelim *ioc)
1699 {
1700 struct pf_statelim *pfstlim;
1701 int error;
1702 size_t namelen;
1703
1704 if (ioc->id < PF_STATELIM_ID_MIN ||
1705 ioc->id > PF_STATELIM_ID_MAX)
1706 return (EINVAL);
1707
1708 if (ioc->limit < PF_STATELIM_LIMIT_MIN ||
1709 ioc->limit > PF_STATELIM_LIMIT_MAX)
1710 return (EINVAL);
1711
1712 if ((ioc->rate.limit == 0) != (ioc->rate.seconds == 0))
1713 return (EINVAL);
1714
1715 namelen = strnlen(ioc->name, sizeof(ioc->name));
1716 /* is the name from userland nul terminated? */
1717 if (namelen == sizeof(ioc->name))
1718 return (EINVAL);
1719
1720 pfstlim = malloc(sizeof(*pfstlim), M_PF_STATE_LIM, M_WAITOK | M_ZERO);
1721 if (pfstlim == NULL)
1722 return (ENOMEM);
1723
1724 pfstlim->pfstlim_id = ioc->id;
1725 if (strlcpy(pfstlim->pfstlim_nm, ioc->name,
1726 sizeof(pfstlim->pfstlim_nm)) >= sizeof(pfstlim->pfstlim_nm)) {
1727 error = EINVAL;
1728 goto free;
1729 }
1730 pfstlim->pfstlim_limit = ioc->limit;
1731 pfstlim->pfstlim_rate.limit = ioc->rate.limit;
1732 pfstlim->pfstlim_rate.seconds = ioc->rate.seconds;
1733
1734 if (pfstlim->pfstlim_rate.limit) {
1735 uint64_t bucket = SEC_TO_NSEC(pfstlim->pfstlim_rate.seconds);
1736 struct timespec ts;
1737
1738 getnanouptime(&ts);
1739
1740 pfstlim->pfstlim_rate_ts = SEC_TO_NSEC(ts.tv_sec) + ts.tv_nsec -
1741 bucket;
1742 pfstlim->pfstlim_rate_token = bucket /
1743 pfstlim->pfstlim_rate.limit;
1744 pfstlim->pfstlim_rate_bucket = bucket;
1745 }
1746
1747 TAILQ_INIT(&pfstlim->pfstlim_states);
1748 mtx_init(&pfstlim->pfstlim_lock, "pf state limit", NULL, MTX_DEF);
1749
1750 PF_RULES_WLOCK();
1751 if (ioc->ticket != pf_main_ruleset.rules[PF_RULESET_FILTER].inactive.ticket) {
1752 error = EBUSY;
1753 goto unlock;
1754 }
1755
1756 if (RB_INSERT(pf_statelim_id_tree, &V_pf_statelim_id_tree_inactive,
1757 pfstlim) != NULL) {
1758 error = EBUSY;
1759 goto unlock;
1760 }
1761
1762 if (RB_INSERT(pf_statelim_nm_tree, &V_pf_statelim_nm_tree_inactive,
1763 pfstlim) != NULL) {
1764 RB_REMOVE(pf_statelim_id_tree, &V_pf_statelim_id_tree_inactive,
1765 pfstlim);
1766 error = EBUSY;
1767 goto unlock;
1768 }
1769
1770 TAILQ_INSERT_HEAD(&V_pf_statelim_list_inactive, pfstlim, pfstlim_list);
1771
1772 PF_RULES_WUNLOCK();
1773
1774 return (0);
1775
1776 unlock:
1777 PF_RULES_WUNLOCK();
1778
1779 free:
1780 free(pfstlim, M_PF_STATE_LIM);
1781
1782 return (error);
1783 }
1784
1785 static void
pf_statelim_unlink(struct pf_statelim * pfstlim,struct pf_state_link_list * garbage)1786 pf_statelim_unlink(struct pf_statelim *pfstlim,
1787 struct pf_state_link_list *garbage)
1788 {
1789 struct pf_state_link *pfl;
1790
1791
1792 /* unwire the links */
1793 TAILQ_FOREACH(pfl, &pfstlim->pfstlim_states, pfl_link) {
1794 struct pf_kstate *s = pfl->pfl_state;
1795
1796 /* if !rmst */
1797 PF_STATE_LOCK(s);
1798 s->statelim = 0;
1799 SLIST_REMOVE(&s->linkage, pfl, pf_state_link, pfl_linkage);
1800 PF_STATE_UNLOCK(s);
1801 }
1802
1803 /* take the list away */
1804 TAILQ_CONCAT(garbage, &pfstlim->pfstlim_states, pfl_link);
1805 pfstlim->pfstlim_inuse = 0;
1806 }
1807
1808 void
pf_statelim_commit(void)1809 pf_statelim_commit(void)
1810 {
1811 struct pf_statelim *pfstlim, *npfstlim, *opfstlim;
1812 struct pf_statelim_list l = TAILQ_HEAD_INITIALIZER(l);
1813 struct pf_state_link_list garbage = TAILQ_HEAD_INITIALIZER(garbage);
1814 struct pf_state_link *pfl, *npfl;
1815
1816 PF_RULES_WASSERT();
1817
1818 /* merge the new statelims into the current set */
1819
1820 /* start with an empty active list */
1821 TAILQ_CONCAT(&l, &V_pf_statelim_list_active, pfstlim_list);
1822
1823 /* beware, the inactive bits gets messed up here */
1824
1825 /* try putting pending statelims into the active tree */
1826 TAILQ_FOREACH_SAFE(pfstlim, &V_pf_statelim_list_inactive, pfstlim_list,
1827 npfstlim) {
1828 opfstlim = RB_INSERT(pf_statelim_id_tree,
1829 &V_pf_statelim_id_tree_active, pfstlim);
1830 if (opfstlim != NULL) {
1831 /* this statelim already exists, merge */
1832 opfstlim->pfstlim_limit = pfstlim->pfstlim_limit;
1833 opfstlim->pfstlim_rate.limit =
1834 pfstlim->pfstlim_rate.limit;
1835 opfstlim->pfstlim_rate.seconds =
1836 pfstlim->pfstlim_rate.seconds;
1837
1838 opfstlim->pfstlim_rate_ts = pfstlim->pfstlim_rate_ts;
1839 opfstlim->pfstlim_rate_token =
1840 pfstlim->pfstlim_rate_token;
1841 opfstlim->pfstlim_rate_bucket =
1842 pfstlim->pfstlim_rate_bucket;
1843
1844 memcpy(opfstlim->pfstlim_nm, pfstlim->pfstlim_nm,
1845 sizeof(opfstlim->pfstlim_nm));
1846
1847 /* use the existing statelim instead */
1848 free(pfstlim, M_PF_STATE_LIM);
1849 TAILQ_REMOVE(&l, opfstlim, pfstlim_list);
1850 pfstlim = opfstlim;
1851 }
1852
1853 TAILQ_INSERT_TAIL(&V_pf_statelim_list_active, pfstlim,
1854 pfstlim_list);
1855 }
1856
1857 /* clean up the now unused statelims from the old set */
1858 TAILQ_FOREACH_SAFE(pfstlim, &l, pfstlim_list, npfstlim) {
1859 pf_statelim_unlink(pfstlim, &garbage);
1860
1861 RB_REMOVE(pf_statelim_id_tree, &V_pf_statelim_id_tree_active,
1862 pfstlim);
1863
1864 free(pfstlim, M_PF_STATE_LIM);
1865 }
1866
1867 /* fix up the inactive tree */
1868 RB_INIT(&V_pf_statelim_id_tree_inactive);
1869 RB_INIT(&V_pf_statelim_nm_tree_inactive);
1870 TAILQ_INIT(&V_pf_statelim_list_inactive);
1871
1872 TAILQ_FOREACH_SAFE(pfl, &garbage, pfl_link, npfl)
1873 free(pfl, M_PF_STATE_LINK);
1874 }
1875
1876 static void
pf_sourcelim_unlink(struct pf_sourcelim * pfsrlim,struct pf_state_link_list * garbage)1877 pf_sourcelim_unlink(struct pf_sourcelim *pfsrlim,
1878 struct pf_state_link_list *garbage)
1879 {
1880 extern struct pf_source_list pf_source_gc;
1881 struct pf_source *pfsr;
1882 struct pf_state_link *pfl;
1883
1884 PF_RULES_WASSERT();
1885
1886 while ((pfsr = RB_ROOT(&pfsrlim->pfsrlim_sources)) != NULL) {
1887 RB_REMOVE(pf_source_tree, &pfsrlim->pfsrlim_sources, pfsr);
1888 RB_REMOVE(pf_source_ioc_tree, &pfsrlim->pfsrlim_ioc_sources,
1889 pfsr);
1890 if (pfsr->pfsr_inuse == 0)
1891 TAILQ_REMOVE(&pf_source_gc, pfsr, pfsr_empty_gc);
1892
1893 /* unwire the links */
1894 TAILQ_FOREACH(pfl, &pfsr->pfsr_states, pfl_link) {
1895 struct pf_kstate *s = pfl->pfl_state;
1896
1897 PF_STATE_LOCK(s);
1898 /* if !rmst */
1899 s->sourcelim = 0;
1900 SLIST_REMOVE(&s->linkage, pfl, pf_state_link,
1901 pfl_linkage);
1902 PF_STATE_UNLOCK(s);
1903 }
1904
1905 /* take the list away */
1906 TAILQ_CONCAT(garbage, &pfsr->pfsr_states, pfl_link);
1907
1908 free(pfsr, M_PF_SOURCE_LIM);
1909 }
1910 }
1911
1912 int
pf_sourcelim_check(void)1913 pf_sourcelim_check(void)
1914 {
1915 struct pf_sourcelim *pfsrlim, *npfsrlim;
1916
1917 PF_RULES_WASSERT();
1918
1919 /* check if we can merge */
1920
1921 TAILQ_FOREACH(pfsrlim, &V_pf_sourcelim_list_inactive, pfsrlim_list) {
1922 npfsrlim = RB_FIND(pf_sourcelim_id_tree,
1923 &V_pf_sourcelim_id_tree_active, pfsrlim);
1924
1925 /* new config, no conflict */
1926 if (npfsrlim == NULL)
1927 continue;
1928
1929 /* nothing is tracked at the moment, no conflict */
1930 if (RB_EMPTY(&npfsrlim->pfsrlim_sources))
1931 continue;
1932
1933 if (strcmp(npfsrlim->pfsrlim_overload.name,
1934 pfsrlim->pfsrlim_overload.name) != 0)
1935 return (EBUSY);
1936
1937 /*
1938 * we should allow the prefixlens to get shorter
1939 * and merge pf_source entries.
1940 */
1941
1942 if ((npfsrlim->pfsrlim_ipv4_prefix !=
1943 pfsrlim->pfsrlim_ipv4_prefix) ||
1944 (npfsrlim->pfsrlim_ipv6_prefix !=
1945 pfsrlim->pfsrlim_ipv6_prefix))
1946 return (EBUSY);
1947 }
1948
1949 return (0);
1950 }
1951
1952 void
pf_sourcelim_commit(void)1953 pf_sourcelim_commit(void)
1954 {
1955 struct pf_sourcelim *pfsrlim, *npfsrlim, *opfsrlim;
1956 struct pf_sourcelim_list l = TAILQ_HEAD_INITIALIZER(l);
1957 struct pf_state_link_list garbage = TAILQ_HEAD_INITIALIZER(garbage);
1958 struct pf_state_link *pfl, *npfl;
1959
1960 PF_RULES_WASSERT();
1961
1962 /* merge the new sourcelims into the current set */
1963
1964 /* start with an empty active list */
1965 TAILQ_CONCAT(&l, &V_pf_sourcelim_list_active, pfsrlim_list);
1966
1967 /* beware, the inactive bits gets messed up here */
1968
1969 /* try putting pending sourcelims into the active tree */
1970 TAILQ_FOREACH_SAFE(pfsrlim, &V_pf_sourcelim_list_inactive, pfsrlim_list,
1971 npfsrlim) {
1972 opfsrlim = RB_INSERT(pf_sourcelim_id_tree,
1973 &V_pf_sourcelim_id_tree_active, pfsrlim);
1974 if (opfsrlim != NULL) {
1975 /* this sourcelim already exists, merge */
1976 opfsrlim->pfsrlim_entries = pfsrlim->pfsrlim_entries;
1977 opfsrlim->pfsrlim_limit = pfsrlim->pfsrlim_limit;
1978 opfsrlim->pfsrlim_ipv4_prefix =
1979 pfsrlim->pfsrlim_ipv4_prefix;
1980 opfsrlim->pfsrlim_ipv6_prefix =
1981 pfsrlim->pfsrlim_ipv6_prefix;
1982 opfsrlim->pfsrlim_rate.limit =
1983 pfsrlim->pfsrlim_rate.limit;
1984 opfsrlim->pfsrlim_rate.seconds =
1985 pfsrlim->pfsrlim_rate.seconds;
1986
1987 opfsrlim->pfsrlim_ipv4_mask =
1988 pfsrlim->pfsrlim_ipv4_mask;
1989 opfsrlim->pfsrlim_ipv6_mask =
1990 pfsrlim->pfsrlim_ipv6_mask;
1991
1992 /* keep the existing pfstlim_rate_ts */
1993
1994 opfsrlim->pfsrlim_rate_token =
1995 pfsrlim->pfsrlim_rate_token;
1996 opfsrlim->pfsrlim_rate_bucket =
1997 pfsrlim->pfsrlim_rate_bucket;
1998
1999 if (opfsrlim->pfsrlim_overload.table != NULL) {
2000 pfr_detach_table(
2001 opfsrlim->pfsrlim_overload.table);
2002 }
2003
2004 strlcpy(opfsrlim->pfsrlim_overload.name,
2005 pfsrlim->pfsrlim_overload.name,
2006 sizeof(opfsrlim->pfsrlim_overload.name));
2007 opfsrlim->pfsrlim_overload.hwm =
2008 pfsrlim->pfsrlim_overload.hwm;
2009 opfsrlim->pfsrlim_overload.lwm =
2010 pfsrlim->pfsrlim_overload.lwm;
2011 opfsrlim->pfsrlim_overload.table =
2012 pfsrlim->pfsrlim_overload.table,
2013
2014 memcpy(opfsrlim->pfsrlim_nm, pfsrlim->pfsrlim_nm,
2015 sizeof(opfsrlim->pfsrlim_nm));
2016
2017 /* use the existing sourcelim instead */
2018 free(pfsrlim, M_PF_SOURCE_LIM);
2019 TAILQ_REMOVE(&l, opfsrlim, pfsrlim_list);
2020 pfsrlim = opfsrlim;
2021 }
2022
2023 TAILQ_INSERT_TAIL(&V_pf_sourcelim_list_active, pfsrlim,
2024 pfsrlim_list);
2025 }
2026
2027 /* clean up the now unused sourcelims from the old set */
2028 TAILQ_FOREACH_SAFE(pfsrlim, &l, pfsrlim_list, npfsrlim) {
2029 pf_sourcelim_unlink(pfsrlim, &garbage);
2030
2031 RB_REMOVE(pf_sourcelim_id_tree, &V_pf_sourcelim_id_tree_active,
2032 pfsrlim);
2033
2034 if (pfsrlim->pfsrlim_overload.table != NULL)
2035 pfr_detach_table(pfsrlim->pfsrlim_overload.table);
2036
2037 free(pfsrlim, M_PF_SOURCE_LIM);
2038 }
2039
2040 /* fix up the inactive tree */
2041 RB_INIT(&V_pf_sourcelim_id_tree_inactive);
2042 RB_INIT(&V_pf_sourcelim_nm_tree_inactive);
2043 TAILQ_INIT(&V_pf_sourcelim_list_inactive);
2044
2045 TAILQ_FOREACH_SAFE(pfl, &garbage, pfl_link, npfl)
2046 free(pfl, M_PF_STATE_LINK);
2047 }
2048
2049 void
pf_statelim_rollback(void)2050 pf_statelim_rollback(void)
2051 {
2052 struct pf_statelim *pfstlim, *npfstlim;
2053
2054 PF_RULES_WASSERT();
2055
2056 TAILQ_FOREACH_SAFE(pfstlim, &V_pf_statelim_list_inactive, pfstlim_list,
2057 npfstlim)
2058 free(pfstlim, M_PF_STATE_LIM);
2059
2060 TAILQ_INIT(&V_pf_statelim_list_inactive);
2061 RB_INIT(&V_pf_statelim_id_tree_inactive);
2062 RB_INIT(&V_pf_statelim_nm_tree_inactive);
2063 }
2064
2065 struct pf_statelim *
pf_statelim_rb_find(struct pf_statelim_id_tree * tree,struct pf_statelim * key)2066 pf_statelim_rb_find(struct pf_statelim_id_tree *tree, struct pf_statelim *key)
2067 {
2068 PF_RULES_ASSERT();
2069
2070 return (RB_FIND(pf_statelim_id_tree, tree, key));
2071 }
2072
2073 struct pf_statelim *
pf_statelim_rb_nfind(struct pf_statelim_id_tree * tree,struct pf_statelim * key)2074 pf_statelim_rb_nfind(struct pf_statelim_id_tree *tree, struct pf_statelim *key)
2075 {
2076 PF_RULES_ASSERT();
2077
2078 return (RB_NFIND(pf_statelim_id_tree, tree, key));
2079 }
2080
2081 int
pf_statelim_get(struct pfioc_statelim * ioc,struct pf_statelim * (* rbt_op)(struct pf_statelim_id_tree *,struct pf_statelim *))2082 pf_statelim_get(struct pfioc_statelim *ioc,
2083 struct pf_statelim *(*rbt_op)(struct pf_statelim_id_tree *,
2084 struct pf_statelim *))
2085 {
2086 struct pf_statelim key = { .pfstlim_id = ioc->id };
2087 struct pf_statelim *pfstlim;
2088 int error = 0;
2089 PF_RULES_RLOCK_TRACKER;
2090
2091 PF_RULES_RLOCK();
2092
2093 pfstlim = (*rbt_op)(&V_pf_statelim_id_tree_active, &key);
2094 if (pfstlim == NULL) {
2095 error = ENOENT;
2096 goto unlock;
2097 }
2098
2099 ioc->id = pfstlim->pfstlim_id;
2100 ioc->limit = pfstlim->pfstlim_limit;
2101 ioc->rate.limit = pfstlim->pfstlim_rate.limit;
2102 ioc->rate.seconds = pfstlim->pfstlim_rate.seconds;
2103 CTASSERT(sizeof(ioc->name) == sizeof(pfstlim->pfstlim_nm));
2104 memcpy(ioc->name, pfstlim->pfstlim_nm, sizeof(ioc->name));
2105
2106 ioc->inuse = pfstlim->pfstlim_inuse;
2107 ioc->admitted = pfstlim->pfstlim_counters.admitted;
2108 ioc->hardlimited = pfstlim->pfstlim_counters.hardlimited;
2109 ioc->ratelimited = pfstlim->pfstlim_counters.ratelimited;
2110
2111 unlock:
2112 PF_RULES_RUNLOCK();
2113
2114 return (error);
2115 }
2116
2117 int
pf_sourcelim_add(const struct pfioc_sourcelim * ioc)2118 pf_sourcelim_add(const struct pfioc_sourcelim *ioc)
2119 {
2120 struct pf_sourcelim *pfsrlim;
2121 int error;
2122 size_t namelen, tablelen;
2123 unsigned int prefix;
2124 size_t i;
2125
2126 if (ioc->id < PF_SOURCELIM_ID_MIN ||
2127 ioc->id > PF_SOURCELIM_ID_MAX)
2128 return (EINVAL);
2129
2130 if (ioc->entries < 1)
2131 return (EINVAL);
2132
2133 if (ioc->limit < 1)
2134 return (EINVAL);
2135
2136 if ((ioc->rate.limit == 0) != (ioc->rate.seconds == 0))
2137 return (EINVAL);
2138
2139 if (ioc->inet_prefix > 32)
2140 return (EINVAL);
2141 if (ioc->inet6_prefix > 128)
2142 return (EINVAL);
2143
2144 namelen = strnlen(ioc->name, sizeof(ioc->name));
2145 /* is the name from userland nul terminated? */
2146 if (namelen == sizeof(ioc->name))
2147 return (EINVAL);
2148
2149 tablelen = strnlen(ioc->overload_tblname,
2150 sizeof(ioc->overload_tblname));
2151 /* is the name from userland nul terminated? */
2152 if (tablelen == sizeof(ioc->overload_tblname))
2153 return (EINVAL);
2154 if (tablelen != 0) {
2155 if (ioc->overload_hwm == 0)
2156 return (EINVAL);
2157
2158 if (ioc->overload_hwm < ioc->overload_lwm)
2159 return (EINVAL);
2160 }
2161
2162 pfsrlim = malloc(sizeof(*pfsrlim), M_PF_SOURCE_LIM, M_WAITOK | M_ZERO);
2163 if (pfsrlim == NULL)
2164 return (ENOMEM);
2165
2166 pfsrlim->pfsrlim_id = ioc->id;
2167 pfsrlim->pfsrlim_entries = ioc->entries;
2168 pfsrlim->pfsrlim_limit = ioc->limit;
2169 pfsrlim->pfsrlim_ipv4_prefix = ioc->inet_prefix;
2170 pfsrlim->pfsrlim_ipv6_prefix = ioc->inet6_prefix;
2171 pfsrlim->pfsrlim_rate.limit = ioc->rate.limit;
2172 pfsrlim->pfsrlim_rate.seconds = ioc->rate.seconds;
2173 if (strlcpy(pfsrlim->pfsrlim_overload.name, ioc->overload_tblname,
2174 sizeof(pfsrlim->pfsrlim_overload.name)) >=
2175 sizeof(pfsrlim->pfsrlim_overload.name)) {
2176 error = EINVAL;
2177 goto free;
2178 }
2179 pfsrlim->pfsrlim_overload.hwm = ioc->overload_hwm;
2180 pfsrlim->pfsrlim_overload.lwm = ioc->overload_lwm;
2181 if (strlcpy(pfsrlim->pfsrlim_nm, ioc->name,
2182 sizeof(pfsrlim->pfsrlim_nm)) >= sizeof(pfsrlim->pfsrlim_nm)) {
2183 error = EINVAL;
2184 goto free;
2185 }
2186
2187 if (pfsrlim->pfsrlim_rate.limit) {
2188 uint64_t bucket = pfsrlim->pfsrlim_rate.seconds * 1000000000ULL;
2189
2190 pfsrlim->pfsrlim_rate_token = bucket /
2191 pfsrlim->pfsrlim_rate.limit;
2192 pfsrlim->pfsrlim_rate_bucket = bucket;
2193 }
2194
2195 pfsrlim->pfsrlim_ipv4_mask.v4.s_addr = htonl(
2196 0xffffffff << (32 - pfsrlim->pfsrlim_ipv4_prefix));
2197
2198 prefix = pfsrlim->pfsrlim_ipv6_prefix;
2199 for (i = 0; i < nitems(pfsrlim->pfsrlim_ipv6_mask.addr32); i++) {
2200 if (prefix == 0) {
2201 /* the memory is already zeroed */
2202 break;
2203 }
2204 if (prefix < 32) {
2205 pfsrlim->pfsrlim_ipv6_mask.addr32[i] = htonl(
2206 0xffffffff << (32 - prefix));
2207 break;
2208 }
2209
2210 pfsrlim->pfsrlim_ipv6_mask.addr32[i] = htonl(0xffffffff);
2211 prefix -= 32;
2212 }
2213
2214 RB_INIT(&pfsrlim->pfsrlim_sources);
2215 mtx_init(&pfsrlim->pfsrlim_lock, "pf source limit", NULL, MTX_DEF);
2216
2217 PF_RULES_WLOCK();
2218 if (ioc->ticket != pf_main_ruleset.rules[PF_RULESET_FILTER].inactive.ticket) {
2219 error = EBUSY;
2220 goto unlock;
2221 }
2222
2223 if (pfsrlim->pfsrlim_overload.name[0] != '\0') {
2224 pfsrlim->pfsrlim_overload.table = pfr_attach_table(
2225 &pf_main_ruleset, pfsrlim->pfsrlim_overload.name);
2226 if (pfsrlim->pfsrlim_overload.table == NULL) {
2227 error = EINVAL;
2228 goto unlock;
2229 }
2230 }
2231
2232 if (RB_INSERT(pf_sourcelim_id_tree, &V_pf_sourcelim_id_tree_inactive,
2233 pfsrlim) != NULL) {
2234 error = EBUSY;
2235 goto unlock;
2236 }
2237
2238 if (RB_INSERT(pf_sourcelim_nm_tree, &V_pf_sourcelim_nm_tree_inactive,
2239 pfsrlim) != NULL) {
2240 RB_REMOVE(pf_sourcelim_nm_tree, &V_pf_sourcelim_nm_tree_inactive,
2241 pfsrlim);
2242 error = EBUSY;
2243 goto unlock;
2244 }
2245
2246 TAILQ_INSERT_HEAD(&V_pf_sourcelim_list_inactive, pfsrlim, pfsrlim_list);
2247
2248 PF_RULES_WUNLOCK();
2249
2250 return (0);
2251
2252 unlock:
2253 if (pfsrlim->pfsrlim_overload.table != NULL)
2254 pfr_detach_table(pfsrlim->pfsrlim_overload.table);
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 /**
3230 * Note: rule hashes may collide. Accept this, because the worst that can
3231 * happen is that we get counter preservation wrong.
3232 * Failing to insert here would be worse.
3233 **/
3234 RB_INSERT(pf_krule_global, ruleset->rules[rs_num].inactive.tree, rule);
3235 PF_CONFIG_UNLOCK();
3236
3237 return (0);
3238
3239 #undef ERROUT
3240 #undef ERROUT_UNLOCKED
3241 errout:
3242 PF_RULES_WUNLOCK();
3243 PF_CONFIG_UNLOCK();
3244 errout_unlocked:
3245 pf_kkif_free(rcv_kif);
3246 pf_kkif_free(kif);
3247 pf_krule_free(rule);
3248 return (error);
3249 }
3250
3251 static bool
pf_label_match(const struct pf_krule * rule,const char * label)3252 pf_label_match(const struct pf_krule *rule, const char *label)
3253 {
3254 int i = 0;
3255
3256 while (*rule->label[i]) {
3257 if (strcmp(rule->label[i], label) == 0)
3258 return (true);
3259 i++;
3260 }
3261
3262 return (false);
3263 }
3264
3265 static unsigned int
pf_kill_matching_state(struct pf_state_key_cmp * key,int dir)3266 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
3267 {
3268 struct pf_kstate *s;
3269 int more = 0;
3270
3271 s = pf_find_state_all(key, dir, &more);
3272 if (s == NULL)
3273 return (0);
3274
3275 if (more) {
3276 PF_STATE_UNLOCK(s);
3277 return (0);
3278 }
3279
3280 pf_remove_state(s);
3281 return (1);
3282 }
3283
3284 static int
pf_killstates_row(struct pf_kstate_kill * psk,struct pf_idhash * ih)3285 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
3286 {
3287 struct pf_kstate *s;
3288 struct pf_state_key *sk;
3289 struct pf_addr *srcaddr, *dstaddr;
3290 struct pf_state_key_cmp match_key;
3291 int idx, killed = 0;
3292 unsigned int dir;
3293 u_int16_t srcport, dstport;
3294 struct pfi_kkif *kif;
3295
3296 relock_DIOCKILLSTATES:
3297 PF_HASHROW_LOCK(ih);
3298 LIST_FOREACH(s, &ih->states, entry) {
3299 /* For floating states look at the original kif. */
3300 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
3301
3302 sk = s->key[psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE];
3303 if (s->direction == PF_OUT) {
3304 srcaddr = &sk->addr[1];
3305 dstaddr = &sk->addr[0];
3306 srcport = sk->port[1];
3307 dstport = sk->port[0];
3308 } else {
3309 srcaddr = &sk->addr[0];
3310 dstaddr = &sk->addr[1];
3311 srcport = sk->port[0];
3312 dstport = sk->port[1];
3313 }
3314
3315 if (psk->psk_af && sk->af != psk->psk_af)
3316 continue;
3317
3318 if (psk->psk_proto && psk->psk_proto != sk->proto)
3319 continue;
3320
3321 if (! pf_match_addr(psk->psk_src.neg,
3322 &psk->psk_src.addr.v.a.addr,
3323 &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
3324 continue;
3325
3326 if (! pf_match_addr(psk->psk_dst.neg,
3327 &psk->psk_dst.addr.v.a.addr,
3328 &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
3329 continue;
3330
3331 if (! pf_match_addr(psk->psk_rt_addr.neg,
3332 &psk->psk_rt_addr.addr.v.a.addr,
3333 &psk->psk_rt_addr.addr.v.a.mask,
3334 &s->act.rt_addr, sk->af))
3335 continue;
3336
3337 if (psk->psk_src.port_op != 0 &&
3338 ! pf_match_port(psk->psk_src.port_op,
3339 psk->psk_src.port[0], psk->psk_src.port[1], srcport))
3340 continue;
3341
3342 if (psk->psk_dst.port_op != 0 &&
3343 ! pf_match_port(psk->psk_dst.port_op,
3344 psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
3345 continue;
3346
3347 if (psk->psk_label[0] &&
3348 ! pf_label_match(s->rule, psk->psk_label))
3349 continue;
3350
3351 if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
3352 kif->pfik_name))
3353 continue;
3354
3355 if (psk->psk_kill_match) {
3356 /* Create the key to find matching states, with lock
3357 * held. */
3358
3359 bzero(&match_key, sizeof(match_key));
3360
3361 if (s->direction == PF_OUT) {
3362 dir = PF_IN;
3363 idx = psk->psk_nat ? PF_SK_WIRE : PF_SK_STACK;
3364 } else {
3365 dir = PF_OUT;
3366 idx = psk->psk_nat ? PF_SK_STACK : PF_SK_WIRE;
3367 }
3368
3369 match_key.af = s->key[idx]->af;
3370 match_key.proto = s->key[idx]->proto;
3371 pf_addrcpy(&match_key.addr[0],
3372 &s->key[idx]->addr[1], match_key.af);
3373 match_key.port[0] = s->key[idx]->port[1];
3374 pf_addrcpy(&match_key.addr[1],
3375 &s->key[idx]->addr[0], match_key.af);
3376 match_key.port[1] = s->key[idx]->port[0];
3377 }
3378
3379 pf_remove_state(s);
3380 killed++;
3381
3382 if (psk->psk_kill_match)
3383 killed += pf_kill_matching_state(&match_key, dir);
3384
3385 goto relock_DIOCKILLSTATES;
3386 }
3387 PF_HASHROW_UNLOCK(ih);
3388
3389 return (killed);
3390 }
3391
3392 int
pf_start(void)3393 pf_start(void)
3394 {
3395 int error = 0;
3396
3397 sx_xlock(&V_pf_ioctl_lock);
3398 if (V_pf_status.running)
3399 error = EEXIST;
3400 else {
3401 hook_pf();
3402 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
3403 hook_pf_eth();
3404 V_pf_status.running = 1;
3405 V_pf_status.since = time_uptime;
3406 new_unrhdr64(&V_pf_stateid, time_second);
3407
3408 DPFPRINTF(PF_DEBUG_MISC, "pf: started");
3409 }
3410 sx_xunlock(&V_pf_ioctl_lock);
3411
3412 return (error);
3413 }
3414
3415 int
pf_stop(void)3416 pf_stop(void)
3417 {
3418 int error = 0;
3419
3420 sx_xlock(&V_pf_ioctl_lock);
3421 if (!V_pf_status.running)
3422 error = ENOENT;
3423 else {
3424 V_pf_status.running = 0;
3425 dehook_pf();
3426 dehook_pf_eth();
3427 V_pf_status.since = time_uptime;
3428 DPFPRINTF(PF_DEBUG_MISC, "pf: stopped");
3429 }
3430 sx_xunlock(&V_pf_ioctl_lock);
3431
3432 return (error);
3433 }
3434
3435 void
pf_ioctl_clear_status(void)3436 pf_ioctl_clear_status(void)
3437 {
3438 PF_RULES_WLOCK();
3439 for (int i = 0; i < PFRES_MAX; i++)
3440 counter_u64_zero(V_pf_status.counters[i]);
3441 for (int i = 0; i < FCNT_MAX; i++)
3442 pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3443 for (int i = 0; i < SCNT_MAX; i++)
3444 counter_u64_zero(V_pf_status.scounters[i]);
3445 for (int i = 0; i < NCNT_MAX; i++)
3446 counter_u64_zero(V_pf_status.ncounters[i]);
3447 for (int i = 0; i < KLCNT_MAX; i++)
3448 counter_u64_zero(V_pf_status.lcounters[i]);
3449 V_pf_status.since = time_uptime;
3450 if (*V_pf_status.ifname)
3451 pfi_update_status(V_pf_status.ifname, NULL);
3452 PF_RULES_WUNLOCK();
3453 }
3454
3455 int
pf_ioctl_set_timeout(int timeout,int seconds,int * prev_seconds)3456 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
3457 {
3458 uint32_t old;
3459
3460 if (timeout < 0 || timeout >= PFTM_MAX ||
3461 seconds < 0)
3462 return (EINVAL);
3463
3464 PF_RULES_WLOCK();
3465 old = V_pf_default_rule.timeout[timeout];
3466 if (timeout == PFTM_INTERVAL && seconds == 0)
3467 seconds = 1;
3468 V_pf_default_rule.timeout[timeout] = seconds;
3469 if (timeout == PFTM_INTERVAL && seconds < old)
3470 wakeup(pf_purge_thread);
3471
3472 if (prev_seconds != NULL)
3473 *prev_seconds = old;
3474
3475 PF_RULES_WUNLOCK();
3476
3477 return (0);
3478 }
3479
3480 int
pf_ioctl_get_timeout(int timeout,int * seconds)3481 pf_ioctl_get_timeout(int timeout, int *seconds)
3482 {
3483 PF_RULES_RLOCK_TRACKER;
3484
3485 if (timeout < 0 || timeout >= PFTM_MAX)
3486 return (EINVAL);
3487
3488 PF_RULES_RLOCK();
3489 *seconds = V_pf_default_rule.timeout[timeout];
3490 PF_RULES_RUNLOCK();
3491
3492 return (0);
3493 }
3494
3495 int
pf_ioctl_set_limit(int index,unsigned int limit,unsigned int * old_limit)3496 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
3497 {
3498
3499 PF_RULES_WLOCK();
3500 if (index < 0 || index >= PF_LIMIT_MAX ||
3501 V_pf_limits[index].zone == NULL) {
3502 PF_RULES_WUNLOCK();
3503 return (EINVAL);
3504 }
3505 uma_zone_set_max(V_pf_limits[index].zone,
3506 limit == 0 ? INT_MAX : limit);
3507 if (old_limit != NULL)
3508 *old_limit = V_pf_limits[index].limit;
3509 V_pf_limits[index].limit = limit;
3510 PF_RULES_WUNLOCK();
3511
3512 return (0);
3513 }
3514
3515 int
pf_ioctl_get_limit(int index,unsigned int * limit)3516 pf_ioctl_get_limit(int index, unsigned int *limit)
3517 {
3518 PF_RULES_RLOCK_TRACKER;
3519
3520 if (index < 0 || index >= PF_LIMIT_MAX)
3521 return (EINVAL);
3522
3523 PF_RULES_RLOCK();
3524 *limit = V_pf_limits[index].limit;
3525 PF_RULES_RUNLOCK();
3526
3527 return (0);
3528 }
3529
3530 int
pf_ioctl_begin_addrs(uint32_t * ticket)3531 pf_ioctl_begin_addrs(uint32_t *ticket)
3532 {
3533 PF_RULES_WLOCK();
3534 pf_empty_kpool(&V_pf_pabuf[0]);
3535 pf_empty_kpool(&V_pf_pabuf[1]);
3536 pf_empty_kpool(&V_pf_pabuf[2]);
3537 *ticket = ++V_ticket_pabuf;
3538 PF_RULES_WUNLOCK();
3539
3540 return (0);
3541 }
3542
3543 int
pf_ioctl_add_addr(struct pf_nl_pooladdr * pp)3544 pf_ioctl_add_addr(struct pf_nl_pooladdr *pp)
3545 {
3546 struct pf_kpooladdr *pa = NULL;
3547 struct pfi_kkif *kif = NULL;
3548 int error;
3549
3550 if (pp->which != PF_RDR && pp->which != PF_NAT &&
3551 pp->which != PF_RT)
3552 return (EINVAL);
3553
3554 switch (pp->af) {
3555 #ifdef INET
3556 case AF_INET:
3557 /* FALLTHROUGH */
3558 #endif /* INET */
3559 #ifdef INET6
3560 case AF_INET6:
3561 /* FALLTHROUGH */
3562 #endif /* INET6 */
3563 case AF_UNSPEC:
3564 break;
3565 default:
3566 return (EAFNOSUPPORT);
3567 }
3568
3569 if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
3570 pp->addr.addr.type != PF_ADDR_DYNIFTL &&
3571 pp->addr.addr.type != PF_ADDR_TABLE)
3572 return (EINVAL);
3573
3574 if (pp->addr.addr.p.dyn != NULL)
3575 return (EINVAL);
3576
3577 pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
3578 error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
3579 if (error != 0)
3580 goto out;
3581 if (pa->ifname[0])
3582 kif = pf_kkif_create(M_WAITOK);
3583 PF_RULES_WLOCK();
3584 if (pp->ticket != V_ticket_pabuf) {
3585 PF_RULES_WUNLOCK();
3586 if (pa->ifname[0])
3587 pf_kkif_free(kif);
3588 error = EBUSY;
3589 goto out;
3590 }
3591 if (pa->ifname[0]) {
3592 pa->kif = pfi_kkif_attach(kif, pa->ifname);
3593 kif = NULL;
3594 pfi_kkif_ref(pa->kif);
3595 } else
3596 pa->kif = NULL;
3597 if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
3598 pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
3599 if (pa->ifname[0])
3600 pfi_kkif_unref(pa->kif);
3601 PF_RULES_WUNLOCK();
3602 goto out;
3603 }
3604 pa->af = pp->af;
3605 switch (pp->which) {
3606 case PF_NAT:
3607 TAILQ_INSERT_TAIL(&V_pf_pabuf[0], pa, entries);
3608 break;
3609 case PF_RDR:
3610 TAILQ_INSERT_TAIL(&V_pf_pabuf[1], pa, entries);
3611 break;
3612 case PF_RT:
3613 TAILQ_INSERT_TAIL(&V_pf_pabuf[2], pa, entries);
3614 break;
3615 }
3616 PF_RULES_WUNLOCK();
3617
3618 return (0);
3619
3620 out:
3621 free(pa, M_PFRULE);
3622 return (error);
3623 }
3624
3625 int
pf_ioctl_get_addrs(struct pf_nl_pooladdr * pp)3626 pf_ioctl_get_addrs(struct pf_nl_pooladdr *pp)
3627 {
3628 struct pf_kpool *pool;
3629 struct pf_kpooladdr *pa;
3630
3631 PF_RULES_RLOCK_TRACKER;
3632
3633 if (pp->which != PF_RDR && pp->which != PF_NAT &&
3634 pp->which != PF_RT)
3635 return (EINVAL);
3636
3637 pp->anchor[sizeof(pp->anchor) - 1] = 0;
3638 pp->nr = 0;
3639
3640 PF_RULES_RLOCK();
3641 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3642 pp->r_num, 0, 1, 0, pp->which);
3643 if (pool == NULL) {
3644 PF_RULES_RUNLOCK();
3645 return (EBUSY);
3646 }
3647 TAILQ_FOREACH(pa, &pool->list, entries)
3648 pp->nr++;
3649 PF_RULES_RUNLOCK();
3650
3651 return (0);
3652 }
3653
3654 int
pf_ioctl_get_addr(struct pf_nl_pooladdr * pp)3655 pf_ioctl_get_addr(struct pf_nl_pooladdr *pp)
3656 {
3657 struct pf_kpool *pool;
3658 struct pf_kpooladdr *pa;
3659 u_int32_t nr = 0;
3660
3661 if (pp->which != PF_RDR && pp->which != PF_NAT &&
3662 pp->which != PF_RT)
3663 return (EINVAL);
3664
3665 PF_RULES_RLOCK_TRACKER;
3666
3667 pp->anchor[sizeof(pp->anchor) - 1] = '\0';
3668
3669 PF_RULES_RLOCK();
3670 pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3671 pp->r_num, 0, 1, 1, pp->which);
3672 if (pool == NULL) {
3673 PF_RULES_RUNLOCK();
3674 return (EBUSY);
3675 }
3676 pa = TAILQ_FIRST(&pool->list);
3677 while ((pa != NULL) && (nr < pp->nr)) {
3678 pa = TAILQ_NEXT(pa, entries);
3679 nr++;
3680 }
3681 if (pa == NULL) {
3682 PF_RULES_RUNLOCK();
3683 return (EBUSY);
3684 }
3685 pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3686 pp->af = pa->af;
3687 pf_addr_copyout(&pp->addr.addr);
3688 PF_RULES_RUNLOCK();
3689
3690 return (0);
3691 }
3692
3693 int
pf_ioctl_get_rulesets(struct pfioc_ruleset * pr)3694 pf_ioctl_get_rulesets(struct pfioc_ruleset *pr)
3695 {
3696 struct pf_kruleset *ruleset;
3697 struct pf_kanchor *anchor;
3698
3699 PF_RULES_RLOCK_TRACKER;
3700
3701 pr->path[sizeof(pr->path) - 1] = '\0';
3702
3703 PF_RULES_RLOCK();
3704 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3705 PF_RULES_RUNLOCK();
3706 return (ENOENT);
3707 }
3708 pr->nr = 0;
3709 if (ruleset == &pf_main_ruleset) {
3710 /* XXX kludge for pf_main_ruleset */
3711 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3712 if (anchor->parent == NULL)
3713 pr->nr++;
3714 } else {
3715 RB_FOREACH(anchor, pf_kanchor_node,
3716 &ruleset->anchor->children)
3717 pr->nr++;
3718 }
3719 PF_RULES_RUNLOCK();
3720
3721 return (0);
3722 }
3723
3724 int
pf_ioctl_get_ruleset(struct pfioc_ruleset * pr)3725 pf_ioctl_get_ruleset(struct pfioc_ruleset *pr)
3726 {
3727 struct pf_kruleset *ruleset;
3728 struct pf_kanchor *anchor;
3729 u_int32_t nr = 0;
3730 int error = 0;
3731
3732 PF_RULES_RLOCK_TRACKER;
3733
3734 PF_RULES_RLOCK();
3735 if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3736 PF_RULES_RUNLOCK();
3737 return (ENOENT);
3738 }
3739
3740 pr->name[0] = '\0';
3741 if (ruleset == &pf_main_ruleset) {
3742 /* XXX kludge for pf_main_ruleset */
3743 RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3744 if (anchor->parent == NULL && nr++ == pr->nr) {
3745 strlcpy(pr->name, anchor->name,
3746 sizeof(pr->name));
3747 break;
3748 }
3749 } else {
3750 RB_FOREACH(anchor, pf_kanchor_node,
3751 &ruleset->anchor->children)
3752 if (nr++ == pr->nr) {
3753 strlcpy(pr->name, anchor->name,
3754 sizeof(pr->name));
3755 break;
3756 }
3757 }
3758 if (!pr->name[0])
3759 error = EBUSY;
3760 PF_RULES_RUNLOCK();
3761
3762 return (error);
3763 }
3764
3765 int
pf_ioctl_natlook(struct pfioc_natlook * pnl)3766 pf_ioctl_natlook(struct pfioc_natlook *pnl)
3767 {
3768 struct pf_state_key *sk;
3769 struct pf_kstate *state;
3770 struct pf_state_key_cmp key;
3771 int m = 0, direction = pnl->direction;
3772 int sidx, didx;
3773
3774 /* NATLOOK src and dst are reversed, so reverse sidx/didx */
3775 sidx = (direction == PF_IN) ? 1 : 0;
3776 didx = (direction == PF_IN) ? 0 : 1;
3777
3778 if (!pnl->proto ||
3779 PF_AZERO(&pnl->saddr, pnl->af) ||
3780 PF_AZERO(&pnl->daddr, pnl->af) ||
3781 ((pnl->proto == IPPROTO_TCP ||
3782 pnl->proto == IPPROTO_UDP) &&
3783 (!pnl->dport || !pnl->sport)))
3784 return (EINVAL);
3785
3786 switch (pnl->direction) {
3787 case PF_IN:
3788 case PF_OUT:
3789 case PF_INOUT:
3790 break;
3791 default:
3792 return (EINVAL);
3793 }
3794
3795 switch (pnl->af) {
3796 #ifdef INET
3797 case AF_INET:
3798 break;
3799 #endif /* INET */
3800 #ifdef INET6
3801 case AF_INET6:
3802 break;
3803 #endif /* INET6 */
3804 default:
3805 return (EAFNOSUPPORT);
3806 }
3807
3808 bzero(&key, sizeof(key));
3809 key.af = pnl->af;
3810 key.proto = pnl->proto;
3811 pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af);
3812 key.port[sidx] = pnl->sport;
3813 pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af);
3814 key.port[didx] = pnl->dport;
3815
3816 state = pf_find_state_all(&key, direction, &m);
3817 if (state == NULL)
3818 return (ENOENT);
3819
3820 if (m > 1) {
3821 PF_STATE_UNLOCK(state);
3822 return (E2BIG); /* more than one state */
3823 }
3824
3825 sk = state->key[sidx];
3826 pf_addrcpy(&pnl->rsaddr,
3827 &sk->addr[sidx], sk->af);
3828 pnl->rsport = sk->port[sidx];
3829 pf_addrcpy(&pnl->rdaddr,
3830 &sk->addr[didx], sk->af);
3831 pnl->rdport = sk->port[didx];
3832 PF_STATE_UNLOCK(state);
3833
3834 return (0);
3835 }
3836
3837 static int
pfioctl(struct cdev * dev,u_long cmd,caddr_t addr,int flags,struct thread * td)3838 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
3839 {
3840 int error = 0;
3841 PF_RULES_RLOCK_TRACKER;
3842
3843 #define ERROUT_IOCTL(target, x) \
3844 do { \
3845 error = (x); \
3846 SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__); \
3847 goto target; \
3848 } while (0)
3849
3850
3851 /* XXX keep in sync with switch() below */
3852 if (securelevel_gt(td->td_ucred, 2))
3853 switch (cmd) {
3854 case DIOCGETRULES:
3855 case DIOCGETRULENV:
3856 case DIOCGETADDRS:
3857 case DIOCGETADDR:
3858 case DIOCGETSTATE:
3859 case DIOCGETSTATENV:
3860 case DIOCSETSTATUSIF:
3861 case DIOCGETSTATUSNV:
3862 case DIOCCLRSTATUS:
3863 case DIOCNATLOOK:
3864 case DIOCSETDEBUG:
3865 #ifdef COMPAT_FREEBSD14
3866 case DIOCGETSTATES:
3867 case DIOCGETSTATESV2:
3868 #endif
3869 case DIOCGETTIMEOUT:
3870 case DIOCCLRRULECTRS:
3871 case DIOCGETLIMIT:
3872 case DIOCGETALTQSV0:
3873 case DIOCGETALTQSV1:
3874 case DIOCGETALTQV0:
3875 case DIOCGETALTQV1:
3876 case DIOCGETQSTATSV0:
3877 case DIOCGETQSTATSV1:
3878 case DIOCGETRULESETS:
3879 case DIOCGETRULESET:
3880 case DIOCRGETTABLES:
3881 case DIOCRGETTSTATS:
3882 case DIOCRCLRTSTATS:
3883 case DIOCRCLRADDRS:
3884 case DIOCRADDADDRS:
3885 case DIOCRDELADDRS:
3886 case DIOCRSETADDRS:
3887 case DIOCRGETADDRS:
3888 case DIOCRGETASTATS:
3889 case DIOCRCLRASTATS:
3890 case DIOCRTSTADDRS:
3891 case DIOCOSFPGET:
3892 case DIOCGETSRCNODES:
3893 case DIOCCLRSRCNODES:
3894 case DIOCGETSYNCOOKIES:
3895 case DIOCIGETIFACES:
3896 case DIOCGIFSPEEDV0:
3897 case DIOCGIFSPEEDV1:
3898 case DIOCGETETHRULES:
3899 case DIOCGETETHRULE:
3900 case DIOCGETETHRULESETS:
3901 case DIOCGETETHRULESET:
3902 break;
3903 case DIOCRCLRTABLES:
3904 case DIOCRADDTABLES:
3905 case DIOCRDELTABLES:
3906 case DIOCRSETTFLAGS:
3907 if (((struct pfioc_table *)addr)->pfrio_flags &
3908 PFR_FLAG_DUMMY)
3909 break; /* dummy operation ok */
3910 return (EPERM);
3911 default:
3912 return (EPERM);
3913 }
3914
3915 if (!(flags & FWRITE))
3916 switch (cmd) {
3917 case DIOCGETRULES:
3918 case DIOCGETADDRS:
3919 case DIOCGETADDR:
3920 case DIOCGETSTATE:
3921 case DIOCGETSTATENV:
3922 case DIOCGETSTATUSNV:
3923 #ifdef COMPAT_FREEBSD14
3924 case DIOCGETSTATES:
3925 case DIOCGETSTATESV2:
3926 #endif
3927 case DIOCGETTIMEOUT:
3928 case DIOCGETLIMIT:
3929 case DIOCGETALTQSV0:
3930 case DIOCGETALTQSV1:
3931 case DIOCGETALTQV0:
3932 case DIOCGETALTQV1:
3933 case DIOCGETQSTATSV0:
3934 case DIOCGETQSTATSV1:
3935 case DIOCGETRULESETS:
3936 case DIOCGETRULESET:
3937 case DIOCNATLOOK:
3938 case DIOCRGETTABLES:
3939 case DIOCRGETTSTATS:
3940 case DIOCRGETADDRS:
3941 case DIOCRGETASTATS:
3942 case DIOCRTSTADDRS:
3943 case DIOCOSFPGET:
3944 case DIOCGETSRCNODES:
3945 case DIOCGETSYNCOOKIES:
3946 case DIOCIGETIFACES:
3947 case DIOCGIFSPEEDV1:
3948 case DIOCGIFSPEEDV0:
3949 case DIOCGETRULENV:
3950 case DIOCGETETHRULES:
3951 case DIOCGETETHRULE:
3952 case DIOCGETETHRULESETS:
3953 case DIOCGETETHRULESET:
3954 break;
3955 case DIOCRCLRTABLES:
3956 case DIOCRADDTABLES:
3957 case DIOCRDELTABLES:
3958 case DIOCRCLRTSTATS:
3959 case DIOCRCLRADDRS:
3960 case DIOCRADDADDRS:
3961 case DIOCRDELADDRS:
3962 case DIOCRSETADDRS:
3963 case DIOCRSETTFLAGS:
3964 if (((struct pfioc_table *)addr)->pfrio_flags &
3965 PFR_FLAG_DUMMY) {
3966 flags |= FWRITE; /* need write lock for dummy */
3967 break; /* dummy operation ok */
3968 }
3969 return (EACCES);
3970 default:
3971 return (EACCES);
3972 }
3973
3974 CURVNET_SET(TD_TO_VNET(td));
3975
3976 switch (cmd) {
3977 #ifdef COMPAT_FREEBSD14
3978 case DIOCSTART:
3979 error = pf_start();
3980 break;
3981
3982 case DIOCSTOP:
3983 error = pf_stop();
3984 break;
3985 #endif
3986
3987 case DIOCGETETHRULES: {
3988 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
3989 nvlist_t *nvl;
3990 void *packed;
3991 struct pf_keth_rule *tail;
3992 struct pf_keth_ruleset *rs;
3993 u_int32_t ticket, nr;
3994 const char *anchor = "";
3995
3996 nvl = NULL;
3997 packed = NULL;
3998
3999 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULES_error, x)
4000
4001 if (nv->len > pf_ioctl_maxcount)
4002 ERROUT(ENOMEM);
4003
4004 /* Copy the request in */
4005 packed = malloc(nv->len, M_NVLIST, M_WAITOK);
4006 error = copyin(nv->data, packed, nv->len);
4007 if (error)
4008 ERROUT(error);
4009
4010 nvl = nvlist_unpack(packed, nv->len, 0);
4011 if (nvl == NULL)
4012 ERROUT(EBADMSG);
4013
4014 if (! nvlist_exists_string(nvl, "anchor"))
4015 ERROUT(EBADMSG);
4016
4017 anchor = nvlist_get_string(nvl, "anchor");
4018
4019 rs = pf_find_keth_ruleset(anchor);
4020
4021 nvlist_destroy(nvl);
4022 nvl = NULL;
4023 free(packed, M_NVLIST);
4024 packed = NULL;
4025
4026 if (rs == NULL)
4027 ERROUT(ENOENT);
4028
4029 /* Reply */
4030 nvl = nvlist_create(0);
4031 if (nvl == NULL)
4032 ERROUT(ENOMEM);
4033
4034 PF_RULES_RLOCK();
4035
4036 ticket = rs->active.ticket;
4037 tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
4038 if (tail)
4039 nr = tail->nr + 1;
4040 else
4041 nr = 0;
4042
4043 PF_RULES_RUNLOCK();
4044
4045 nvlist_add_number(nvl, "ticket", ticket);
4046 nvlist_add_number(nvl, "nr", nr);
4047
4048 packed = nvlist_pack(nvl, &nv->len);
4049 if (packed == NULL)
4050 ERROUT(ENOMEM);
4051
4052 if (nv->size == 0)
4053 ERROUT(0);
4054 else if (nv->size < nv->len)
4055 ERROUT(ENOSPC);
4056
4057 error = copyout(packed, nv->data, nv->len);
4058
4059 #undef ERROUT
4060 DIOCGETETHRULES_error:
4061 free(packed, M_NVLIST);
4062 nvlist_destroy(nvl);
4063 break;
4064 }
4065
4066 case DIOCGETETHRULE: {
4067 struct epoch_tracker et;
4068 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4069 nvlist_t *nvl = NULL;
4070 void *nvlpacked = NULL;
4071 struct pf_keth_rule *rule = NULL;
4072 struct pf_keth_ruleset *rs;
4073 u_int32_t ticket, nr;
4074 bool clear = false;
4075 const char *anchor;
4076
4077 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULE_error, x)
4078
4079 if (nv->len > pf_ioctl_maxcount)
4080 ERROUT(ENOMEM);
4081
4082 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4083 error = copyin(nv->data, nvlpacked, nv->len);
4084 if (error)
4085 ERROUT(error);
4086
4087 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4088 if (nvl == NULL)
4089 ERROUT(EBADMSG);
4090 if (! nvlist_exists_number(nvl, "ticket"))
4091 ERROUT(EBADMSG);
4092 ticket = nvlist_get_number(nvl, "ticket");
4093 if (! nvlist_exists_string(nvl, "anchor"))
4094 ERROUT(EBADMSG);
4095 anchor = nvlist_get_string(nvl, "anchor");
4096
4097 if (nvlist_exists_bool(nvl, "clear"))
4098 clear = nvlist_get_bool(nvl, "clear");
4099
4100 if (clear && !(flags & FWRITE))
4101 ERROUT(EACCES);
4102
4103 if (! nvlist_exists_number(nvl, "nr"))
4104 ERROUT(EBADMSG);
4105 nr = nvlist_get_number(nvl, "nr");
4106
4107 PF_RULES_RLOCK();
4108 rs = pf_find_keth_ruleset(anchor);
4109 if (rs == NULL) {
4110 PF_RULES_RUNLOCK();
4111 ERROUT(ENOENT);
4112 }
4113 if (ticket != rs->active.ticket) {
4114 PF_RULES_RUNLOCK();
4115 ERROUT(EBUSY);
4116 }
4117
4118 nvlist_destroy(nvl);
4119 nvl = NULL;
4120 free(nvlpacked, M_NVLIST);
4121 nvlpacked = NULL;
4122
4123 rule = TAILQ_FIRST(rs->active.rules);
4124 while ((rule != NULL) && (rule->nr != nr))
4125 rule = TAILQ_NEXT(rule, entries);
4126 if (rule == NULL) {
4127 PF_RULES_RUNLOCK();
4128 ERROUT(ENOENT);
4129 }
4130 /* Make sure rule can't go away. */
4131 NET_EPOCH_ENTER(et);
4132 PF_RULES_RUNLOCK();
4133 nvl = pf_keth_rule_to_nveth_rule(rule);
4134 if (pf_keth_anchor_nvcopyout(rs, rule, nvl)) {
4135 NET_EPOCH_EXIT(et);
4136 ERROUT(EBUSY);
4137 }
4138 NET_EPOCH_EXIT(et);
4139 if (nvl == NULL)
4140 ERROUT(ENOMEM);
4141
4142 nvlpacked = nvlist_pack(nvl, &nv->len);
4143 if (nvlpacked == NULL)
4144 ERROUT(ENOMEM);
4145
4146 if (nv->size == 0)
4147 ERROUT(0);
4148 else if (nv->size < nv->len)
4149 ERROUT(ENOSPC);
4150
4151 error = copyout(nvlpacked, nv->data, nv->len);
4152 if (error == 0 && clear) {
4153 counter_u64_zero(rule->evaluations);
4154 for (int i = 0; i < 2; i++) {
4155 counter_u64_zero(rule->packets[i]);
4156 counter_u64_zero(rule->bytes[i]);
4157 }
4158 }
4159
4160 #undef ERROUT
4161 DIOCGETETHRULE_error:
4162 free(nvlpacked, M_NVLIST);
4163 nvlist_destroy(nvl);
4164 break;
4165 }
4166
4167 case DIOCADDETHRULE: {
4168 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4169 nvlist_t *nvl = NULL;
4170 void *nvlpacked = NULL;
4171 struct pf_keth_rule *rule = NULL, *tail = NULL;
4172 struct pf_keth_ruleset *ruleset = NULL;
4173 struct pfi_kkif *kif = NULL, *bridge_to_kif = NULL;
4174 const char *anchor = "", *anchor_call = "";
4175
4176 #define ERROUT(x) ERROUT_IOCTL(DIOCADDETHRULE_error, x)
4177
4178 if (nv->len > pf_ioctl_maxcount)
4179 ERROUT(ENOMEM);
4180
4181 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4182 error = copyin(nv->data, nvlpacked, nv->len);
4183 if (error)
4184 ERROUT(error);
4185
4186 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4187 if (nvl == NULL)
4188 ERROUT(EBADMSG);
4189
4190 if (! nvlist_exists_number(nvl, "ticket"))
4191 ERROUT(EBADMSG);
4192
4193 if (nvlist_exists_string(nvl, "anchor"))
4194 anchor = nvlist_get_string(nvl, "anchor");
4195 if (nvlist_exists_string(nvl, "anchor_call"))
4196 anchor_call = nvlist_get_string(nvl, "anchor_call");
4197
4198 ruleset = pf_find_keth_ruleset(anchor);
4199 if (ruleset == NULL)
4200 ERROUT(EINVAL);
4201
4202 if (nvlist_get_number(nvl, "ticket") !=
4203 ruleset->inactive.ticket) {
4204 DPFPRINTF(PF_DEBUG_MISC,
4205 "ticket: %d != %d",
4206 (u_int32_t)nvlist_get_number(nvl, "ticket"),
4207 ruleset->inactive.ticket);
4208 ERROUT(EBUSY);
4209 }
4210
4211 rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
4212 rule->timestamp = NULL;
4213
4214 error = pf_nveth_rule_to_keth_rule(nvl, rule);
4215 if (error != 0)
4216 ERROUT(error);
4217
4218 if (rule->ifname[0])
4219 kif = pf_kkif_create(M_WAITOK);
4220 if (rule->bridge_to_name[0])
4221 bridge_to_kif = pf_kkif_create(M_WAITOK);
4222 rule->evaluations = counter_u64_alloc(M_WAITOK);
4223 for (int i = 0; i < 2; i++) {
4224 rule->packets[i] = counter_u64_alloc(M_WAITOK);
4225 rule->bytes[i] = counter_u64_alloc(M_WAITOK);
4226 }
4227 rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
4228 M_WAITOK | M_ZERO);
4229
4230 PF_RULES_WLOCK();
4231
4232 if (rule->ifname[0]) {
4233 rule->kif = pfi_kkif_attach(kif, rule->ifname);
4234 pfi_kkif_ref(rule->kif);
4235 } else
4236 rule->kif = NULL;
4237 if (rule->bridge_to_name[0]) {
4238 rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
4239 rule->bridge_to_name);
4240 pfi_kkif_ref(rule->bridge_to);
4241 } else
4242 rule->bridge_to = NULL;
4243
4244 #ifdef ALTQ
4245 /* set queue IDs */
4246 if (rule->qname[0] != 0) {
4247 if ((rule->qid = pf_qname2qid(rule->qname, true)) == 0)
4248 error = EBUSY;
4249 else
4250 rule->qid = rule->qid;
4251 }
4252 #endif
4253 if (rule->tagname[0])
4254 if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
4255 error = EBUSY;
4256 if (rule->match_tagname[0])
4257 if ((rule->match_tag = pf_tagname2tag(
4258 rule->match_tagname)) == 0)
4259 error = EBUSY;
4260
4261 if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
4262 error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
4263 if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
4264 error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
4265
4266 if (error) {
4267 pf_free_eth_rule(rule);
4268 PF_RULES_WUNLOCK();
4269 ERROUT(error);
4270 }
4271
4272 if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
4273 pf_free_eth_rule(rule);
4274 PF_RULES_WUNLOCK();
4275 ERROUT(EINVAL);
4276 }
4277
4278 tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
4279 if (tail)
4280 rule->nr = tail->nr + 1;
4281 else
4282 rule->nr = 0;
4283
4284 TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
4285
4286 PF_RULES_WUNLOCK();
4287
4288 #undef ERROUT
4289 DIOCADDETHRULE_error:
4290 nvlist_destroy(nvl);
4291 free(nvlpacked, M_NVLIST);
4292 break;
4293 }
4294
4295 case DIOCGETETHRULESETS: {
4296 struct epoch_tracker et;
4297 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4298 nvlist_t *nvl = NULL;
4299 void *nvlpacked = NULL;
4300 struct pf_keth_ruleset *ruleset;
4301 struct pf_keth_anchor *anchor;
4302 int nr = 0;
4303
4304 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
4305
4306 if (nv->len > pf_ioctl_maxcount)
4307 ERROUT(ENOMEM);
4308
4309 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4310 error = copyin(nv->data, nvlpacked, nv->len);
4311 if (error)
4312 ERROUT(error);
4313
4314 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4315 if (nvl == NULL)
4316 ERROUT(EBADMSG);
4317 if (! nvlist_exists_string(nvl, "path"))
4318 ERROUT(EBADMSG);
4319
4320 NET_EPOCH_ENTER(et);
4321
4322 if ((ruleset = pf_find_keth_ruleset(
4323 nvlist_get_string(nvl, "path"))) == NULL) {
4324 NET_EPOCH_EXIT(et);
4325 ERROUT(ENOENT);
4326 }
4327
4328 if (ruleset->anchor == NULL) {
4329 RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
4330 if (anchor->parent == NULL)
4331 nr++;
4332 } else {
4333 RB_FOREACH(anchor, pf_keth_anchor_node,
4334 &ruleset->anchor->children)
4335 nr++;
4336 }
4337
4338 NET_EPOCH_EXIT(et);
4339
4340 nvlist_destroy(nvl);
4341 nvl = NULL;
4342 free(nvlpacked, M_NVLIST);
4343 nvlpacked = NULL;
4344
4345 nvl = nvlist_create(0);
4346 if (nvl == NULL)
4347 ERROUT(ENOMEM);
4348
4349 nvlist_add_number(nvl, "nr", nr);
4350
4351 nvlpacked = nvlist_pack(nvl, &nv->len);
4352 if (nvlpacked == NULL)
4353 ERROUT(ENOMEM);
4354
4355 if (nv->size == 0)
4356 ERROUT(0);
4357 else if (nv->size < nv->len)
4358 ERROUT(ENOSPC);
4359
4360 error = copyout(nvlpacked, nv->data, nv->len);
4361
4362 #undef ERROUT
4363 DIOCGETETHRULESETS_error:
4364 free(nvlpacked, M_NVLIST);
4365 nvlist_destroy(nvl);
4366 break;
4367 }
4368
4369 case DIOCGETETHRULESET: {
4370 struct epoch_tracker et;
4371 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4372 nvlist_t *nvl = NULL;
4373 void *nvlpacked = NULL;
4374 struct pf_keth_ruleset *ruleset;
4375 struct pf_keth_anchor *anchor;
4376 int nr = 0, req_nr = 0;
4377 bool found = false;
4378
4379 #define ERROUT(x) ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
4380
4381 if (nv->len > pf_ioctl_maxcount)
4382 ERROUT(ENOMEM);
4383
4384 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4385 error = copyin(nv->data, nvlpacked, nv->len);
4386 if (error)
4387 ERROUT(error);
4388
4389 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4390 if (nvl == NULL)
4391 ERROUT(EBADMSG);
4392 if (! nvlist_exists_string(nvl, "path"))
4393 ERROUT(EBADMSG);
4394 if (! nvlist_exists_number(nvl, "nr"))
4395 ERROUT(EBADMSG);
4396
4397 req_nr = nvlist_get_number(nvl, "nr");
4398
4399 NET_EPOCH_ENTER(et);
4400
4401 if ((ruleset = pf_find_keth_ruleset(
4402 nvlist_get_string(nvl, "path"))) == NULL) {
4403 NET_EPOCH_EXIT(et);
4404 ERROUT(ENOENT);
4405 }
4406
4407 nvlist_destroy(nvl);
4408 nvl = NULL;
4409 free(nvlpacked, M_NVLIST);
4410 nvlpacked = NULL;
4411
4412 nvl = nvlist_create(0);
4413 if (nvl == NULL) {
4414 NET_EPOCH_EXIT(et);
4415 ERROUT(ENOMEM);
4416 }
4417
4418 if (ruleset->anchor == NULL) {
4419 RB_FOREACH(anchor, pf_keth_anchor_global,
4420 &V_pf_keth_anchors) {
4421 if (anchor->parent == NULL && nr++ == req_nr) {
4422 found = true;
4423 break;
4424 }
4425 }
4426 } else {
4427 RB_FOREACH(anchor, pf_keth_anchor_node,
4428 &ruleset->anchor->children) {
4429 if (nr++ == req_nr) {
4430 found = true;
4431 break;
4432 }
4433 }
4434 }
4435
4436 NET_EPOCH_EXIT(et);
4437 if (found) {
4438 nvlist_add_number(nvl, "nr", nr);
4439 nvlist_add_string(nvl, "name", anchor->name);
4440 if (ruleset->anchor)
4441 nvlist_add_string(nvl, "path",
4442 ruleset->anchor->path);
4443 else
4444 nvlist_add_string(nvl, "path", "");
4445 } else {
4446 ERROUT(EBUSY);
4447 }
4448
4449 nvlpacked = nvlist_pack(nvl, &nv->len);
4450 if (nvlpacked == NULL)
4451 ERROUT(ENOMEM);
4452
4453 if (nv->size == 0)
4454 ERROUT(0);
4455 else if (nv->size < nv->len)
4456 ERROUT(ENOSPC);
4457
4458 error = copyout(nvlpacked, nv->data, nv->len);
4459
4460 #undef ERROUT
4461 DIOCGETETHRULESET_error:
4462 free(nvlpacked, M_NVLIST);
4463 nvlist_destroy(nvl);
4464 break;
4465 }
4466
4467 case DIOCADDRULENV: {
4468 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4469 nvlist_t *nvl = NULL;
4470 void *nvlpacked = NULL;
4471 struct pf_krule *rule = NULL;
4472 const char *anchor = "", *anchor_call = "";
4473 uint32_t ticket = 0, pool_ticket = 0;
4474
4475 #define ERROUT(x) ERROUT_IOCTL(DIOCADDRULENV_error, x)
4476
4477 if (nv->len > pf_ioctl_maxcount)
4478 ERROUT(ENOMEM);
4479
4480 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4481 error = copyin(nv->data, nvlpacked, nv->len);
4482 if (error)
4483 ERROUT(error);
4484
4485 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4486 if (nvl == NULL)
4487 ERROUT(EBADMSG);
4488
4489 if (! nvlist_exists_number(nvl, "ticket"))
4490 ERROUT(EINVAL);
4491 ticket = nvlist_get_number(nvl, "ticket");
4492
4493 if (! nvlist_exists_number(nvl, "pool_ticket"))
4494 ERROUT(EINVAL);
4495 pool_ticket = nvlist_get_number(nvl, "pool_ticket");
4496
4497 if (! nvlist_exists_nvlist(nvl, "rule"))
4498 ERROUT(EINVAL);
4499
4500 rule = pf_krule_alloc();
4501 error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
4502 rule);
4503 if (error)
4504 ERROUT(error);
4505
4506 if (nvlist_exists_string(nvl, "anchor"))
4507 anchor = nvlist_get_string(nvl, "anchor");
4508 if (nvlist_exists_string(nvl, "anchor_call"))
4509 anchor_call = nvlist_get_string(nvl, "anchor_call");
4510
4511 if ((error = nvlist_error(nvl)))
4512 ERROUT(error);
4513
4514 /* Frees rule on error */
4515 error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
4516 anchor_call, td->td_ucred->cr_ruid,
4517 td->td_proc ? td->td_proc->p_pid : 0);
4518
4519 nvlist_destroy(nvl);
4520 free(nvlpacked, M_NVLIST);
4521 break;
4522 #undef ERROUT
4523 DIOCADDRULENV_error:
4524 pf_krule_free(rule);
4525 nvlist_destroy(nvl);
4526 free(nvlpacked, M_NVLIST);
4527
4528 break;
4529 }
4530 case DIOCADDRULE: {
4531 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
4532 struct pf_krule *rule;
4533
4534 rule = pf_krule_alloc();
4535 error = pf_rule_to_krule(&pr->rule, rule);
4536 if (error != 0) {
4537 pf_krule_free(rule);
4538 goto fail;
4539 }
4540
4541 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
4542
4543 /* Frees rule on error */
4544 error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
4545 pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
4546 td->td_proc ? td->td_proc->p_pid : 0);
4547 break;
4548 }
4549
4550 case DIOCGETRULES: {
4551 struct pfioc_rule *pr = (struct pfioc_rule *)addr;
4552
4553 pr->anchor[sizeof(pr->anchor) - 1] = '\0';
4554
4555 error = pf_ioctl_getrules(pr);
4556
4557 break;
4558 }
4559
4560 case DIOCGETRULENV: {
4561 PF_RULES_RLOCK_TRACKER;
4562 struct pfioc_nv *nv = (struct pfioc_nv *)addr;
4563 nvlist_t *nvrule = NULL;
4564 nvlist_t *nvl = NULL;
4565 struct pf_kruleset *ruleset;
4566 struct pf_krule *rule;
4567 void *nvlpacked = NULL;
4568 int rs_num, nr;
4569 bool clear_counter = false;
4570
4571 #define ERROUT(x) ERROUT_IOCTL(DIOCGETRULENV_error, x)
4572 #define ERROUT_LOCKED(x) do { \
4573 if (clear_counter) \
4574 PF_RULES_WUNLOCK(); \
4575 else \
4576 PF_RULES_RUNLOCK(); \
4577 ERROUT(x); \
4578 } while (0)
4579
4580 if (nv->len > pf_ioctl_maxcount)
4581 ERROUT(ENOMEM);
4582
4583 /* Copy the request in */
4584 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
4585 error = copyin(nv->data, nvlpacked, nv->len);
4586 if (error)
4587 ERROUT(error);
4588
4589 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
4590 if (nvl == NULL)
4591 ERROUT(EBADMSG);
4592
4593 if (! nvlist_exists_string(nvl, "anchor"))
4594 ERROUT(EBADMSG);
4595 if (! nvlist_exists_number(nvl, "ruleset"))
4596 ERROUT(EBADMSG);
4597 if (! nvlist_exists_number(nvl, "ticket"))
4598 ERROUT(EBADMSG);
4599 if (! nvlist_exists_number(nvl, "nr"))
4600 ERROUT(EBADMSG);
4601
4602 if (nvlist_exists_bool(nvl, "clear_counter"))
4603 clear_counter = nvlist_get_bool(nvl, "clear_counter");
4604
4605 if (clear_counter && !(flags & FWRITE))
4606 ERROUT(EACCES);
4607
4608 nr = nvlist_get_number(nvl, "nr");
4609
4610 if (clear_counter)
4611 PF_RULES_WLOCK();
4612 else
4613 PF_RULES_RLOCK();
4614 ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
4615 if (ruleset == NULL)
4616 ERROUT_LOCKED(ENOENT);
4617
4618 rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
4619 if (rs_num >= PF_RULESET_MAX)
4620 ERROUT_LOCKED(EINVAL);
4621
4622 if (nvlist_get_number(nvl, "ticket") !=
4623 ruleset->rules[rs_num].active.ticket)
4624 ERROUT_LOCKED(EBUSY);
4625
4626 if ((error = nvlist_error(nvl)))
4627 ERROUT_LOCKED(error);
4628
4629 rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
4630 while ((rule != NULL) && (rule->nr != nr))
4631 rule = TAILQ_NEXT(rule, entries);
4632 if (rule == NULL)
4633 ERROUT_LOCKED(EBUSY);
4634
4635 nvrule = pf_krule_to_nvrule(rule);
4636
4637 nvlist_destroy(nvl);
4638 nvl = nvlist_create(0);
4639 if (nvl == NULL)
4640 ERROUT_LOCKED(ENOMEM);
4641 nvlist_add_number(nvl, "nr", nr);
4642 nvlist_add_nvlist(nvl, "rule", nvrule);
4643 nvlist_destroy(nvrule);
4644 nvrule = NULL;
4645 if (pf_kanchor_nvcopyout(ruleset, rule, nvl))
4646 ERROUT_LOCKED(EBUSY);
4647
4648 free(nvlpacked, M_NVLIST);
4649 nvlpacked = nvlist_pack(nvl, &nv->len);
4650 if (nvlpacked == NULL)
4651 ERROUT_LOCKED(ENOMEM);
4652
4653 if (nv->size == 0)
4654 ERROUT_LOCKED(0);
4655 else if (nv->size < nv->len)
4656 ERROUT_LOCKED(ENOSPC);
4657
4658 if (clear_counter) {
4659 pf_krule_clear_counters(rule);
4660 PF_RULES_WUNLOCK();
4661 } else {
4662 PF_RULES_RUNLOCK();
4663 }
4664
4665 error = copyout(nvlpacked, nv->data, nv->len);
4666
4667 #undef ERROUT_LOCKED
4668 #undef ERROUT
4669 DIOCGETRULENV_error:
4670 free(nvlpacked, M_NVLIST);
4671 nvlist_destroy(nvrule);
4672 nvlist_destroy(nvl);
4673
4674 break;
4675 }
4676
4677 case DIOCCHANGERULE: {
4678 struct pfioc_rule *pcr = (struct pfioc_rule *)addr;
4679 struct pf_kruleset *ruleset;
4680 struct pf_krule *oldrule = NULL, *newrule = NULL;
4681 struct pfi_kkif *kif = NULL;
4682 struct pf_kpooladdr *pa;
4683 u_int32_t nr = 0;
4684 int rs_num;
4685
4686 pcr->anchor[sizeof(pcr->anchor) - 1] = '\0';
4687
4688 if (pcr->action < PF_CHANGE_ADD_HEAD ||
4689 pcr->action > PF_CHANGE_GET_TICKET) {
4690 error = EINVAL;
4691 goto fail;
4692 }
4693 if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
4694 error = EINVAL;
4695 goto fail;
4696 }
4697
4698 if (pcr->action != PF_CHANGE_REMOVE) {
4699 newrule = pf_krule_alloc();
4700 error = pf_rule_to_krule(&pcr->rule, newrule);
4701 if (error != 0) {
4702 pf_krule_free(newrule);
4703 goto fail;
4704 }
4705
4706 if ((error = pf_rule_checkaf(newrule))) {
4707 pf_krule_free(newrule);
4708 goto fail;
4709 }
4710 if (newrule->ifname[0])
4711 kif = pf_kkif_create(M_WAITOK);
4712 pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
4713 for (int i = 0; i < 2; i++) {
4714 pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
4715 pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
4716 }
4717 newrule->states_cur = counter_u64_alloc(M_WAITOK);
4718 newrule->states_tot = counter_u64_alloc(M_WAITOK);
4719 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
4720 newrule->src_nodes[sn_type] = counter_u64_alloc(M_WAITOK);
4721 newrule->cuid = td->td_ucred->cr_ruid;
4722 newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
4723 TAILQ_INIT(&newrule->nat.list);
4724 TAILQ_INIT(&newrule->rdr.list);
4725 TAILQ_INIT(&newrule->route.list);
4726 }
4727 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGERULE_error, x)
4728
4729 PF_CONFIG_LOCK();
4730 PF_RULES_WLOCK();
4731 #ifdef PF_WANT_32_TO_64_COUNTER
4732 if (newrule != NULL) {
4733 LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
4734 newrule->allrulelinked = true;
4735 V_pf_allrulecount++;
4736 }
4737 #endif
4738
4739 if (!(pcr->action == PF_CHANGE_REMOVE ||
4740 pcr->action == PF_CHANGE_GET_TICKET) &&
4741 pcr->pool_ticket != V_ticket_pabuf)
4742 ERROUT(EBUSY);
4743
4744 ruleset = pf_find_kruleset(pcr->anchor);
4745 if (ruleset == NULL)
4746 ERROUT(EINVAL);
4747
4748 rs_num = pf_get_ruleset_number(pcr->rule.action);
4749 if (rs_num >= PF_RULESET_MAX)
4750 ERROUT(EINVAL);
4751
4752 /*
4753 * XXXMJG: there is no guarantee that the ruleset was
4754 * created by the usual route of calling DIOCXBEGIN.
4755 * As a result it is possible the rule tree will not
4756 * be allocated yet. Hack around it by doing it here.
4757 * Note it is fine to let the tree persist in case of
4758 * error as it will be freed down the road on future
4759 * updates (if need be).
4760 */
4761 if (ruleset->rules[rs_num].active.tree == NULL) {
4762 ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
4763 if (ruleset->rules[rs_num].active.tree == NULL) {
4764 ERROUT(ENOMEM);
4765 }
4766 }
4767
4768 if (pcr->action == PF_CHANGE_GET_TICKET) {
4769 pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
4770 ERROUT(0);
4771 } else if (pcr->ticket !=
4772 ruleset->rules[rs_num].active.ticket)
4773 ERROUT(EINVAL);
4774
4775 if (pcr->action != PF_CHANGE_REMOVE) {
4776 if (newrule->ifname[0]) {
4777 newrule->kif = pfi_kkif_attach(kif,
4778 newrule->ifname);
4779 kif = NULL;
4780 pfi_kkif_ref(newrule->kif);
4781 } else
4782 newrule->kif = NULL;
4783
4784 if (newrule->rtableid > 0 &&
4785 newrule->rtableid >= rt_numfibs)
4786 error = EBUSY;
4787
4788 #ifdef ALTQ
4789 /* set queue IDs */
4790 if (newrule->qname[0] != 0) {
4791 if ((newrule->qid =
4792 pf_qname2qid(newrule->qname, true)) == 0)
4793 error = EBUSY;
4794 else if (newrule->pqname[0] != 0) {
4795 if ((newrule->pqid =
4796 pf_qname2qid(newrule->pqname, true)) == 0)
4797 error = EBUSY;
4798 } else
4799 newrule->pqid = newrule->qid;
4800 }
4801 #endif /* ALTQ */
4802 if (newrule->tagname[0])
4803 if ((newrule->tag =
4804 pf_tagname2tag(newrule->tagname)) == 0)
4805 error = EBUSY;
4806 if (newrule->match_tagname[0])
4807 if ((newrule->match_tag = pf_tagname2tag(
4808 newrule->match_tagname)) == 0)
4809 error = EBUSY;
4810 if (newrule->rt && !newrule->direction)
4811 error = EINVAL;
4812 if (!newrule->log)
4813 newrule->logif = 0;
4814 if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
4815 error = ENOMEM;
4816 if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
4817 error = ENOMEM;
4818 if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
4819 error = EINVAL;
4820 for (int i = 0; i < 3; i++) {
4821 TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries)
4822 if (pa->addr.type == PF_ADDR_TABLE) {
4823 pa->addr.p.tbl =
4824 pfr_attach_table(ruleset,
4825 pa->addr.v.tblname);
4826 if (pa->addr.p.tbl == NULL)
4827 error = ENOMEM;
4828 }
4829 }
4830
4831 newrule->overload_tbl = NULL;
4832 if (newrule->overload_tblname[0]) {
4833 if ((newrule->overload_tbl = pfr_attach_table(
4834 ruleset, newrule->overload_tblname)) ==
4835 NULL)
4836 error = EINVAL;
4837 else
4838 newrule->overload_tbl->pfrkt_flags |=
4839 PFR_TFLAG_ACTIVE;
4840 }
4841
4842 pf_mv_kpool(&V_pf_pabuf[0], &newrule->nat.list);
4843 pf_mv_kpool(&V_pf_pabuf[1], &newrule->rdr.list);
4844 pf_mv_kpool(&V_pf_pabuf[2], &newrule->route.list);
4845 if (((((newrule->action == PF_NAT) ||
4846 (newrule->action == PF_RDR) ||
4847 (newrule->action == PF_BINAT) ||
4848 (newrule->rt > PF_NOPFROUTE)) &&
4849 !newrule->anchor)) &&
4850 (TAILQ_FIRST(&newrule->rdr.list) == NULL))
4851 error = EINVAL;
4852
4853 if (error) {
4854 pf_free_rule(newrule);
4855 PF_RULES_WUNLOCK();
4856 PF_CONFIG_UNLOCK();
4857 goto fail;
4858 }
4859
4860 newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list);
4861 newrule->rdr.cur = TAILQ_FIRST(&newrule->rdr.list);
4862 }
4863 pf_empty_kpool(&V_pf_pabuf[0]);
4864 pf_empty_kpool(&V_pf_pabuf[1]);
4865 pf_empty_kpool(&V_pf_pabuf[2]);
4866
4867 if (pcr->action == PF_CHANGE_ADD_HEAD)
4868 oldrule = TAILQ_FIRST(
4869 ruleset->rules[rs_num].active.ptr);
4870 else if (pcr->action == PF_CHANGE_ADD_TAIL)
4871 oldrule = TAILQ_LAST(
4872 ruleset->rules[rs_num].active.ptr, pf_krulequeue);
4873 else {
4874 oldrule = TAILQ_FIRST(
4875 ruleset->rules[rs_num].active.ptr);
4876 while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
4877 oldrule = TAILQ_NEXT(oldrule, entries);
4878 if (oldrule == NULL) {
4879 if (newrule != NULL)
4880 pf_free_rule(newrule);
4881 PF_RULES_WUNLOCK();
4882 PF_CONFIG_UNLOCK();
4883 error = EINVAL;
4884 goto fail;
4885 }
4886 }
4887
4888 if (pcr->action == PF_CHANGE_REMOVE) {
4889 pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
4890 oldrule);
4891 RB_REMOVE(pf_krule_global,
4892 ruleset->rules[rs_num].active.tree, oldrule);
4893 ruleset->rules[rs_num].active.rcount--;
4894 } else {
4895 pf_hash_rule(newrule);
4896 RB_INSERT(pf_krule_global,
4897 ruleset->rules[rs_num].active.tree, newrule);
4898
4899 if (oldrule == NULL)
4900 TAILQ_INSERT_TAIL(
4901 ruleset->rules[rs_num].active.ptr,
4902 newrule, entries);
4903 else if (pcr->action == PF_CHANGE_ADD_HEAD ||
4904 pcr->action == PF_CHANGE_ADD_BEFORE)
4905 TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
4906 else
4907 TAILQ_INSERT_AFTER(
4908 ruleset->rules[rs_num].active.ptr,
4909 oldrule, newrule, entries);
4910 ruleset->rules[rs_num].active.rcount++;
4911 }
4912
4913 nr = 0;
4914 TAILQ_FOREACH(oldrule,
4915 ruleset->rules[rs_num].active.ptr, entries)
4916 oldrule->nr = nr++;
4917
4918 ruleset->rules[rs_num].active.ticket++;
4919
4920 pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
4921 pf_remove_if_empty_kruleset(ruleset);
4922
4923 PF_RULES_WUNLOCK();
4924 PF_CONFIG_UNLOCK();
4925 break;
4926
4927 #undef ERROUT
4928 DIOCCHANGERULE_error:
4929 PF_RULES_WUNLOCK();
4930 PF_CONFIG_UNLOCK();
4931 pf_krule_free(newrule);
4932 pf_kkif_free(kif);
4933 break;
4934 }
4935
4936 case DIOCCLRSTATESNV: {
4937 error = pf_clearstates_nv((struct pfioc_nv *)addr);
4938 break;
4939 }
4940
4941 case DIOCKILLSTATESNV: {
4942 error = pf_killstates_nv((struct pfioc_nv *)addr);
4943 break;
4944 }
4945
4946 case DIOCADDSTATE: {
4947 struct pfioc_state *ps = (struct pfioc_state *)addr;
4948 struct pfsync_state_1301 *sp = &ps->state;
4949
4950 if (sp->timeout >= PFTM_MAX) {
4951 error = EINVAL;
4952 goto fail;
4953 }
4954 if (V_pfsync_state_import_ptr != NULL) {
4955 PF_RULES_RLOCK();
4956 error = V_pfsync_state_import_ptr(
4957 (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
4958 PFSYNC_MSG_VERSION_1301);
4959 PF_RULES_RUNLOCK();
4960 } else
4961 error = EOPNOTSUPP;
4962 break;
4963 }
4964
4965 case DIOCGETSTATE: {
4966 struct pfioc_state *ps = (struct pfioc_state *)addr;
4967 struct pf_kstate *s;
4968
4969 s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
4970 if (s == NULL) {
4971 error = ENOENT;
4972 goto fail;
4973 }
4974
4975 pfsync_state_export_1301(&ps->state, s);
4976 PF_STATE_UNLOCK(s);
4977 break;
4978 }
4979
4980 case DIOCGETSTATENV: {
4981 error = pf_getstate((struct pfioc_nv *)addr);
4982 break;
4983 }
4984
4985 #ifdef COMPAT_FREEBSD14
4986 case DIOCGETSTATES: {
4987 struct pfioc_states *ps = (struct pfioc_states *)addr;
4988 struct pf_kstate *s;
4989 struct pfsync_state_1301 *pstore, *p;
4990 int i, nr;
4991 size_t slice_count = 16, count;
4992 void *out;
4993
4994 if (ps->ps_len <= 0) {
4995 nr = uma_zone_get_cur(V_pf_state_z);
4996 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
4997 break;
4998 }
4999
5000 out = ps->ps_states;
5001 pstore = mallocarray(slice_count,
5002 sizeof(struct pfsync_state_1301), M_PF, M_WAITOK | M_ZERO);
5003 nr = 0;
5004
5005 for (i = 0; i <= V_pf_hashmask; i++) {
5006 struct pf_idhash *ih = &V_pf_idhash[i];
5007
5008 DIOCGETSTATES_retry:
5009 p = pstore;
5010
5011 if (LIST_EMPTY(&ih->states))
5012 continue;
5013
5014 PF_HASHROW_LOCK(ih);
5015 count = 0;
5016 LIST_FOREACH(s, &ih->states, entry) {
5017 if (s->timeout == PFTM_UNLINKED)
5018 continue;
5019 count++;
5020 }
5021
5022 if (count > slice_count) {
5023 PF_HASHROW_UNLOCK(ih);
5024 free(pstore, M_PF);
5025 slice_count = count * 2;
5026 pstore = mallocarray(slice_count,
5027 sizeof(struct pfsync_state_1301), M_PF,
5028 M_WAITOK | M_ZERO);
5029 goto DIOCGETSTATES_retry;
5030 }
5031
5032 if ((nr+count) * sizeof(*p) > ps->ps_len) {
5033 PF_HASHROW_UNLOCK(ih);
5034 goto DIOCGETSTATES_full;
5035 }
5036
5037 LIST_FOREACH(s, &ih->states, entry) {
5038 if (s->timeout == PFTM_UNLINKED)
5039 continue;
5040
5041 pfsync_state_export_1301(p, s);
5042 p++;
5043 nr++;
5044 }
5045 PF_HASHROW_UNLOCK(ih);
5046 error = copyout(pstore, out,
5047 sizeof(struct pfsync_state_1301) * count);
5048 if (error) {
5049 free(pstore, M_PF);
5050 goto fail;
5051 }
5052 out = ps->ps_states + nr;
5053 }
5054 DIOCGETSTATES_full:
5055 ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
5056 free(pstore, M_PF);
5057
5058 break;
5059 }
5060
5061 case DIOCGETSTATESV2: {
5062 struct pfioc_states_v2 *ps = (struct pfioc_states_v2 *)addr;
5063 struct pf_kstate *s;
5064 struct pf_state_export *pstore, *p;
5065 int i, nr;
5066 size_t slice_count = 16, count;
5067 void *out;
5068
5069 if (ps->ps_req_version > PF_STATE_VERSION) {
5070 error = ENOTSUP;
5071 goto fail;
5072 }
5073
5074 if (ps->ps_len <= 0) {
5075 nr = uma_zone_get_cur(V_pf_state_z);
5076 ps->ps_len = sizeof(struct pf_state_export) * nr;
5077 break;
5078 }
5079
5080 out = ps->ps_states;
5081 pstore = mallocarray(slice_count,
5082 sizeof(struct pf_state_export), M_PF, M_WAITOK | M_ZERO);
5083 nr = 0;
5084
5085 for (i = 0; i <= V_pf_hashmask; i++) {
5086 struct pf_idhash *ih = &V_pf_idhash[i];
5087
5088 DIOCGETSTATESV2_retry:
5089 p = pstore;
5090
5091 if (LIST_EMPTY(&ih->states))
5092 continue;
5093
5094 PF_HASHROW_LOCK(ih);
5095 count = 0;
5096 LIST_FOREACH(s, &ih->states, entry) {
5097 if (s->timeout == PFTM_UNLINKED)
5098 continue;
5099 count++;
5100 }
5101
5102 if (count > slice_count) {
5103 PF_HASHROW_UNLOCK(ih);
5104 free(pstore, M_PF);
5105 slice_count = count * 2;
5106 pstore = mallocarray(slice_count,
5107 sizeof(struct pf_state_export), M_PF,
5108 M_WAITOK | M_ZERO);
5109 goto DIOCGETSTATESV2_retry;
5110 }
5111
5112 if ((nr+count) * sizeof(*p) > ps->ps_len) {
5113 PF_HASHROW_UNLOCK(ih);
5114 goto DIOCGETSTATESV2_full;
5115 }
5116
5117 LIST_FOREACH(s, &ih->states, entry) {
5118 if (s->timeout == PFTM_UNLINKED)
5119 continue;
5120
5121 pf_state_export(p, s);
5122 p++;
5123 nr++;
5124 }
5125 PF_HASHROW_UNLOCK(ih);
5126 error = copyout(pstore, out,
5127 sizeof(struct pf_state_export) * count);
5128 if (error) {
5129 free(pstore, M_PF);
5130 goto fail;
5131 }
5132 out = ps->ps_states + nr;
5133 }
5134 DIOCGETSTATESV2_full:
5135 ps->ps_len = nr * sizeof(struct pf_state_export);
5136 free(pstore, M_PF);
5137
5138 break;
5139 }
5140 #endif
5141 case DIOCGETSTATUSNV: {
5142 error = pf_getstatus((struct pfioc_nv *)addr);
5143 break;
5144 }
5145
5146 case DIOCSETSTATUSIF: {
5147 struct pfioc_if *pi = (struct pfioc_if *)addr;
5148
5149 if (pi->ifname[0] == 0) {
5150 bzero(V_pf_status.ifname, IFNAMSIZ);
5151 break;
5152 }
5153 PF_RULES_WLOCK();
5154 error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
5155 PF_RULES_WUNLOCK();
5156 break;
5157 }
5158
5159 case DIOCCLRSTATUS: {
5160 pf_ioctl_clear_status();
5161 break;
5162 }
5163
5164 case DIOCNATLOOK: {
5165 struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr;
5166
5167 error = pf_ioctl_natlook(pnl);
5168 break;
5169 }
5170
5171 case DIOCSETTIMEOUT: {
5172 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
5173
5174 error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
5175 &pt->seconds);
5176 break;
5177 }
5178
5179 case DIOCGETTIMEOUT: {
5180 struct pfioc_tm *pt = (struct pfioc_tm *)addr;
5181
5182 error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
5183 break;
5184 }
5185
5186 case DIOCGETLIMIT: {
5187 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
5188
5189 error = pf_ioctl_get_limit(pl->index, &pl->limit);
5190 break;
5191 }
5192
5193 case DIOCSETLIMIT: {
5194 struct pfioc_limit *pl = (struct pfioc_limit *)addr;
5195 unsigned int old_limit;
5196
5197 error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
5198 pl->limit = old_limit;
5199 break;
5200 }
5201
5202 case DIOCSETDEBUG: {
5203 u_int32_t *level = (u_int32_t *)addr;
5204
5205 PF_RULES_WLOCK();
5206 V_pf_status.debug = *level;
5207 PF_RULES_WUNLOCK();
5208 break;
5209 }
5210
5211 case DIOCCLRRULECTRS: {
5212 /* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
5213 struct pf_kruleset *ruleset = &pf_main_ruleset;
5214 struct pf_krule *rule;
5215
5216 PF_RULES_WLOCK();
5217 TAILQ_FOREACH(rule,
5218 ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
5219 pf_counter_u64_zero(&rule->evaluations);
5220 for (int i = 0; i < 2; i++) {
5221 pf_counter_u64_zero(&rule->packets[i]);
5222 pf_counter_u64_zero(&rule->bytes[i]);
5223 }
5224 }
5225 PF_RULES_WUNLOCK();
5226 break;
5227 }
5228
5229 case DIOCGIFSPEEDV0:
5230 case DIOCGIFSPEEDV1: {
5231 struct pf_ifspeed_v1 *psp = (struct pf_ifspeed_v1 *)addr;
5232 struct pf_ifspeed_v1 ps;
5233 struct ifnet *ifp;
5234
5235 if (psp->ifname[0] == '\0') {
5236 error = EINVAL;
5237 goto fail;
5238 }
5239
5240 error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
5241 if (error != 0)
5242 goto fail;
5243 ifp = ifunit(ps.ifname);
5244 if (ifp != NULL) {
5245 psp->baudrate32 =
5246 (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
5247 if (cmd == DIOCGIFSPEEDV1)
5248 psp->baudrate = ifp->if_baudrate;
5249 } else {
5250 error = EINVAL;
5251 }
5252 break;
5253 }
5254
5255 #ifdef ALTQ
5256 case DIOCSTARTALTQ: {
5257 struct pf_altq *altq;
5258
5259 PF_RULES_WLOCK();
5260 /* enable all altq interfaces on active list */
5261 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
5262 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
5263 error = pf_enable_altq(altq);
5264 if (error != 0)
5265 break;
5266 }
5267 }
5268 if (error == 0)
5269 V_pf_altq_running = 1;
5270 PF_RULES_WUNLOCK();
5271 DPFPRINTF(PF_DEBUG_MISC, "altq: started");
5272 break;
5273 }
5274
5275 case DIOCSTOPALTQ: {
5276 struct pf_altq *altq;
5277
5278 PF_RULES_WLOCK();
5279 /* disable all altq interfaces on active list */
5280 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
5281 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
5282 error = pf_disable_altq(altq);
5283 if (error != 0)
5284 break;
5285 }
5286 }
5287 if (error == 0)
5288 V_pf_altq_running = 0;
5289 PF_RULES_WUNLOCK();
5290 DPFPRINTF(PF_DEBUG_MISC, "altq: stopped");
5291 break;
5292 }
5293
5294 case DIOCADDALTQV0:
5295 case DIOCADDALTQV1: {
5296 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
5297 struct pf_altq *altq, *a;
5298 struct ifnet *ifp;
5299
5300 altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
5301 error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
5302 if (error)
5303 goto fail;
5304 altq->local_flags = 0;
5305
5306 PF_RULES_WLOCK();
5307 if (pa->ticket != V_ticket_altqs_inactive) {
5308 PF_RULES_WUNLOCK();
5309 free(altq, M_PFALTQ);
5310 error = EBUSY;
5311 goto fail;
5312 }
5313
5314 /*
5315 * if this is for a queue, find the discipline and
5316 * copy the necessary fields
5317 */
5318 if (altq->qname[0] != 0) {
5319 if ((altq->qid = pf_qname2qid(altq->qname, true)) == 0) {
5320 PF_RULES_WUNLOCK();
5321 error = EBUSY;
5322 free(altq, M_PFALTQ);
5323 goto fail;
5324 }
5325 altq->altq_disc = NULL;
5326 TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
5327 if (strncmp(a->ifname, altq->ifname,
5328 IFNAMSIZ) == 0) {
5329 altq->altq_disc = a->altq_disc;
5330 break;
5331 }
5332 }
5333 }
5334
5335 if ((ifp = ifunit(altq->ifname)) == NULL)
5336 altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
5337 else
5338 error = altq_add(ifp, altq);
5339
5340 if (error) {
5341 PF_RULES_WUNLOCK();
5342 free(altq, M_PFALTQ);
5343 goto fail;
5344 }
5345
5346 if (altq->qname[0] != 0)
5347 TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
5348 else
5349 TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
5350 /* version error check done on import above */
5351 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
5352 PF_RULES_WUNLOCK();
5353 break;
5354 }
5355
5356 case DIOCGETALTQSV0:
5357 case DIOCGETALTQSV1: {
5358 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
5359 struct pf_altq *altq;
5360
5361 PF_RULES_RLOCK();
5362 pa->nr = 0;
5363 TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
5364 pa->nr++;
5365 TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
5366 pa->nr++;
5367 pa->ticket = V_ticket_altqs_active;
5368 PF_RULES_RUNLOCK();
5369 break;
5370 }
5371
5372 case DIOCGETALTQV0:
5373 case DIOCGETALTQV1: {
5374 struct pfioc_altq_v1 *pa = (struct pfioc_altq_v1 *)addr;
5375 struct pf_altq *altq;
5376
5377 PF_RULES_RLOCK();
5378 if (pa->ticket != V_ticket_altqs_active) {
5379 PF_RULES_RUNLOCK();
5380 error = EBUSY;
5381 goto fail;
5382 }
5383 altq = pf_altq_get_nth_active(pa->nr);
5384 if (altq == NULL) {
5385 PF_RULES_RUNLOCK();
5386 error = EBUSY;
5387 goto fail;
5388 }
5389 pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
5390 PF_RULES_RUNLOCK();
5391 break;
5392 }
5393
5394 case DIOCCHANGEALTQV0:
5395 case DIOCCHANGEALTQV1:
5396 /* CHANGEALTQ not supported yet! */
5397 error = ENODEV;
5398 break;
5399
5400 case DIOCGETQSTATSV0:
5401 case DIOCGETQSTATSV1: {
5402 struct pfioc_qstats_v1 *pq = (struct pfioc_qstats_v1 *)addr;
5403 struct pf_altq *altq;
5404 int nbytes;
5405 u_int32_t version;
5406
5407 PF_RULES_RLOCK();
5408 if (pq->ticket != V_ticket_altqs_active) {
5409 PF_RULES_RUNLOCK();
5410 error = EBUSY;
5411 goto fail;
5412 }
5413 nbytes = pq->nbytes;
5414 altq = pf_altq_get_nth_active(pq->nr);
5415 if (altq == NULL) {
5416 PF_RULES_RUNLOCK();
5417 error = EBUSY;
5418 goto fail;
5419 }
5420
5421 if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
5422 PF_RULES_RUNLOCK();
5423 error = ENXIO;
5424 goto fail;
5425 }
5426 PF_RULES_RUNLOCK();
5427 if (cmd == DIOCGETQSTATSV0)
5428 version = 0; /* DIOCGETQSTATSV0 means stats struct v0 */
5429 else
5430 version = pq->version;
5431 error = altq_getqstats(altq, pq->buf, &nbytes, version);
5432 if (error == 0) {
5433 pq->scheduler = altq->scheduler;
5434 pq->nbytes = nbytes;
5435 }
5436 break;
5437 }
5438 #endif /* ALTQ */
5439
5440 case DIOCBEGINADDRS: {
5441 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
5442
5443 error = pf_ioctl_begin_addrs(&pp->ticket);
5444 break;
5445 }
5446
5447 case DIOCADDADDR: {
5448 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
5449 struct pf_nl_pooladdr npp = {};
5450
5451 npp.which = PF_RDR;
5452 memcpy(&npp, pp, sizeof(*pp));
5453 error = pf_ioctl_add_addr(&npp);
5454 break;
5455 }
5456
5457 case DIOCGETADDRS: {
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_get_addrs(&npp);
5464 memcpy(pp, &npp, sizeof(*pp));
5465
5466 break;
5467 }
5468
5469 case DIOCGETADDR: {
5470 struct pfioc_pooladdr *pp = (struct pfioc_pooladdr *)addr;
5471 struct pf_nl_pooladdr npp = {};
5472
5473 npp.which = PF_RDR;
5474 memcpy(&npp, pp, sizeof(*pp));
5475 error = pf_ioctl_get_addr(&npp);
5476 memcpy(pp, &npp, sizeof(*pp));
5477
5478 break;
5479 }
5480
5481 case DIOCCHANGEADDR: {
5482 struct pfioc_pooladdr *pca = (struct pfioc_pooladdr *)addr;
5483 struct pf_kpool *pool;
5484 struct pf_kpooladdr *oldpa = NULL, *newpa = NULL;
5485 struct pf_kruleset *ruleset;
5486 struct pfi_kkif *kif = NULL;
5487
5488 pca->anchor[sizeof(pca->anchor) - 1] = '\0';
5489
5490 if (pca->action < PF_CHANGE_ADD_HEAD ||
5491 pca->action > PF_CHANGE_REMOVE) {
5492 error = EINVAL;
5493 goto fail;
5494 }
5495 if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
5496 pca->addr.addr.type != PF_ADDR_DYNIFTL &&
5497 pca->addr.addr.type != PF_ADDR_TABLE) {
5498 error = EINVAL;
5499 goto fail;
5500 }
5501 if (pca->addr.addr.p.dyn != NULL) {
5502 error = EINVAL;
5503 goto fail;
5504 }
5505
5506 if (pca->action != PF_CHANGE_REMOVE) {
5507 #ifndef INET
5508 if (pca->af == AF_INET) {
5509 error = EAFNOSUPPORT;
5510 goto fail;
5511 }
5512 #endif /* INET */
5513 #ifndef INET6
5514 if (pca->af == AF_INET6) {
5515 error = EAFNOSUPPORT;
5516 goto fail;
5517 }
5518 #endif /* INET6 */
5519 newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
5520 bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
5521 if (newpa->ifname[0])
5522 kif = pf_kkif_create(M_WAITOK);
5523 newpa->kif = NULL;
5524 }
5525 #define ERROUT(x) ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
5526 PF_RULES_WLOCK();
5527 ruleset = pf_find_kruleset(pca->anchor);
5528 if (ruleset == NULL)
5529 ERROUT(EBUSY);
5530
5531 pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
5532 pca->r_num, pca->r_last, 1, 1, PF_RDR);
5533 if (pool == NULL)
5534 ERROUT(EBUSY);
5535
5536 if (pca->action != PF_CHANGE_REMOVE) {
5537 if (newpa->ifname[0]) {
5538 newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
5539 pfi_kkif_ref(newpa->kif);
5540 kif = NULL;
5541 }
5542
5543 switch (newpa->addr.type) {
5544 case PF_ADDR_DYNIFTL:
5545 error = pfi_dynaddr_setup(&newpa->addr,
5546 pca->af);
5547 break;
5548 case PF_ADDR_TABLE:
5549 newpa->addr.p.tbl = pfr_attach_table(ruleset,
5550 newpa->addr.v.tblname);
5551 if (newpa->addr.p.tbl == NULL)
5552 error = ENOMEM;
5553 break;
5554 }
5555 if (error)
5556 goto DIOCCHANGEADDR_error;
5557 }
5558
5559 switch (pca->action) {
5560 case PF_CHANGE_ADD_HEAD:
5561 oldpa = TAILQ_FIRST(&pool->list);
5562 break;
5563 case PF_CHANGE_ADD_TAIL:
5564 oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
5565 break;
5566 default:
5567 oldpa = TAILQ_FIRST(&pool->list);
5568 for (int i = 0; oldpa && i < pca->nr; i++)
5569 oldpa = TAILQ_NEXT(oldpa, entries);
5570
5571 if (oldpa == NULL)
5572 ERROUT(EINVAL);
5573 }
5574
5575 if (pca->action == PF_CHANGE_REMOVE) {
5576 TAILQ_REMOVE(&pool->list, oldpa, entries);
5577 switch (oldpa->addr.type) {
5578 case PF_ADDR_DYNIFTL:
5579 pfi_dynaddr_remove(oldpa->addr.p.dyn);
5580 break;
5581 case PF_ADDR_TABLE:
5582 pfr_detach_table(oldpa->addr.p.tbl);
5583 break;
5584 }
5585 if (oldpa->kif)
5586 pfi_kkif_unref(oldpa->kif);
5587 free(oldpa, M_PFRULE);
5588 } else {
5589 if (oldpa == NULL)
5590 TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
5591 else if (pca->action == PF_CHANGE_ADD_HEAD ||
5592 pca->action == PF_CHANGE_ADD_BEFORE)
5593 TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
5594 else
5595 TAILQ_INSERT_AFTER(&pool->list, oldpa,
5596 newpa, entries);
5597 }
5598
5599 pool->cur = TAILQ_FIRST(&pool->list);
5600 pf_addrcpy(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
5601 PF_RULES_WUNLOCK();
5602 break;
5603
5604 #undef ERROUT
5605 DIOCCHANGEADDR_error:
5606 if (newpa != NULL) {
5607 if (newpa->kif)
5608 pfi_kkif_unref(newpa->kif);
5609 free(newpa, M_PFRULE);
5610 }
5611 PF_RULES_WUNLOCK();
5612 pf_kkif_free(kif);
5613 break;
5614 }
5615
5616 case DIOCGETRULESETS: {
5617 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
5618
5619 pr->path[sizeof(pr->path) - 1] = '\0';
5620
5621 error = pf_ioctl_get_rulesets(pr);
5622 break;
5623 }
5624
5625 case DIOCGETRULESET: {
5626 struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr;
5627
5628 pr->path[sizeof(pr->path) - 1] = '\0';
5629
5630 error = pf_ioctl_get_ruleset(pr);
5631 break;
5632 }
5633
5634 case DIOCRCLRTABLES: {
5635 struct pfioc_table *io = (struct pfioc_table *)addr;
5636
5637 if (io->pfrio_esize != 0) {
5638 error = ENODEV;
5639 goto fail;
5640 }
5641 if (strnlen(io->pfrio_table.pfrt_anchor, MAXPATHLEN)
5642 == MAXPATHLEN) {
5643 error = EINVAL;
5644 goto fail;
5645 }
5646 if (strnlen(io->pfrio_table.pfrt_name, PF_TABLE_NAME_SIZE)
5647 == PF_TABLE_NAME_SIZE) {
5648 error = EINVAL;
5649 goto fail;
5650 }
5651
5652 PF_RULES_WLOCK();
5653 error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
5654 io->pfrio_flags | PFR_FLAG_USERIOCTL);
5655 PF_RULES_WUNLOCK();
5656 break;
5657 }
5658
5659 case DIOCRADDTABLES: {
5660 struct pfioc_table *io = (struct pfioc_table *)addr;
5661 struct pfr_table *pfrts;
5662 size_t totlen;
5663
5664 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5665 error = ENODEV;
5666 goto fail;
5667 }
5668
5669 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
5670 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
5671 error = ENOMEM;
5672 goto fail;
5673 }
5674
5675 totlen = io->pfrio_size * sizeof(struct pfr_table);
5676 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5677 M_PF, M_WAITOK);
5678 error = copyin(io->pfrio_buffer, pfrts, totlen);
5679 if (error) {
5680 free(pfrts, M_PF);
5681 goto fail;
5682 }
5683 PF_RULES_WLOCK();
5684 error = pfr_add_tables(pfrts, io->pfrio_size,
5685 &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5686 PF_RULES_WUNLOCK();
5687 free(pfrts, M_PF);
5688 break;
5689 }
5690
5691 case DIOCRDELTABLES: {
5692 struct pfioc_table *io = (struct pfioc_table *)addr;
5693 struct pfr_table *pfrts;
5694 size_t totlen;
5695
5696 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5697 error = ENODEV;
5698 goto fail;
5699 }
5700
5701 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
5702 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
5703 error = ENOMEM;
5704 goto fail;
5705 }
5706
5707 totlen = io->pfrio_size * sizeof(struct pfr_table);
5708 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5709 M_PF, M_WAITOK);
5710 error = copyin(io->pfrio_buffer, pfrts, totlen);
5711 if (error) {
5712 free(pfrts, M_PF);
5713 goto fail;
5714 }
5715 PF_RULES_WLOCK();
5716 error = pfr_del_tables(pfrts, io->pfrio_size,
5717 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5718 PF_RULES_WUNLOCK();
5719 free(pfrts, M_PF);
5720 break;
5721 }
5722
5723 case DIOCRGETTABLES: {
5724 struct pfioc_table *io = (struct pfioc_table *)addr;
5725 struct pfr_table *pfrts;
5726 size_t totlen;
5727 int n;
5728
5729 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5730 error = ENODEV;
5731 goto fail;
5732 }
5733 PF_RULES_RLOCK();
5734 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
5735 if (n < 0) {
5736 PF_RULES_RUNLOCK();
5737 error = EINVAL;
5738 goto fail;
5739 }
5740 io->pfrio_size = min(io->pfrio_size, n);
5741
5742 totlen = io->pfrio_size * sizeof(struct pfr_table);
5743
5744 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5745 M_PF, M_NOWAIT | M_ZERO);
5746 if (pfrts == NULL) {
5747 error = ENOMEM;
5748 PF_RULES_RUNLOCK();
5749 goto fail;
5750 }
5751 error = pfr_get_tables(&io->pfrio_table, pfrts,
5752 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5753 PF_RULES_RUNLOCK();
5754 if (error == 0)
5755 error = copyout(pfrts, io->pfrio_buffer, totlen);
5756 free(pfrts, M_PF);
5757 break;
5758 }
5759
5760 case DIOCRGETTSTATS: {
5761 struct pfioc_table *io = (struct pfioc_table *)addr;
5762 struct pfr_tstats *pfrtstats;
5763 size_t totlen;
5764 int n;
5765
5766 if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
5767 error = ENODEV;
5768 goto fail;
5769 }
5770 PF_TABLE_STATS_LOCK();
5771 PF_RULES_RLOCK();
5772 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
5773 if (n < 0) {
5774 PF_RULES_RUNLOCK();
5775 PF_TABLE_STATS_UNLOCK();
5776 error = EINVAL;
5777 goto fail;
5778 }
5779 io->pfrio_size = min(io->pfrio_size, n);
5780
5781 totlen = io->pfrio_size * sizeof(struct pfr_tstats);
5782 pfrtstats = mallocarray(io->pfrio_size,
5783 sizeof(struct pfr_tstats), M_PF, M_NOWAIT | M_ZERO);
5784 if (pfrtstats == NULL) {
5785 error = ENOMEM;
5786 PF_RULES_RUNLOCK();
5787 PF_TABLE_STATS_UNLOCK();
5788 goto fail;
5789 }
5790 error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
5791 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5792 PF_RULES_RUNLOCK();
5793 PF_TABLE_STATS_UNLOCK();
5794 if (error == 0)
5795 error = copyout(pfrtstats, io->pfrio_buffer, totlen);
5796 free(pfrtstats, M_PF);
5797 break;
5798 }
5799
5800 case DIOCRCLRTSTATS: {
5801 struct pfioc_table *io = (struct pfioc_table *)addr;
5802 struct pfr_table *pfrts;
5803 size_t totlen;
5804
5805 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5806 error = ENODEV;
5807 goto fail;
5808 }
5809
5810 if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
5811 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
5812 /* We used to count tables and use the minimum required
5813 * size, so we didn't fail on overly large requests.
5814 * Keep doing so. */
5815 io->pfrio_size = pf_ioctl_maxcount;
5816 goto fail;
5817 }
5818
5819 totlen = io->pfrio_size * sizeof(struct pfr_table);
5820 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5821 M_PF, M_WAITOK);
5822 error = copyin(io->pfrio_buffer, pfrts, totlen);
5823 if (error) {
5824 free(pfrts, M_PF);
5825 goto fail;
5826 }
5827
5828 PF_TABLE_STATS_LOCK();
5829 PF_RULES_RLOCK();
5830 error = pfr_clr_tstats(pfrts, io->pfrio_size,
5831 &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5832 PF_RULES_RUNLOCK();
5833 PF_TABLE_STATS_UNLOCK();
5834 free(pfrts, M_PF);
5835 break;
5836 }
5837
5838 case DIOCRSETTFLAGS: {
5839 struct pfioc_table *io = (struct pfioc_table *)addr;
5840 struct pfr_table *pfrts;
5841 size_t totlen;
5842 int n;
5843
5844 if (io->pfrio_esize != sizeof(struct pfr_table)) {
5845 error = ENODEV;
5846 goto fail;
5847 }
5848
5849 PF_RULES_RLOCK();
5850 n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
5851 if (n < 0) {
5852 PF_RULES_RUNLOCK();
5853 error = EINVAL;
5854 goto fail;
5855 }
5856
5857 io->pfrio_size = min(io->pfrio_size, n);
5858 PF_RULES_RUNLOCK();
5859
5860 totlen = io->pfrio_size * sizeof(struct pfr_table);
5861 pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
5862 M_PF, M_WAITOK);
5863 error = copyin(io->pfrio_buffer, pfrts, totlen);
5864 if (error) {
5865 free(pfrts, M_PF);
5866 goto fail;
5867 }
5868 PF_RULES_WLOCK();
5869 error = pfr_set_tflags(pfrts, io->pfrio_size,
5870 io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
5871 &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5872 PF_RULES_WUNLOCK();
5873 free(pfrts, M_PF);
5874 break;
5875 }
5876
5877 case DIOCRCLRADDRS: {
5878 struct pfioc_table *io = (struct pfioc_table *)addr;
5879
5880 if (io->pfrio_esize != 0) {
5881 error = ENODEV;
5882 goto fail;
5883 }
5884 PF_RULES_WLOCK();
5885 error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
5886 io->pfrio_flags | PFR_FLAG_USERIOCTL);
5887 PF_RULES_WUNLOCK();
5888 break;
5889 }
5890
5891 case DIOCRADDADDRS: {
5892 struct pfioc_table *io = (struct pfioc_table *)addr;
5893 struct pfr_addr *pfras;
5894 size_t totlen;
5895
5896 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5897 error = ENODEV;
5898 goto fail;
5899 }
5900 if (io->pfrio_size < 0 ||
5901 io->pfrio_size > pf_ioctl_maxcount ||
5902 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5903 error = EINVAL;
5904 goto fail;
5905 }
5906 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5907 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5908 M_PF, M_WAITOK);
5909 error = copyin(io->pfrio_buffer, pfras, totlen);
5910 if (error) {
5911 free(pfras, M_PF);
5912 goto fail;
5913 }
5914 PF_RULES_WLOCK();
5915 io->pfrio_nadd = 0;
5916 error = pfr_add_addrs(&io->pfrio_table, pfras,
5917 io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
5918 PFR_FLAG_USERIOCTL);
5919 PF_RULES_WUNLOCK();
5920 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5921 error = copyout(pfras, io->pfrio_buffer, totlen);
5922 free(pfras, M_PF);
5923 break;
5924 }
5925
5926 case DIOCRDELADDRS: {
5927 struct pfioc_table *io = (struct pfioc_table *)addr;
5928 struct pfr_addr *pfras;
5929 size_t totlen;
5930
5931 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5932 error = ENODEV;
5933 goto fail;
5934 }
5935 if (io->pfrio_size < 0 ||
5936 io->pfrio_size > pf_ioctl_maxcount ||
5937 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5938 error = EINVAL;
5939 goto fail;
5940 }
5941 totlen = io->pfrio_size * sizeof(struct pfr_addr);
5942 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5943 M_PF, M_WAITOK);
5944 error = copyin(io->pfrio_buffer, pfras, totlen);
5945 if (error) {
5946 free(pfras, M_PF);
5947 goto fail;
5948 }
5949 PF_RULES_WLOCK();
5950 error = pfr_del_addrs(&io->pfrio_table, pfras,
5951 io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
5952 PFR_FLAG_USERIOCTL);
5953 PF_RULES_WUNLOCK();
5954 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5955 error = copyout(pfras, io->pfrio_buffer, totlen);
5956 free(pfras, M_PF);
5957 break;
5958 }
5959
5960 case DIOCRSETADDRS: {
5961 struct pfioc_table *io = (struct pfioc_table *)addr;
5962 struct pfr_addr *pfras;
5963 size_t totlen, count;
5964
5965 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5966 error = ENODEV;
5967 goto fail;
5968 }
5969 if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
5970 error = EINVAL;
5971 goto fail;
5972 }
5973 count = max(io->pfrio_size, io->pfrio_size2);
5974 if (count > pf_ioctl_maxcount ||
5975 WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
5976 error = EINVAL;
5977 goto fail;
5978 }
5979 totlen = count * sizeof(struct pfr_addr);
5980 pfras = mallocarray(count, sizeof(struct pfr_addr), M_PF,
5981 M_WAITOK);
5982 error = copyin(io->pfrio_buffer, pfras, totlen);
5983 if (error) {
5984 free(pfras, M_PF);
5985 goto fail;
5986 }
5987 PF_RULES_WLOCK();
5988 error = pfr_set_addrs(&io->pfrio_table, pfras,
5989 io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
5990 &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
5991 PFR_FLAG_START | PFR_FLAG_DONE | PFR_FLAG_USERIOCTL, 0);
5992 PF_RULES_WUNLOCK();
5993 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
5994 error = copyout(pfras, io->pfrio_buffer, totlen);
5995 free(pfras, M_PF);
5996 break;
5997 }
5998
5999 case DIOCRGETADDRS: {
6000 struct pfioc_table *io = (struct pfioc_table *)addr;
6001 struct pfr_addr *pfras;
6002 size_t totlen;
6003
6004 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
6005 error = ENODEV;
6006 goto fail;
6007 }
6008 if (io->pfrio_size < 0 ||
6009 io->pfrio_size > pf_ioctl_maxcount ||
6010 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
6011 error = EINVAL;
6012 goto fail;
6013 }
6014 totlen = io->pfrio_size * sizeof(struct pfr_addr);
6015 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
6016 M_PF, M_WAITOK | M_ZERO);
6017 PF_RULES_RLOCK();
6018 error = pfr_get_addrs(&io->pfrio_table, pfras,
6019 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
6020 PF_RULES_RUNLOCK();
6021 if (error == 0)
6022 error = copyout(pfras, io->pfrio_buffer, totlen);
6023 free(pfras, M_PF);
6024 break;
6025 }
6026
6027 case DIOCRGETASTATS: {
6028 struct pfioc_table *io = (struct pfioc_table *)addr;
6029 struct pfr_astats *pfrastats;
6030 size_t totlen;
6031
6032 if (io->pfrio_esize != sizeof(struct pfr_astats)) {
6033 error = ENODEV;
6034 goto fail;
6035 }
6036 if (io->pfrio_size < 0 ||
6037 io->pfrio_size > pf_ioctl_maxcount ||
6038 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
6039 error = EINVAL;
6040 goto fail;
6041 }
6042 totlen = io->pfrio_size * sizeof(struct pfr_astats);
6043 pfrastats = mallocarray(io->pfrio_size,
6044 sizeof(struct pfr_astats), M_PF, M_WAITOK | M_ZERO);
6045 PF_RULES_RLOCK();
6046 error = pfr_get_astats(&io->pfrio_table, pfrastats,
6047 &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
6048 PF_RULES_RUNLOCK();
6049 if (error == 0)
6050 error = copyout(pfrastats, io->pfrio_buffer, totlen);
6051 free(pfrastats, M_PF);
6052 break;
6053 }
6054
6055 case DIOCRCLRASTATS: {
6056 struct pfioc_table *io = (struct pfioc_table *)addr;
6057 struct pfr_addr *pfras;
6058 size_t totlen;
6059
6060 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
6061 error = ENODEV;
6062 goto fail;
6063 }
6064 if (io->pfrio_size < 0 ||
6065 io->pfrio_size > pf_ioctl_maxcount ||
6066 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
6067 error = EINVAL;
6068 goto fail;
6069 }
6070 totlen = io->pfrio_size * sizeof(struct pfr_addr);
6071 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
6072 M_PF, M_WAITOK);
6073 error = copyin(io->pfrio_buffer, pfras, totlen);
6074 if (error) {
6075 free(pfras, M_PF);
6076 goto fail;
6077 }
6078 PF_RULES_WLOCK();
6079 error = pfr_clr_astats(&io->pfrio_table, pfras,
6080 io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
6081 PFR_FLAG_USERIOCTL);
6082 PF_RULES_WUNLOCK();
6083 if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
6084 error = copyout(pfras, io->pfrio_buffer, totlen);
6085 free(pfras, M_PF);
6086 break;
6087 }
6088
6089 case DIOCRTSTADDRS: {
6090 struct pfioc_table *io = (struct pfioc_table *)addr;
6091 struct pfr_addr *pfras;
6092 size_t totlen;
6093
6094 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
6095 error = ENODEV;
6096 goto fail;
6097 }
6098 if (io->pfrio_size < 0 ||
6099 io->pfrio_size > pf_ioctl_maxcount ||
6100 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
6101 error = EINVAL;
6102 goto fail;
6103 }
6104 totlen = io->pfrio_size * sizeof(struct pfr_addr);
6105 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
6106 M_PF, M_WAITOK);
6107 error = copyin(io->pfrio_buffer, pfras, totlen);
6108 if (error) {
6109 free(pfras, M_PF);
6110 goto fail;
6111 }
6112 PF_RULES_RLOCK();
6113 error = pfr_tst_addrs(&io->pfrio_table, pfras,
6114 io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
6115 PFR_FLAG_USERIOCTL);
6116 PF_RULES_RUNLOCK();
6117 if (error == 0)
6118 error = copyout(pfras, io->pfrio_buffer, totlen);
6119 free(pfras, M_PF);
6120 break;
6121 }
6122
6123 case DIOCRINADEFINE: {
6124 struct pfioc_table *io = (struct pfioc_table *)addr;
6125 struct pfr_addr *pfras;
6126 size_t totlen;
6127
6128 if (io->pfrio_esize != sizeof(struct pfr_addr)) {
6129 error = ENODEV;
6130 goto fail;
6131 }
6132 if (io->pfrio_size < 0 ||
6133 io->pfrio_size > pf_ioctl_maxcount ||
6134 WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
6135 error = EINVAL;
6136 goto fail;
6137 }
6138 totlen = io->pfrio_size * sizeof(struct pfr_addr);
6139 pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
6140 M_PF, M_WAITOK);
6141 error = copyin(io->pfrio_buffer, pfras, totlen);
6142 if (error) {
6143 free(pfras, M_PF);
6144 goto fail;
6145 }
6146 PF_RULES_WLOCK();
6147 error = pfr_ina_define(&io->pfrio_table, pfras,
6148 io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
6149 io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
6150 PF_RULES_WUNLOCK();
6151 free(pfras, M_PF);
6152 break;
6153 }
6154
6155 case DIOCOSFPADD: {
6156 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
6157 PF_RULES_WLOCK();
6158 error = pf_osfp_add(io);
6159 PF_RULES_WUNLOCK();
6160 break;
6161 }
6162
6163 case DIOCOSFPGET: {
6164 struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
6165 PF_RULES_RLOCK();
6166 error = pf_osfp_get(io);
6167 PF_RULES_RUNLOCK();
6168 break;
6169 }
6170
6171 case DIOCXBEGIN: {
6172 struct pfioc_trans *io = (struct pfioc_trans *)addr;
6173 struct pfioc_trans_e *ioes, *ioe;
6174 size_t totlen;
6175 int i;
6176
6177 if (io->esize != sizeof(*ioe)) {
6178 error = ENODEV;
6179 goto fail;
6180 }
6181 if (io->size < 0 ||
6182 io->size > pf_ioctl_maxcount ||
6183 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
6184 error = EINVAL;
6185 goto fail;
6186 }
6187 totlen = sizeof(struct pfioc_trans_e) * io->size;
6188 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
6189 M_PF, M_WAITOK);
6190 error = copyin(io->array, ioes, totlen);
6191 if (error) {
6192 free(ioes, M_PF);
6193 goto fail;
6194 }
6195 PF_RULES_WLOCK();
6196 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
6197 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
6198 switch (ioe->rs_num) {
6199 case PF_RULESET_ETH:
6200 if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
6201 PF_RULES_WUNLOCK();
6202 free(ioes, M_PF);
6203 goto fail;
6204 }
6205 break;
6206 #ifdef ALTQ
6207 case PF_RULESET_ALTQ:
6208 if (ioe->anchor[0]) {
6209 PF_RULES_WUNLOCK();
6210 free(ioes, M_PF);
6211 error = EINVAL;
6212 goto fail;
6213 }
6214 if ((error = pf_begin_altq(&ioe->ticket))) {
6215 PF_RULES_WUNLOCK();
6216 free(ioes, M_PF);
6217 goto fail;
6218 }
6219 break;
6220 #endif /* ALTQ */
6221 case PF_RULESET_TABLE:
6222 {
6223 struct pfr_table table;
6224
6225 bzero(&table, sizeof(table));
6226 strlcpy(table.pfrt_anchor, ioe->anchor,
6227 sizeof(table.pfrt_anchor));
6228 if ((error = pfr_ina_begin(&table,
6229 &ioe->ticket, NULL, 0))) {
6230 PF_RULES_WUNLOCK();
6231 free(ioes, M_PF);
6232 goto fail;
6233 }
6234 break;
6235 }
6236 default:
6237 if ((error = pf_begin_rules(&ioe->ticket,
6238 ioe->rs_num, ioe->anchor))) {
6239 PF_RULES_WUNLOCK();
6240 free(ioes, M_PF);
6241 goto fail;
6242 }
6243 break;
6244 }
6245 }
6246 PF_RULES_WUNLOCK();
6247 error = copyout(ioes, io->array, totlen);
6248 free(ioes, M_PF);
6249 break;
6250 }
6251
6252 case DIOCXROLLBACK: {
6253 struct pfioc_trans *io = (struct pfioc_trans *)addr;
6254 struct pfioc_trans_e *ioe, *ioes;
6255 size_t totlen;
6256 int i;
6257
6258 if (io->esize != sizeof(*ioe)) {
6259 error = ENODEV;
6260 goto fail;
6261 }
6262 if (io->size < 0 ||
6263 io->size > pf_ioctl_maxcount ||
6264 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
6265 error = EINVAL;
6266 goto fail;
6267 }
6268 totlen = sizeof(struct pfioc_trans_e) * io->size;
6269 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
6270 M_PF, M_WAITOK);
6271 error = copyin(io->array, ioes, totlen);
6272 if (error) {
6273 free(ioes, M_PF);
6274 goto fail;
6275 }
6276 PF_RULES_WLOCK();
6277 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
6278 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
6279 switch (ioe->rs_num) {
6280 case PF_RULESET_ETH:
6281 if ((error = pf_rollback_eth(ioe->ticket,
6282 ioe->anchor))) {
6283 PF_RULES_WUNLOCK();
6284 free(ioes, M_PF);
6285 goto fail; /* really bad */
6286 }
6287 break;
6288 #ifdef ALTQ
6289 case PF_RULESET_ALTQ:
6290 if (ioe->anchor[0]) {
6291 PF_RULES_WUNLOCK();
6292 free(ioes, M_PF);
6293 error = EINVAL;
6294 goto fail;
6295 }
6296 if ((error = pf_rollback_altq(ioe->ticket))) {
6297 PF_RULES_WUNLOCK();
6298 free(ioes, M_PF);
6299 goto fail; /* really bad */
6300 }
6301 break;
6302 #endif /* ALTQ */
6303 case PF_RULESET_TABLE:
6304 {
6305 struct pfr_table table;
6306
6307 bzero(&table, sizeof(table));
6308 strlcpy(table.pfrt_anchor, ioe->anchor,
6309 sizeof(table.pfrt_anchor));
6310 if ((error = pfr_ina_rollback(&table,
6311 ioe->ticket, NULL, 0))) {
6312 PF_RULES_WUNLOCK();
6313 free(ioes, M_PF);
6314 goto fail; /* really bad */
6315 }
6316 break;
6317 }
6318 default:
6319 if ((error = pf_rollback_rules(ioe->ticket,
6320 ioe->rs_num, ioe->anchor))) {
6321 PF_RULES_WUNLOCK();
6322 free(ioes, M_PF);
6323 goto fail; /* really bad */
6324 }
6325 break;
6326 }
6327 }
6328 PF_RULES_WUNLOCK();
6329 free(ioes, M_PF);
6330 break;
6331 }
6332
6333 case DIOCXCOMMIT: {
6334 struct pfioc_trans *io = (struct pfioc_trans *)addr;
6335 struct pfioc_trans_e *ioe, *ioes;
6336 struct pf_kruleset *rs;
6337 struct pf_keth_ruleset *ers;
6338 size_t totlen;
6339 int i;
6340
6341 if (io->esize != sizeof(*ioe)) {
6342 error = ENODEV;
6343 goto fail;
6344 }
6345
6346 if (io->size < 0 ||
6347 io->size > pf_ioctl_maxcount ||
6348 WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
6349 error = EINVAL;
6350 goto fail;
6351 }
6352
6353 totlen = sizeof(struct pfioc_trans_e) * io->size;
6354 ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
6355 M_PF, M_WAITOK);
6356 error = copyin(io->array, ioes, totlen);
6357 if (error) {
6358 free(ioes, M_PF);
6359 goto fail;
6360 }
6361 PF_RULES_WLOCK();
6362 /* First makes sure everything will succeed. */
6363 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
6364 ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
6365 switch (ioe->rs_num) {
6366 case PF_RULESET_ETH:
6367 ers = pf_find_keth_ruleset(ioe->anchor);
6368 if (ers == NULL || ioe->ticket == 0 ||
6369 ioe->ticket != ers->inactive.ticket) {
6370 PF_RULES_WUNLOCK();
6371 free(ioes, M_PF);
6372 error = EINVAL;
6373 goto fail;
6374 }
6375 break;
6376 #ifdef ALTQ
6377 case PF_RULESET_ALTQ:
6378 if (ioe->anchor[0]) {
6379 PF_RULES_WUNLOCK();
6380 free(ioes, M_PF);
6381 error = EINVAL;
6382 goto fail;
6383 }
6384 if (!V_altqs_inactive_open || ioe->ticket !=
6385 V_ticket_altqs_inactive) {
6386 PF_RULES_WUNLOCK();
6387 free(ioes, M_PF);
6388 error = EBUSY;
6389 goto fail;
6390 }
6391 break;
6392 #endif /* ALTQ */
6393 case PF_RULESET_TABLE:
6394 rs = pf_find_kruleset(ioe->anchor);
6395 if (rs == NULL || !rs->topen || ioe->ticket !=
6396 rs->tticket) {
6397 PF_RULES_WUNLOCK();
6398 free(ioes, M_PF);
6399 error = EBUSY;
6400 goto fail;
6401 }
6402 break;
6403 default:
6404 if (ioe->rs_num < 0 || ioe->rs_num >=
6405 PF_RULESET_MAX) {
6406 PF_RULES_WUNLOCK();
6407 free(ioes, M_PF);
6408 error = EINVAL;
6409 goto fail;
6410 }
6411 rs = pf_find_kruleset(ioe->anchor);
6412 if (rs == NULL ||
6413 !rs->rules[ioe->rs_num].inactive.open ||
6414 rs->rules[ioe->rs_num].inactive.ticket !=
6415 ioe->ticket) {
6416 PF_RULES_WUNLOCK();
6417 free(ioes, M_PF);
6418 error = EBUSY;
6419 goto fail;
6420 }
6421 break;
6422 }
6423 }
6424 /* Now do the commit - no errors should happen here. */
6425 for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
6426 switch (ioe->rs_num) {
6427 case PF_RULESET_ETH:
6428 if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
6429 PF_RULES_WUNLOCK();
6430 free(ioes, M_PF);
6431 goto fail; /* really bad */
6432 }
6433 break;
6434 #ifdef ALTQ
6435 case PF_RULESET_ALTQ:
6436 if ((error = pf_commit_altq(ioe->ticket))) {
6437 PF_RULES_WUNLOCK();
6438 free(ioes, M_PF);
6439 goto fail; /* really bad */
6440 }
6441 break;
6442 #endif /* ALTQ */
6443 case PF_RULESET_TABLE:
6444 {
6445 struct pfr_table table;
6446
6447 bzero(&table, sizeof(table));
6448 (void)strlcpy(table.pfrt_anchor, ioe->anchor,
6449 sizeof(table.pfrt_anchor));
6450 if ((error = pfr_ina_commit(&table,
6451 ioe->ticket, NULL, NULL, 0))) {
6452 PF_RULES_WUNLOCK();
6453 free(ioes, M_PF);
6454 goto fail; /* really bad */
6455 }
6456 break;
6457 }
6458 default:
6459 if ((error = pf_commit_rules(ioe->ticket,
6460 ioe->rs_num, ioe->anchor))) {
6461 PF_RULES_WUNLOCK();
6462 free(ioes, M_PF);
6463 goto fail; /* really bad */
6464 }
6465 break;
6466 }
6467 }
6468 PF_RULES_WUNLOCK();
6469
6470 /* Only hook into EtherNet taffic if we've got rules for it. */
6471 if (! TAILQ_EMPTY(V_pf_keth->active.rules))
6472 hook_pf_eth();
6473 else
6474 dehook_pf_eth();
6475
6476 free(ioes, M_PF);
6477 break;
6478 }
6479
6480 case DIOCGETSRCNODES: {
6481 struct pfioc_src_nodes *psn = (struct pfioc_src_nodes *)addr;
6482 struct pf_srchash *sh;
6483 struct pf_ksrc_node *n;
6484 struct pf_src_node *p, *pstore;
6485 uint32_t i, nr = 0;
6486
6487 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
6488 i++, sh++) {
6489 PF_HASHROW_LOCK(sh);
6490 LIST_FOREACH(n, &sh->nodes, entry)
6491 nr++;
6492 PF_HASHROW_UNLOCK(sh);
6493 }
6494
6495 psn->psn_len = min(psn->psn_len,
6496 sizeof(struct pf_src_node) * nr);
6497
6498 if (psn->psn_len == 0) {
6499 psn->psn_len = sizeof(struct pf_src_node) * nr;
6500 goto fail;
6501 }
6502
6503 nr = 0;
6504
6505 p = pstore = malloc(psn->psn_len, M_PF, M_WAITOK | M_ZERO);
6506 for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
6507 i++, sh++) {
6508 PF_HASHROW_LOCK(sh);
6509 LIST_FOREACH(n, &sh->nodes, entry) {
6510
6511 if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
6512 break;
6513
6514 pf_src_node_copy(n, p);
6515
6516 p++;
6517 nr++;
6518 }
6519 PF_HASHROW_UNLOCK(sh);
6520 }
6521 error = copyout(pstore, psn->psn_src_nodes,
6522 sizeof(struct pf_src_node) * nr);
6523 if (error) {
6524 free(pstore, M_PF);
6525 goto fail;
6526 }
6527 psn->psn_len = sizeof(struct pf_src_node) * nr;
6528 free(pstore, M_PF);
6529 break;
6530 }
6531
6532 case DIOCCLRSRCNODES: {
6533 pf_kill_srcnodes(NULL);
6534 break;
6535 }
6536
6537 case DIOCKILLSRCNODES:
6538 pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
6539 break;
6540
6541 #ifdef COMPAT_FREEBSD13
6542 case DIOCKEEPCOUNTERS_FREEBSD13:
6543 #endif
6544 case DIOCKEEPCOUNTERS:
6545 error = pf_keepcounters((struct pfioc_nv *)addr);
6546 break;
6547
6548 case DIOCGETSYNCOOKIES:
6549 error = pf_get_syncookies((struct pfioc_nv *)addr);
6550 break;
6551
6552 case DIOCSETSYNCOOKIES:
6553 error = pf_set_syncookies((struct pfioc_nv *)addr);
6554 break;
6555
6556 case DIOCSETHOSTID: {
6557 u_int32_t *hostid = (u_int32_t *)addr;
6558
6559 PF_RULES_WLOCK();
6560 if (*hostid == 0)
6561 V_pf_status.hostid = arc4random();
6562 else
6563 V_pf_status.hostid = *hostid;
6564 PF_RULES_WUNLOCK();
6565 break;
6566 }
6567
6568 case DIOCOSFPFLUSH:
6569 PF_RULES_WLOCK();
6570 pf_osfp_flush();
6571 PF_RULES_WUNLOCK();
6572 break;
6573
6574 case DIOCIGETIFACES: {
6575 struct pfioc_iface *io = (struct pfioc_iface *)addr;
6576 struct pfi_kif *ifstore;
6577 size_t bufsiz;
6578
6579 if (io->pfiio_esize != sizeof(struct pfi_kif)) {
6580 error = ENODEV;
6581 goto fail;
6582 }
6583
6584 if (io->pfiio_size < 0 ||
6585 io->pfiio_size > pf_ioctl_maxcount ||
6586 WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
6587 error = EINVAL;
6588 goto fail;
6589 }
6590
6591 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
6592
6593 bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
6594 ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
6595 M_PF, M_WAITOK | M_ZERO);
6596
6597 PF_RULES_RLOCK();
6598 pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
6599 PF_RULES_RUNLOCK();
6600 error = copyout(ifstore, io->pfiio_buffer, bufsiz);
6601 free(ifstore, M_PF);
6602 break;
6603 }
6604
6605 case DIOCSETIFFLAG: {
6606 struct pfioc_iface *io = (struct pfioc_iface *)addr;
6607
6608 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
6609
6610 PF_RULES_WLOCK();
6611 error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
6612 PF_RULES_WUNLOCK();
6613 break;
6614 }
6615
6616 case DIOCCLRIFFLAG: {
6617 struct pfioc_iface *io = (struct pfioc_iface *)addr;
6618
6619 io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
6620
6621 PF_RULES_WLOCK();
6622 error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
6623 PF_RULES_WUNLOCK();
6624 break;
6625 }
6626
6627 case DIOCSETREASS: {
6628 u_int32_t *reass = (u_int32_t *)addr;
6629
6630 V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
6631 /* Removal of DF flag without reassembly enabled is not a
6632 * valid combination. Disable reassembly in such case. */
6633 if (!(V_pf_status.reass & PF_REASS_ENABLED))
6634 V_pf_status.reass = 0;
6635 break;
6636 }
6637
6638 default:
6639 error = ENODEV;
6640 break;
6641 }
6642 fail:
6643 CURVNET_RESTORE();
6644
6645 #undef ERROUT_IOCTL
6646
6647 return (error);
6648 }
6649
6650 static void
pfsync_state_export(union pfsync_state_union * sp,struct pf_kstate * st,int msg_version)6651 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
6652 {
6653 const char *tagname;
6654
6655 /* copy from state key */
6656 sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
6657 sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
6658 sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
6659 sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
6660 sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
6661 sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
6662 sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
6663 sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
6664
6665 /* copy from state */
6666 strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
6667 bcopy(&st->act.rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
6668 sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
6669 sp->pfs_1301.expire = pf_state_expires(st);
6670 if (sp->pfs_1301.expire <= time_uptime)
6671 sp->pfs_1301.expire = htonl(0);
6672 else
6673 sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
6674
6675 switch (msg_version) {
6676 case PFSYNC_MSG_VERSION_1301:
6677 sp->pfs_1301.state_flags = st->state_flags;
6678 sp->pfs_1301.direction = st->direction;
6679 sp->pfs_1301.log = st->act.log;
6680 sp->pfs_1301.timeout = st->timeout;
6681 sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
6682 sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
6683 /*
6684 * XXX Why do we bother pfsyncing source node information if source
6685 * nodes are not synced? Showing users that there is source tracking
6686 * when there is none seems useless.
6687 */
6688 if (st->sns[PF_SN_LIMIT] != NULL)
6689 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
6690 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
6691 sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
6692 break;
6693 case PFSYNC_MSG_VERSION_1400:
6694 sp->pfs_1400.state_flags = htons(st->state_flags);
6695 sp->pfs_1400.direction = st->direction;
6696 sp->pfs_1400.log = st->act.log;
6697 sp->pfs_1400.timeout = st->timeout;
6698 sp->pfs_1400.proto = st->key[PF_SK_WIRE]->proto;
6699 sp->pfs_1400.af = st->key[PF_SK_WIRE]->af;
6700 sp->pfs_1400.qid = htons(st->act.qid);
6701 sp->pfs_1400.pqid = htons(st->act.pqid);
6702 sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
6703 sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
6704 sp->pfs_1400.rtableid = htonl(st->act.rtableid);
6705 sp->pfs_1400.min_ttl = st->act.min_ttl;
6706 sp->pfs_1400.set_tos = st->act.set_tos;
6707 sp->pfs_1400.max_mss = htons(st->act.max_mss);
6708 sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
6709 sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
6710 sp->pfs_1400.rt = st->act.rt;
6711 if (st->act.rt_kif)
6712 strlcpy(sp->pfs_1400.rt_ifname,
6713 st->act.rt_kif->pfik_name,
6714 sizeof(sp->pfs_1400.rt_ifname));
6715 /*
6716 * XXX Why do we bother pfsyncing source node information if source
6717 * nodes are not synced? Showing users that there is source tracking
6718 * when there is none seems useless.
6719 */
6720 if (st->sns[PF_SN_LIMIT] != NULL)
6721 sp->pfs_1400.sync_flags |= PFSYNC_FLAG_SRCNODE;
6722 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE])
6723 sp->pfs_1400.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
6724 break;
6725 case PFSYNC_MSG_VERSION_1500:
6726 sp->pfs_1500.state_flags = htons(st->state_flags);
6727 sp->pfs_1500.direction = st->direction;
6728 sp->pfs_1500.log = st->act.log;
6729 sp->pfs_1500.timeout = st->timeout;
6730 sp->pfs_1500.wire_proto = st->key[PF_SK_WIRE]->proto;
6731 sp->pfs_1500.wire_af = st->key[PF_SK_WIRE]->af;
6732 sp->pfs_1500.stack_proto = st->key[PF_SK_STACK]->proto;
6733 sp->pfs_1500.stack_af = st->key[PF_SK_STACK]->af;
6734 sp->pfs_1500.qid = htons(st->act.qid);
6735 sp->pfs_1500.pqid = htons(st->act.pqid);
6736 sp->pfs_1500.dnpipe = htons(st->act.dnpipe);
6737 sp->pfs_1500.dnrpipe = htons(st->act.dnrpipe);
6738 sp->pfs_1500.rtableid = htonl(st->act.rtableid);
6739 sp->pfs_1500.min_ttl = st->act.min_ttl;
6740 sp->pfs_1500.set_tos = st->act.set_tos;
6741 sp->pfs_1500.max_mss = htons(st->act.max_mss);
6742 sp->pfs_1500.set_prio[0] = st->act.set_prio[0];
6743 sp->pfs_1500.set_prio[1] = st->act.set_prio[1];
6744 sp->pfs_1500.rt = st->act.rt;
6745 sp->pfs_1500.rt_af = st->act.rt_af;
6746 if (st->act.rt_kif)
6747 strlcpy(sp->pfs_1500.rt_ifname,
6748 st->act.rt_kif->pfik_name,
6749 sizeof(sp->pfs_1500.rt_ifname));
6750 strlcpy(sp->pfs_1500.orig_ifname,
6751 st->orig_kif->pfik_name,
6752 sizeof(sp->pfs_1500.orig_ifname));
6753 if ((tagname = pf_tag2tagname(st->tag)) != NULL)
6754 strlcpy(sp->pfs_1500.tagname, tagname,
6755 sizeof(sp->pfs_1500.tagname));
6756 break;
6757 default:
6758 panic("%s: Unsupported pfsync_msg_version %d",
6759 __func__, msg_version);
6760 }
6761
6762 sp->pfs_1301.id = st->id;
6763 sp->pfs_1301.creatorid = st->creatorid;
6764 pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
6765 pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
6766
6767 if (st->rule == NULL)
6768 sp->pfs_1301.rule = htonl(-1);
6769 else
6770 sp->pfs_1301.rule = htonl(st->rule->nr);
6771 if (st->anchor == NULL)
6772 sp->pfs_1301.anchor = htonl(-1);
6773 else
6774 sp->pfs_1301.anchor = htonl(st->anchor->nr);
6775 if (st->nat_rule == NULL)
6776 sp->pfs_1301.nat_rule = htonl(-1);
6777 else
6778 sp->pfs_1301.nat_rule = htonl(st->nat_rule->nr);
6779
6780 pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
6781 pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
6782 pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
6783 pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
6784 }
6785
6786 void
pfsync_state_export_1301(struct pfsync_state_1301 * sp,struct pf_kstate * st)6787 pfsync_state_export_1301(struct pfsync_state_1301 *sp, struct pf_kstate *st)
6788 {
6789 bzero(sp, sizeof(*sp));
6790 pfsync_state_export((union pfsync_state_union *)sp, st,
6791 PFSYNC_MSG_VERSION_1301);
6792 }
6793
6794 void
pfsync_state_export_1400(struct pfsync_state_1400 * sp,struct pf_kstate * st)6795 pfsync_state_export_1400(struct pfsync_state_1400 *sp, struct pf_kstate *st)
6796 {
6797 bzero(sp, sizeof(*sp));
6798 pfsync_state_export((union pfsync_state_union *)sp, st,
6799 PFSYNC_MSG_VERSION_1400);
6800 }
6801
6802 void
pfsync_state_export_1500(struct pfsync_state_1500 * sp,struct pf_kstate * st)6803 pfsync_state_export_1500(struct pfsync_state_1500 *sp, struct pf_kstate *st)
6804 {
6805 bzero(sp, sizeof(*sp));
6806 pfsync_state_export((union pfsync_state_union *)sp, st,
6807 PFSYNC_MSG_VERSION_1500);
6808 }
6809
6810 void
pf_state_export(struct pf_state_export * sp,struct pf_kstate * st)6811 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
6812 {
6813 bzero(sp, sizeof(*sp));
6814
6815 sp->version = PF_STATE_VERSION;
6816
6817 /* copy from state key */
6818 sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
6819 sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
6820 sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
6821 sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
6822 sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
6823 sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
6824 sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
6825 sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
6826 sp->proto = st->key[PF_SK_WIRE]->proto;
6827 sp->af = st->key[PF_SK_WIRE]->af;
6828
6829 /* copy from state */
6830 strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
6831 strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
6832 sizeof(sp->orig_ifname));
6833 memcpy(&sp->rt_addr, &st->act.rt_addr, sizeof(sp->rt_addr));
6834 sp->creation = htonl(time_uptime - (st->creation / 1000));
6835 sp->expire = pf_state_expires(st);
6836 if (sp->expire <= time_uptime)
6837 sp->expire = htonl(0);
6838 else
6839 sp->expire = htonl(sp->expire - time_uptime);
6840
6841 sp->direction = st->direction;
6842 sp->log = st->act.log;
6843 sp->timeout = st->timeout;
6844 /* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
6845 sp->state_flags_compat = st->state_flags;
6846 sp->state_flags = htons(st->state_flags);
6847 if (st->sns[PF_SN_LIMIT] != NULL)
6848 sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
6849 if (st->sns[PF_SN_NAT] != NULL || st->sns[PF_SN_ROUTE] != NULL)
6850 sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
6851 sp->id = st->id;
6852 sp->creatorid = st->creatorid;
6853 pf_state_peer_hton(&st->src, &sp->src);
6854 pf_state_peer_hton(&st->dst, &sp->dst);
6855
6856 if (st->rule == NULL)
6857 sp->rule = htonl(-1);
6858 else
6859 sp->rule = htonl(st->rule->nr);
6860 if (st->anchor == NULL)
6861 sp->anchor = htonl(-1);
6862 else
6863 sp->anchor = htonl(st->anchor->nr);
6864 if (st->nat_rule == NULL)
6865 sp->nat_rule = htonl(-1);
6866 else
6867 sp->nat_rule = htonl(st->nat_rule->nr);
6868
6869 sp->packets[0] = st->packets[0];
6870 sp->packets[1] = st->packets[1];
6871 sp->bytes[0] = st->bytes[0];
6872 sp->bytes[1] = st->bytes[1];
6873
6874 sp->qid = htons(st->act.qid);
6875 sp->pqid = htons(st->act.pqid);
6876 sp->dnpipe = htons(st->act.dnpipe);
6877 sp->dnrpipe = htons(st->act.dnrpipe);
6878 sp->rtableid = htonl(st->act.rtableid);
6879 sp->min_ttl = st->act.min_ttl;
6880 sp->set_tos = st->act.set_tos;
6881 sp->max_mss = htons(st->act.max_mss);
6882 sp->rt = st->act.rt;
6883 if (st->act.rt_kif)
6884 strlcpy(sp->rt_ifname, st->act.rt_kif->pfik_name,
6885 sizeof(sp->rt_ifname));
6886 sp->set_prio[0] = st->act.set_prio[0];
6887 sp->set_prio[1] = st->act.set_prio[1];
6888
6889 }
6890
6891 static void
pf_tbladdr_copyout(struct pf_addr_wrap * aw)6892 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
6893 {
6894 struct pfr_ktable *kt;
6895
6896 KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
6897
6898 kt = aw->p.tbl;
6899 if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
6900 kt = kt->pfrkt_root;
6901 aw->p.tbl = NULL;
6902 aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
6903 kt->pfrkt_cnt : -1;
6904 }
6905
6906 static int
pf_add_status_counters(nvlist_t * nvl,const char * name,counter_u64_t * counters,size_t number,char ** names)6907 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
6908 size_t number, char **names)
6909 {
6910 nvlist_t *nvc;
6911
6912 nvc = nvlist_create(0);
6913 if (nvc == NULL)
6914 return (ENOMEM);
6915
6916 for (int i = 0; i < number; i++) {
6917 nvlist_append_number_array(nvc, "counters",
6918 counter_u64_fetch(counters[i]));
6919 nvlist_append_string_array(nvc, "names",
6920 names[i]);
6921 nvlist_append_number_array(nvc, "ids",
6922 i);
6923 }
6924 nvlist_add_nvlist(nvl, name, nvc);
6925 nvlist_destroy(nvc);
6926
6927 return (0);
6928 }
6929
6930 static int
pf_getstatus(struct pfioc_nv * nv)6931 pf_getstatus(struct pfioc_nv *nv)
6932 {
6933 nvlist_t *nvl = NULL, *nvc = NULL;
6934 void *nvlpacked = NULL;
6935 int error;
6936 struct pf_status s;
6937 char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
6938 char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
6939 char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
6940 time_t since;
6941
6942 PF_RULES_RLOCK_TRACKER;
6943
6944 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
6945
6946 PF_RULES_RLOCK();
6947
6948 nvl = nvlist_create(0);
6949 if (nvl == NULL)
6950 ERROUT(ENOMEM);
6951
6952 since = time_second - (time_uptime - V_pf_status.since);
6953
6954 nvlist_add_bool(nvl, "running", V_pf_status.running);
6955 nvlist_add_number(nvl, "since", since);
6956 nvlist_add_number(nvl, "debug", V_pf_status.debug);
6957 nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
6958 nvlist_add_number(nvl, "states", V_pf_status.states);
6959 nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
6960 nvlist_add_number(nvl, "reass", V_pf_status.reass);
6961 nvlist_add_bool(nvl, "syncookies_active",
6962 V_pf_status.syncookies_active);
6963 nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
6964
6965 /* counters */
6966 error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
6967 PFRES_MAX, pf_reasons);
6968 if (error != 0)
6969 ERROUT(error);
6970
6971 /* lcounters */
6972 error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
6973 KLCNT_MAX, pf_lcounter);
6974 if (error != 0)
6975 ERROUT(error);
6976
6977 /* fcounters */
6978 nvc = nvlist_create(0);
6979 if (nvc == NULL)
6980 ERROUT(ENOMEM);
6981
6982 for (int i = 0; i < FCNT_MAX; i++) {
6983 nvlist_append_number_array(nvc, "counters",
6984 pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
6985 nvlist_append_string_array(nvc, "names",
6986 pf_fcounter[i]);
6987 nvlist_append_number_array(nvc, "ids",
6988 i);
6989 }
6990 nvlist_add_nvlist(nvl, "fcounters", nvc);
6991 nvlist_destroy(nvc);
6992 nvc = NULL;
6993
6994 /* scounters */
6995 error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
6996 SCNT_MAX, pf_fcounter);
6997 if (error != 0)
6998 ERROUT(error);
6999
7000 nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
7001 nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
7002 PF_MD5_DIGEST_LENGTH);
7003
7004 pfi_update_status(V_pf_status.ifname, &s);
7005
7006 /* pcounters / bcounters */
7007 for (int i = 0; i < 2; i++) {
7008 for (int j = 0; j < 2; j++) {
7009 for (int k = 0; k < 2; k++) {
7010 nvlist_append_number_array(nvl, "pcounters",
7011 s.pcounters[i][j][k]);
7012 }
7013 nvlist_append_number_array(nvl, "bcounters",
7014 s.bcounters[i][j]);
7015 }
7016 }
7017
7018 nvlpacked = nvlist_pack(nvl, &nv->len);
7019 if (nvlpacked == NULL)
7020 ERROUT(ENOMEM);
7021
7022 if (nv->size == 0)
7023 ERROUT(0);
7024 else if (nv->size < nv->len)
7025 ERROUT(ENOSPC);
7026
7027 PF_RULES_RUNLOCK();
7028 error = copyout(nvlpacked, nv->data, nv->len);
7029 goto done;
7030
7031 #undef ERROUT
7032 errout:
7033 PF_RULES_RUNLOCK();
7034 done:
7035 free(nvlpacked, M_NVLIST);
7036 nvlist_destroy(nvc);
7037 nvlist_destroy(nvl);
7038
7039 return (error);
7040 }
7041
7042 /*
7043 * XXX - Check for version mismatch!!!
7044 */
7045 static void
pf_clear_all_states(void)7046 pf_clear_all_states(void)
7047 {
7048 struct epoch_tracker et;
7049 struct pf_kstate *s;
7050 u_int i;
7051
7052 NET_EPOCH_ENTER(et);
7053 for (i = 0; i <= V_pf_hashmask; i++) {
7054 struct pf_idhash *ih = &V_pf_idhash[i];
7055 relock:
7056 PF_HASHROW_LOCK(ih);
7057 LIST_FOREACH(s, &ih->states, entry) {
7058 s->timeout = PFTM_PURGE;
7059 /* Don't send out individual delete messages. */
7060 s->state_flags |= PFSTATE_NOSYNC;
7061 pf_remove_state(s);
7062 goto relock;
7063 }
7064 PF_HASHROW_UNLOCK(ih);
7065 }
7066 NET_EPOCH_EXIT(et);
7067 }
7068
7069 static int
pf_clear_tables(void)7070 pf_clear_tables(void)
7071 {
7072 struct pfioc_table io;
7073 int error;
7074
7075 bzero(&io, sizeof(io));
7076 io.pfrio_flags |= PFR_FLAG_ALLRSETS;
7077
7078 error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
7079 io.pfrio_flags);
7080
7081 return (error);
7082 }
7083
7084 static void
pf_kill_srcnodes(struct pfioc_src_node_kill * psnk)7085 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
7086 {
7087 struct pf_ksrc_node_list kill;
7088 u_int killed;
7089
7090 LIST_INIT(&kill);
7091 for (int i = 0; i <= V_pf_srchashmask; i++) {
7092 struct pf_srchash *sh = &V_pf_srchash[i];
7093 struct pf_ksrc_node *sn, *tmp;
7094
7095 PF_HASHROW_LOCK(sh);
7096 LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
7097 if (psnk == NULL ||
7098 (pf_match_addr(psnk->psnk_src.neg,
7099 &psnk->psnk_src.addr.v.a.addr,
7100 &psnk->psnk_src.addr.v.a.mask,
7101 &sn->addr, sn->af) &&
7102 pf_match_addr(psnk->psnk_dst.neg,
7103 &psnk->psnk_dst.addr.v.a.addr,
7104 &psnk->psnk_dst.addr.v.a.mask,
7105 &sn->raddr, sn->af))) {
7106 pf_unlink_src_node(sn);
7107 LIST_INSERT_HEAD(&kill, sn, entry);
7108 sn->expire = 1;
7109 }
7110 PF_HASHROW_UNLOCK(sh);
7111 }
7112
7113 for (int i = 0; i <= V_pf_hashmask; i++) {
7114 struct pf_idhash *ih = &V_pf_idhash[i];
7115 struct pf_kstate *s;
7116
7117 PF_HASHROW_LOCK(ih);
7118 LIST_FOREACH(s, &ih->states, entry) {
7119 for(pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX;
7120 sn_type++) {
7121 if (s->sns[sn_type] &&
7122 s->sns[sn_type]->expire == 1) {
7123 s->sns[sn_type] = NULL;
7124 }
7125 }
7126 }
7127 PF_HASHROW_UNLOCK(ih);
7128 }
7129
7130 killed = pf_free_src_nodes(&kill);
7131
7132 if (psnk != NULL)
7133 psnk->psnk_killed = killed;
7134 }
7135
7136 static int
pf_keepcounters(struct pfioc_nv * nv)7137 pf_keepcounters(struct pfioc_nv *nv)
7138 {
7139 nvlist_t *nvl = NULL;
7140 void *nvlpacked = NULL;
7141 int error = 0;
7142
7143 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
7144
7145 if (nv->len > pf_ioctl_maxcount)
7146 ERROUT(ENOMEM);
7147
7148 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
7149 error = copyin(nv->data, nvlpacked, nv->len);
7150 if (error)
7151 ERROUT(error);
7152
7153 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
7154 if (nvl == NULL)
7155 ERROUT(EBADMSG);
7156
7157 if (! nvlist_exists_bool(nvl, "keep_counters"))
7158 ERROUT(EBADMSG);
7159
7160 V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
7161
7162 on_error:
7163 nvlist_destroy(nvl);
7164 free(nvlpacked, M_NVLIST);
7165 return (error);
7166 }
7167
7168 unsigned int
pf_clear_states(const struct pf_kstate_kill * kill)7169 pf_clear_states(const struct pf_kstate_kill *kill)
7170 {
7171 struct pf_state_key_cmp match_key;
7172 struct pf_kstate *s;
7173 struct pfi_kkif *kif;
7174 int idx;
7175 unsigned int killed = 0, dir;
7176
7177 NET_EPOCH_ASSERT();
7178
7179 for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
7180 struct pf_idhash *ih = &V_pf_idhash[i];
7181
7182 relock_DIOCCLRSTATES:
7183 PF_HASHROW_LOCK(ih);
7184 LIST_FOREACH(s, &ih->states, entry) {
7185 /* For floating states look at the original kif. */
7186 kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
7187
7188 if (kill->psk_ifname[0] &&
7189 strcmp(kill->psk_ifname,
7190 kif->pfik_name))
7191 continue;
7192
7193 if (kill->psk_kill_match) {
7194 bzero(&match_key, sizeof(match_key));
7195
7196 if (s->direction == PF_OUT) {
7197 dir = PF_IN;
7198 idx = PF_SK_STACK;
7199 } else {
7200 dir = PF_OUT;
7201 idx = PF_SK_WIRE;
7202 }
7203
7204 match_key.af = s->key[idx]->af;
7205 match_key.proto = s->key[idx]->proto;
7206 pf_addrcpy(&match_key.addr[0],
7207 &s->key[idx]->addr[1], match_key.af);
7208 match_key.port[0] = s->key[idx]->port[1];
7209 pf_addrcpy(&match_key.addr[1],
7210 &s->key[idx]->addr[0], match_key.af);
7211 match_key.port[1] = s->key[idx]->port[0];
7212 }
7213
7214 /*
7215 * Don't send out individual
7216 * delete messages.
7217 */
7218 s->state_flags |= PFSTATE_NOSYNC;
7219 pf_remove_state(s);
7220 killed++;
7221
7222 if (kill->psk_kill_match)
7223 killed += pf_kill_matching_state(&match_key,
7224 dir);
7225
7226 goto relock_DIOCCLRSTATES;
7227 }
7228 PF_HASHROW_UNLOCK(ih);
7229 }
7230
7231 if (V_pfsync_clear_states_ptr != NULL)
7232 V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
7233
7234 return (killed);
7235 }
7236
7237 void
pf_killstates(struct pf_kstate_kill * kill,unsigned int * killed)7238 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
7239 {
7240 struct pf_kstate *s;
7241
7242 NET_EPOCH_ASSERT();
7243 if (kill->psk_pfcmp.id) {
7244 if (kill->psk_pfcmp.creatorid == 0)
7245 kill->psk_pfcmp.creatorid = V_pf_status.hostid;
7246 if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
7247 kill->psk_pfcmp.creatorid))) {
7248 pf_remove_state(s);
7249 *killed = 1;
7250 }
7251 return;
7252 }
7253
7254 for (unsigned int i = 0; i <= V_pf_hashmask; i++)
7255 *killed += pf_killstates_row(kill, &V_pf_idhash[i]);
7256 }
7257
7258 static int
pf_killstates_nv(struct pfioc_nv * nv)7259 pf_killstates_nv(struct pfioc_nv *nv)
7260 {
7261 struct pf_kstate_kill kill;
7262 struct epoch_tracker et;
7263 nvlist_t *nvl = NULL;
7264 void *nvlpacked = NULL;
7265 int error = 0;
7266 unsigned int killed = 0;
7267
7268 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
7269
7270 if (nv->len > pf_ioctl_maxcount)
7271 ERROUT(ENOMEM);
7272
7273 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
7274 error = copyin(nv->data, nvlpacked, nv->len);
7275 if (error)
7276 ERROUT(error);
7277
7278 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
7279 if (nvl == NULL)
7280 ERROUT(EBADMSG);
7281
7282 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
7283 if (error)
7284 ERROUT(error);
7285
7286 NET_EPOCH_ENTER(et);
7287 pf_killstates(&kill, &killed);
7288 NET_EPOCH_EXIT(et);
7289
7290 free(nvlpacked, M_NVLIST);
7291 nvlpacked = NULL;
7292 nvlist_destroy(nvl);
7293 nvl = nvlist_create(0);
7294 if (nvl == NULL)
7295 ERROUT(ENOMEM);
7296
7297 nvlist_add_number(nvl, "killed", killed);
7298
7299 nvlpacked = nvlist_pack(nvl, &nv->len);
7300 if (nvlpacked == NULL)
7301 ERROUT(ENOMEM);
7302
7303 if (nv->size == 0)
7304 ERROUT(0);
7305 else if (nv->size < nv->len)
7306 ERROUT(ENOSPC);
7307
7308 error = copyout(nvlpacked, nv->data, nv->len);
7309
7310 on_error:
7311 nvlist_destroy(nvl);
7312 free(nvlpacked, M_NVLIST);
7313 return (error);
7314 }
7315
7316 static int
pf_clearstates_nv(struct pfioc_nv * nv)7317 pf_clearstates_nv(struct pfioc_nv *nv)
7318 {
7319 struct pf_kstate_kill kill;
7320 struct epoch_tracker et;
7321 nvlist_t *nvl = NULL;
7322 void *nvlpacked = NULL;
7323 int error = 0;
7324 unsigned int killed;
7325
7326 #define ERROUT(x) ERROUT_FUNCTION(on_error, x)
7327
7328 if (nv->len > pf_ioctl_maxcount)
7329 ERROUT(ENOMEM);
7330
7331 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
7332 error = copyin(nv->data, nvlpacked, nv->len);
7333 if (error)
7334 ERROUT(error);
7335
7336 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
7337 if (nvl == NULL)
7338 ERROUT(EBADMSG);
7339
7340 error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
7341 if (error)
7342 ERROUT(error);
7343
7344 NET_EPOCH_ENTER(et);
7345 killed = pf_clear_states(&kill);
7346 NET_EPOCH_EXIT(et);
7347
7348 free(nvlpacked, M_NVLIST);
7349 nvlpacked = NULL;
7350 nvlist_destroy(nvl);
7351 nvl = nvlist_create(0);
7352 if (nvl == NULL)
7353 ERROUT(ENOMEM);
7354
7355 nvlist_add_number(nvl, "killed", killed);
7356
7357 nvlpacked = nvlist_pack(nvl, &nv->len);
7358 if (nvlpacked == NULL)
7359 ERROUT(ENOMEM);
7360
7361 if (nv->size == 0)
7362 ERROUT(0);
7363 else if (nv->size < nv->len)
7364 ERROUT(ENOSPC);
7365
7366 error = copyout(nvlpacked, nv->data, nv->len);
7367
7368 #undef ERROUT
7369 on_error:
7370 nvlist_destroy(nvl);
7371 free(nvlpacked, M_NVLIST);
7372 return (error);
7373 }
7374
7375 static int
pf_getstate(struct pfioc_nv * nv)7376 pf_getstate(struct pfioc_nv *nv)
7377 {
7378 nvlist_t *nvl = NULL, *nvls;
7379 void *nvlpacked = NULL;
7380 struct pf_kstate *s = NULL;
7381 int error = 0;
7382 uint64_t id, creatorid;
7383
7384 #define ERROUT(x) ERROUT_FUNCTION(errout, x)
7385
7386 if (nv->len > pf_ioctl_maxcount)
7387 ERROUT(ENOMEM);
7388
7389 nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
7390 error = copyin(nv->data, nvlpacked, nv->len);
7391 if (error)
7392 ERROUT(error);
7393
7394 nvl = nvlist_unpack(nvlpacked, nv->len, 0);
7395 if (nvl == NULL)
7396 ERROUT(EBADMSG);
7397
7398 PFNV_CHK(pf_nvuint64(nvl, "id", &id));
7399 PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
7400
7401 s = pf_find_state_byid(id, creatorid);
7402 if (s == NULL)
7403 ERROUT(ENOENT);
7404
7405 free(nvlpacked, M_NVLIST);
7406 nvlpacked = NULL;
7407 nvlist_destroy(nvl);
7408 nvl = nvlist_create(0);
7409 if (nvl == NULL)
7410 ERROUT(ENOMEM);
7411
7412 nvls = pf_state_to_nvstate(s);
7413 if (nvls == NULL)
7414 ERROUT(ENOMEM);
7415
7416 nvlist_add_nvlist(nvl, "state", nvls);
7417 nvlist_destroy(nvls);
7418
7419 nvlpacked = nvlist_pack(nvl, &nv->len);
7420 if (nvlpacked == NULL)
7421 ERROUT(ENOMEM);
7422
7423 if (nv->size == 0)
7424 ERROUT(0);
7425 else if (nv->size < nv->len)
7426 ERROUT(ENOSPC);
7427
7428 error = copyout(nvlpacked, nv->data, nv->len);
7429
7430 #undef ERROUT
7431 errout:
7432 if (s != NULL)
7433 PF_STATE_UNLOCK(s);
7434 free(nvlpacked, M_NVLIST);
7435 nvlist_destroy(nvl);
7436 return (error);
7437 }
7438
7439 /*
7440 * XXX - Check for version mismatch!!!
7441 */
7442
7443 /*
7444 * Duplicate pfctl -Fa operation to get rid of as much as we can.
7445 */
7446 static int
shutdown_pf(void)7447 shutdown_pf(void)
7448 {
7449 int error = 0;
7450 u_int32_t t[5];
7451 char nn = '\0';
7452 struct pf_kanchor *anchor, *tmp_anchor;
7453 struct pf_keth_anchor *eth_anchor, *tmp_eth_anchor;
7454 int rs_num;
7455
7456 do {
7457 /* Unlink rules of all user defined anchors */
7458 RB_FOREACH_SAFE(anchor, pf_kanchor_global, &V_pf_anchors,
7459 tmp_anchor) {
7460 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
7461 if ((error = pf_begin_rules(&t[rs_num], rs_num,
7462 anchor->path)) != 0) {
7463 DPFPRINTF(PF_DEBUG_MISC, "%s: "
7464 "anchor.path=%s rs_num=%d",
7465 __func__, anchor->path, rs_num);
7466 goto error; /* XXX: rollback? */
7467 }
7468 }
7469 for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
7470 error = pf_commit_rules(t[rs_num], rs_num,
7471 anchor->path);
7472 MPASS(error == 0);
7473 }
7474 }
7475
7476 /* Unlink rules of all user defined ether anchors */
7477 RB_FOREACH_SAFE(eth_anchor, pf_keth_anchor_global,
7478 &V_pf_keth_anchors, tmp_eth_anchor) {
7479 if ((error = pf_begin_eth(&t[0], eth_anchor->path))
7480 != 0) {
7481 DPFPRINTF(PF_DEBUG_MISC, "%s: eth "
7482 "anchor.path=%s", __func__,
7483 eth_anchor->path);
7484 goto error;
7485 }
7486 error = pf_commit_eth(t[0], eth_anchor->path);
7487 MPASS(error == 0);
7488 }
7489
7490 if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
7491 != 0) {
7492 DPFPRINTF(PF_DEBUG_MISC, "%s: SCRUB", __func__);
7493 break;
7494 }
7495 if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
7496 != 0) {
7497 DPFPRINTF(PF_DEBUG_MISC, "%s: FILTER", __func__);
7498 break; /* XXX: rollback? */
7499 }
7500 if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
7501 != 0) {
7502 DPFPRINTF(PF_DEBUG_MISC, "%s: NAT", __func__);
7503 break; /* XXX: rollback? */
7504 }
7505 if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
7506 != 0) {
7507 DPFPRINTF(PF_DEBUG_MISC, "%s: BINAT", __func__);
7508 break; /* XXX: rollback? */
7509 }
7510 if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
7511 != 0) {
7512 DPFPRINTF(PF_DEBUG_MISC, "%s: RDR", __func__);
7513 break; /* XXX: rollback? */
7514 }
7515
7516 error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
7517 MPASS(error == 0);
7518 error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
7519 MPASS(error == 0);
7520 error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
7521 MPASS(error == 0);
7522 error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
7523 MPASS(error == 0);
7524 error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
7525 MPASS(error == 0);
7526
7527 if ((error = pf_clear_tables()) != 0)
7528 break;
7529
7530 if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
7531 DPFPRINTF(PF_DEBUG_MISC, "%s: eth", __func__);
7532 break;
7533 }
7534 error = pf_commit_eth(t[0], &nn);
7535 MPASS(error == 0);
7536
7537 #ifdef ALTQ
7538 if ((error = pf_begin_altq(&t[0])) != 0) {
7539 DPFPRINTF(PF_DEBUG_MISC, "%s: ALTQ", __func__);
7540 break;
7541 }
7542 pf_commit_altq(t[0]);
7543 #endif
7544
7545 pf_clear_all_states();
7546
7547 pf_kill_srcnodes(NULL);
7548
7549 for (int i = 0; i < PF_RULESET_MAX; i++) {
7550 pf_rule_tree_free(pf_main_ruleset.rules[i].active.tree);
7551 pf_rule_tree_free(pf_main_ruleset.rules[i].inactive.tree);
7552 }
7553
7554 /* status does not use malloced mem so no need to cleanup */
7555 /* fingerprints and interfaces have their own cleanup code */
7556 } while(0);
7557
7558 error:
7559 return (error);
7560 }
7561
7562 static pfil_return_t
pf_check_return(int chk,struct mbuf ** m)7563 pf_check_return(int chk, struct mbuf **m)
7564 {
7565
7566 switch (chk) {
7567 case PF_PASS:
7568 if (*m == NULL)
7569 return (PFIL_CONSUMED);
7570 else
7571 return (PFIL_PASS);
7572 break;
7573 default:
7574 if (*m != NULL) {
7575 m_freem(*m);
7576 *m = NULL;
7577 }
7578 return (PFIL_DROPPED);
7579 }
7580 }
7581
7582 static pfil_return_t
pf_eth_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7583 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
7584 void *ruleset __unused, struct inpcb *inp)
7585 {
7586 int chk;
7587
7588 CURVNET_ASSERT_SET();
7589
7590 chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
7591
7592 return (pf_check_return(chk, m));
7593 }
7594
7595 static pfil_return_t
pf_eth_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7596 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
7597 void *ruleset __unused, struct inpcb *inp)
7598 {
7599 int chk;
7600
7601 CURVNET_ASSERT_SET();
7602
7603 chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
7604
7605 return (pf_check_return(chk, m));
7606 }
7607
7608 #ifdef INET
7609 static pfil_return_t
pf_check_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7610 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
7611 void *ruleset __unused, struct inpcb *inp)
7612 {
7613 int chk;
7614
7615 CURVNET_ASSERT_SET();
7616
7617 chk = pf_test(AF_INET, PF_IN, flags, ifp, m, inp, NULL);
7618
7619 return (pf_check_return(chk, m));
7620 }
7621
7622 static pfil_return_t
pf_check_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7623 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
7624 void *ruleset __unused, struct inpcb *inp)
7625 {
7626 int chk;
7627
7628 CURVNET_ASSERT_SET();
7629
7630 chk = pf_test(AF_INET, PF_OUT, flags, ifp, m, inp, NULL);
7631
7632 return (pf_check_return(chk, m));
7633 }
7634 #endif
7635
7636 #ifdef INET6
7637 static pfil_return_t
pf_check6_in(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7638 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
7639 void *ruleset __unused, struct inpcb *inp)
7640 {
7641 int chk;
7642
7643 CURVNET_ASSERT_SET();
7644
7645 /*
7646 * In case of loopback traffic IPv6 uses the real interface in
7647 * order to support scoped addresses. In order to support stateful
7648 * filtering we have change this to lo0 as it is the case in IPv4.
7649 */
7650 chk = pf_test(AF_INET6, PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
7651 m, inp, NULL);
7652
7653 return (pf_check_return(chk, m));
7654 }
7655
7656 static pfil_return_t
pf_check6_out(struct mbuf ** m,struct ifnet * ifp,int flags,void * ruleset __unused,struct inpcb * inp)7657 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
7658 void *ruleset __unused, struct inpcb *inp)
7659 {
7660 int chk;
7661
7662 CURVNET_ASSERT_SET();
7663
7664 chk = pf_test(AF_INET6, PF_OUT, flags, ifp, m, inp, NULL);
7665
7666 return (pf_check_return(chk, m));
7667 }
7668 #endif /* INET6 */
7669
7670 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
7671 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
7672 #define V_pf_eth_in_hook VNET(pf_eth_in_hook)
7673 #define V_pf_eth_out_hook VNET(pf_eth_out_hook)
7674
7675 #ifdef INET
7676 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
7677 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
7678 #define V_pf_ip4_in_hook VNET(pf_ip4_in_hook)
7679 #define V_pf_ip4_out_hook VNET(pf_ip4_out_hook)
7680 #endif
7681 #ifdef INET6
7682 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
7683 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
7684 #define V_pf_ip6_in_hook VNET(pf_ip6_in_hook)
7685 #define V_pf_ip6_out_hook VNET(pf_ip6_out_hook)
7686 #endif
7687
7688 static void
hook_pf_eth(void)7689 hook_pf_eth(void)
7690 {
7691 struct pfil_hook_args pha = {
7692 .pa_version = PFIL_VERSION,
7693 .pa_modname = "pf",
7694 .pa_type = PFIL_TYPE_ETHERNET,
7695 };
7696 struct pfil_link_args pla = {
7697 .pa_version = PFIL_VERSION,
7698 };
7699 int ret __diagused;
7700
7701 if (atomic_load_bool(&V_pf_pfil_eth_hooked))
7702 return;
7703
7704 pha.pa_mbuf_chk = pf_eth_check_in;
7705 pha.pa_flags = PFIL_IN;
7706 pha.pa_rulname = "eth-in";
7707 V_pf_eth_in_hook = pfil_add_hook(&pha);
7708 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
7709 pla.pa_head = V_link_pfil_head;
7710 pla.pa_hook = V_pf_eth_in_hook;
7711 ret = pfil_link(&pla);
7712 MPASS(ret == 0);
7713 pha.pa_mbuf_chk = pf_eth_check_out;
7714 pha.pa_flags = PFIL_OUT;
7715 pha.pa_rulname = "eth-out";
7716 V_pf_eth_out_hook = pfil_add_hook(&pha);
7717 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7718 pla.pa_head = V_link_pfil_head;
7719 pla.pa_hook = V_pf_eth_out_hook;
7720 ret = pfil_link(&pla);
7721 MPASS(ret == 0);
7722
7723 atomic_store_bool(&V_pf_pfil_eth_hooked, true);
7724 }
7725
7726 static void
hook_pf(void)7727 hook_pf(void)
7728 {
7729 struct pfil_hook_args pha = {
7730 .pa_version = PFIL_VERSION,
7731 .pa_modname = "pf",
7732 };
7733 struct pfil_link_args pla = {
7734 .pa_version = PFIL_VERSION,
7735 };
7736 int ret __diagused;
7737
7738 if (atomic_load_bool(&V_pf_pfil_hooked))
7739 return;
7740
7741 #ifdef INET
7742 pha.pa_type = PFIL_TYPE_IP4;
7743 pha.pa_mbuf_chk = pf_check_in;
7744 pha.pa_flags = PFIL_IN;
7745 pha.pa_rulname = "default-in";
7746 V_pf_ip4_in_hook = pfil_add_hook(&pha);
7747 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
7748 pla.pa_head = V_inet_pfil_head;
7749 pla.pa_hook = V_pf_ip4_in_hook;
7750 ret = pfil_link(&pla);
7751 MPASS(ret == 0);
7752 pha.pa_mbuf_chk = pf_check_out;
7753 pha.pa_flags = PFIL_OUT;
7754 pha.pa_rulname = "default-out";
7755 V_pf_ip4_out_hook = pfil_add_hook(&pha);
7756 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7757 pla.pa_head = V_inet_pfil_head;
7758 pla.pa_hook = V_pf_ip4_out_hook;
7759 ret = pfil_link(&pla);
7760 MPASS(ret == 0);
7761 if (V_pf_filter_local) {
7762 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7763 pla.pa_head = V_inet_local_pfil_head;
7764 pla.pa_hook = V_pf_ip4_out_hook;
7765 ret = pfil_link(&pla);
7766 MPASS(ret == 0);
7767 }
7768 #endif
7769 #ifdef INET6
7770 pha.pa_type = PFIL_TYPE_IP6;
7771 pha.pa_mbuf_chk = pf_check6_in;
7772 pha.pa_flags = PFIL_IN;
7773 pha.pa_rulname = "default-in6";
7774 V_pf_ip6_in_hook = pfil_add_hook(&pha);
7775 pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
7776 pla.pa_head = V_inet6_pfil_head;
7777 pla.pa_hook = V_pf_ip6_in_hook;
7778 ret = pfil_link(&pla);
7779 MPASS(ret == 0);
7780 pha.pa_mbuf_chk = pf_check6_out;
7781 pha.pa_rulname = "default-out6";
7782 pha.pa_flags = PFIL_OUT;
7783 V_pf_ip6_out_hook = pfil_add_hook(&pha);
7784 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7785 pla.pa_head = V_inet6_pfil_head;
7786 pla.pa_hook = V_pf_ip6_out_hook;
7787 ret = pfil_link(&pla);
7788 MPASS(ret == 0);
7789 if (V_pf_filter_local) {
7790 pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
7791 pla.pa_head = V_inet6_local_pfil_head;
7792 pla.pa_hook = V_pf_ip6_out_hook;
7793 ret = pfil_link(&pla);
7794 MPASS(ret == 0);
7795 }
7796 #endif
7797
7798 atomic_store_bool(&V_pf_pfil_hooked, true);
7799 }
7800
7801 static void
dehook_pf_eth(void)7802 dehook_pf_eth(void)
7803 {
7804
7805 if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
7806 return;
7807
7808 pfil_remove_hook(V_pf_eth_in_hook);
7809 pfil_remove_hook(V_pf_eth_out_hook);
7810
7811 atomic_store_bool(&V_pf_pfil_eth_hooked, false);
7812 }
7813
7814 static void
dehook_pf(void)7815 dehook_pf(void)
7816 {
7817
7818 if (!atomic_load_bool(&V_pf_pfil_hooked))
7819 return;
7820
7821 #ifdef INET
7822 pfil_remove_hook(V_pf_ip4_in_hook);
7823 pfil_remove_hook(V_pf_ip4_out_hook);
7824 #endif
7825 #ifdef INET6
7826 pfil_remove_hook(V_pf_ip6_in_hook);
7827 pfil_remove_hook(V_pf_ip6_out_hook);
7828 #endif
7829
7830 atomic_store_bool(&V_pf_pfil_hooked, false);
7831 }
7832
7833 static void
pf_load_vnet(void)7834 pf_load_vnet(void)
7835 {
7836 V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
7837 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
7838
7839 rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
7840 rm_init_flags(&V_pf_tags_lock, "pf tags and queues", RM_RECURSE);
7841 sx_init(&V_pf_ioctl_lock, "pf ioctl");
7842
7843 pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
7844 PF_RULE_TAG_HASH_SIZE_DEFAULT);
7845 #ifdef ALTQ
7846 pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
7847 PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
7848 #endif
7849
7850 V_pf_keth = &V_pf_main_keth_anchor.ruleset;
7851
7852 pfattach_vnet();
7853 V_pf_vnet_active = 1;
7854 }
7855
7856 static int
pf_load(void)7857 pf_load(void)
7858 {
7859 int error;
7860
7861 sx_init(&pf_end_lock, "pf end thread");
7862
7863 pf_mtag_initialize();
7864
7865 pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
7866 if (pf_dev == NULL)
7867 return (ENOMEM);
7868
7869 pf_end_threads = 0;
7870 error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
7871 if (error != 0)
7872 return (error);
7873
7874 pfi_initialize();
7875
7876 return (0);
7877 }
7878
7879 static void
pf_unload_vnet(void)7880 pf_unload_vnet(void)
7881 {
7882 int ret __diagused;
7883
7884 V_pf_vnet_active = 0;
7885 V_pf_status.running = 0;
7886 dehook_pf();
7887 dehook_pf_eth();
7888
7889 PF_RULES_WLOCK();
7890 pf_syncookies_cleanup();
7891 shutdown_pf();
7892 PF_RULES_WUNLOCK();
7893
7894 ret = swi_remove(V_pf_swi_cookie);
7895 MPASS(ret == 0);
7896 ret = intr_event_destroy(V_pf_swi_ie);
7897 MPASS(ret == 0);
7898
7899 pf_unload_vnet_purge();
7900
7901 pf_normalize_cleanup();
7902 PF_RULES_WLOCK();
7903 pfi_cleanup_vnet();
7904 PF_RULES_WUNLOCK();
7905 pfr_cleanup();
7906 pf_osfp_flush();
7907 pf_cleanup();
7908 if (IS_DEFAULT_VNET(curvnet))
7909 pf_mtag_cleanup();
7910
7911 pf_cleanup_tagset(&V_pf_tags);
7912 #ifdef ALTQ
7913 pf_cleanup_tagset(&V_pf_qids);
7914 #endif
7915 uma_zdestroy(V_pf_tag_z);
7916
7917 #ifdef PF_WANT_32_TO_64_COUNTER
7918 PF_RULES_WLOCK();
7919 LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
7920
7921 MPASS(LIST_EMPTY(&V_pf_allkiflist));
7922 MPASS(V_pf_allkifcount == 0);
7923
7924 LIST_REMOVE(&V_pf_default_rule, allrulelist);
7925 V_pf_allrulecount--;
7926 LIST_REMOVE(V_pf_rulemarker, allrulelist);
7927
7928 MPASS(LIST_EMPTY(&V_pf_allrulelist));
7929 MPASS(V_pf_allrulecount == 0);
7930
7931 PF_RULES_WUNLOCK();
7932
7933 free(V_pf_kifmarker, PFI_MTYPE);
7934 free(V_pf_rulemarker, M_PFRULE);
7935 #endif
7936
7937 /* Free counters last as we updated them during shutdown. */
7938 pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
7939 for (int i = 0; i < 2; i++) {
7940 pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
7941 pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
7942 }
7943 counter_u64_free(V_pf_default_rule.states_cur);
7944 counter_u64_free(V_pf_default_rule.states_tot);
7945 for (pf_sn_types_t sn_type=0; sn_type<PF_SN_MAX; sn_type++)
7946 counter_u64_free(V_pf_default_rule.src_nodes[sn_type]);
7947 uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
7948
7949 for (int i = 0; i < PFRES_MAX; i++)
7950 counter_u64_free(V_pf_status.counters[i]);
7951 for (int i = 0; i < KLCNT_MAX; i++)
7952 counter_u64_free(V_pf_status.lcounters[i]);
7953 for (int i = 0; i < FCNT_MAX; i++)
7954 pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
7955 for (int i = 0; i < SCNT_MAX; i++)
7956 counter_u64_free(V_pf_status.scounters[i]);
7957 for (int i = 0; i < NCNT_MAX; i++)
7958 counter_u64_free(V_pf_status.ncounters[i]);
7959
7960 rm_destroy(&V_pf_rules_lock);
7961 sx_destroy(&V_pf_ioctl_lock);
7962 }
7963
7964 static void
pf_unload(void * dummy __unused)7965 pf_unload(void *dummy __unused)
7966 {
7967
7968 sx_xlock(&pf_end_lock);
7969 pf_end_threads = 1;
7970 while (pf_end_threads < 2) {
7971 wakeup_one(pf_purge_thread);
7972 sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
7973 }
7974 sx_xunlock(&pf_end_lock);
7975
7976 pf_nl_unregister();
7977
7978 if (pf_dev != NULL)
7979 destroy_dev(pf_dev);
7980
7981 pfi_cleanup();
7982
7983 sx_destroy(&pf_end_lock);
7984 }
7985
7986 static void
vnet_pf_init(void * unused __unused)7987 vnet_pf_init(void *unused __unused)
7988 {
7989
7990 pf_load_vnet();
7991 }
7992 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
7993 vnet_pf_init, NULL);
7994
7995 static void
vnet_pf_uninit(const void * unused __unused)7996 vnet_pf_uninit(const void *unused __unused)
7997 {
7998
7999 pf_unload_vnet();
8000 }
8001 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
8002 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
8003 vnet_pf_uninit, NULL);
8004
8005 static int
pf_modevent(module_t mod,int type,void * data)8006 pf_modevent(module_t mod, int type, void *data)
8007 {
8008 int error = 0;
8009
8010 switch(type) {
8011 case MOD_LOAD:
8012 error = pf_load();
8013 pf_nl_register();
8014 break;
8015 case MOD_UNLOAD:
8016 /* Handled in SYSUNINIT(pf_unload) to ensure it's done after
8017 * the vnet_pf_uninit()s */
8018 break;
8019 default:
8020 error = EINVAL;
8021 break;
8022 }
8023
8024 return (error);
8025 }
8026
8027 static moduledata_t pf_mod = {
8028 "pf",
8029 pf_modevent,
8030 0
8031 };
8032
8033 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
8034 MODULE_DEPEND(pf, netlink, 1, 1, 1);
8035 MODULE_DEPEND(pf, crypto, 1, 1, 1);
8036 MODULE_VERSION(pf, PF_MODVER);
8037