npf.c revision 1.8 1 /* $NetBSD: npf.c,v 1.8 2012/02/20 00:18:19 rmind Exp $ */
2
3 /*-
4 * Copyright (c) 2009-2010 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This material is based upon work partially supported by The
8 * NetBSD Foundation under a contract with Mindaugas Rasiukevicius.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
32 /*
33 * NPF main: dynamic load/initialisation and unload routines.
34 */
35
36 #include <sys/cdefs.h>
37 __KERNEL_RCSID(0, "$NetBSD: npf.c,v 1.8 2012/02/20 00:18:19 rmind Exp $");
38
39 #include <sys/param.h>
40 #include <sys/types.h>
41
42 #include <sys/atomic.h>
43 #include <sys/conf.h>
44 #include <sys/kauth.h>
45 #include <sys/kmem.h>
46 #include <sys/lwp.h>
47 #include <sys/module.h>
48 #include <sys/percpu.h>
49 #include <sys/rwlock.h>
50 #include <sys/socketvar.h>
51 #include <sys/uio.h>
52
53 #include "npf_impl.h"
54
55 /*
56 * Module and device structures.
57 */
58 MODULE(MODULE_CLASS_DRIVER, npf, NULL);
59
60 void npfattach(int);
61
62 static int npf_fini(void);
63 static int npf_dev_open(dev_t, int, int, lwp_t *);
64 static int npf_dev_close(dev_t, int, int, lwp_t *);
65 static int npf_dev_ioctl(dev_t, u_long, void *, int, lwp_t *);
66 static int npf_dev_poll(dev_t, int, lwp_t *);
67 static int npf_dev_read(dev_t, struct uio *, int);
68
69 typedef struct {
70 npf_ruleset_t * n_rules;
71 npf_tableset_t * n_tables;
72 npf_ruleset_t * n_nat_rules;
73 bool n_default_pass;
74 } npf_core_t;
75
76 static void npf_core_destroy(npf_core_t *);
77 static int npfctl_stats(void *);
78
79 static krwlock_t npf_lock __cacheline_aligned;
80 static npf_core_t * npf_core __cacheline_aligned;
81 static percpu_t * npf_stats_percpu __read_mostly;
82
83 const struct cdevsw npf_cdevsw = {
84 npf_dev_open, npf_dev_close, npf_dev_read, nowrite, npf_dev_ioctl,
85 nostop, notty, npf_dev_poll, nommap, nokqfilter, D_OTHER | D_MPSAFE
86 };
87
88 static int
89 npf_init(void)
90 {
91 #ifdef _MODULE
92 devmajor_t bmajor = NODEVMAJOR, cmajor = NODEVMAJOR;
93 #endif
94 npf_ruleset_t *rset, *nset;
95 npf_tableset_t *tset;
96 int error = 0;
97
98 rw_init(&npf_lock);
99 npf_stats_percpu = percpu_alloc(NPF_STATS_SIZE);
100 npf_tableset_sysinit();
101 npf_session_sysinit();
102 npf_nat_sysinit();
103 npf_alg_sysinit();
104 npflogattach(1);
105
106 /* Load empty configuration. */
107 rset = npf_ruleset_create();
108 tset = npf_tableset_create();
109 nset = npf_ruleset_create();
110 npf_reload(rset, tset, nset, true);
111 KASSERT(npf_core != NULL);
112
113 #ifdef _MODULE
114 /* Attach /dev/npf device. */
115 error = devsw_attach("npf", NULL, &bmajor, &npf_cdevsw, &cmajor);
116 if (error) {
117 /* It will call devsw_detach(), which is safe. */
118 (void)npf_fini();
119 }
120 #endif
121 return error;
122 }
123
124 static int
125 npf_fini(void)
126 {
127
128 /*
129 * At first, detach device, remove pfil hooks and unload existing
130 * configuration, destroy structures.
131 */
132 #ifdef _MODULE
133 devsw_detach(NULL, &npf_cdevsw);
134 #endif
135 npf_unregister_pfil();
136 npf_core_destroy(npf_core);
137 npflogdetach();
138
139 /* Note: order is particular. */
140 npf_nat_sysfini();
141 npf_alg_sysfini();
142 npf_session_sysfini();
143 npf_tableset_sysfini();
144 percpu_free(npf_stats_percpu, NPF_STATS_SIZE);
145 rw_destroy(&npf_lock);
146
147 return 0;
148 }
149
150 /*
151 * Module interface.
152 */
153 static int
154 npf_modcmd(modcmd_t cmd, void *arg)
155 {
156
157 switch (cmd) {
158 case MODULE_CMD_INIT:
159 return npf_init();
160 case MODULE_CMD_FINI:
161 return npf_fini();
162 default:
163 return ENOTTY;
164 }
165 return 0;
166 }
167
168 void
169 npfattach(int nunits)
170 {
171
172 /* Void. */
173 }
174
175 static int
176 npf_dev_open(dev_t dev, int flag, int mode, lwp_t *l)
177 {
178
179 /* Available only for super-user. */
180 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
181 return EPERM;
182 }
183 return 0;
184 }
185
186 static int
187 npf_dev_close(dev_t dev, int flag, int mode, lwp_t *l)
188 {
189
190 return 0;
191 }
192
193 static int
194 npf_dev_ioctl(dev_t dev, u_long cmd, void *data, int flag, lwp_t *l)
195 {
196 int error;
197
198 /* Available only for super-user. */
199 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) {
200 return EPERM;
201 }
202
203 switch (cmd) {
204 case IOC_NPF_VERSION:
205 *(int *)data = NPF_VERSION;
206 error = 0;
207 break;
208 case IOC_NPF_SWITCH:
209 error = npfctl_switch(data);
210 break;
211 case IOC_NPF_RELOAD:
212 error = npfctl_reload(cmd, data);
213 break;
214 case IOC_NPF_TABLE:
215 error = npfctl_table(data);
216 break;
217 case IOC_NPF_STATS:
218 error = npfctl_stats(data);
219 break;
220 case IOC_NPF_SESSIONS_SAVE:
221 error = npfctl_sessions_save(cmd, data);
222 break;
223 case IOC_NPF_SESSIONS_LOAD:
224 error = npfctl_sessions_load(cmd, data);
225 break;
226 case IOC_NPF_UPDATE_RULE:
227 error = npfctl_update_rule(cmd, data);
228 break;
229 default:
230 error = ENOTTY;
231 break;
232 }
233 return error;
234 }
235
236 static int
237 npf_dev_poll(dev_t dev, int events, lwp_t *l)
238 {
239
240 return ENOTSUP;
241 }
242
243 static int
244 npf_dev_read(dev_t dev, struct uio *uio, int flag)
245 {
246
247 return ENOTSUP;
248 }
249
250 /*
251 * NPF core loading/reloading/unloading mechanism.
252 */
253
254 static void
255 npf_core_destroy(npf_core_t *nc)
256 {
257
258 npf_ruleset_destroy(nc->n_rules);
259 npf_ruleset_destroy(nc->n_nat_rules);
260 npf_tableset_destroy(nc->n_tables);
261 kmem_free(nc, sizeof(npf_core_t));
262 }
263
264 /*
265 * npf_reload: atomically load new ruleset, tableset and NAT policies.
266 * Then destroy old (unloaded) structures.
267 */
268 void
269 npf_reload(npf_ruleset_t *rset, npf_tableset_t *tset, npf_ruleset_t *nset,
270 bool flush)
271 {
272 npf_core_t *nc, *onc;
273
274 /* Setup a new core structure. */
275 nc = kmem_zalloc(sizeof(npf_core_t), KM_SLEEP);
276 nc->n_default_pass = flush;
277 nc->n_rules = rset;
278 nc->n_tables = tset;
279 nc->n_nat_rules = nset;
280
281 /* Lock and load the core structure. */
282 rw_enter(&npf_lock, RW_WRITER);
283 onc = atomic_swap_ptr(&npf_core, nc);
284 if (onc) {
285 /* Reload only necessary NAT policies. */
286 npf_ruleset_natreload(nset, onc->n_nat_rules);
287 }
288 /* Unlock. Everything goes "live" now. */
289 rw_exit(&npf_lock);
290
291 if (onc) {
292 /* Destroy unloaded structures. */
293 npf_core_destroy(onc);
294 }
295 }
296
297 void
298 npf_core_enter(void)
299 {
300 rw_enter(&npf_lock, RW_READER);
301 }
302
303 npf_ruleset_t *
304 npf_core_ruleset(void)
305 {
306 KASSERT(rw_lock_held(&npf_lock));
307 return npf_core->n_rules;
308 }
309
310 npf_ruleset_t *
311 npf_core_natset(void)
312 {
313 KASSERT(rw_lock_held(&npf_lock));
314 return npf_core->n_nat_rules;
315 }
316
317 npf_tableset_t *
318 npf_core_tableset(void)
319 {
320 KASSERT(rw_lock_held(&npf_lock));
321 return npf_core->n_tables;
322 }
323
324 void
325 npf_core_exit(void)
326 {
327 rw_exit(&npf_lock);
328 }
329
330 bool
331 npf_core_locked(void)
332 {
333 return rw_lock_held(&npf_lock);
334 }
335
336 bool
337 npf_default_pass(void)
338 {
339 KASSERT(rw_lock_held(&npf_lock));
340 return npf_core->n_default_pass;
341 }
342
343 /*
344 * NPF statistics interface.
345 */
346
347 void
348 npf_stats_inc(npf_stats_t st)
349 {
350 uint64_t *stats = percpu_getref(npf_stats_percpu);
351 stats[st]++;
352 percpu_putref(npf_stats_percpu);
353 }
354
355 void
356 npf_stats_dec(npf_stats_t st)
357 {
358 uint64_t *stats = percpu_getref(npf_stats_percpu);
359 stats[st]--;
360 percpu_putref(npf_stats_percpu);
361 }
362
363 static void
364 npf_stats_collect(void *mem, void *arg, struct cpu_info *ci)
365 {
366 uint64_t *percpu_stats = mem, *full_stats = arg;
367 int i;
368
369 for (i = 0; i < NPF_STATS_COUNT; i++) {
370 full_stats[i] += percpu_stats[i];
371 }
372 }
373
374 /*
375 * npfctl_stats: export collected statistics.
376 */
377 static int
378 npfctl_stats(void *data)
379 {
380 uint64_t *fullst, *uptr = *(uint64_t **)data;
381 int error;
382
383 fullst = kmem_zalloc(NPF_STATS_SIZE, KM_SLEEP);
384 percpu_foreach(npf_stats_percpu, npf_stats_collect, fullst);
385 error = copyout(fullst, uptr, NPF_STATS_SIZE);
386 kmem_free(fullst, NPF_STATS_SIZE);
387 return error;
388 }
389