1 /* $NetBSD: npf.c,v 1.5 2011/04/25 22:16:21 yamt 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 1.5 2011/04/25 22:16:21 yamt 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_MISC, 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 } npf_core_t; 74 75 static void npf_core_destroy(npf_core_t *); 76 static int npfctl_stats(void *); 77 78 static krwlock_t npf_lock __cacheline_aligned; 79 static npf_core_t * npf_core __cacheline_aligned; 80 static percpu_t * npf_stats_percpu __read_mostly; 81 82 const struct cdevsw npf_cdevsw = { 83 npf_dev_open, npf_dev_close, npf_dev_read, nowrite, npf_dev_ioctl, 84 nostop, notty, npf_dev_poll, nommap, nokqfilter, D_OTHER | D_MPSAFE 85 }; 86 87 static int 88 npf_init(void) 89 { 90 #ifdef _MODULE 91 devmajor_t bmajor = NODEVMAJOR, cmajor = NODEVMAJOR; 92 #endif 93 npf_ruleset_t *rset, *nset; 94 npf_tableset_t *tset; 95 int error = 0; 96 97 rw_init(&npf_lock); 98 npf_stats_percpu = percpu_alloc(NPF_STATS_SIZE); 99 npf_tableset_sysinit(); 100 npf_session_sysinit(); 101 npf_nat_sysinit(); 102 npf_alg_sysinit(); 103 npflogattach(1); 104 105 /* Load empty configuration. */ 106 rset = npf_ruleset_create(); 107 tset = npf_tableset_create(); 108 nset = npf_ruleset_create(); 109 npf_reload(rset, tset, nset); 110 KASSERT(npf_core != NULL); 111 112 #ifdef _MODULE 113 /* Attach /dev/npf device. */ 114 error = devsw_attach("npf", NULL, &bmajor, &npf_cdevsw, &cmajor); 115 if (error) { 116 /* It will call devsw_detach(), which is safe. */ 117 (void)npf_fini(); 118 } 119 #endif 120 return error; 121 } 122 123 static int 124 npf_fini(void) 125 { 126 127 /* 128 * At first, detach device, remove pfil hooks and unload existing 129 * configuration, destroy structures. 130 */ 131 #ifdef _MODULE 132 devsw_detach(NULL, &npf_cdevsw); 133 #endif 134 npf_unregister_pfil(); 135 npf_core_destroy(npf_core); 136 npflogdetach(); 137 138 /* Note: order is particular. */ 139 npf_nat_sysfini(); 140 npf_alg_sysfini(); 141 npf_session_sysfini(); 142 npf_tableset_sysfini(); 143 percpu_free(npf_stats_percpu, NPF_STATS_SIZE); 144 rw_destroy(&npf_lock); 145 146 return 0; 147 } 148 149 /* 150 * Module interface. 151 */ 152 static int 153 npf_modcmd(modcmd_t cmd, void *arg) 154 { 155 156 switch (cmd) { 157 case MODULE_CMD_INIT: 158 return npf_init(); 159 case MODULE_CMD_FINI: 160 return npf_fini(); 161 default: 162 return ENOTTY; 163 } 164 return 0; 165 } 166 167 void 168 npfattach(int nunits) 169 { 170 171 /* Void. */ 172 } 173 174 static int 175 npf_dev_open(dev_t dev, int flag, int mode, lwp_t *l) 176 { 177 178 /* Available only for super-user. */ 179 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 180 return EPERM; 181 } 182 return 0; 183 } 184 185 static int 186 npf_dev_close(dev_t dev, int flag, int mode, lwp_t *l) 187 { 188 189 return 0; 190 } 191 192 static int 193 npf_dev_ioctl(dev_t dev, u_long cmd, void *data, int flag, lwp_t *l) 194 { 195 int error; 196 197 /* Available only for super-user. */ 198 if (kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER, NULL)) { 199 return EPERM; 200 } 201 202 switch (cmd) { 203 case IOC_NPF_VERSION: 204 *(int *)data = NPF_VERSION; 205 error = 0; 206 break; 207 case IOC_NPF_SWITCH: 208 error = npfctl_switch(data); 209 break; 210 case IOC_NPF_RELOAD: 211 error = npfctl_reload(cmd, data); 212 break; 213 case IOC_NPF_TABLE: 214 error = npfctl_table(data); 215 break; 216 case IOC_NPF_STATS: 217 error = npfctl_stats(data); 218 break; 219 case IOC_NPF_SESSIONS_SAVE: 220 error = npfctl_sessions_save(cmd, data); 221 break; 222 case IOC_NPF_SESSIONS_LOAD: 223 error = npfctl_sessions_load(cmd, data); 224 break; 225 case IOC_NPF_UPDATE_RULE: 226 error = npfctl_update_rule(cmd, data); 227 break; 228 default: 229 error = ENOTTY; 230 break; 231 } 232 return error; 233 } 234 235 static int 236 npf_dev_poll(dev_t dev, int events, lwp_t *l) 237 { 238 239 return ENOTSUP; 240 } 241 242 static int 243 npf_dev_read(dev_t dev, struct uio *uio, int flag) 244 { 245 246 return ENOTSUP; 247 } 248 249 /* 250 * NPF core loading/reloading/unloading mechanism. 251 */ 252 253 static void 254 npf_core_destroy(npf_core_t *nc) 255 { 256 257 npf_ruleset_destroy(nc->n_rules); 258 npf_ruleset_destroy(nc->n_nat_rules); 259 npf_tableset_destroy(nc->n_tables); 260 kmem_free(nc, sizeof(npf_core_t)); 261 } 262 263 /* 264 * npf_reload: atomically load new ruleset, tableset and NAT policies. 265 * Then destroy old (unloaded) structures. 266 */ 267 void 268 npf_reload(npf_ruleset_t *rset, npf_tableset_t *tset, npf_ruleset_t *nset) 269 { 270 npf_core_t *nc, *onc; 271 272 /* Setup a new core structure. */ 273 nc = kmem_alloc(sizeof(npf_core_t), KM_SLEEP); 274 nc->n_rules = rset; 275 nc->n_tables = tset; 276 nc->n_nat_rules = nset; 277 278 /* Lock and load the core structure. */ 279 rw_enter(&npf_lock, RW_WRITER); 280 onc = atomic_swap_ptr(&npf_core, nc); 281 if (onc) { 282 /* Reload only necessary NAT policies. */ 283 npf_ruleset_natreload(nset, onc->n_nat_rules); 284 } 285 /* Unlock. Everything goes "live" now. */ 286 rw_exit(&npf_lock); 287 288 /* Turn on/off session tracking accordingly. */ 289 npf_session_tracking(true); 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 /* 337 * NPF statistics interface. 338 */ 339 340 void 341 npf_stats_inc(npf_stats_t st) 342 { 343 uint64_t *stats = percpu_getref(npf_stats_percpu); 344 stats[st]++; 345 percpu_putref(npf_stats_percpu); 346 } 347 348 void 349 npf_stats_dec(npf_stats_t st) 350 { 351 uint64_t *stats = percpu_getref(npf_stats_percpu); 352 stats[st]--; 353 percpu_putref(npf_stats_percpu); 354 } 355 356 static void 357 npf_stats_collect(void *mem, void *arg, struct cpu_info *ci) 358 { 359 uint64_t *percpu_stats = mem, *full_stats = arg; 360 int i; 361 362 for (i = 0; i < NPF_STATS_COUNT; i++) { 363 full_stats[i] += percpu_stats[i]; 364 } 365 } 366 367 /* 368 * npfctl_stats: export collected statistics. 369 */ 370 static int 371 npfctl_stats(void *data) 372 { 373 uint64_t *fullst, *uptr = *(uint64_t **)data; 374 int error; 375 376 fullst = kmem_zalloc(NPF_STATS_SIZE, KM_SLEEP); 377 percpu_foreach(npf_stats_percpu, npf_stats_collect, fullst); 378 error = copyout(fullst, uptr, NPF_STATS_SIZE); 379 kmem_free(fullst, NPF_STATS_SIZE); 380 return error; 381 } 382