1 /*- 2 * Copyright (c) 2009-2013 The NetBSD Foundation, Inc. 3 * All rights reserved. 4 * 5 * This material is based upon work partially supported by The 6 * NetBSD Foundation under a contract with Mindaugas Rasiukevicius. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 18 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 19 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 20 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 21 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 22 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 23 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 24 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 25 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 26 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 27 * POSSIBILITY OF SUCH DAMAGE. 28 */ 29 30 /* 31 * NPF packet handler. 32 * 33 * Note: pfil(9) hooks are currently locked by softnet_lock and kernel-lock. 34 */ 35 36 #ifdef _KERNEL 37 #include <sys/cdefs.h> 38 __KERNEL_RCSID(0, "$NetBSD: npf_handler.c,v 1.48 2019/08/25 13:21:03 rmind Exp $"); 39 40 #include <sys/types.h> 41 #include <sys/param.h> 42 43 #include <sys/mbuf.h> 44 #include <sys/mutex.h> 45 #include <net/if.h> 46 #include <net/pfil.h> 47 #include <sys/socketvar.h> 48 49 #include <netinet/in_systm.h> 50 #include <netinet/in.h> 51 #include <netinet/ip_var.h> 52 #include <netinet/ip6.h> 53 #include <netinet6/ip6_var.h> 54 #endif 55 56 #include "npf_impl.h" 57 #include "npf_conn.h" 58 59 #if defined(_NPF_STANDALONE) 60 #define m_freem(m) npf->mbufops->free(m) 61 #define m_clear_flag(m,f) 62 #else 63 #define m_clear_flag(m,f) (m)->m_flags &= ~(f) 64 #endif 65 66 #ifndef INET6 67 #define ip6_reass_packet(x, y) ENOTSUP 68 #endif 69 70 static int 71 npf_reassembly(npf_t *npf, npf_cache_t *npc, bool *mff) 72 { 73 nbuf_t *nbuf = npc->npc_nbuf; 74 int error = EINVAL; 75 struct mbuf *m; 76 77 *mff = false; 78 m = nbuf_head_mbuf(nbuf); 79 80 if (npf_iscached(npc, NPC_IP4)) { 81 error = ip_reass_packet(&m); 82 } else if (npf_iscached(npc, NPC_IP6)) { 83 error = ip6_reass_packet(&m, npc->npc_hlen); 84 } 85 86 if (error) { 87 /* Reass failed. Free the mbuf, clear the nbuf. */ 88 npf_stats_inc(npf, NPF_STAT_REASSFAIL); 89 m_freem(m); 90 memset(nbuf, 0, sizeof(nbuf_t)); 91 return error; 92 } 93 if (m == NULL) { 94 /* More fragments should come. */ 95 npf_stats_inc(npf, NPF_STAT_FRAGMENTS); 96 *mff = true; 97 return 0; 98 } 99 100 /* 101 * Reassembly is complete, we have the final packet. 102 * Cache again, since layer 4 data is accessible now. 103 */ 104 nbuf_init(npf, nbuf, m, nbuf->nb_ifp); 105 npc->npc_info = 0; 106 107 if (npf_cache_all(npc) & (NPC_IPFRAG|NPC_FMTERR)) { 108 return EINVAL; 109 } 110 npf_stats_inc(npf, NPF_STAT_REASSEMBLY); 111 return 0; 112 } 113 114 /* 115 * npfk_packet_handler: main packet handling routine for layer 3. 116 * 117 * Note: packet flow and inspection logic is in strict order. 118 */ 119 __dso_public int 120 npfk_packet_handler(npf_t *npf, struct mbuf **mp, ifnet_t *ifp, int di) 121 { 122 nbuf_t nbuf; 123 npf_cache_t npc; 124 npf_conn_t *con; 125 npf_rule_t *rl; 126 npf_rproc_t *rp; 127 int error, decision, flags; 128 uint32_t ntag; 129 npf_match_info_t mi; 130 bool mff; 131 132 KASSERT(ifp != NULL); 133 134 /* 135 * Initialise packet information cache. 136 * Note: it is enough to clear the info bits. 137 */ 138 npc.npc_ctx = npf; 139 nbuf_init(npf, &nbuf, *mp, ifp); 140 npc.npc_nbuf = &nbuf; 141 npc.npc_info = 0; 142 143 mi.mi_di = di; 144 mi.mi_rid = 0; 145 mi.mi_retfl = 0; 146 147 *mp = NULL; 148 decision = NPF_DECISION_BLOCK; 149 error = 0; 150 rp = NULL; 151 con = NULL; 152 153 /* Cache everything. */ 154 flags = npf_cache_all(&npc); 155 156 /* If error on the format, leave quickly. */ 157 if (flags & NPC_FMTERR) { 158 error = EINVAL; 159 goto out; 160 } 161 162 /* Determine whether it is an IP fragment. */ 163 if (__predict_false(flags & NPC_IPFRAG)) { 164 /* Pass to IPv4/IPv6 reassembly mechanism. */ 165 error = npf_reassembly(npf, &npc, &mff); 166 if (error) { 167 goto out; 168 } 169 if (mff) { 170 /* More fragments should come. */ 171 return 0; 172 } 173 } 174 175 /* Just pass-through if specially tagged. */ 176 if (nbuf_find_tag(&nbuf, &ntag) == 0 && (ntag & NPF_NTAG_PASS) != 0) { 177 goto pass; 178 } 179 180 /* Inspect the list of connections (if found, acquires a reference). */ 181 con = npf_conn_inspect(&npc, di, &error); 182 183 /* If "passing" connection found - skip the ruleset inspection. */ 184 if (con && npf_conn_pass(con, &mi, &rp)) { 185 npf_stats_inc(npf, NPF_STAT_PASS_CONN); 186 KASSERT(error == 0); 187 goto pass; 188 } 189 if (__predict_false(error)) { 190 if (error == ENETUNREACH) 191 goto block; 192 goto out; 193 } 194 195 /* Acquire the lock, inspect the ruleset using this packet. */ 196 int slock = npf_config_read_enter(npf); 197 npf_ruleset_t *rlset = npf_config_ruleset(npf); 198 199 rl = npf_ruleset_inspect(&npc, rlset, di, NPF_LAYER_3); 200 if (__predict_false(rl == NULL)) { 201 const bool pass = npf_default_pass(npf); 202 npf_config_read_exit(npf, slock); 203 204 if (pass) { 205 npf_stats_inc(npf, NPF_STAT_PASS_DEFAULT); 206 goto pass; 207 } 208 npf_stats_inc(npf, NPF_STAT_BLOCK_DEFAULT); 209 goto block; 210 } 211 212 /* 213 * Get the rule procedure (acquires a reference) for association 214 * with a connection (if any) and execution. 215 */ 216 KASSERT(rp == NULL); 217 rp = npf_rule_getrproc(rl); 218 219 /* Conclude with the rule and release the lock. */ 220 error = npf_rule_conclude(rl, &mi); 221 npf_config_read_exit(npf, slock); 222 223 if (error) { 224 npf_stats_inc(npf, NPF_STAT_BLOCK_RULESET); 225 goto block; 226 } 227 npf_stats_inc(npf, NPF_STAT_PASS_RULESET); 228 229 /* 230 * Establish a "pass" connection, if required. Just proceed if 231 * connection creation fails (e.g. due to unsupported protocol). 232 */ 233 if ((mi.mi_retfl & NPF_RULE_STATEFUL) != 0 && !con) { 234 con = npf_conn_establish(&npc, di, 235 (mi.mi_retfl & NPF_RULE_GSTATEFUL) == 0); 236 if (con) { 237 /* 238 * Note: the reference on the rule procedure is 239 * transfered to the connection. It will be 240 * released on connection destruction. 241 */ 242 npf_conn_setpass(con, &mi, rp); 243 } 244 } 245 246 pass: 247 decision = NPF_DECISION_PASS; 248 KASSERT(error == 0); 249 250 /* 251 * Perform NAT. 252 */ 253 error = npf_do_nat(&npc, con, di); 254 255 block: 256 /* 257 * Execute the rule procedure, if any is associated. 258 * It may reverse the decision from pass to block. 259 */ 260 if (rp && !npf_rproc_run(&npc, rp, &mi, &decision)) { 261 if (con) { 262 npf_conn_release(con); 263 } 264 npf_rproc_release(rp); 265 /* mbuf already freed */ 266 return 0; 267 } 268 269 out: 270 /* 271 * Release the reference on a connection. Release the reference 272 * on a rule procedure only if there was no association. 273 */ 274 if (con) { 275 npf_conn_release(con); 276 } else if (rp) { 277 npf_rproc_release(rp); 278 } 279 280 /* Get the new mbuf pointer. */ 281 if ((*mp = nbuf_head_mbuf(&nbuf)) == NULL) { 282 return error ? error : ENOMEM; 283 } 284 285 /* Pass the packet if decided and there is no error. */ 286 if (decision == NPF_DECISION_PASS && !error) { 287 /* 288 * XXX: Disable for now, it will be set accordingly later, 289 * for optimisations (to reduce inspection). 290 */ 291 m_clear_flag(*mp, M_CANFASTFWD); 292 return 0; 293 } 294 295 /* 296 * Block the packet. ENETUNREACH is used to indicate blocking. 297 * Depending on the flags and protocol, return TCP reset (RST) or 298 * ICMP destination unreachable. 299 */ 300 if (mi.mi_retfl && npf_return_block(&npc, mi.mi_retfl)) { 301 *mp = NULL; 302 } 303 304 if (!error) { 305 error = ENETUNREACH; 306 } 307 308 if (*mp) { 309 /* Free the mbuf chain. */ 310 m_freem(*mp); 311 *mp = NULL; 312 } 313 return error; 314 } 315