1 /* 2 * DCCP connection tracking protocol helper 3 * 4 * Copyright (c) 2005, 2006, 2008 Patrick McHardy <kaber@trash.net> 5 * 6 * This program is free software; you can redistribute it and/or modify 7 * it under the terms of the GNU General Public License version 2 as 8 * published by the Free Software Foundation. 9 * 10 */ 11 #include <linux/kernel.h> 12 #include <linux/init.h> 13 #include <linux/sysctl.h> 14 #include <linux/spinlock.h> 15 #include <linux/skbuff.h> 16 #include <linux/dccp.h> 17 #include <linux/slab.h> 18 19 #include <net/net_namespace.h> 20 #include <net/netns/generic.h> 21 22 #include <linux/netfilter/nfnetlink_conntrack.h> 23 #include <net/netfilter/nf_conntrack.h> 24 #include <net/netfilter/nf_conntrack_l4proto.h> 25 #include <net/netfilter/nf_conntrack_ecache.h> 26 #include <net/netfilter/nf_conntrack_timeout.h> 27 #include <net/netfilter/nf_log.h> 28 29 /* Timeouts are based on values from RFC4340: 30 * 31 * - REQUEST: 32 * 33 * 8.1.2. Client Request 34 * 35 * A client MAY give up on its DCCP-Requests after some time 36 * (3 minutes, for example). 37 * 38 * - RESPOND: 39 * 40 * 8.1.3. Server Response 41 * 42 * It MAY also leave the RESPOND state for CLOSED after a timeout of 43 * not less than 4MSL (8 minutes); 44 * 45 * - PARTOPEN: 46 * 47 * 8.1.5. Handshake Completion 48 * 49 * If the client remains in PARTOPEN for more than 4MSL (8 minutes), 50 * it SHOULD reset the connection with Reset Code 2, "Aborted". 51 * 52 * - OPEN: 53 * 54 * The DCCP timestamp overflows after 11.9 hours. If the connection 55 * stays idle this long the sequence number won't be recognized 56 * as valid anymore. 57 * 58 * - CLOSEREQ/CLOSING: 59 * 60 * 8.3. Termination 61 * 62 * The retransmission timer should initially be set to go off in two 63 * round-trip times and should back off to not less than once every 64 * 64 seconds ... 65 * 66 * - TIMEWAIT: 67 * 68 * 4.3. States 69 * 70 * A server or client socket remains in this state for 2MSL (4 minutes) 71 * after the connection has been town down, ... 72 */ 73 74 #define DCCP_MSL (2 * 60 * HZ) 75 76 static const char * const dccp_state_names[] = { 77 [CT_DCCP_NONE] = "NONE", 78 [CT_DCCP_REQUEST] = "REQUEST", 79 [CT_DCCP_RESPOND] = "RESPOND", 80 [CT_DCCP_PARTOPEN] = "PARTOPEN", 81 [CT_DCCP_OPEN] = "OPEN", 82 [CT_DCCP_CLOSEREQ] = "CLOSEREQ", 83 [CT_DCCP_CLOSING] = "CLOSING", 84 [CT_DCCP_TIMEWAIT] = "TIMEWAIT", 85 [CT_DCCP_IGNORE] = "IGNORE", 86 [CT_DCCP_INVALID] = "INVALID", 87 }; 88 89 #define sNO CT_DCCP_NONE 90 #define sRQ CT_DCCP_REQUEST 91 #define sRS CT_DCCP_RESPOND 92 #define sPO CT_DCCP_PARTOPEN 93 #define sOP CT_DCCP_OPEN 94 #define sCR CT_DCCP_CLOSEREQ 95 #define sCG CT_DCCP_CLOSING 96 #define sTW CT_DCCP_TIMEWAIT 97 #define sIG CT_DCCP_IGNORE 98 #define sIV CT_DCCP_INVALID 99 100 /* 101 * DCCP state transition table 102 * 103 * The assumption is the same as for TCP tracking: 104 * 105 * We are the man in the middle. All the packets go through us but might 106 * get lost in transit to the destination. It is assumed that the destination 107 * can't receive segments we haven't seen. 108 * 109 * The following states exist: 110 * 111 * NONE: Initial state, expecting Request 112 * REQUEST: Request seen, waiting for Response from server 113 * RESPOND: Response from server seen, waiting for Ack from client 114 * PARTOPEN: Ack after Response seen, waiting for packet other than Response, 115 * Reset or Sync from server 116 * OPEN: Packet other than Response, Reset or Sync seen 117 * CLOSEREQ: CloseReq from server seen, expecting Close from client 118 * CLOSING: Close seen, expecting Reset 119 * TIMEWAIT: Reset seen 120 * IGNORE: Not determinable whether packet is valid 121 * 122 * Some states exist only on one side of the connection: REQUEST, RESPOND, 123 * PARTOPEN, CLOSEREQ. For the other side these states are equivalent to 124 * the one it was in before. 125 * 126 * Packets are marked as ignored (sIG) if we don't know if they're valid 127 * (for example a reincarnation of a connection we didn't notice is dead 128 * already) and the server may send back a connection closing Reset or a 129 * Response. They're also used for Sync/SyncAck packets, which we don't 130 * care about. 131 */ 132 static const u_int8_t 133 dccp_state_table[CT_DCCP_ROLE_MAX + 1][DCCP_PKT_SYNCACK + 1][CT_DCCP_MAX + 1] = { 134 [CT_DCCP_ROLE_CLIENT] = { 135 [DCCP_PKT_REQUEST] = { 136 /* 137 * sNO -> sRQ Regular Request 138 * sRQ -> sRQ Retransmitted Request or reincarnation 139 * sRS -> sRS Retransmitted Request (apparently Response 140 * got lost after we saw it) or reincarnation 141 * sPO -> sIG Ignore, conntrack might be out of sync 142 * sOP -> sIG Ignore, conntrack might be out of sync 143 * sCR -> sIG Ignore, conntrack might be out of sync 144 * sCG -> sIG Ignore, conntrack might be out of sync 145 * sTW -> sRQ Reincarnation 146 * 147 * sNO, sRQ, sRS, sPO. sOP, sCR, sCG, sTW, */ 148 sRQ, sRQ, sRS, sIG, sIG, sIG, sIG, sRQ, 149 }, 150 [DCCP_PKT_RESPONSE] = { 151 /* 152 * sNO -> sIV Invalid 153 * sRQ -> sIG Ignore, might be response to ignored Request 154 * sRS -> sIG Ignore, might be response to ignored Request 155 * sPO -> sIG Ignore, might be response to ignored Request 156 * sOP -> sIG Ignore, might be response to ignored Request 157 * sCR -> sIG Ignore, might be response to ignored Request 158 * sCG -> sIG Ignore, might be response to ignored Request 159 * sTW -> sIV Invalid, reincarnation in reverse direction 160 * goes through sRQ 161 * 162 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 163 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIV, 164 }, 165 [DCCP_PKT_ACK] = { 166 /* 167 * sNO -> sIV No connection 168 * sRQ -> sIV No connection 169 * sRS -> sPO Ack for Response, move to PARTOPEN (8.1.5.) 170 * sPO -> sPO Retransmitted Ack for Response, remain in PARTOPEN 171 * sOP -> sOP Regular ACK, remain in OPEN 172 * sCR -> sCR Ack in CLOSEREQ MAY be processed (8.3.) 173 * sCG -> sCG Ack in CLOSING MAY be processed (8.3.) 174 * sTW -> sIV 175 * 176 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 177 sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV 178 }, 179 [DCCP_PKT_DATA] = { 180 /* 181 * sNO -> sIV No connection 182 * sRQ -> sIV No connection 183 * sRS -> sIV No connection 184 * sPO -> sIV MUST use DataAck in PARTOPEN state (8.1.5.) 185 * sOP -> sOP Regular Data packet 186 * sCR -> sCR Data in CLOSEREQ MAY be processed (8.3.) 187 * sCG -> sCG Data in CLOSING MAY be processed (8.3.) 188 * sTW -> sIV 189 * 190 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 191 sIV, sIV, sIV, sIV, sOP, sCR, sCG, sIV, 192 }, 193 [DCCP_PKT_DATAACK] = { 194 /* 195 * sNO -> sIV No connection 196 * sRQ -> sIV No connection 197 * sRS -> sPO Ack for Response, move to PARTOPEN (8.1.5.) 198 * sPO -> sPO Remain in PARTOPEN state 199 * sOP -> sOP Regular DataAck packet in OPEN state 200 * sCR -> sCR DataAck in CLOSEREQ MAY be processed (8.3.) 201 * sCG -> sCG DataAck in CLOSING MAY be processed (8.3.) 202 * sTW -> sIV 203 * 204 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 205 sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV 206 }, 207 [DCCP_PKT_CLOSEREQ] = { 208 /* 209 * CLOSEREQ may only be sent by the server. 210 * 211 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 212 sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV 213 }, 214 [DCCP_PKT_CLOSE] = { 215 /* 216 * sNO -> sIV No connection 217 * sRQ -> sIV No connection 218 * sRS -> sIV No connection 219 * sPO -> sCG Client-initiated close 220 * sOP -> sCG Client-initiated close 221 * sCR -> sCG Close in response to CloseReq (8.3.) 222 * sCG -> sCG Retransmit 223 * sTW -> sIV Late retransmit, already in TIME_WAIT 224 * 225 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 226 sIV, sIV, sIV, sCG, sCG, sCG, sIV, sIV 227 }, 228 [DCCP_PKT_RESET] = { 229 /* 230 * sNO -> sIV No connection 231 * sRQ -> sTW Sync received or timeout, SHOULD send Reset (8.1.1.) 232 * sRS -> sTW Response received without Request 233 * sPO -> sTW Timeout, SHOULD send Reset (8.1.5.) 234 * sOP -> sTW Connection reset 235 * sCR -> sTW Connection reset 236 * sCG -> sTW Connection reset 237 * sTW -> sIG Ignore (don't refresh timer) 238 * 239 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 240 sIV, sTW, sTW, sTW, sTW, sTW, sTW, sIG 241 }, 242 [DCCP_PKT_SYNC] = { 243 /* 244 * We currently ignore Sync packets 245 * 246 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 247 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG, 248 }, 249 [DCCP_PKT_SYNCACK] = { 250 /* 251 * We currently ignore SyncAck packets 252 * 253 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 254 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG, 255 }, 256 }, 257 [CT_DCCP_ROLE_SERVER] = { 258 [DCCP_PKT_REQUEST] = { 259 /* 260 * sNO -> sIV Invalid 261 * sRQ -> sIG Ignore, conntrack might be out of sync 262 * sRS -> sIG Ignore, conntrack might be out of sync 263 * sPO -> sIG Ignore, conntrack might be out of sync 264 * sOP -> sIG Ignore, conntrack might be out of sync 265 * sCR -> sIG Ignore, conntrack might be out of sync 266 * sCG -> sIG Ignore, conntrack might be out of sync 267 * sTW -> sRQ Reincarnation, must reverse roles 268 * 269 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 270 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sRQ 271 }, 272 [DCCP_PKT_RESPONSE] = { 273 /* 274 * sNO -> sIV Response without Request 275 * sRQ -> sRS Response to clients Request 276 * sRS -> sRS Retransmitted Response (8.1.3. SHOULD NOT) 277 * sPO -> sIG Response to an ignored Request or late retransmit 278 * sOP -> sIG Ignore, might be response to ignored Request 279 * sCR -> sIG Ignore, might be response to ignored Request 280 * sCG -> sIG Ignore, might be response to ignored Request 281 * sTW -> sIV Invalid, Request from client in sTW moves to sRQ 282 * 283 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 284 sIV, sRS, sRS, sIG, sIG, sIG, sIG, sIV 285 }, 286 [DCCP_PKT_ACK] = { 287 /* 288 * sNO -> sIV No connection 289 * sRQ -> sIV No connection 290 * sRS -> sIV No connection 291 * sPO -> sOP Enter OPEN state (8.1.5.) 292 * sOP -> sOP Regular Ack in OPEN state 293 * sCR -> sIV Waiting for Close from client 294 * sCG -> sCG Ack in CLOSING MAY be processed (8.3.) 295 * sTW -> sIV 296 * 297 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 298 sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV 299 }, 300 [DCCP_PKT_DATA] = { 301 /* 302 * sNO -> sIV No connection 303 * sRQ -> sIV No connection 304 * sRS -> sIV No connection 305 * sPO -> sOP Enter OPEN state (8.1.5.) 306 * sOP -> sOP Regular Data packet in OPEN state 307 * sCR -> sIV Waiting for Close from client 308 * sCG -> sCG Data in CLOSING MAY be processed (8.3.) 309 * sTW -> sIV 310 * 311 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 312 sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV 313 }, 314 [DCCP_PKT_DATAACK] = { 315 /* 316 * sNO -> sIV No connection 317 * sRQ -> sIV No connection 318 * sRS -> sIV No connection 319 * sPO -> sOP Enter OPEN state (8.1.5.) 320 * sOP -> sOP Regular DataAck in OPEN state 321 * sCR -> sIV Waiting for Close from client 322 * sCG -> sCG Data in CLOSING MAY be processed (8.3.) 323 * sTW -> sIV 324 * 325 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 326 sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV 327 }, 328 [DCCP_PKT_CLOSEREQ] = { 329 /* 330 * sNO -> sIV No connection 331 * sRQ -> sIV No connection 332 * sRS -> sIV No connection 333 * sPO -> sOP -> sCR Move directly to CLOSEREQ (8.1.5.) 334 * sOP -> sCR CloseReq in OPEN state 335 * sCR -> sCR Retransmit 336 * sCG -> sCR Simultaneous close, client sends another Close 337 * sTW -> sIV Already closed 338 * 339 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 340 sIV, sIV, sIV, sCR, sCR, sCR, sCR, sIV 341 }, 342 [DCCP_PKT_CLOSE] = { 343 /* 344 * sNO -> sIV No connection 345 * sRQ -> sIV No connection 346 * sRS -> sIV No connection 347 * sPO -> sOP -> sCG Move direcly to CLOSING 348 * sOP -> sCG Move to CLOSING 349 * sCR -> sIV Close after CloseReq is invalid 350 * sCG -> sCG Retransmit 351 * sTW -> sIV Already closed 352 * 353 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 354 sIV, sIV, sIV, sCG, sCG, sIV, sCG, sIV 355 }, 356 [DCCP_PKT_RESET] = { 357 /* 358 * sNO -> sIV No connection 359 * sRQ -> sTW Reset in response to Request 360 * sRS -> sTW Timeout, SHOULD send Reset (8.1.3.) 361 * sPO -> sTW Timeout, SHOULD send Reset (8.1.3.) 362 * sOP -> sTW 363 * sCR -> sTW 364 * sCG -> sTW 365 * sTW -> sIG Ignore (don't refresh timer) 366 * 367 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW, sTW */ 368 sIV, sTW, sTW, sTW, sTW, sTW, sTW, sTW, sIG 369 }, 370 [DCCP_PKT_SYNC] = { 371 /* 372 * We currently ignore Sync packets 373 * 374 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 375 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG, 376 }, 377 [DCCP_PKT_SYNCACK] = { 378 /* 379 * We currently ignore SyncAck packets 380 * 381 * sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */ 382 sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIG, 383 }, 384 }, 385 }; 386 387 static inline struct nf_dccp_net *dccp_pernet(struct net *net) 388 { 389 return &net->ct.nf_ct_proto.dccp; 390 } 391 392 static bool dccp_new(struct nf_conn *ct, const struct sk_buff *skb, 393 unsigned int dataoff) 394 { 395 struct net *net = nf_ct_net(ct); 396 struct nf_dccp_net *dn; 397 struct dccp_hdr _dh, *dh; 398 const char *msg; 399 u_int8_t state; 400 401 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &_dh); 402 BUG_ON(dh == NULL); 403 404 state = dccp_state_table[CT_DCCP_ROLE_CLIENT][dh->dccph_type][CT_DCCP_NONE]; 405 switch (state) { 406 default: 407 dn = dccp_pernet(net); 408 if (dn->dccp_loose == 0) { 409 msg = "not picking up existing connection "; 410 goto out_invalid; 411 } 412 case CT_DCCP_REQUEST: 413 break; 414 case CT_DCCP_INVALID: 415 msg = "invalid state transition "; 416 goto out_invalid; 417 } 418 419 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT; 420 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER; 421 ct->proto.dccp.state = CT_DCCP_NONE; 422 ct->proto.dccp.last_pkt = DCCP_PKT_REQUEST; 423 ct->proto.dccp.last_dir = IP_CT_DIR_ORIGINAL; 424 ct->proto.dccp.handshake_seq = 0; 425 return true; 426 427 out_invalid: 428 nf_ct_l4proto_log_invalid(skb, ct, "%s", msg); 429 return false; 430 } 431 432 static u64 dccp_ack_seq(const struct dccp_hdr *dh) 433 { 434 const struct dccp_hdr_ack_bits *dhack; 435 436 dhack = (void *)dh + __dccp_basic_hdr_len(dh); 437 return ((u64)ntohs(dhack->dccph_ack_nr_high) << 32) + 438 ntohl(dhack->dccph_ack_nr_low); 439 } 440 441 static int dccp_packet(struct nf_conn *ct, const struct sk_buff *skb, 442 unsigned int dataoff, enum ip_conntrack_info ctinfo) 443 { 444 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); 445 struct dccp_hdr _dh, *dh; 446 u_int8_t type, old_state, new_state; 447 enum ct_dccp_roles role; 448 unsigned int *timeouts; 449 450 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &_dh); 451 BUG_ON(dh == NULL); 452 type = dh->dccph_type; 453 454 if (type == DCCP_PKT_RESET && 455 !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { 456 /* Tear down connection immediately if only reply is a RESET */ 457 nf_ct_kill_acct(ct, ctinfo, skb); 458 return NF_ACCEPT; 459 } 460 461 spin_lock_bh(&ct->lock); 462 463 role = ct->proto.dccp.role[dir]; 464 old_state = ct->proto.dccp.state; 465 new_state = dccp_state_table[role][type][old_state]; 466 467 switch (new_state) { 468 case CT_DCCP_REQUEST: 469 if (old_state == CT_DCCP_TIMEWAIT && 470 role == CT_DCCP_ROLE_SERVER) { 471 /* Reincarnation in the reverse direction: reopen and 472 * reverse client/server roles. */ 473 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_CLIENT; 474 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_SERVER; 475 } 476 break; 477 case CT_DCCP_RESPOND: 478 if (old_state == CT_DCCP_REQUEST) 479 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh); 480 break; 481 case CT_DCCP_PARTOPEN: 482 if (old_state == CT_DCCP_RESPOND && 483 type == DCCP_PKT_ACK && 484 dccp_ack_seq(dh) == ct->proto.dccp.handshake_seq) 485 set_bit(IPS_ASSURED_BIT, &ct->status); 486 break; 487 case CT_DCCP_IGNORE: 488 /* 489 * Connection tracking might be out of sync, so we ignore 490 * packets that might establish a new connection and resync 491 * if the server responds with a valid Response. 492 */ 493 if (ct->proto.dccp.last_dir == !dir && 494 ct->proto.dccp.last_pkt == DCCP_PKT_REQUEST && 495 type == DCCP_PKT_RESPONSE) { 496 ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_CLIENT; 497 ct->proto.dccp.role[dir] = CT_DCCP_ROLE_SERVER; 498 ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh); 499 new_state = CT_DCCP_RESPOND; 500 break; 501 } 502 ct->proto.dccp.last_dir = dir; 503 ct->proto.dccp.last_pkt = type; 504 505 spin_unlock_bh(&ct->lock); 506 nf_ct_l4proto_log_invalid(skb, ct, "%s", "invalid packet"); 507 return NF_ACCEPT; 508 case CT_DCCP_INVALID: 509 spin_unlock_bh(&ct->lock); 510 nf_ct_l4proto_log_invalid(skb, ct, "%s", "invalid state transition"); 511 return -NF_ACCEPT; 512 } 513 514 ct->proto.dccp.last_dir = dir; 515 ct->proto.dccp.last_pkt = type; 516 ct->proto.dccp.state = new_state; 517 spin_unlock_bh(&ct->lock); 518 519 if (new_state != old_state) 520 nf_conntrack_event_cache(IPCT_PROTOINFO, ct); 521 522 timeouts = nf_ct_timeout_lookup(ct); 523 if (!timeouts) 524 timeouts = dccp_pernet(nf_ct_net(ct))->dccp_timeout; 525 nf_ct_refresh_acct(ct, ctinfo, skb, timeouts[new_state]); 526 527 return NF_ACCEPT; 528 } 529 530 static int dccp_error(struct net *net, struct nf_conn *tmpl, 531 struct sk_buff *skb, unsigned int dataoff, 532 u_int8_t pf, unsigned int hooknum) 533 { 534 struct dccp_hdr _dh, *dh; 535 unsigned int dccp_len = skb->len - dataoff; 536 unsigned int cscov; 537 const char *msg; 538 539 dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &_dh); 540 if (dh == NULL) { 541 msg = "nf_ct_dccp: short packet "; 542 goto out_invalid; 543 } 544 545 if (dh->dccph_doff * 4 < sizeof(struct dccp_hdr) || 546 dh->dccph_doff * 4 > dccp_len) { 547 msg = "nf_ct_dccp: truncated/malformed packet "; 548 goto out_invalid; 549 } 550 551 cscov = dccp_len; 552 if (dh->dccph_cscov) { 553 cscov = (dh->dccph_cscov - 1) * 4; 554 if (cscov > dccp_len) { 555 msg = "nf_ct_dccp: bad checksum coverage "; 556 goto out_invalid; 557 } 558 } 559 560 if (net->ct.sysctl_checksum && hooknum == NF_INET_PRE_ROUTING && 561 nf_checksum_partial(skb, hooknum, dataoff, cscov, IPPROTO_DCCP, 562 pf)) { 563 msg = "nf_ct_dccp: bad checksum "; 564 goto out_invalid; 565 } 566 567 if (dh->dccph_type >= DCCP_PKT_INVALID) { 568 msg = "nf_ct_dccp: reserved packet type "; 569 goto out_invalid; 570 } 571 572 return NF_ACCEPT; 573 574 out_invalid: 575 nf_l4proto_log_invalid(skb, net, pf, IPPROTO_DCCP, "%s", msg); 576 return -NF_ACCEPT; 577 } 578 579 static bool dccp_can_early_drop(const struct nf_conn *ct) 580 { 581 switch (ct->proto.dccp.state) { 582 case CT_DCCP_CLOSEREQ: 583 case CT_DCCP_CLOSING: 584 case CT_DCCP_TIMEWAIT: 585 return true; 586 default: 587 break; 588 } 589 590 return false; 591 } 592 593 #ifdef CONFIG_NF_CONNTRACK_PROCFS 594 static void dccp_print_conntrack(struct seq_file *s, struct nf_conn *ct) 595 { 596 seq_printf(s, "%s ", dccp_state_names[ct->proto.dccp.state]); 597 } 598 #endif 599 600 #if IS_ENABLED(CONFIG_NF_CT_NETLINK) 601 static int dccp_to_nlattr(struct sk_buff *skb, struct nlattr *nla, 602 struct nf_conn *ct) 603 { 604 struct nlattr *nest_parms; 605 606 spin_lock_bh(&ct->lock); 607 nest_parms = nla_nest_start(skb, CTA_PROTOINFO_DCCP | NLA_F_NESTED); 608 if (!nest_parms) 609 goto nla_put_failure; 610 if (nla_put_u8(skb, CTA_PROTOINFO_DCCP_STATE, ct->proto.dccp.state) || 611 nla_put_u8(skb, CTA_PROTOINFO_DCCP_ROLE, 612 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL]) || 613 nla_put_be64(skb, CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ, 614 cpu_to_be64(ct->proto.dccp.handshake_seq), 615 CTA_PROTOINFO_DCCP_PAD)) 616 goto nla_put_failure; 617 nla_nest_end(skb, nest_parms); 618 spin_unlock_bh(&ct->lock); 619 return 0; 620 621 nla_put_failure: 622 spin_unlock_bh(&ct->lock); 623 return -1; 624 } 625 626 static const struct nla_policy dccp_nla_policy[CTA_PROTOINFO_DCCP_MAX + 1] = { 627 [CTA_PROTOINFO_DCCP_STATE] = { .type = NLA_U8 }, 628 [CTA_PROTOINFO_DCCP_ROLE] = { .type = NLA_U8 }, 629 [CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ] = { .type = NLA_U64 }, 630 [CTA_PROTOINFO_DCCP_PAD] = { .type = NLA_UNSPEC }, 631 }; 632 633 #define DCCP_NLATTR_SIZE ( \ 634 NLA_ALIGN(NLA_HDRLEN + 1) + \ 635 NLA_ALIGN(NLA_HDRLEN + 1) + \ 636 NLA_ALIGN(NLA_HDRLEN + sizeof(u64)) + \ 637 NLA_ALIGN(NLA_HDRLEN + 0)) 638 639 static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct) 640 { 641 struct nlattr *attr = cda[CTA_PROTOINFO_DCCP]; 642 struct nlattr *tb[CTA_PROTOINFO_DCCP_MAX + 1]; 643 int err; 644 645 if (!attr) 646 return 0; 647 648 err = nla_parse_nested(tb, CTA_PROTOINFO_DCCP_MAX, attr, 649 dccp_nla_policy, NULL); 650 if (err < 0) 651 return err; 652 653 if (!tb[CTA_PROTOINFO_DCCP_STATE] || 654 !tb[CTA_PROTOINFO_DCCP_ROLE] || 655 nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) > CT_DCCP_ROLE_MAX || 656 nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]) >= CT_DCCP_IGNORE) { 657 return -EINVAL; 658 } 659 660 spin_lock_bh(&ct->lock); 661 ct->proto.dccp.state = nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]); 662 if (nla_get_u8(tb[CTA_PROTOINFO_DCCP_ROLE]) == CT_DCCP_ROLE_CLIENT) { 663 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT; 664 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER; 665 } else { 666 ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_SERVER; 667 ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_CLIENT; 668 } 669 if (tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ]) { 670 ct->proto.dccp.handshake_seq = 671 be64_to_cpu(nla_get_be64(tb[CTA_PROTOINFO_DCCP_HANDSHAKE_SEQ])); 672 } 673 spin_unlock_bh(&ct->lock); 674 return 0; 675 } 676 #endif 677 678 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT 679 680 #include <linux/netfilter/nfnetlink.h> 681 #include <linux/netfilter/nfnetlink_cttimeout.h> 682 683 static int dccp_timeout_nlattr_to_obj(struct nlattr *tb[], 684 struct net *net, void *data) 685 { 686 struct nf_dccp_net *dn = dccp_pernet(net); 687 unsigned int *timeouts = data; 688 int i; 689 690 /* set default DCCP timeouts. */ 691 for (i=0; i<CT_DCCP_MAX; i++) 692 timeouts[i] = dn->dccp_timeout[i]; 693 694 /* there's a 1:1 mapping between attributes and protocol states. */ 695 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) { 696 if (tb[i]) { 697 timeouts[i] = ntohl(nla_get_be32(tb[i])) * HZ; 698 } 699 } 700 701 timeouts[CTA_TIMEOUT_DCCP_UNSPEC] = timeouts[CTA_TIMEOUT_DCCP_REQUEST]; 702 return 0; 703 } 704 705 static int 706 dccp_timeout_obj_to_nlattr(struct sk_buff *skb, const void *data) 707 { 708 const unsigned int *timeouts = data; 709 int i; 710 711 for (i=CTA_TIMEOUT_DCCP_UNSPEC+1; i<CTA_TIMEOUT_DCCP_MAX+1; i++) { 712 if (nla_put_be32(skb, i, htonl(timeouts[i] / HZ))) 713 goto nla_put_failure; 714 } 715 return 0; 716 717 nla_put_failure: 718 return -ENOSPC; 719 } 720 721 static const struct nla_policy 722 dccp_timeout_nla_policy[CTA_TIMEOUT_DCCP_MAX+1] = { 723 [CTA_TIMEOUT_DCCP_REQUEST] = { .type = NLA_U32 }, 724 [CTA_TIMEOUT_DCCP_RESPOND] = { .type = NLA_U32 }, 725 [CTA_TIMEOUT_DCCP_PARTOPEN] = { .type = NLA_U32 }, 726 [CTA_TIMEOUT_DCCP_OPEN] = { .type = NLA_U32 }, 727 [CTA_TIMEOUT_DCCP_CLOSEREQ] = { .type = NLA_U32 }, 728 [CTA_TIMEOUT_DCCP_CLOSING] = { .type = NLA_U32 }, 729 [CTA_TIMEOUT_DCCP_TIMEWAIT] = { .type = NLA_U32 }, 730 }; 731 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */ 732 733 #ifdef CONFIG_SYSCTL 734 /* template, data assigned later */ 735 static struct ctl_table dccp_sysctl_table[] = { 736 { 737 .procname = "nf_conntrack_dccp_timeout_request", 738 .maxlen = sizeof(unsigned int), 739 .mode = 0644, 740 .proc_handler = proc_dointvec_jiffies, 741 }, 742 { 743 .procname = "nf_conntrack_dccp_timeout_respond", 744 .maxlen = sizeof(unsigned int), 745 .mode = 0644, 746 .proc_handler = proc_dointvec_jiffies, 747 }, 748 { 749 .procname = "nf_conntrack_dccp_timeout_partopen", 750 .maxlen = sizeof(unsigned int), 751 .mode = 0644, 752 .proc_handler = proc_dointvec_jiffies, 753 }, 754 { 755 .procname = "nf_conntrack_dccp_timeout_open", 756 .maxlen = sizeof(unsigned int), 757 .mode = 0644, 758 .proc_handler = proc_dointvec_jiffies, 759 }, 760 { 761 .procname = "nf_conntrack_dccp_timeout_closereq", 762 .maxlen = sizeof(unsigned int), 763 .mode = 0644, 764 .proc_handler = proc_dointvec_jiffies, 765 }, 766 { 767 .procname = "nf_conntrack_dccp_timeout_closing", 768 .maxlen = sizeof(unsigned int), 769 .mode = 0644, 770 .proc_handler = proc_dointvec_jiffies, 771 }, 772 { 773 .procname = "nf_conntrack_dccp_timeout_timewait", 774 .maxlen = sizeof(unsigned int), 775 .mode = 0644, 776 .proc_handler = proc_dointvec_jiffies, 777 }, 778 { 779 .procname = "nf_conntrack_dccp_loose", 780 .maxlen = sizeof(int), 781 .mode = 0644, 782 .proc_handler = proc_dointvec, 783 }, 784 { } 785 }; 786 #endif /* CONFIG_SYSCTL */ 787 788 static int dccp_kmemdup_sysctl_table(struct net *net, struct nf_proto_net *pn, 789 struct nf_dccp_net *dn) 790 { 791 #ifdef CONFIG_SYSCTL 792 if (pn->ctl_table) 793 return 0; 794 795 pn->ctl_table = kmemdup(dccp_sysctl_table, 796 sizeof(dccp_sysctl_table), 797 GFP_KERNEL); 798 if (!pn->ctl_table) 799 return -ENOMEM; 800 801 pn->ctl_table[0].data = &dn->dccp_timeout[CT_DCCP_REQUEST]; 802 pn->ctl_table[1].data = &dn->dccp_timeout[CT_DCCP_RESPOND]; 803 pn->ctl_table[2].data = &dn->dccp_timeout[CT_DCCP_PARTOPEN]; 804 pn->ctl_table[3].data = &dn->dccp_timeout[CT_DCCP_OPEN]; 805 pn->ctl_table[4].data = &dn->dccp_timeout[CT_DCCP_CLOSEREQ]; 806 pn->ctl_table[5].data = &dn->dccp_timeout[CT_DCCP_CLOSING]; 807 pn->ctl_table[6].data = &dn->dccp_timeout[CT_DCCP_TIMEWAIT]; 808 pn->ctl_table[7].data = &dn->dccp_loose; 809 810 /* Don't export sysctls to unprivileged users */ 811 if (net->user_ns != &init_user_ns) 812 pn->ctl_table[0].procname = NULL; 813 #endif 814 return 0; 815 } 816 817 static int dccp_init_net(struct net *net, u_int16_t proto) 818 { 819 struct nf_dccp_net *dn = dccp_pernet(net); 820 struct nf_proto_net *pn = &dn->pn; 821 822 if (!pn->users) { 823 /* default values */ 824 dn->dccp_loose = 1; 825 dn->dccp_timeout[CT_DCCP_REQUEST] = 2 * DCCP_MSL; 826 dn->dccp_timeout[CT_DCCP_RESPOND] = 4 * DCCP_MSL; 827 dn->dccp_timeout[CT_DCCP_PARTOPEN] = 4 * DCCP_MSL; 828 dn->dccp_timeout[CT_DCCP_OPEN] = 12 * 3600 * HZ; 829 dn->dccp_timeout[CT_DCCP_CLOSEREQ] = 64 * HZ; 830 dn->dccp_timeout[CT_DCCP_CLOSING] = 64 * HZ; 831 dn->dccp_timeout[CT_DCCP_TIMEWAIT] = 2 * DCCP_MSL; 832 833 /* timeouts[0] is unused, make it same as SYN_SENT so 834 * ->timeouts[0] contains 'new' timeout, like udp or icmp. 835 */ 836 dn->dccp_timeout[CT_DCCP_NONE] = dn->dccp_timeout[CT_DCCP_REQUEST]; 837 } 838 839 return dccp_kmemdup_sysctl_table(net, pn, dn); 840 } 841 842 static struct nf_proto_net *dccp_get_net_proto(struct net *net) 843 { 844 return &net->ct.nf_ct_proto.dccp.pn; 845 } 846 847 const struct nf_conntrack_l4proto nf_conntrack_l4proto_dccp4 = { 848 .l3proto = AF_INET, 849 .l4proto = IPPROTO_DCCP, 850 .new = dccp_new, 851 .packet = dccp_packet, 852 .error = dccp_error, 853 .can_early_drop = dccp_can_early_drop, 854 #ifdef CONFIG_NF_CONNTRACK_PROCFS 855 .print_conntrack = dccp_print_conntrack, 856 #endif 857 #if IS_ENABLED(CONFIG_NF_CT_NETLINK) 858 .nlattr_size = DCCP_NLATTR_SIZE, 859 .to_nlattr = dccp_to_nlattr, 860 .from_nlattr = nlattr_to_dccp, 861 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr, 862 .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size, 863 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple, 864 .nla_policy = nf_ct_port_nla_policy, 865 #endif 866 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT 867 .ctnl_timeout = { 868 .nlattr_to_obj = dccp_timeout_nlattr_to_obj, 869 .obj_to_nlattr = dccp_timeout_obj_to_nlattr, 870 .nlattr_max = CTA_TIMEOUT_DCCP_MAX, 871 .obj_size = sizeof(unsigned int) * CT_DCCP_MAX, 872 .nla_policy = dccp_timeout_nla_policy, 873 }, 874 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */ 875 .init_net = dccp_init_net, 876 .get_net_proto = dccp_get_net_proto, 877 }; 878 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_dccp4); 879 880 const struct nf_conntrack_l4proto nf_conntrack_l4proto_dccp6 = { 881 .l3proto = AF_INET6, 882 .l4proto = IPPROTO_DCCP, 883 .new = dccp_new, 884 .packet = dccp_packet, 885 .error = dccp_error, 886 .can_early_drop = dccp_can_early_drop, 887 #ifdef CONFIG_NF_CONNTRACK_PROCFS 888 .print_conntrack = dccp_print_conntrack, 889 #endif 890 #if IS_ENABLED(CONFIG_NF_CT_NETLINK) 891 .nlattr_size = DCCP_NLATTR_SIZE, 892 .to_nlattr = dccp_to_nlattr, 893 .from_nlattr = nlattr_to_dccp, 894 .tuple_to_nlattr = nf_ct_port_tuple_to_nlattr, 895 .nlattr_tuple_size = nf_ct_port_nlattr_tuple_size, 896 .nlattr_to_tuple = nf_ct_port_nlattr_to_tuple, 897 .nla_policy = nf_ct_port_nla_policy, 898 #endif 899 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT 900 .ctnl_timeout = { 901 .nlattr_to_obj = dccp_timeout_nlattr_to_obj, 902 .obj_to_nlattr = dccp_timeout_obj_to_nlattr, 903 .nlattr_max = CTA_TIMEOUT_DCCP_MAX, 904 .obj_size = sizeof(unsigned int) * CT_DCCP_MAX, 905 .nla_policy = dccp_timeout_nla_policy, 906 }, 907 #endif /* CONFIG_NF_CONNTRACK_TIMEOUT */ 908 .init_net = dccp_init_net, 909 .get_net_proto = dccp_get_net_proto, 910 }; 911 EXPORT_SYMBOL_GPL(nf_conntrack_l4proto_dccp6); 912