1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * inet_diag.c Module for monitoring INET transport protocols sockets. 4 * 5 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 6 */ 7 8 #include <linux/kernel.h> 9 #include <linux/module.h> 10 #include <linux/types.h> 11 #include <linux/fcntl.h> 12 #include <linux/random.h> 13 #include <linux/slab.h> 14 #include <linux/cache.h> 15 #include <linux/init.h> 16 #include <linux/time.h> 17 18 #include <net/icmp.h> 19 #include <net/tcp.h> 20 #include <net/ipv6.h> 21 #include <net/inet_common.h> 22 #include <net/inet_connection_sock.h> 23 #include <net/bpf_sk_storage.h> 24 #include <net/netlink.h> 25 26 #include <linux/inet.h> 27 #include <linux/stddef.h> 28 29 #include <linux/inet_diag.h> 30 #include <linux/sock_diag.h> 31 32 static const struct inet_diag_handler __rcu **inet_diag_table; 33 34 struct inet_diag_entry { 35 const __be32 *saddr; 36 const __be32 *daddr; 37 u16 sport; 38 u16 dport; 39 u16 family; 40 u16 userlocks; 41 u32 ifindex; 42 u32 mark; 43 #ifdef CONFIG_SOCK_CGROUP_DATA 44 u64 cgroup_id; 45 #endif 46 }; 47 48 static const struct inet_diag_handler *inet_diag_lock_handler(int proto) 49 { 50 const struct inet_diag_handler *handler; 51 52 if (proto < 0 || proto >= IPPROTO_MAX) 53 return NULL; 54 55 if (!READ_ONCE(inet_diag_table[proto])) 56 sock_load_diag_module(AF_INET, proto); 57 58 rcu_read_lock(); 59 handler = rcu_dereference(inet_diag_table[proto]); 60 if (handler && !try_module_get(handler->owner)) 61 handler = NULL; 62 rcu_read_unlock(); 63 64 return handler; 65 } 66 67 static void inet_diag_unlock_handler(const struct inet_diag_handler *handler) 68 { 69 module_put(handler->owner); 70 } 71 72 void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk) 73 { 74 r->idiag_family = READ_ONCE(sk->sk_family); 75 76 r->id.idiag_sport = htons(READ_ONCE(sk->sk_num)); 77 r->id.idiag_dport = READ_ONCE(sk->sk_dport); 78 r->id.idiag_if = READ_ONCE(sk->sk_bound_dev_if); 79 sock_diag_save_cookie(sk, r->id.idiag_cookie); 80 81 #if IS_ENABLED(CONFIG_IPV6) 82 if (r->idiag_family == AF_INET6) { 83 data_race(*(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr); 84 data_race(*(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr); 85 } else 86 #endif 87 { 88 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src)); 89 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst)); 90 91 r->id.idiag_src[0] = READ_ONCE(sk->sk_rcv_saddr); 92 r->id.idiag_dst[0] = READ_ONCE(sk->sk_daddr); 93 } 94 } 95 EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill); 96 97 int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb, 98 struct inet_diag_msg *r, int ext, 99 struct user_namespace *user_ns, 100 bool net_admin) 101 { 102 const struct inet_sock *inet = inet_sk(sk); 103 struct inet_diag_sockopt inet_sockopt; 104 105 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown)) 106 goto errout; 107 108 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections, 109 * hence this needs to be included regardless of socket family. 110 */ 111 if (ext & (1 << (INET_DIAG_TOS - 1))) 112 if (nla_put_u8(skb, INET_DIAG_TOS, READ_ONCE(inet->tos)) < 0) 113 goto errout; 114 115 #if IS_ENABLED(CONFIG_IPV6) 116 if (r->idiag_family == AF_INET6) { 117 if (ext & (1 << (INET_DIAG_TCLASS - 1))) 118 if (nla_put_u8(skb, INET_DIAG_TCLASS, 119 inet6_sk(sk)->tclass) < 0) 120 goto errout; 121 122 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) && 123 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk))) 124 goto errout; 125 } 126 #endif 127 128 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, READ_ONCE(sk->sk_mark))) 129 goto errout; 130 131 if (ext & (1 << (INET_DIAG_CLASS_ID - 1)) || 132 ext & (1 << (INET_DIAG_TCLASS - 1))) { 133 u32 classid = 0; 134 135 #ifdef CONFIG_CGROUP_NET_CLASSID 136 classid = sock_cgroup_classid(&sk->sk_cgrp_data); 137 #endif 138 /* Fallback to socket priority if class id isn't set. 139 * Classful qdiscs use it as direct reference to class. 140 * For cgroup2 classid is always zero. 141 */ 142 if (!classid) 143 classid = READ_ONCE(sk->sk_priority); 144 145 if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid)) 146 goto errout; 147 } 148 149 #ifdef CONFIG_SOCK_CGROUP_DATA 150 if (nla_put_u64_64bit(skb, INET_DIAG_CGROUP_ID, 151 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)), 152 INET_DIAG_PAD)) 153 goto errout; 154 #endif 155 156 r->idiag_uid = from_kuid_munged(user_ns, sk_uid(sk)); 157 r->idiag_inode = sock_i_ino(sk); 158 159 memset(&inet_sockopt, 0, sizeof(inet_sockopt)); 160 inet_sockopt.recverr = inet_test_bit(RECVERR, sk); 161 inet_sockopt.is_icsk = inet_test_bit(IS_ICSK, sk); 162 inet_sockopt.freebind = inet_test_bit(FREEBIND, sk); 163 inet_sockopt.hdrincl = inet_test_bit(HDRINCL, sk); 164 inet_sockopt.mc_loop = inet_test_bit(MC_LOOP, sk); 165 inet_sockopt.transparent = inet_test_bit(TRANSPARENT, sk); 166 inet_sockopt.mc_all = inet_test_bit(MC_ALL, sk); 167 inet_sockopt.nodefrag = inet_test_bit(NODEFRAG, sk); 168 inet_sockopt.bind_address_no_port = inet_test_bit(BIND_ADDRESS_NO_PORT, sk); 169 inet_sockopt.recverr_rfc4884 = inet_test_bit(RECVERR_RFC4884, sk); 170 inet_sockopt.defer_connect = inet_test_bit(DEFER_CONNECT, sk); 171 if (nla_put(skb, INET_DIAG_SOCKOPT, sizeof(inet_sockopt), 172 &inet_sockopt)) 173 goto errout; 174 175 return 0; 176 errout: 177 return 1; 178 } 179 EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill); 180 181 static int inet_diag_parse_attrs(const struct nlmsghdr *nlh, int hdrlen, 182 struct nlattr **req_nlas) 183 { 184 struct nlattr *nla; 185 int remaining; 186 187 nlmsg_for_each_attr(nla, nlh, hdrlen, remaining) { 188 int type = nla_type(nla); 189 190 if (type == INET_DIAG_REQ_PROTOCOL && nla_len(nla) != sizeof(u32)) 191 return -EINVAL; 192 193 if (type < __INET_DIAG_REQ_MAX) 194 req_nlas[type] = nla; 195 } 196 return 0; 197 } 198 199 static int inet_diag_get_protocol(const struct inet_diag_req_v2 *req, 200 const struct inet_diag_dump_data *data) 201 { 202 if (data->req_nlas[INET_DIAG_REQ_PROTOCOL]) 203 return nla_get_u32(data->req_nlas[INET_DIAG_REQ_PROTOCOL]); 204 return req->sdiag_protocol; 205 } 206 207 #define MAX_DUMP_ALLOC_SIZE (KMALLOC_MAX_SIZE - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) 208 209 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk, 210 struct sk_buff *skb, struct netlink_callback *cb, 211 const struct inet_diag_req_v2 *req, 212 u16 nlmsg_flags, bool net_admin) 213 { 214 const struct tcp_congestion_ops *ca_ops; 215 const struct inet_diag_handler *handler; 216 struct inet_diag_dump_data *cb_data; 217 int ext = req->idiag_ext; 218 struct inet_diag_msg *r; 219 struct nlmsghdr *nlh; 220 struct nlattr *attr; 221 void *info = NULL; 222 u8 icsk_pending; 223 int protocol; 224 225 cb_data = cb->data; 226 protocol = inet_diag_get_protocol(req, cb_data); 227 228 /* inet_diag_lock_handler() made sure inet_diag_table[] is stable. */ 229 handler = rcu_dereference_protected(inet_diag_table[protocol], 1); 230 DEBUG_NET_WARN_ON_ONCE(!handler); 231 if (!handler) 232 return -ENXIO; 233 234 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, 235 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags); 236 if (!nlh) 237 return -EMSGSIZE; 238 239 r = nlmsg_data(nlh); 240 BUG_ON(!sk_fullsock(sk)); 241 242 inet_diag_msg_common_fill(r, sk); 243 r->idiag_state = sk->sk_state; 244 r->idiag_timer = 0; 245 r->idiag_retrans = 0; 246 r->idiag_expires = 0; 247 248 if (inet_diag_msg_attrs_fill(sk, skb, r, ext, 249 sk_user_ns(NETLINK_CB(cb->skb).sk), 250 net_admin)) 251 goto errout; 252 253 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) { 254 struct inet_diag_meminfo minfo = { 255 .idiag_rmem = sk_rmem_alloc_get(sk), 256 .idiag_wmem = READ_ONCE(sk->sk_wmem_queued), 257 .idiag_fmem = READ_ONCE(sk->sk_forward_alloc), 258 .idiag_tmem = sk_wmem_alloc_get(sk), 259 }; 260 261 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0) 262 goto errout; 263 } 264 265 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1))) 266 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO)) 267 goto errout; 268 269 /* 270 * RAW sockets might have user-defined protocols assigned, 271 * so report the one supplied on socket creation. 272 */ 273 if (sk->sk_type == SOCK_RAW) { 274 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol)) 275 goto errout; 276 } 277 278 if (!icsk) { 279 handler->idiag_get_info(sk, r, NULL); 280 goto out; 281 } 282 283 icsk_pending = smp_load_acquire(&icsk->icsk_pending); 284 if (icsk_pending == ICSK_TIME_RETRANS || 285 icsk_pending == ICSK_TIME_REO_TIMEOUT || 286 icsk_pending == ICSK_TIME_LOSS_PROBE) { 287 r->idiag_timer = 1; 288 r->idiag_retrans = READ_ONCE(icsk->icsk_retransmits); 289 r->idiag_expires = 290 jiffies_delta_to_msecs(icsk_timeout(icsk) - jiffies); 291 } else if (icsk_pending == ICSK_TIME_PROBE0) { 292 r->idiag_timer = 4; 293 r->idiag_retrans = READ_ONCE(icsk->icsk_probes_out); 294 r->idiag_expires = 295 jiffies_delta_to_msecs(icsk_timeout(icsk) - jiffies); 296 } else if (timer_pending(&sk->sk_timer)) { 297 r->idiag_timer = 2; 298 r->idiag_retrans = READ_ONCE(icsk->icsk_probes_out); 299 r->idiag_expires = 300 jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies); 301 } 302 303 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) { 304 attr = nla_reserve_64bit(skb, INET_DIAG_INFO, 305 handler->idiag_info_size, 306 INET_DIAG_PAD); 307 if (!attr) 308 goto errout; 309 310 info = nla_data(attr); 311 } 312 313 if (ext & (1 << (INET_DIAG_CONG - 1))) { 314 int err = 0; 315 316 rcu_read_lock(); 317 ca_ops = READ_ONCE(icsk->icsk_ca_ops); 318 if (ca_ops) 319 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name); 320 rcu_read_unlock(); 321 if (err < 0) 322 goto errout; 323 } 324 325 handler->idiag_get_info(sk, r, info); 326 327 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux) 328 if (handler->idiag_get_aux(sk, net_admin, skb) < 0) 329 goto errout; 330 331 if (sk->sk_state < TCP_TIME_WAIT) { 332 union tcp_cc_info info; 333 size_t sz = 0; 334 int attr; 335 336 rcu_read_lock(); 337 ca_ops = READ_ONCE(icsk->icsk_ca_ops); 338 if (ca_ops && ca_ops->get_info) 339 sz = ca_ops->get_info(sk, ext, &attr, &info); 340 rcu_read_unlock(); 341 if (sz && nla_put(skb, attr, sz, &info) < 0) 342 goto errout; 343 } 344 345 /* Keep it at the end for potential retry with a larger skb, 346 * or else do best-effort fitting, which is only done for the 347 * first_nlmsg. 348 */ 349 if (cb_data->bpf_stg_diag) { 350 bool first_nlmsg = ((unsigned char *)nlh == skb->data); 351 unsigned int prev_min_dump_alloc; 352 unsigned int total_nla_size = 0; 353 unsigned int msg_len; 354 int err; 355 356 msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh; 357 err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb, 358 INET_DIAG_SK_BPF_STORAGES, 359 &total_nla_size); 360 361 if (!err) 362 goto out; 363 364 total_nla_size += msg_len; 365 prev_min_dump_alloc = cb->min_dump_alloc; 366 if (total_nla_size > prev_min_dump_alloc) 367 cb->min_dump_alloc = min_t(u32, total_nla_size, 368 MAX_DUMP_ALLOC_SIZE); 369 370 if (!first_nlmsg) 371 goto errout; 372 373 if (cb->min_dump_alloc > prev_min_dump_alloc) 374 /* Retry with pskb_expand_head() with 375 * __GFP_DIRECT_RECLAIM 376 */ 377 goto errout; 378 379 WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc); 380 381 /* Send what we have for this sk 382 * and move on to the next sk in the following 383 * dump() 384 */ 385 } 386 387 out: 388 nlmsg_end(skb, nlh); 389 return 0; 390 391 errout: 392 nlmsg_cancel(skb, nlh); 393 return -EMSGSIZE; 394 } 395 EXPORT_SYMBOL_GPL(inet_sk_diag_fill); 396 397 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb, 398 const struct nlmsghdr *nlh, 399 int hdrlen, 400 const struct inet_diag_req_v2 *req) 401 { 402 const struct inet_diag_handler *handler; 403 struct inet_diag_dump_data dump_data; 404 int err, protocol; 405 406 memset(&dump_data, 0, sizeof(dump_data)); 407 err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas); 408 if (err) 409 return err; 410 411 protocol = inet_diag_get_protocol(req, &dump_data); 412 413 handler = inet_diag_lock_handler(protocol); 414 if (!handler) 415 return -ENOENT; 416 417 if (cmd == SOCK_DIAG_BY_FAMILY) { 418 struct netlink_callback cb = { 419 .nlh = nlh, 420 .skb = in_skb, 421 .data = &dump_data, 422 }; 423 err = handler->dump_one(&cb, req); 424 } else if (cmd == SOCK_DESTROY && handler->destroy) { 425 err = handler->destroy(in_skb, req); 426 } else { 427 err = -EOPNOTSUPP; 428 } 429 inet_diag_unlock_handler(handler); 430 431 return err; 432 } 433 434 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits) 435 { 436 int words = bits >> 5; 437 438 bits &= 0x1f; 439 440 if (words) { 441 if (memcmp(a1, a2, words << 2)) 442 return 0; 443 } 444 if (bits) { 445 __be32 w1, w2; 446 __be32 mask; 447 448 w1 = a1[words]; 449 w2 = a2[words]; 450 451 mask = htonl((0xffffffff) << (32 - bits)); 452 453 if ((w1 ^ w2) & mask) 454 return 0; 455 } 456 457 return 1; 458 } 459 460 static int inet_diag_bc_run(const struct nlattr *_bc, 461 const struct inet_diag_entry *entry) 462 { 463 const void *bc = nla_data(_bc); 464 int len = nla_len(_bc); 465 466 while (len > 0) { 467 int yes = 1; 468 const struct inet_diag_bc_op *op = bc; 469 470 switch (op->code) { 471 case INET_DIAG_BC_NOP: 472 break; 473 case INET_DIAG_BC_JMP: 474 yes = 0; 475 break; 476 case INET_DIAG_BC_S_EQ: 477 yes = entry->sport == op[1].no; 478 break; 479 case INET_DIAG_BC_S_GE: 480 yes = entry->sport >= op[1].no; 481 break; 482 case INET_DIAG_BC_S_LE: 483 yes = entry->sport <= op[1].no; 484 break; 485 case INET_DIAG_BC_D_EQ: 486 yes = entry->dport == op[1].no; 487 break; 488 case INET_DIAG_BC_D_GE: 489 yes = entry->dport >= op[1].no; 490 break; 491 case INET_DIAG_BC_D_LE: 492 yes = entry->dport <= op[1].no; 493 break; 494 case INET_DIAG_BC_AUTO: 495 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK); 496 break; 497 case INET_DIAG_BC_S_COND: 498 case INET_DIAG_BC_D_COND: { 499 const struct inet_diag_hostcond *cond; 500 const __be32 *addr; 501 502 cond = (const struct inet_diag_hostcond *)(op + 1); 503 if (cond->port != -1 && 504 cond->port != (op->code == INET_DIAG_BC_S_COND ? 505 entry->sport : entry->dport)) { 506 yes = 0; 507 break; 508 } 509 510 if (op->code == INET_DIAG_BC_S_COND) 511 addr = entry->saddr; 512 else 513 addr = entry->daddr; 514 515 if (cond->family != AF_UNSPEC && 516 cond->family != entry->family) { 517 if (entry->family == AF_INET6 && 518 cond->family == AF_INET) { 519 if (addr[0] == 0 && addr[1] == 0 && 520 addr[2] == htonl(0xffff) && 521 bitstring_match(addr + 3, 522 cond->addr, 523 cond->prefix_len)) 524 break; 525 } 526 yes = 0; 527 break; 528 } 529 530 if (cond->prefix_len == 0) 531 break; 532 if (bitstring_match(addr, cond->addr, 533 cond->prefix_len)) 534 break; 535 yes = 0; 536 break; 537 } 538 case INET_DIAG_BC_DEV_COND: { 539 u32 ifindex; 540 541 ifindex = *((const u32 *)(op + 1)); 542 if (ifindex != entry->ifindex) 543 yes = 0; 544 break; 545 } 546 case INET_DIAG_BC_MARK_COND: { 547 struct inet_diag_markcond *cond; 548 549 cond = (struct inet_diag_markcond *)(op + 1); 550 if ((entry->mark & cond->mask) != cond->mark) 551 yes = 0; 552 break; 553 } 554 #ifdef CONFIG_SOCK_CGROUP_DATA 555 case INET_DIAG_BC_CGROUP_COND: { 556 u64 cgroup_id; 557 558 cgroup_id = get_unaligned((const u64 *)(op + 1)); 559 if (cgroup_id != entry->cgroup_id) 560 yes = 0; 561 break; 562 } 563 #endif 564 } 565 566 if (yes) { 567 len -= op->yes; 568 bc += op->yes; 569 } else { 570 len -= op->no; 571 bc += op->no; 572 } 573 } 574 return len == 0; 575 } 576 577 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV) 578 */ 579 static void entry_fill_addrs(struct inet_diag_entry *entry, 580 const struct sock *sk) 581 { 582 #if IS_ENABLED(CONFIG_IPV6) 583 if (entry->family == AF_INET6) { 584 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32; 585 entry->daddr = sk->sk_v6_daddr.s6_addr32; 586 } else 587 #endif 588 { 589 entry->saddr = &sk->sk_rcv_saddr; 590 entry->daddr = &sk->sk_daddr; 591 } 592 } 593 594 int inet_diag_bc_sk(const struct inet_diag_dump_data *cb_data, struct sock *sk) 595 { 596 const struct nlattr *bc = cb_data->inet_diag_nla_bc; 597 const struct inet_sock *inet = inet_sk(sk); 598 struct inet_diag_entry entry; 599 600 if (!bc) 601 return 1; 602 603 entry.family = READ_ONCE(sk->sk_family); 604 entry_fill_addrs(&entry, sk); 605 entry.sport = READ_ONCE(inet->inet_num); 606 entry.dport = ntohs(READ_ONCE(inet->inet_dport)); 607 entry.ifindex = READ_ONCE(sk->sk_bound_dev_if); 608 if (cb_data->userlocks_needed) 609 entry.userlocks = sk_fullsock(sk) ? READ_ONCE(sk->sk_userlocks) : 0; 610 if (cb_data->mark_needed) { 611 if (sk_fullsock(sk)) 612 entry.mark = READ_ONCE(sk->sk_mark); 613 else if (sk->sk_state == TCP_NEW_SYN_RECV) 614 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark; 615 else if (sk->sk_state == TCP_TIME_WAIT) 616 entry.mark = inet_twsk(sk)->tw_mark; 617 else 618 entry.mark = 0; 619 } 620 #ifdef CONFIG_SOCK_CGROUP_DATA 621 if (cb_data->cgroup_needed) 622 entry.cgroup_id = sk_fullsock(sk) ? 623 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0; 624 #endif 625 626 return inet_diag_bc_run(bc, &entry); 627 } 628 EXPORT_SYMBOL_GPL(inet_diag_bc_sk); 629 630 static int valid_cc(const void *bc, int len, int cc) 631 { 632 while (len >= 0) { 633 const struct inet_diag_bc_op *op = bc; 634 635 if (cc > len) 636 return 0; 637 if (cc == len) 638 return 1; 639 if (op->yes < 4 || op->yes & 3) 640 return 0; 641 len -= op->yes; 642 bc += op->yes; 643 } 644 return 0; 645 } 646 647 /* data is u32 ifindex */ 648 static bool valid_devcond(const struct inet_diag_bc_op *op, int len, 649 int *min_len) 650 { 651 /* Check ifindex space. */ 652 *min_len += sizeof(u32); 653 if (len < *min_len) 654 return false; 655 656 return true; 657 } 658 /* Validate an inet_diag_hostcond. */ 659 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len, 660 int *min_len) 661 { 662 struct inet_diag_hostcond *cond; 663 int addr_len; 664 665 /* Check hostcond space. */ 666 *min_len += sizeof(struct inet_diag_hostcond); 667 if (len < *min_len) 668 return false; 669 cond = (struct inet_diag_hostcond *)(op + 1); 670 671 /* Check address family and address length. */ 672 switch (cond->family) { 673 case AF_UNSPEC: 674 addr_len = 0; 675 break; 676 case AF_INET: 677 addr_len = sizeof(struct in_addr); 678 break; 679 case AF_INET6: 680 addr_len = sizeof(struct in6_addr); 681 break; 682 default: 683 return false; 684 } 685 *min_len += addr_len; 686 if (len < *min_len) 687 return false; 688 689 /* Check prefix length (in bits) vs address length (in bytes). */ 690 if (cond->prefix_len > 8 * addr_len) 691 return false; 692 693 return true; 694 } 695 696 /* Validate a port comparison operator. */ 697 static bool valid_port_comparison(const struct inet_diag_bc_op *op, 698 int len, int *min_len) 699 { 700 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */ 701 *min_len += sizeof(struct inet_diag_bc_op); 702 if (len < *min_len) 703 return false; 704 return true; 705 } 706 707 static bool valid_markcond(const struct inet_diag_bc_op *op, int len, 708 int *min_len) 709 { 710 *min_len += sizeof(struct inet_diag_markcond); 711 return len >= *min_len; 712 } 713 714 #ifdef CONFIG_SOCK_CGROUP_DATA 715 static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len, 716 int *min_len) 717 { 718 *min_len += sizeof(u64); 719 return len >= *min_len; 720 } 721 #endif 722 723 static int inet_diag_bc_audit(struct inet_diag_dump_data *cb_data, 724 const struct sk_buff *skb) 725 { 726 const struct nlattr *attr = cb_data->inet_diag_nla_bc; 727 const void *bytecode, *bc; 728 int bytecode_len, len; 729 bool net_admin; 730 731 if (!attr) 732 return 0; 733 734 if (nla_len(attr) < sizeof(struct inet_diag_bc_op)) 735 return -EINVAL; 736 737 net_admin = netlink_net_capable(skb, CAP_NET_ADMIN); 738 bytecode = bc = nla_data(attr); 739 len = bytecode_len = nla_len(attr); 740 741 while (len > 0) { 742 int min_len = sizeof(struct inet_diag_bc_op); 743 const struct inet_diag_bc_op *op = bc; 744 745 switch (op->code) { 746 case INET_DIAG_BC_S_COND: 747 case INET_DIAG_BC_D_COND: 748 if (!valid_hostcond(bc, len, &min_len)) 749 return -EINVAL; 750 break; 751 case INET_DIAG_BC_DEV_COND: 752 if (!valid_devcond(bc, len, &min_len)) 753 return -EINVAL; 754 break; 755 case INET_DIAG_BC_S_EQ: 756 case INET_DIAG_BC_S_GE: 757 case INET_DIAG_BC_S_LE: 758 case INET_DIAG_BC_D_EQ: 759 case INET_DIAG_BC_D_GE: 760 case INET_DIAG_BC_D_LE: 761 if (!valid_port_comparison(bc, len, &min_len)) 762 return -EINVAL; 763 break; 764 case INET_DIAG_BC_MARK_COND: 765 if (!net_admin) 766 return -EPERM; 767 if (!valid_markcond(bc, len, &min_len)) 768 return -EINVAL; 769 cb_data->mark_needed = true; 770 break; 771 #ifdef CONFIG_SOCK_CGROUP_DATA 772 case INET_DIAG_BC_CGROUP_COND: 773 if (!valid_cgroupcond(bc, len, &min_len)) 774 return -EINVAL; 775 cb_data->cgroup_needed = true; 776 break; 777 #endif 778 case INET_DIAG_BC_AUTO: 779 cb_data->userlocks_needed = true; 780 fallthrough; 781 case INET_DIAG_BC_JMP: 782 case INET_DIAG_BC_NOP: 783 break; 784 default: 785 return -EINVAL; 786 } 787 788 if (op->code != INET_DIAG_BC_NOP) { 789 if (op->no < min_len || op->no > len + 4 || op->no & 3) 790 return -EINVAL; 791 if (op->no < len && 792 !valid_cc(bytecode, bytecode_len, len - op->no)) 793 return -EINVAL; 794 } 795 796 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3) 797 return -EINVAL; 798 bc += op->yes; 799 len -= op->yes; 800 } 801 return len == 0 ? 0 : -EINVAL; 802 } 803 804 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb, 805 const struct inet_diag_req_v2 *r) 806 { 807 struct inet_diag_dump_data *cb_data = cb->data; 808 const struct inet_diag_handler *handler; 809 u32 prev_min_dump_alloc; 810 int protocol, err = 0; 811 812 protocol = inet_diag_get_protocol(r, cb_data); 813 814 again: 815 prev_min_dump_alloc = cb->min_dump_alloc; 816 handler = inet_diag_lock_handler(protocol); 817 if (handler) { 818 handler->dump(skb, cb, r); 819 inet_diag_unlock_handler(handler); 820 } else { 821 err = -ENOENT; 822 } 823 /* The skb is not large enough to fit one sk info and 824 * inet_sk_diag_fill() has requested for a larger skb. 825 */ 826 if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) { 827 err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL); 828 if (!err) 829 goto again; 830 } 831 832 return err ? : skb->len; 833 } 834 835 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb) 836 { 837 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh)); 838 } 839 840 static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen) 841 { 842 const struct nlmsghdr *nlh = cb->nlh; 843 struct inet_diag_dump_data *cb_data; 844 struct sk_buff *skb = cb->skb; 845 struct nlattr *nla; 846 int err; 847 848 cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL); 849 if (!cb_data) 850 return -ENOMEM; 851 852 err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas); 853 if (err) { 854 kfree(cb_data); 855 return err; 856 } 857 err = inet_diag_bc_audit(cb_data, skb); 858 if (err) { 859 kfree(cb_data); 860 return err; 861 } 862 863 nla = cb_data->inet_diag_nla_bpf_stgs; 864 if (nla) { 865 struct bpf_sk_storage_diag *bpf_stg_diag; 866 867 bpf_stg_diag = bpf_sk_storage_diag_alloc(nla); 868 if (IS_ERR(bpf_stg_diag)) { 869 kfree(cb_data); 870 return PTR_ERR(bpf_stg_diag); 871 } 872 cb_data->bpf_stg_diag = bpf_stg_diag; 873 } 874 875 cb->data = cb_data; 876 return 0; 877 } 878 879 static int inet_diag_dump_start(struct netlink_callback *cb) 880 { 881 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2)); 882 } 883 884 static int inet_diag_dump_start_compat(struct netlink_callback *cb) 885 { 886 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req)); 887 } 888 889 static int inet_diag_dump_done(struct netlink_callback *cb) 890 { 891 struct inet_diag_dump_data *cb_data = cb->data; 892 893 bpf_sk_storage_diag_free(cb_data->bpf_stg_diag); 894 kfree(cb->data); 895 896 return 0; 897 } 898 899 static int inet_diag_type2proto(int type) 900 { 901 switch (type) { 902 case TCPDIAG_GETSOCK: 903 return IPPROTO_TCP; 904 default: 905 return 0; 906 } 907 } 908 909 static int inet_diag_dump_compat(struct sk_buff *skb, 910 struct netlink_callback *cb) 911 { 912 struct inet_diag_req *rc = nlmsg_data(cb->nlh); 913 struct inet_diag_req_v2 req; 914 915 req.sdiag_family = AF_UNSPEC; /* compatibility */ 916 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type); 917 req.idiag_ext = rc->idiag_ext; 918 req.pad = 0; 919 req.idiag_states = rc->idiag_states; 920 req.id = rc->id; 921 922 return __inet_diag_dump(skb, cb, &req); 923 } 924 925 static int inet_diag_get_exact_compat(struct sk_buff *in_skb, 926 const struct nlmsghdr *nlh) 927 { 928 struct inet_diag_req *rc = nlmsg_data(nlh); 929 struct inet_diag_req_v2 req; 930 931 req.sdiag_family = rc->idiag_family; 932 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type); 933 req.idiag_ext = rc->idiag_ext; 934 req.pad = 0; 935 req.idiag_states = rc->idiag_states; 936 req.id = rc->id; 937 938 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh, 939 sizeof(struct inet_diag_req), &req); 940 } 941 942 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh) 943 { 944 int hdrlen = sizeof(struct inet_diag_req); 945 struct net *net = sock_net(skb->sk); 946 947 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX || 948 nlmsg_len(nlh) < hdrlen) 949 return -EINVAL; 950 951 if (nlh->nlmsg_flags & NLM_F_DUMP) { 952 struct netlink_dump_control c = { 953 .start = inet_diag_dump_start_compat, 954 .done = inet_diag_dump_done, 955 .dump = inet_diag_dump_compat, 956 }; 957 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c); 958 } 959 960 return inet_diag_get_exact_compat(skb, nlh); 961 } 962 963 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h) 964 { 965 int hdrlen = sizeof(struct inet_diag_req_v2); 966 struct net *net = sock_net(skb->sk); 967 968 if (nlmsg_len(h) < hdrlen) 969 return -EINVAL; 970 971 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY && 972 h->nlmsg_flags & NLM_F_DUMP) { 973 struct netlink_dump_control c = { 974 .start = inet_diag_dump_start, 975 .done = inet_diag_dump_done, 976 .dump = inet_diag_dump, 977 }; 978 return netlink_dump_start(net->diag_nlsk, skb, h, &c); 979 } 980 981 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen, 982 nlmsg_data(h)); 983 } 984 985 static 986 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk) 987 { 988 const struct inet_diag_handler *handler; 989 struct nlmsghdr *nlh; 990 struct nlattr *attr; 991 struct inet_diag_msg *r; 992 void *info = NULL; 993 int err = 0; 994 995 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0); 996 if (!nlh) 997 return -ENOMEM; 998 999 r = nlmsg_data(nlh); 1000 memset(r, 0, sizeof(*r)); 1001 inet_diag_msg_common_fill(r, sk); 1002 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM) 1003 r->id.idiag_sport = inet_sk(sk)->inet_sport; 1004 r->idiag_state = sk->sk_state; 1005 1006 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) { 1007 nlmsg_cancel(skb, nlh); 1008 return err; 1009 } 1010 1011 handler = inet_diag_lock_handler(sk->sk_protocol); 1012 if (!handler) { 1013 nlmsg_cancel(skb, nlh); 1014 return -ENOENT; 1015 } 1016 1017 attr = handler->idiag_info_size 1018 ? nla_reserve_64bit(skb, INET_DIAG_INFO, 1019 handler->idiag_info_size, 1020 INET_DIAG_PAD) 1021 : NULL; 1022 if (attr) 1023 info = nla_data(attr); 1024 1025 handler->idiag_get_info(sk, r, info); 1026 inet_diag_unlock_handler(handler); 1027 1028 nlmsg_end(skb, nlh); 1029 return 0; 1030 } 1031 1032 static const struct sock_diag_handler inet_diag_handler = { 1033 .owner = THIS_MODULE, 1034 .family = AF_INET, 1035 .dump = inet_diag_handler_cmd, 1036 .get_info = inet_diag_handler_get_info, 1037 .destroy = inet_diag_handler_cmd, 1038 }; 1039 1040 static const struct sock_diag_handler inet6_diag_handler = { 1041 .owner = THIS_MODULE, 1042 .family = AF_INET6, 1043 .dump = inet_diag_handler_cmd, 1044 .get_info = inet_diag_handler_get_info, 1045 .destroy = inet_diag_handler_cmd, 1046 }; 1047 1048 int inet_diag_register(const struct inet_diag_handler *h) 1049 { 1050 const __u16 type = h->idiag_type; 1051 1052 if (type >= IPPROTO_MAX) 1053 return -EINVAL; 1054 1055 return !cmpxchg((const struct inet_diag_handler **)&inet_diag_table[type], 1056 NULL, h) ? 0 : -EEXIST; 1057 } 1058 EXPORT_SYMBOL_GPL(inet_diag_register); 1059 1060 void inet_diag_unregister(const struct inet_diag_handler *h) 1061 { 1062 const __u16 type = h->idiag_type; 1063 1064 if (type >= IPPROTO_MAX) 1065 return; 1066 1067 xchg((const struct inet_diag_handler **)&inet_diag_table[type], 1068 NULL); 1069 } 1070 EXPORT_SYMBOL_GPL(inet_diag_unregister); 1071 1072 static const struct sock_diag_inet_compat inet_diag_compat = { 1073 .owner = THIS_MODULE, 1074 .fn = inet_diag_rcv_msg_compat, 1075 }; 1076 1077 static int __init inet_diag_init(void) 1078 { 1079 const int inet_diag_table_size = (IPPROTO_MAX * 1080 sizeof(struct inet_diag_handler *)); 1081 int err = -ENOMEM; 1082 1083 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL); 1084 if (!inet_diag_table) 1085 goto out; 1086 1087 err = sock_diag_register(&inet_diag_handler); 1088 if (err) 1089 goto out_free_nl; 1090 1091 err = sock_diag_register(&inet6_diag_handler); 1092 if (err) 1093 goto out_free_inet; 1094 1095 sock_diag_register_inet_compat(&inet_diag_compat); 1096 out: 1097 return err; 1098 1099 out_free_inet: 1100 sock_diag_unregister(&inet_diag_handler); 1101 out_free_nl: 1102 kfree(inet_diag_table); 1103 goto out; 1104 } 1105 1106 static void __exit inet_diag_exit(void) 1107 { 1108 sock_diag_unregister(&inet6_diag_handler); 1109 sock_diag_unregister(&inet_diag_handler); 1110 sock_diag_unregister_inet_compat(&inet_diag_compat); 1111 kfree(inet_diag_table); 1112 } 1113 1114 module_init(inet_diag_init); 1115 module_exit(inet_diag_exit); 1116 MODULE_LICENSE("GPL"); 1117 MODULE_DESCRIPTION("INET/INET6: socket monitoring via SOCK_DIAG"); 1118 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */); 1119 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */); 1120