1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * This is a module which is used for logging packets to userspace via 4 * nfetlink. 5 * 6 * (C) 2005 by Harald Welte <laforge@netfilter.org> 7 * (C) 2006-2012 Patrick McHardy <kaber@trash.net> 8 * 9 * Based on the old ipv4-only ipt_ULOG.c: 10 * (C) 2000-2004 by Harald Welte <laforge@netfilter.org> 11 */ 12 13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 14 15 #include <linux/module.h> 16 #include <linux/skbuff.h> 17 #include <linux/if_arp.h> 18 #include <linux/init.h> 19 #include <linux/ip.h> 20 #include <linux/ipv6.h> 21 #include <linux/netdevice.h> 22 #include <linux/netfilter.h> 23 #include <linux/netfilter_bridge.h> 24 #include <net/netlink.h> 25 #include <linux/netfilter/nfnetlink.h> 26 #include <linux/netfilter/nfnetlink_log.h> 27 #include <linux/netfilter/nf_conntrack_common.h> 28 #include <linux/spinlock.h> 29 #include <linux/sysctl.h> 30 #include <linux/proc_fs.h> 31 #include <linux/security.h> 32 #include <linux/list.h> 33 #include <linux/slab.h> 34 #include <net/sock.h> 35 #include <net/netfilter/nf_log.h> 36 #include <net/netns/generic.h> 37 38 #include <linux/atomic.h> 39 #include <linux/refcount.h> 40 41 42 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 43 #include "../bridge/br_private.h" 44 #endif 45 46 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 47 #include <net/netfilter/nf_conntrack.h> 48 #endif 49 50 #define NFULNL_COPY_DISABLED 0xff 51 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE 52 #define NFULNL_TIMEOUT_DEFAULT 100 /* every second */ 53 #define NFULNL_QTHRESH_DEFAULT 100 /* 100 packets */ 54 /* max packet size is limited by 16-bit struct nfattr nfa_len field */ 55 #define NFULNL_COPY_RANGE_MAX (0xFFFF - NLA_HDRLEN) 56 57 #define PRINTR(x, args...) do { if (net_ratelimit()) \ 58 printk(x, ## args); } while (0); 59 60 struct nfulnl_instance { 61 struct hlist_node hlist; /* global list of instances */ 62 spinlock_t lock; 63 refcount_t use; /* use count */ 64 65 unsigned int qlen; /* number of nlmsgs in skb */ 66 struct sk_buff *skb; /* pre-allocatd skb */ 67 struct timer_list timer; 68 struct net *net; 69 netns_tracker ns_tracker; 70 struct user_namespace *peer_user_ns; /* User namespace of the peer process */ 71 u32 peer_portid; /* PORTID of the peer process */ 72 73 /* configurable parameters */ 74 unsigned int flushtimeout; /* timeout until queue flush */ 75 unsigned int nlbufsiz; /* netlink buffer allocation size */ 76 unsigned int qthreshold; /* threshold of the queue */ 77 u_int32_t copy_range; 78 u_int32_t seq; /* instance-local sequential counter */ 79 u_int16_t group_num; /* number of this queue */ 80 u_int16_t flags; 81 u_int8_t copy_mode; 82 struct rcu_head rcu; 83 }; 84 85 #define INSTANCE_BUCKETS 16 86 87 static unsigned int nfnl_log_net_id __read_mostly; 88 89 struct nfnl_log_net { 90 spinlock_t instances_lock; 91 struct hlist_head instance_table[INSTANCE_BUCKETS]; 92 atomic_t global_seq; 93 }; 94 95 static struct nfnl_log_net *nfnl_log_pernet(struct net *net) 96 { 97 return net_generic(net, nfnl_log_net_id); 98 } 99 100 static inline u_int8_t instance_hashfn(u_int16_t group_num) 101 { 102 return ((group_num & 0xff) % INSTANCE_BUCKETS); 103 } 104 105 static struct nfulnl_instance * 106 __instance_lookup(const struct nfnl_log_net *log, u16 group_num) 107 { 108 const struct hlist_head *head; 109 struct nfulnl_instance *inst; 110 111 head = &log->instance_table[instance_hashfn(group_num)]; 112 hlist_for_each_entry_rcu(inst, head, hlist) { 113 if (inst->group_num == group_num) 114 return inst; 115 } 116 return NULL; 117 } 118 119 static inline void 120 instance_get(struct nfulnl_instance *inst) 121 { 122 refcount_inc(&inst->use); 123 } 124 125 static struct nfulnl_instance * 126 instance_lookup_get_rcu(const struct nfnl_log_net *log, u16 group_num) 127 { 128 struct nfulnl_instance *inst; 129 130 inst = __instance_lookup(log, group_num); 131 if (inst && !refcount_inc_not_zero(&inst->use)) 132 inst = NULL; 133 134 return inst; 135 } 136 137 static struct nfulnl_instance * 138 instance_lookup_get(const struct nfnl_log_net *log, u16 group_num) 139 { 140 struct nfulnl_instance *inst; 141 142 rcu_read_lock(); 143 inst = instance_lookup_get_rcu(log, group_num); 144 rcu_read_unlock(); 145 146 return inst; 147 } 148 149 static void nfulnl_instance_free_rcu(struct rcu_head *head) 150 { 151 struct nfulnl_instance *inst = 152 container_of(head, struct nfulnl_instance, rcu); 153 154 put_net_track(inst->net, &inst->ns_tracker); 155 kfree(inst); 156 module_put(THIS_MODULE); 157 } 158 159 static void 160 instance_put(struct nfulnl_instance *inst) 161 { 162 if (inst && refcount_dec_and_test(&inst->use)) 163 call_rcu(&inst->rcu, nfulnl_instance_free_rcu); 164 } 165 166 static void nfulnl_timer(struct timer_list *t); 167 168 static struct nfulnl_instance * 169 instance_create(struct net *net, u_int16_t group_num, 170 u32 portid, struct user_namespace *user_ns) 171 { 172 struct nfulnl_instance *inst; 173 struct nfnl_log_net *log = nfnl_log_pernet(net); 174 int err; 175 176 spin_lock_bh(&log->instances_lock); 177 if (__instance_lookup(log, group_num)) { 178 err = -EEXIST; 179 goto out_unlock; 180 } 181 182 inst = kzalloc_obj(*inst, GFP_ATOMIC); 183 if (!inst) { 184 err = -ENOMEM; 185 goto out_unlock; 186 } 187 188 if (!try_module_get(THIS_MODULE)) { 189 kfree(inst); 190 err = -EAGAIN; 191 goto out_unlock; 192 } 193 194 INIT_HLIST_NODE(&inst->hlist); 195 spin_lock_init(&inst->lock); 196 /* needs to be two, since we _put() after creation */ 197 refcount_set(&inst->use, 2); 198 199 timer_setup(&inst->timer, nfulnl_timer, 0); 200 201 inst->net = get_net_track(net, &inst->ns_tracker, GFP_ATOMIC); 202 inst->peer_user_ns = user_ns; 203 inst->peer_portid = portid; 204 inst->group_num = group_num; 205 206 inst->qthreshold = NFULNL_QTHRESH_DEFAULT; 207 inst->flushtimeout = NFULNL_TIMEOUT_DEFAULT; 208 inst->nlbufsiz = NFULNL_NLBUFSIZ_DEFAULT; 209 inst->copy_mode = NFULNL_COPY_PACKET; 210 inst->copy_range = NFULNL_COPY_RANGE_MAX; 211 212 hlist_add_head_rcu(&inst->hlist, 213 &log->instance_table[instance_hashfn(group_num)]); 214 215 216 spin_unlock_bh(&log->instances_lock); 217 218 return inst; 219 220 out_unlock: 221 spin_unlock_bh(&log->instances_lock); 222 return ERR_PTR(err); 223 } 224 225 static void __nfulnl_flush(struct nfulnl_instance *inst); 226 227 /* called with BH disabled */ 228 static void 229 __instance_destroy(struct nfulnl_instance *inst) 230 { 231 /* first pull it out of the global list */ 232 hlist_del_rcu(&inst->hlist); 233 234 /* then flush all pending packets from skb */ 235 236 spin_lock(&inst->lock); 237 238 /* lockless readers wont be able to use us */ 239 inst->copy_mode = NFULNL_COPY_DISABLED; 240 241 if (inst->skb) 242 __nfulnl_flush(inst); 243 spin_unlock(&inst->lock); 244 245 /* and finally put the refcount */ 246 instance_put(inst); 247 } 248 249 static inline void 250 instance_destroy(struct nfnl_log_net *log, 251 struct nfulnl_instance *inst) 252 { 253 spin_lock_bh(&log->instances_lock); 254 __instance_destroy(inst); 255 spin_unlock_bh(&log->instances_lock); 256 } 257 258 static int 259 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode, 260 unsigned int range) 261 { 262 int status = 0; 263 264 spin_lock_bh(&inst->lock); 265 266 switch (mode) { 267 case NFULNL_COPY_NONE: 268 case NFULNL_COPY_META: 269 inst->copy_mode = mode; 270 inst->copy_range = 0; 271 break; 272 273 case NFULNL_COPY_PACKET: 274 inst->copy_mode = mode; 275 if (range == 0) 276 range = NFULNL_COPY_RANGE_MAX; 277 inst->copy_range = min_t(unsigned int, 278 range, NFULNL_COPY_RANGE_MAX); 279 break; 280 281 default: 282 status = -EINVAL; 283 break; 284 } 285 286 spin_unlock_bh(&inst->lock); 287 288 return status; 289 } 290 291 static int 292 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz) 293 { 294 int status; 295 296 spin_lock_bh(&inst->lock); 297 if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT) 298 status = -ERANGE; 299 else if (nlbufsiz > 131072) 300 status = -ERANGE; 301 else { 302 inst->nlbufsiz = nlbufsiz; 303 status = 0; 304 } 305 spin_unlock_bh(&inst->lock); 306 307 return status; 308 } 309 310 static void 311 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout) 312 { 313 spin_lock_bh(&inst->lock); 314 inst->flushtimeout = timeout; 315 spin_unlock_bh(&inst->lock); 316 } 317 318 static void 319 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh) 320 { 321 spin_lock_bh(&inst->lock); 322 inst->qthreshold = qthresh; 323 spin_unlock_bh(&inst->lock); 324 } 325 326 static int 327 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags) 328 { 329 spin_lock_bh(&inst->lock); 330 inst->flags = flags; 331 spin_unlock_bh(&inst->lock); 332 333 return 0; 334 } 335 336 static struct sk_buff * 337 nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size, 338 unsigned int pkt_size) 339 { 340 struct sk_buff *skb; 341 unsigned int n; 342 343 /* alloc skb which should be big enough for a whole multipart 344 * message. WARNING: has to be <= 128k due to slab restrictions */ 345 346 n = max(inst_size, pkt_size); 347 skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN); 348 if (!skb) { 349 if (n > pkt_size) { 350 /* try to allocate only as much as we need for current 351 * packet */ 352 353 skb = alloc_skb(pkt_size, GFP_ATOMIC); 354 } 355 } 356 357 return skb; 358 } 359 360 static void 361 __nfulnl_send(struct nfulnl_instance *inst) 362 { 363 if (inst->qlen > 1) { 364 struct nlmsghdr *nlh = nfnl_msg_put(inst->skb, 0, 0, 365 NLMSG_DONE, 0, 366 AF_UNSPEC, NFNETLINK_V0, 367 htons(inst->group_num)); 368 if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n", 369 inst->skb->len, skb_tailroom(inst->skb))) { 370 kfree_skb(inst->skb); 371 goto out; 372 } 373 } 374 nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid); 375 out: 376 inst->qlen = 0; 377 inst->skb = NULL; 378 } 379 380 static void 381 __nfulnl_flush(struct nfulnl_instance *inst) 382 { 383 /* timer holds a reference */ 384 if (timer_delete(&inst->timer)) 385 instance_put(inst); 386 if (inst->skb) 387 __nfulnl_send(inst); 388 } 389 390 static void 391 nfulnl_timer(struct timer_list *t) 392 { 393 struct nfulnl_instance *inst = timer_container_of(inst, t, timer); 394 395 spin_lock_bh(&inst->lock); 396 if (inst->skb) 397 __nfulnl_send(inst); 398 spin_unlock_bh(&inst->lock); 399 instance_put(inst); 400 } 401 402 static u32 nfulnl_get_bridge_size(const struct sk_buff *skb) 403 { 404 u32 mac_len, size = 0; 405 406 if (!skb_mac_header_was_set(skb)) 407 return 0; 408 409 if (skb_vlan_tag_present(skb)) { 410 size += nla_total_size(0); /* nested */ 411 size += nla_total_size(sizeof(u16)); /* id */ 412 size += nla_total_size(sizeof(u16)); /* tag */ 413 } 414 415 mac_len = skb_mac_header_len(skb); 416 if (mac_len > 0) 417 size += nla_total_size(mac_len); 418 419 return size; 420 } 421 422 static int nfulnl_put_bridge(struct nfulnl_instance *inst, const struct sk_buff *skb) 423 { 424 u32 mac_len; 425 426 if (!skb_mac_header_was_set(skb)) 427 return 0; 428 429 if (skb_vlan_tag_present(skb)) { 430 struct nlattr *nest; 431 432 nest = nla_nest_start(inst->skb, NFULA_VLAN); 433 if (!nest) 434 goto nla_put_failure; 435 436 if (nla_put_be16(inst->skb, NFULA_VLAN_TCI, htons(skb->vlan_tci)) || 437 nla_put_be16(inst->skb, NFULA_VLAN_PROTO, skb->vlan_proto)) 438 goto nla_put_failure; 439 440 nla_nest_end(inst->skb, nest); 441 } 442 443 mac_len = skb_mac_header_len(skb); 444 if (mac_len > 0 && 445 nla_put(inst->skb, NFULA_L2HDR, mac_len, skb_mac_header(skb))) 446 goto nla_put_failure; 447 448 return 0; 449 450 nla_put_failure: 451 return -1; 452 } 453 454 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 455 static int nflog_put_master_ifindex(struct sk_buff *nlskb, int attr, 456 const struct net_device *dev) 457 { 458 const struct net_device *upper; 459 460 if (dev && !netif_is_bridge_port(dev)) 461 return 0; 462 463 upper = netdev_master_upper_dev_get_rcu((struct net_device *)dev); 464 if (upper && nla_put_be32(nlskb, attr, htonl(upper->ifindex))) 465 return -EMSGSIZE; 466 467 return 0; 468 } 469 #endif 470 471 /* This is an inline function, we don't really care about a long 472 * list of arguments */ 473 static inline int 474 __build_packet_message(struct nfnl_log_net *log, 475 struct nfulnl_instance *inst, 476 const struct sk_buff *skb, 477 unsigned int data_len, 478 u_int8_t pf, 479 unsigned int hooknum, 480 const struct net_device *indev, 481 const struct net_device *outdev, 482 const char *prefix, unsigned int plen, 483 const struct nfnl_ct_hook *nfnl_ct, 484 struct nf_conn *ct, enum ip_conntrack_info ctinfo) 485 { 486 struct nfulnl_msg_packet_hdr pmsg; 487 struct nlmsghdr *nlh; 488 sk_buff_data_t old_tail = inst->skb->tail; 489 struct sock *sk; 490 const unsigned char *hwhdrp; 491 492 nlh = nfnl_msg_put(inst->skb, 0, 0, 493 nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET), 494 0, pf, NFNETLINK_V0, htons(inst->group_num)); 495 if (!nlh) 496 return -1; 497 498 memset(&pmsg, 0, sizeof(pmsg)); 499 pmsg.hw_protocol = skb->protocol; 500 pmsg.hook = hooknum; 501 502 if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg)) 503 goto nla_put_failure; 504 505 if (prefix && 506 nla_put(inst->skb, NFULA_PREFIX, plen, prefix)) 507 goto nla_put_failure; 508 509 if (indev) { 510 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 511 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV, 512 htonl(indev->ifindex))) 513 goto nla_put_failure; 514 #else 515 if (pf == PF_BRIDGE) { 516 /* Case 1: outdev is physical input device, we need to 517 * look for bridge group (when called from 518 * netfilter_bridge) */ 519 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV, 520 htonl(indev->ifindex)) || 521 /* this is the bridge group "brX" */ 522 /* rcu_read_lock()ed by nf_hook_thresh or 523 * nf_log_packet. 524 */ 525 nflog_put_master_ifindex(inst->skb, NFULA_IFINDEX_INDEV, indev)) 526 goto nla_put_failure; 527 } else { 528 int physinif; 529 530 /* Case 2: indev is bridge group, we need to look for 531 * physical device (when called from ipv4) */ 532 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV, 533 htonl(indev->ifindex))) 534 goto nla_put_failure; 535 536 physinif = nf_bridge_get_physinif(skb); 537 if (physinif && 538 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV, 539 htonl(physinif))) 540 goto nla_put_failure; 541 } 542 #endif 543 } 544 545 if (outdev) { 546 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 547 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV, 548 htonl(outdev->ifindex))) 549 goto nla_put_failure; 550 #else 551 if (pf == PF_BRIDGE) { 552 /* Case 1: outdev is physical output device, we need to 553 * look for bridge group (when called from 554 * netfilter_bridge) */ 555 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV, 556 htonl(outdev->ifindex)) || 557 /* this is the bridge group "brX" */ 558 /* rcu_read_lock()ed by nf_hook_thresh or 559 * nf_log_packet. 560 */ 561 nflog_put_master_ifindex(inst->skb, NFULA_IFINDEX_OUTDEV, outdev)) 562 goto nla_put_failure; 563 } else { 564 struct net_device *physoutdev; 565 566 /* Case 2: indev is a bridge group, we need to look 567 * for physical device (when called from ipv4) */ 568 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV, 569 htonl(outdev->ifindex))) 570 goto nla_put_failure; 571 572 physoutdev = nf_bridge_get_physoutdev(skb); 573 if (physoutdev && 574 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV, 575 htonl(physoutdev->ifindex))) 576 goto nla_put_failure; 577 } 578 #endif 579 } 580 581 if (skb->mark && 582 nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark))) 583 goto nla_put_failure; 584 585 if (indev && skb->dev && 586 skb_mac_header_was_set(skb) && 587 skb_mac_header_len(skb) != 0) { 588 struct nfulnl_msg_packet_hw phw; 589 int len; 590 591 memset(&phw, 0, sizeof(phw)); 592 len = dev_parse_header(skb, phw.hw_addr); 593 if (len > 0) { 594 phw.hw_addrlen = htons(len); 595 if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw)) 596 goto nla_put_failure; 597 } 598 } 599 600 if (indev && skb_mac_header_was_set(skb)) { 601 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) || 602 nla_put_be16(inst->skb, NFULA_HWLEN, 603 htons(skb->dev->hard_header_len))) 604 goto nla_put_failure; 605 606 hwhdrp = skb_mac_header(skb); 607 608 if (skb->dev->type == ARPHRD_SIT) 609 hwhdrp -= ETH_HLEN; 610 611 if (hwhdrp >= skb->head && 612 nla_put(inst->skb, NFULA_HWHEADER, 613 skb->dev->hard_header_len, hwhdrp)) 614 goto nla_put_failure; 615 } 616 617 if (hooknum <= NF_INET_FORWARD) { 618 struct timespec64 kts = ktime_to_timespec64(skb_tstamp_cond(skb, true)); 619 struct nfulnl_msg_packet_timestamp ts; 620 ts.sec = cpu_to_be64(kts.tv_sec); 621 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC); 622 623 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts)) 624 goto nla_put_failure; 625 } 626 627 /* UID */ 628 sk = skb->sk; 629 if (sk && sk_fullsock(sk)) { 630 const struct socket *sock; 631 const struct file *file; 632 633 /* The sk pointer remains valid as long as the skb is. 634 * The sk_socket and file pointer may become NULL 635 * if the socket is closed. 636 * Both structures (including file->cred) are RCU freed 637 * which means they can be accessed within a RCU read section. 638 */ 639 sock = READ_ONCE(sk->sk_socket); 640 file = sock ? READ_ONCE(sock->file) : NULL; 641 if (file) { 642 const struct cred *cred = file->f_cred; 643 struct user_namespace *user_ns = inst->peer_user_ns; 644 __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid)); 645 __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid)); 646 if (nla_put_be32(inst->skb, NFULA_UID, uid) || 647 nla_put_be32(inst->skb, NFULA_GID, gid)) 648 goto nla_put_failure; 649 } 650 } 651 652 /* local sequence number */ 653 if ((inst->flags & NFULNL_CFG_F_SEQ) && 654 nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++))) 655 goto nla_put_failure; 656 657 /* global sequence number */ 658 if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) && 659 nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL, 660 htonl(atomic_inc_return(&log->global_seq)))) 661 goto nla_put_failure; 662 663 if (ct && nfnl_ct->build(inst->skb, ct, ctinfo, 664 NFULA_CT, NFULA_CT_INFO) < 0) 665 goto nla_put_failure; 666 667 if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) && 668 nfulnl_put_bridge(inst, skb) < 0) 669 goto nla_put_failure; 670 671 if (data_len) { 672 struct nlattr *nla; 673 674 nla = nla_reserve(inst->skb, NFULA_PAYLOAD, data_len); 675 if (!nla) 676 goto nla_put_failure; 677 678 if (skb_copy_bits(skb, 0, nla_data(nla), data_len)) 679 BUG(); 680 } 681 682 nlh->nlmsg_len = inst->skb->tail - old_tail; 683 return 0; 684 685 nla_put_failure: 686 PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n"); 687 return -1; 688 } 689 690 static const struct nf_loginfo default_loginfo = { 691 .type = NF_LOG_TYPE_ULOG, 692 .u = { 693 .ulog = { 694 .copy_len = 0xffff, 695 .group = 0, 696 .qthreshold = 1, 697 }, 698 }, 699 }; 700 701 /* log handler for internal netfilter logging api */ 702 static void 703 nfulnl_log_packet(struct net *net, 704 u_int8_t pf, 705 unsigned int hooknum, 706 const struct sk_buff *skb, 707 const struct net_device *in, 708 const struct net_device *out, 709 const struct nf_loginfo *li_user, 710 const char *prefix) 711 { 712 size_t size; 713 unsigned int data_len; 714 struct nfulnl_instance *inst; 715 const struct nf_loginfo *li; 716 unsigned int qthreshold; 717 unsigned int plen = 0; 718 struct nfnl_log_net *log = nfnl_log_pernet(net); 719 const struct nfnl_ct_hook *nfnl_ct = NULL; 720 enum ip_conntrack_info ctinfo = 0; 721 struct nf_conn *ct = NULL; 722 723 if (li_user && li_user->type == NF_LOG_TYPE_ULOG) 724 li = li_user; 725 else 726 li = &default_loginfo; 727 728 inst = instance_lookup_get_rcu(log, li->u.ulog.group); 729 if (!inst) 730 return; 731 732 if (prefix) 733 plen = strlen(prefix) + 1; 734 735 /* FIXME: do we want to make the size calculation conditional based on 736 * what is actually present? way more branches and checks, but more 737 * memory efficient... */ 738 size = nlmsg_total_size(sizeof(struct nfgenmsg)) 739 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr)) 740 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 741 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 742 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 743 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 744 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 745 #endif 746 + nla_total_size(sizeof(u_int32_t)) /* mark */ 747 + nla_total_size(sizeof(u_int32_t)) /* uid */ 748 + nla_total_size(sizeof(u_int32_t)) /* gid */ 749 + nla_total_size(plen) /* prefix */ 750 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw)) 751 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp)) 752 + nlmsg_total_size(sizeof(struct nfgenmsg)); /* NLMSG_DONE */ 753 754 if (in && skb_mac_header_was_set(skb)) { 755 size += nla_total_size(skb->dev->hard_header_len) 756 + nla_total_size(sizeof(u_int16_t)) /* hwtype */ 757 + nla_total_size(sizeof(u_int16_t)); /* hwlen */ 758 } 759 760 spin_lock_bh(&inst->lock); 761 762 if (inst->flags & NFULNL_CFG_F_SEQ) 763 size += nla_total_size(sizeof(u_int32_t)); 764 if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) 765 size += nla_total_size(sizeof(u_int32_t)); 766 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 767 if (inst->flags & NFULNL_CFG_F_CONNTRACK) { 768 nfnl_ct = rcu_dereference(nfnl_ct_hook); 769 if (nfnl_ct != NULL) { 770 ct = nf_ct_get(skb, &ctinfo); 771 if (ct != NULL) 772 size += nfnl_ct->build_size(ct); 773 } 774 } 775 #endif 776 if (pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) 777 size += nfulnl_get_bridge_size(skb); 778 779 qthreshold = inst->qthreshold; 780 /* per-rule qthreshold overrides per-instance */ 781 if (li->u.ulog.qthreshold) 782 if (qthreshold > li->u.ulog.qthreshold) 783 qthreshold = li->u.ulog.qthreshold; 784 785 786 switch (inst->copy_mode) { 787 case NFULNL_COPY_META: 788 case NFULNL_COPY_NONE: 789 data_len = 0; 790 break; 791 792 case NFULNL_COPY_PACKET: 793 data_len = inst->copy_range; 794 if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) && 795 (li->u.ulog.copy_len < data_len)) 796 data_len = li->u.ulog.copy_len; 797 798 if (data_len > skb->len) 799 data_len = skb->len; 800 801 size += nla_total_size(data_len); 802 break; 803 804 case NFULNL_COPY_DISABLED: 805 default: 806 goto unlock_and_release; 807 } 808 809 if (inst->skb && size > skb_tailroom(inst->skb)) { 810 /* either the queue len is too high or we don't have 811 * enough room in the skb left. flush to userspace. */ 812 __nfulnl_flush(inst); 813 } 814 815 if (!inst->skb) { 816 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid, 817 inst->nlbufsiz, size); 818 if (!inst->skb) 819 goto alloc_failure; 820 } 821 822 inst->qlen++; 823 824 __build_packet_message(log, inst, skb, data_len, pf, 825 hooknum, in, out, prefix, plen, 826 nfnl_ct, ct, ctinfo); 827 828 if (inst->qlen >= qthreshold) 829 __nfulnl_flush(inst); 830 /* timer_pending always called within inst->lock, so there 831 * is no chance of a race here */ 832 else if (!timer_pending(&inst->timer)) { 833 instance_get(inst); 834 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100); 835 add_timer(&inst->timer); 836 } 837 838 unlock_and_release: 839 spin_unlock_bh(&inst->lock); 840 instance_put(inst); 841 return; 842 843 alloc_failure: 844 /* FIXME: statistics */ 845 goto unlock_and_release; 846 } 847 848 static int 849 nfulnl_rcv_nl_event(struct notifier_block *this, 850 unsigned long event, void *ptr) 851 { 852 struct netlink_notify *n = ptr; 853 struct nfnl_log_net *log = nfnl_log_pernet(n->net); 854 855 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) { 856 int i; 857 858 /* destroy all instances for this portid */ 859 spin_lock_bh(&log->instances_lock); 860 for (i = 0; i < INSTANCE_BUCKETS; i++) { 861 struct hlist_node *t2; 862 struct nfulnl_instance *inst; 863 struct hlist_head *head = &log->instance_table[i]; 864 865 hlist_for_each_entry_safe(inst, t2, head, hlist) { 866 if (n->portid == inst->peer_portid) 867 __instance_destroy(inst); 868 } 869 } 870 spin_unlock_bh(&log->instances_lock); 871 } 872 return NOTIFY_DONE; 873 } 874 875 static struct notifier_block nfulnl_rtnl_notifier = { 876 .notifier_call = nfulnl_rcv_nl_event, 877 }; 878 879 static int nfulnl_recv_unsupp(struct sk_buff *skb, const struct nfnl_info *info, 880 const struct nlattr * const nfula[]) 881 { 882 return -ENOTSUPP; 883 } 884 885 static struct nf_logger nfulnl_logger __read_mostly = { 886 .name = "nfnetlink_log", 887 .type = NF_LOG_TYPE_ULOG, 888 .logfn = nfulnl_log_packet, 889 .me = THIS_MODULE, 890 }; 891 892 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = { 893 [NFULA_CFG_CMD] = { .len = sizeof(struct nfulnl_msg_config_cmd) }, 894 [NFULA_CFG_MODE] = { .len = sizeof(struct nfulnl_msg_config_mode) }, 895 [NFULA_CFG_TIMEOUT] = { .type = NLA_U32 }, 896 [NFULA_CFG_QTHRESH] = { .type = NLA_U32 }, 897 [NFULA_CFG_NLBUFSIZ] = { .type = NLA_U32 }, 898 [NFULA_CFG_FLAGS] = NLA_POLICY_MASK(NLA_BE16, NFULNL_CFG_F_SEQ | 899 NFULNL_CFG_F_SEQ_GLOBAL | 900 NFULNL_CFG_F_CONNTRACK), 901 }; 902 903 static int nfulnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info, 904 const struct nlattr * const nfula[]) 905 { 906 struct nfnl_log_net *log = nfnl_log_pernet(info->net); 907 u_int16_t group_num = ntohs(info->nfmsg->res_id); 908 struct nfulnl_msg_config_cmd *cmd = NULL; 909 struct nfulnl_instance *inst; 910 u16 flags = 0; 911 int ret = 0; 912 913 if (nfula[NFULA_CFG_CMD]) { 914 u_int8_t pf = info->nfmsg->nfgen_family; 915 cmd = nla_data(nfula[NFULA_CFG_CMD]); 916 917 /* Commands without queue context */ 918 switch (cmd->command) { 919 case NFULNL_CFG_CMD_PF_BIND: 920 return nf_log_bind_pf(info->net, pf, &nfulnl_logger); 921 case NFULNL_CFG_CMD_PF_UNBIND: 922 nf_log_unbind_pf(info->net, pf); 923 return 0; 924 } 925 } 926 927 inst = instance_lookup_get(log, group_num); 928 if (inst && inst->peer_portid != NETLINK_CB(skb).portid) { 929 ret = -EPERM; 930 goto out_put; 931 } 932 933 /* Check if we support these flags in first place, dependencies should 934 * be there too not to break atomicity. 935 */ 936 if (nfula[NFULA_CFG_FLAGS]) { 937 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS])); 938 939 if ((flags & NFULNL_CFG_F_CONNTRACK) && 940 !rcu_access_pointer(nfnl_ct_hook)) { 941 #ifdef CONFIG_MODULES 942 nfnl_unlock(NFNL_SUBSYS_ULOG); 943 request_module("ip_conntrack_netlink"); 944 nfnl_lock(NFNL_SUBSYS_ULOG); 945 if (rcu_access_pointer(nfnl_ct_hook)) { 946 ret = -EAGAIN; 947 goto out_put; 948 } 949 #endif 950 ret = -EOPNOTSUPP; 951 goto out_put; 952 } 953 } 954 955 if (cmd != NULL) { 956 switch (cmd->command) { 957 case NFULNL_CFG_CMD_BIND: 958 if (inst) { 959 ret = -EBUSY; 960 goto out_put; 961 } 962 963 inst = instance_create(info->net, group_num, 964 NETLINK_CB(skb).portid, 965 sk_user_ns(NETLINK_CB(skb).sk)); 966 if (IS_ERR(inst)) { 967 ret = PTR_ERR(inst); 968 goto out; 969 } 970 break; 971 case NFULNL_CFG_CMD_UNBIND: 972 if (!inst) { 973 ret = -ENODEV; 974 goto out; 975 } 976 977 instance_destroy(log, inst); 978 goto out_put; 979 default: 980 ret = -ENOTSUPP; 981 goto out_put; 982 } 983 } else if (!inst) { 984 ret = -ENODEV; 985 goto out; 986 } 987 988 if (nfula[NFULA_CFG_MODE]) { 989 struct nfulnl_msg_config_mode *params = 990 nla_data(nfula[NFULA_CFG_MODE]); 991 992 nfulnl_set_mode(inst, params->copy_mode, 993 ntohl(params->copy_range)); 994 } 995 996 if (nfula[NFULA_CFG_TIMEOUT]) { 997 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]); 998 999 nfulnl_set_timeout(inst, ntohl(timeout)); 1000 } 1001 1002 if (nfula[NFULA_CFG_NLBUFSIZ]) { 1003 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]); 1004 1005 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz)); 1006 } 1007 1008 if (nfula[NFULA_CFG_QTHRESH]) { 1009 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]); 1010 1011 nfulnl_set_qthresh(inst, ntohl(qthresh)); 1012 } 1013 1014 if (nfula[NFULA_CFG_FLAGS]) 1015 nfulnl_set_flags(inst, flags); 1016 1017 out_put: 1018 instance_put(inst); 1019 out: 1020 return ret; 1021 } 1022 1023 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = { 1024 [NFULNL_MSG_PACKET] = { 1025 .call = nfulnl_recv_unsupp, 1026 .type = NFNL_CB_MUTEX, 1027 .attr_count = NFULA_MAX, 1028 }, 1029 [NFULNL_MSG_CONFIG] = { 1030 .call = nfulnl_recv_config, 1031 .type = NFNL_CB_MUTEX, 1032 .attr_count = NFULA_CFG_MAX, 1033 .policy = nfula_cfg_policy 1034 }, 1035 }; 1036 1037 static const struct nfnetlink_subsystem nfulnl_subsys = { 1038 .name = "log", 1039 .subsys_id = NFNL_SUBSYS_ULOG, 1040 .cb_count = NFULNL_MSG_MAX, 1041 .cb = nfulnl_cb, 1042 }; 1043 1044 #ifdef CONFIG_PROC_FS 1045 struct iter_state { 1046 struct seq_net_private p; 1047 unsigned int bucket; 1048 }; 1049 1050 static struct hlist_node *get_first(struct net *net, struct iter_state *st) 1051 { 1052 struct nfnl_log_net *log; 1053 if (!st) 1054 return NULL; 1055 1056 log = nfnl_log_pernet(net); 1057 1058 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) { 1059 struct hlist_head *head = &log->instance_table[st->bucket]; 1060 1061 if (!hlist_empty(head)) 1062 return rcu_dereference(hlist_first_rcu(head)); 1063 } 1064 return NULL; 1065 } 1066 1067 static struct hlist_node *get_next(struct net *net, struct iter_state *st, 1068 struct hlist_node *h) 1069 { 1070 h = rcu_dereference(hlist_next_rcu(h)); 1071 while (!h) { 1072 struct nfnl_log_net *log; 1073 struct hlist_head *head; 1074 1075 if (++st->bucket >= INSTANCE_BUCKETS) 1076 return NULL; 1077 1078 log = nfnl_log_pernet(net); 1079 head = &log->instance_table[st->bucket]; 1080 h = rcu_dereference(hlist_first_rcu(head)); 1081 } 1082 return h; 1083 } 1084 1085 static struct hlist_node *get_idx(struct net *net, struct iter_state *st, 1086 loff_t pos) 1087 { 1088 struct hlist_node *head; 1089 head = get_first(net, st); 1090 1091 if (head) 1092 while (pos && (head = get_next(net, st, head))) 1093 pos--; 1094 return pos ? NULL : head; 1095 } 1096 1097 static void *seq_start(struct seq_file *s, loff_t *pos) 1098 __acquires(rcu) 1099 { 1100 rcu_read_lock(); 1101 return get_idx(seq_file_net(s), s->private, *pos); 1102 } 1103 1104 static void *seq_next(struct seq_file *s, void *v, loff_t *pos) 1105 { 1106 (*pos)++; 1107 return get_next(seq_file_net(s), s->private, v); 1108 } 1109 1110 static void seq_stop(struct seq_file *s, void *v) 1111 __releases(rcu) 1112 { 1113 rcu_read_unlock(); 1114 } 1115 1116 static int seq_show(struct seq_file *s, void *v) 1117 { 1118 const struct nfulnl_instance *inst = v; 1119 1120 seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n", 1121 inst->group_num, 1122 inst->peer_portid, inst->qlen, 1123 inst->copy_mode, inst->copy_range, 1124 inst->flushtimeout, refcount_read(&inst->use)); 1125 1126 return 0; 1127 } 1128 1129 static const struct seq_operations nful_seq_ops = { 1130 .start = seq_start, 1131 .next = seq_next, 1132 .stop = seq_stop, 1133 .show = seq_show, 1134 }; 1135 #endif /* PROC_FS */ 1136 1137 static int __net_init nfnl_log_net_init(struct net *net) 1138 { 1139 unsigned int i; 1140 struct nfnl_log_net *log = nfnl_log_pernet(net); 1141 #ifdef CONFIG_PROC_FS 1142 struct proc_dir_entry *proc; 1143 kuid_t root_uid; 1144 kgid_t root_gid; 1145 #endif 1146 1147 for (i = 0; i < INSTANCE_BUCKETS; i++) 1148 INIT_HLIST_HEAD(&log->instance_table[i]); 1149 spin_lock_init(&log->instances_lock); 1150 1151 #ifdef CONFIG_PROC_FS 1152 proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter, 1153 &nful_seq_ops, sizeof(struct iter_state)); 1154 if (!proc) 1155 return -ENOMEM; 1156 1157 root_uid = make_kuid(net->user_ns, 0); 1158 root_gid = make_kgid(net->user_ns, 0); 1159 if (uid_valid(root_uid) && gid_valid(root_gid)) 1160 proc_set_user(proc, root_uid, root_gid); 1161 #endif 1162 return 0; 1163 } 1164 1165 static void __net_exit nfnl_log_net_exit(struct net *net) 1166 { 1167 struct nfnl_log_net *log = nfnl_log_pernet(net); 1168 unsigned int i; 1169 1170 #ifdef CONFIG_PROC_FS 1171 remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter); 1172 #endif 1173 nf_log_unset(net, &nfulnl_logger); 1174 for (i = 0; i < INSTANCE_BUCKETS; i++) 1175 WARN_ON_ONCE(!hlist_empty(&log->instance_table[i])); 1176 } 1177 1178 static struct pernet_operations nfnl_log_net_ops = { 1179 .init = nfnl_log_net_init, 1180 .exit = nfnl_log_net_exit, 1181 .id = &nfnl_log_net_id, 1182 .size = sizeof(struct nfnl_log_net), 1183 }; 1184 1185 static int __init nfnetlink_log_init(void) 1186 { 1187 int status; 1188 1189 status = register_pernet_subsys(&nfnl_log_net_ops); 1190 if (status < 0) { 1191 pr_err("failed to register pernet ops\n"); 1192 goto out; 1193 } 1194 1195 netlink_register_notifier(&nfulnl_rtnl_notifier); 1196 status = nfnetlink_subsys_register(&nfulnl_subsys); 1197 if (status < 0) { 1198 pr_err("failed to create netlink socket\n"); 1199 goto cleanup_netlink_notifier; 1200 } 1201 1202 status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger); 1203 if (status < 0) { 1204 pr_err("failed to register logger\n"); 1205 goto cleanup_subsys; 1206 } 1207 1208 return status; 1209 1210 cleanup_subsys: 1211 nfnetlink_subsys_unregister(&nfulnl_subsys); 1212 cleanup_netlink_notifier: 1213 netlink_unregister_notifier(&nfulnl_rtnl_notifier); 1214 unregister_pernet_subsys(&nfnl_log_net_ops); 1215 out: 1216 return status; 1217 } 1218 1219 static void __exit nfnetlink_log_fini(void) 1220 { 1221 nfnetlink_subsys_unregister(&nfulnl_subsys); 1222 netlink_unregister_notifier(&nfulnl_rtnl_notifier); 1223 unregister_pernet_subsys(&nfnl_log_net_ops); 1224 nf_log_unregister(&nfulnl_logger); 1225 } 1226 1227 MODULE_DESCRIPTION("netfilter userspace logging"); 1228 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); 1229 MODULE_LICENSE("GPL"); 1230 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG); 1231 MODULE_ALIAS_NF_LOGGER(AF_INET, 1); 1232 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1); 1233 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1); 1234 MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */ 1235 MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */ 1236 1237 module_init(nfnetlink_log_init); 1238 module_exit(nfnetlink_log_fini); 1239