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 struct user_namespace *peer_user_ns; /* User namespace of the peer process */ 70 u32 peer_portid; /* PORTID of the peer process */ 71 72 /* configurable parameters */ 73 unsigned int flushtimeout; /* timeout until queue flush */ 74 unsigned int nlbufsiz; /* netlink buffer allocation size */ 75 unsigned int qthreshold; /* threshold of the queue */ 76 u_int32_t copy_range; 77 u_int32_t seq; /* instance-local sequential counter */ 78 u_int16_t group_num; /* number of this queue */ 79 u_int16_t flags; 80 u_int8_t copy_mode; 81 struct rcu_head rcu; 82 }; 83 84 #define INSTANCE_BUCKETS 16 85 86 static unsigned int nfnl_log_net_id __read_mostly; 87 88 struct nfnl_log_net { 89 spinlock_t instances_lock; 90 struct hlist_head instance_table[INSTANCE_BUCKETS]; 91 atomic_t global_seq; 92 }; 93 94 static struct nfnl_log_net *nfnl_log_pernet(struct net *net) 95 { 96 return net_generic(net, nfnl_log_net_id); 97 } 98 99 static inline u_int8_t instance_hashfn(u_int16_t group_num) 100 { 101 return ((group_num & 0xff) % INSTANCE_BUCKETS); 102 } 103 104 static struct nfulnl_instance * 105 __instance_lookup(struct nfnl_log_net *log, u_int16_t group_num) 106 { 107 struct hlist_head *head; 108 struct nfulnl_instance *inst; 109 110 head = &log->instance_table[instance_hashfn(group_num)]; 111 hlist_for_each_entry_rcu(inst, head, hlist) { 112 if (inst->group_num == group_num) 113 return inst; 114 } 115 return NULL; 116 } 117 118 static inline void 119 instance_get(struct nfulnl_instance *inst) 120 { 121 refcount_inc(&inst->use); 122 } 123 124 static struct nfulnl_instance * 125 instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num) 126 { 127 struct nfulnl_instance *inst; 128 129 rcu_read_lock_bh(); 130 inst = __instance_lookup(log, group_num); 131 if (inst && !refcount_inc_not_zero(&inst->use)) 132 inst = NULL; 133 rcu_read_unlock_bh(); 134 135 return inst; 136 } 137 138 static void nfulnl_instance_free_rcu(struct rcu_head *head) 139 { 140 struct nfulnl_instance *inst = 141 container_of(head, struct nfulnl_instance, rcu); 142 143 put_net(inst->net); 144 kfree(inst); 145 module_put(THIS_MODULE); 146 } 147 148 static void 149 instance_put(struct nfulnl_instance *inst) 150 { 151 if (inst && refcount_dec_and_test(&inst->use)) 152 call_rcu(&inst->rcu, nfulnl_instance_free_rcu); 153 } 154 155 static void nfulnl_timer(struct timer_list *t); 156 157 static struct nfulnl_instance * 158 instance_create(struct net *net, u_int16_t group_num, 159 u32 portid, struct user_namespace *user_ns) 160 { 161 struct nfulnl_instance *inst; 162 struct nfnl_log_net *log = nfnl_log_pernet(net); 163 int err; 164 165 spin_lock_bh(&log->instances_lock); 166 if (__instance_lookup(log, group_num)) { 167 err = -EEXIST; 168 goto out_unlock; 169 } 170 171 inst = kzalloc(sizeof(*inst), GFP_ATOMIC); 172 if (!inst) { 173 err = -ENOMEM; 174 goto out_unlock; 175 } 176 177 if (!try_module_get(THIS_MODULE)) { 178 kfree(inst); 179 err = -EAGAIN; 180 goto out_unlock; 181 } 182 183 INIT_HLIST_NODE(&inst->hlist); 184 spin_lock_init(&inst->lock); 185 /* needs to be two, since we _put() after creation */ 186 refcount_set(&inst->use, 2); 187 188 timer_setup(&inst->timer, nfulnl_timer, 0); 189 190 inst->net = get_net(net); 191 inst->peer_user_ns = user_ns; 192 inst->peer_portid = portid; 193 inst->group_num = group_num; 194 195 inst->qthreshold = NFULNL_QTHRESH_DEFAULT; 196 inst->flushtimeout = NFULNL_TIMEOUT_DEFAULT; 197 inst->nlbufsiz = NFULNL_NLBUFSIZ_DEFAULT; 198 inst->copy_mode = NFULNL_COPY_PACKET; 199 inst->copy_range = NFULNL_COPY_RANGE_MAX; 200 201 hlist_add_head_rcu(&inst->hlist, 202 &log->instance_table[instance_hashfn(group_num)]); 203 204 205 spin_unlock_bh(&log->instances_lock); 206 207 return inst; 208 209 out_unlock: 210 spin_unlock_bh(&log->instances_lock); 211 return ERR_PTR(err); 212 } 213 214 static void __nfulnl_flush(struct nfulnl_instance *inst); 215 216 /* called with BH disabled */ 217 static void 218 __instance_destroy(struct nfulnl_instance *inst) 219 { 220 /* first pull it out of the global list */ 221 hlist_del_rcu(&inst->hlist); 222 223 /* then flush all pending packets from skb */ 224 225 spin_lock(&inst->lock); 226 227 /* lockless readers wont be able to use us */ 228 inst->copy_mode = NFULNL_COPY_DISABLED; 229 230 if (inst->skb) 231 __nfulnl_flush(inst); 232 spin_unlock(&inst->lock); 233 234 /* and finally put the refcount */ 235 instance_put(inst); 236 } 237 238 static inline void 239 instance_destroy(struct nfnl_log_net *log, 240 struct nfulnl_instance *inst) 241 { 242 spin_lock_bh(&log->instances_lock); 243 __instance_destroy(inst); 244 spin_unlock_bh(&log->instances_lock); 245 } 246 247 static int 248 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode, 249 unsigned int range) 250 { 251 int status = 0; 252 253 spin_lock_bh(&inst->lock); 254 255 switch (mode) { 256 case NFULNL_COPY_NONE: 257 case NFULNL_COPY_META: 258 inst->copy_mode = mode; 259 inst->copy_range = 0; 260 break; 261 262 case NFULNL_COPY_PACKET: 263 inst->copy_mode = mode; 264 if (range == 0) 265 range = NFULNL_COPY_RANGE_MAX; 266 inst->copy_range = min_t(unsigned int, 267 range, NFULNL_COPY_RANGE_MAX); 268 break; 269 270 default: 271 status = -EINVAL; 272 break; 273 } 274 275 spin_unlock_bh(&inst->lock); 276 277 return status; 278 } 279 280 static int 281 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz) 282 { 283 int status; 284 285 spin_lock_bh(&inst->lock); 286 if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT) 287 status = -ERANGE; 288 else if (nlbufsiz > 131072) 289 status = -ERANGE; 290 else { 291 inst->nlbufsiz = nlbufsiz; 292 status = 0; 293 } 294 spin_unlock_bh(&inst->lock); 295 296 return status; 297 } 298 299 static void 300 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout) 301 { 302 spin_lock_bh(&inst->lock); 303 inst->flushtimeout = timeout; 304 spin_unlock_bh(&inst->lock); 305 } 306 307 static void 308 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh) 309 { 310 spin_lock_bh(&inst->lock); 311 inst->qthreshold = qthresh; 312 spin_unlock_bh(&inst->lock); 313 } 314 315 static int 316 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags) 317 { 318 spin_lock_bh(&inst->lock); 319 inst->flags = flags; 320 spin_unlock_bh(&inst->lock); 321 322 return 0; 323 } 324 325 static struct sk_buff * 326 nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size, 327 unsigned int pkt_size) 328 { 329 struct sk_buff *skb; 330 unsigned int n; 331 332 /* alloc skb which should be big enough for a whole multipart 333 * message. WARNING: has to be <= 128k due to slab restrictions */ 334 335 n = max(inst_size, pkt_size); 336 skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN); 337 if (!skb) { 338 if (n > pkt_size) { 339 /* try to allocate only as much as we need for current 340 * packet */ 341 342 skb = alloc_skb(pkt_size, GFP_ATOMIC); 343 } 344 } 345 346 return skb; 347 } 348 349 static void 350 __nfulnl_send(struct nfulnl_instance *inst) 351 { 352 if (inst->qlen > 1) { 353 struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0, 354 NLMSG_DONE, 355 sizeof(struct nfgenmsg), 356 0); 357 if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n", 358 inst->skb->len, skb_tailroom(inst->skb))) { 359 kfree_skb(inst->skb); 360 goto out; 361 } 362 } 363 nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid); 364 out: 365 inst->qlen = 0; 366 inst->skb = NULL; 367 } 368 369 static void 370 __nfulnl_flush(struct nfulnl_instance *inst) 371 { 372 /* timer holds a reference */ 373 if (del_timer(&inst->timer)) 374 instance_put(inst); 375 if (inst->skb) 376 __nfulnl_send(inst); 377 } 378 379 static void 380 nfulnl_timer(struct timer_list *t) 381 { 382 struct nfulnl_instance *inst = from_timer(inst, t, timer); 383 384 spin_lock_bh(&inst->lock); 385 if (inst->skb) 386 __nfulnl_send(inst); 387 spin_unlock_bh(&inst->lock); 388 instance_put(inst); 389 } 390 391 static u32 nfulnl_get_bridge_size(const struct sk_buff *skb) 392 { 393 u32 size = 0; 394 395 if (!skb_mac_header_was_set(skb)) 396 return 0; 397 398 if (skb_vlan_tag_present(skb)) { 399 size += nla_total_size(0); /* nested */ 400 size += nla_total_size(sizeof(u16)); /* id */ 401 size += nla_total_size(sizeof(u16)); /* tag */ 402 } 403 404 if (skb->network_header > skb->mac_header) 405 size += nla_total_size(skb->network_header - skb->mac_header); 406 407 return size; 408 } 409 410 static int nfulnl_put_bridge(struct nfulnl_instance *inst, const struct sk_buff *skb) 411 { 412 if (!skb_mac_header_was_set(skb)) 413 return 0; 414 415 if (skb_vlan_tag_present(skb)) { 416 struct nlattr *nest; 417 418 nest = nla_nest_start(inst->skb, NFULA_VLAN); 419 if (!nest) 420 goto nla_put_failure; 421 422 if (nla_put_be16(inst->skb, NFULA_VLAN_TCI, htons(skb->vlan_tci)) || 423 nla_put_be16(inst->skb, NFULA_VLAN_PROTO, skb->vlan_proto)) 424 goto nla_put_failure; 425 426 nla_nest_end(inst->skb, nest); 427 } 428 429 if (skb->mac_header < skb->network_header) { 430 int len = (int)(skb->network_header - skb->mac_header); 431 432 if (nla_put(inst->skb, NFULA_L2HDR, len, skb_mac_header(skb))) 433 goto nla_put_failure; 434 } 435 436 return 0; 437 438 nla_put_failure: 439 return -1; 440 } 441 442 /* This is an inline function, we don't really care about a long 443 * list of arguments */ 444 static inline int 445 __build_packet_message(struct nfnl_log_net *log, 446 struct nfulnl_instance *inst, 447 const struct sk_buff *skb, 448 unsigned int data_len, 449 u_int8_t pf, 450 unsigned int hooknum, 451 const struct net_device *indev, 452 const struct net_device *outdev, 453 const char *prefix, unsigned int plen, 454 const struct nfnl_ct_hook *nfnl_ct, 455 struct nf_conn *ct, enum ip_conntrack_info ctinfo) 456 { 457 struct nfulnl_msg_packet_hdr pmsg; 458 struct nlmsghdr *nlh; 459 struct nfgenmsg *nfmsg; 460 sk_buff_data_t old_tail = inst->skb->tail; 461 struct sock *sk; 462 const unsigned char *hwhdrp; 463 464 nlh = nlmsg_put(inst->skb, 0, 0, 465 nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET), 466 sizeof(struct nfgenmsg), 0); 467 if (!nlh) 468 return -1; 469 nfmsg = nlmsg_data(nlh); 470 nfmsg->nfgen_family = pf; 471 nfmsg->version = NFNETLINK_V0; 472 nfmsg->res_id = htons(inst->group_num); 473 474 memset(&pmsg, 0, sizeof(pmsg)); 475 pmsg.hw_protocol = skb->protocol; 476 pmsg.hook = hooknum; 477 478 if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg)) 479 goto nla_put_failure; 480 481 if (prefix && 482 nla_put(inst->skb, NFULA_PREFIX, plen, prefix)) 483 goto nla_put_failure; 484 485 if (indev) { 486 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 487 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV, 488 htonl(indev->ifindex))) 489 goto nla_put_failure; 490 #else 491 if (pf == PF_BRIDGE) { 492 /* Case 1: outdev is physical input device, we need to 493 * look for bridge group (when called from 494 * netfilter_bridge) */ 495 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV, 496 htonl(indev->ifindex)) || 497 /* this is the bridge group "brX" */ 498 /* rcu_read_lock()ed by nf_hook_thresh or 499 * nf_log_packet. 500 */ 501 nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV, 502 htonl(br_port_get_rcu(indev)->br->dev->ifindex))) 503 goto nla_put_failure; 504 } else { 505 struct net_device *physindev; 506 507 /* Case 2: indev is bridge group, we need to look for 508 * physical device (when called from ipv4) */ 509 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV, 510 htonl(indev->ifindex))) 511 goto nla_put_failure; 512 513 physindev = nf_bridge_get_physindev(skb); 514 if (physindev && 515 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV, 516 htonl(physindev->ifindex))) 517 goto nla_put_failure; 518 } 519 #endif 520 } 521 522 if (outdev) { 523 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 524 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV, 525 htonl(outdev->ifindex))) 526 goto nla_put_failure; 527 #else 528 if (pf == PF_BRIDGE) { 529 /* Case 1: outdev is physical output device, we need to 530 * look for bridge group (when called from 531 * netfilter_bridge) */ 532 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV, 533 htonl(outdev->ifindex)) || 534 /* this is the bridge group "brX" */ 535 /* rcu_read_lock()ed by nf_hook_thresh or 536 * nf_log_packet. 537 */ 538 nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV, 539 htonl(br_port_get_rcu(outdev)->br->dev->ifindex))) 540 goto nla_put_failure; 541 } else { 542 struct net_device *physoutdev; 543 544 /* Case 2: indev is a bridge group, we need to look 545 * for physical device (when called from ipv4) */ 546 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV, 547 htonl(outdev->ifindex))) 548 goto nla_put_failure; 549 550 physoutdev = nf_bridge_get_physoutdev(skb); 551 if (physoutdev && 552 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV, 553 htonl(physoutdev->ifindex))) 554 goto nla_put_failure; 555 } 556 #endif 557 } 558 559 if (skb->mark && 560 nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark))) 561 goto nla_put_failure; 562 563 if (indev && skb->dev && 564 skb->mac_header != skb->network_header) { 565 struct nfulnl_msg_packet_hw phw; 566 int len; 567 568 memset(&phw, 0, sizeof(phw)); 569 len = dev_parse_header(skb, phw.hw_addr); 570 if (len > 0) { 571 phw.hw_addrlen = htons(len); 572 if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw)) 573 goto nla_put_failure; 574 } 575 } 576 577 if (indev && skb_mac_header_was_set(skb)) { 578 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) || 579 nla_put_be16(inst->skb, NFULA_HWLEN, 580 htons(skb->dev->hard_header_len))) 581 goto nla_put_failure; 582 583 hwhdrp = skb_mac_header(skb); 584 585 if (skb->dev->type == ARPHRD_SIT) 586 hwhdrp -= ETH_HLEN; 587 588 if (hwhdrp >= skb->head && 589 nla_put(inst->skb, NFULA_HWHEADER, 590 skb->dev->hard_header_len, hwhdrp)) 591 goto nla_put_failure; 592 } 593 594 if (hooknum <= NF_INET_FORWARD && skb->tstamp) { 595 struct nfulnl_msg_packet_timestamp ts; 596 struct timespec64 kts = ktime_to_timespec64(skb->tstamp); 597 ts.sec = cpu_to_be64(kts.tv_sec); 598 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC); 599 600 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts)) 601 goto nla_put_failure; 602 } 603 604 /* UID */ 605 sk = skb->sk; 606 if (sk && sk_fullsock(sk)) { 607 read_lock_bh(&sk->sk_callback_lock); 608 if (sk->sk_socket && sk->sk_socket->file) { 609 struct file *file = sk->sk_socket->file; 610 const struct cred *cred = file->f_cred; 611 struct user_namespace *user_ns = inst->peer_user_ns; 612 __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid)); 613 __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid)); 614 read_unlock_bh(&sk->sk_callback_lock); 615 if (nla_put_be32(inst->skb, NFULA_UID, uid) || 616 nla_put_be32(inst->skb, NFULA_GID, gid)) 617 goto nla_put_failure; 618 } else 619 read_unlock_bh(&sk->sk_callback_lock); 620 } 621 622 /* local sequence number */ 623 if ((inst->flags & NFULNL_CFG_F_SEQ) && 624 nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++))) 625 goto nla_put_failure; 626 627 /* global sequence number */ 628 if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) && 629 nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL, 630 htonl(atomic_inc_return(&log->global_seq)))) 631 goto nla_put_failure; 632 633 if (ct && nfnl_ct->build(inst->skb, ct, ctinfo, 634 NFULA_CT, NFULA_CT_INFO) < 0) 635 goto nla_put_failure; 636 637 if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) && 638 nfulnl_put_bridge(inst, skb) < 0) 639 goto nla_put_failure; 640 641 if (data_len) { 642 struct nlattr *nla; 643 int size = nla_attr_size(data_len); 644 645 if (skb_tailroom(inst->skb) < nla_total_size(data_len)) 646 goto nla_put_failure; 647 648 nla = skb_put(inst->skb, nla_total_size(data_len)); 649 nla->nla_type = NFULA_PAYLOAD; 650 nla->nla_len = size; 651 652 if (skb_copy_bits(skb, 0, nla_data(nla), data_len)) 653 BUG(); 654 } 655 656 nlh->nlmsg_len = inst->skb->tail - old_tail; 657 return 0; 658 659 nla_put_failure: 660 PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n"); 661 return -1; 662 } 663 664 static const struct nf_loginfo default_loginfo = { 665 .type = NF_LOG_TYPE_ULOG, 666 .u = { 667 .ulog = { 668 .copy_len = 0xffff, 669 .group = 0, 670 .qthreshold = 1, 671 }, 672 }, 673 }; 674 675 /* log handler for internal netfilter logging api */ 676 static void 677 nfulnl_log_packet(struct net *net, 678 u_int8_t pf, 679 unsigned int hooknum, 680 const struct sk_buff *skb, 681 const struct net_device *in, 682 const struct net_device *out, 683 const struct nf_loginfo *li_user, 684 const char *prefix) 685 { 686 size_t size; 687 unsigned int data_len; 688 struct nfulnl_instance *inst; 689 const struct nf_loginfo *li; 690 unsigned int qthreshold; 691 unsigned int plen = 0; 692 struct nfnl_log_net *log = nfnl_log_pernet(net); 693 const struct nfnl_ct_hook *nfnl_ct = NULL; 694 struct nf_conn *ct = NULL; 695 enum ip_conntrack_info ctinfo; 696 697 if (li_user && li_user->type == NF_LOG_TYPE_ULOG) 698 li = li_user; 699 else 700 li = &default_loginfo; 701 702 inst = instance_lookup_get(log, li->u.ulog.group); 703 if (!inst) 704 return; 705 706 if (prefix) 707 plen = strlen(prefix) + 1; 708 709 /* FIXME: do we want to make the size calculation conditional based on 710 * what is actually present? way more branches and checks, but more 711 * memory efficient... */ 712 size = nlmsg_total_size(sizeof(struct nfgenmsg)) 713 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr)) 714 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 715 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 716 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 717 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 718 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 719 #endif 720 + nla_total_size(sizeof(u_int32_t)) /* mark */ 721 + nla_total_size(sizeof(u_int32_t)) /* uid */ 722 + nla_total_size(sizeof(u_int32_t)) /* gid */ 723 + nla_total_size(plen) /* prefix */ 724 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw)) 725 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp)) 726 + nla_total_size(sizeof(struct nfgenmsg)); /* NLMSG_DONE */ 727 728 if (in && skb_mac_header_was_set(skb)) { 729 size += nla_total_size(skb->dev->hard_header_len) 730 + nla_total_size(sizeof(u_int16_t)) /* hwtype */ 731 + nla_total_size(sizeof(u_int16_t)); /* hwlen */ 732 } 733 734 spin_lock_bh(&inst->lock); 735 736 if (inst->flags & NFULNL_CFG_F_SEQ) 737 size += nla_total_size(sizeof(u_int32_t)); 738 if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) 739 size += nla_total_size(sizeof(u_int32_t)); 740 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 741 if (inst->flags & NFULNL_CFG_F_CONNTRACK) { 742 nfnl_ct = rcu_dereference(nfnl_ct_hook); 743 if (nfnl_ct != NULL) { 744 ct = nf_ct_get(skb, &ctinfo); 745 if (ct != NULL) 746 size += nfnl_ct->build_size(ct); 747 } 748 } 749 #endif 750 if (pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) 751 size += nfulnl_get_bridge_size(skb); 752 753 qthreshold = inst->qthreshold; 754 /* per-rule qthreshold overrides per-instance */ 755 if (li->u.ulog.qthreshold) 756 if (qthreshold > li->u.ulog.qthreshold) 757 qthreshold = li->u.ulog.qthreshold; 758 759 760 switch (inst->copy_mode) { 761 case NFULNL_COPY_META: 762 case NFULNL_COPY_NONE: 763 data_len = 0; 764 break; 765 766 case NFULNL_COPY_PACKET: 767 data_len = inst->copy_range; 768 if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) && 769 (li->u.ulog.copy_len < data_len)) 770 data_len = li->u.ulog.copy_len; 771 772 if (data_len > skb->len) 773 data_len = skb->len; 774 775 size += nla_total_size(data_len); 776 break; 777 778 case NFULNL_COPY_DISABLED: 779 default: 780 goto unlock_and_release; 781 } 782 783 if (inst->skb && size > skb_tailroom(inst->skb)) { 784 /* either the queue len is too high or we don't have 785 * enough room in the skb left. flush to userspace. */ 786 __nfulnl_flush(inst); 787 } 788 789 if (!inst->skb) { 790 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid, 791 inst->nlbufsiz, size); 792 if (!inst->skb) 793 goto alloc_failure; 794 } 795 796 inst->qlen++; 797 798 __build_packet_message(log, inst, skb, data_len, pf, 799 hooknum, in, out, prefix, plen, 800 nfnl_ct, ct, ctinfo); 801 802 if (inst->qlen >= qthreshold) 803 __nfulnl_flush(inst); 804 /* timer_pending always called within inst->lock, so there 805 * is no chance of a race here */ 806 else if (!timer_pending(&inst->timer)) { 807 instance_get(inst); 808 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100); 809 add_timer(&inst->timer); 810 } 811 812 unlock_and_release: 813 spin_unlock_bh(&inst->lock); 814 instance_put(inst); 815 return; 816 817 alloc_failure: 818 /* FIXME: statistics */ 819 goto unlock_and_release; 820 } 821 822 static int 823 nfulnl_rcv_nl_event(struct notifier_block *this, 824 unsigned long event, void *ptr) 825 { 826 struct netlink_notify *n = ptr; 827 struct nfnl_log_net *log = nfnl_log_pernet(n->net); 828 829 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) { 830 int i; 831 832 /* destroy all instances for this portid */ 833 spin_lock_bh(&log->instances_lock); 834 for (i = 0; i < INSTANCE_BUCKETS; i++) { 835 struct hlist_node *t2; 836 struct nfulnl_instance *inst; 837 struct hlist_head *head = &log->instance_table[i]; 838 839 hlist_for_each_entry_safe(inst, t2, head, hlist) { 840 if (n->portid == inst->peer_portid) 841 __instance_destroy(inst); 842 } 843 } 844 spin_unlock_bh(&log->instances_lock); 845 } 846 return NOTIFY_DONE; 847 } 848 849 static struct notifier_block nfulnl_rtnl_notifier = { 850 .notifier_call = nfulnl_rcv_nl_event, 851 }; 852 853 static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl, 854 struct sk_buff *skb, const struct nlmsghdr *nlh, 855 const struct nlattr * const nfqa[], 856 struct netlink_ext_ack *extack) 857 { 858 return -ENOTSUPP; 859 } 860 861 static struct nf_logger nfulnl_logger __read_mostly = { 862 .name = "nfnetlink_log", 863 .type = NF_LOG_TYPE_ULOG, 864 .logfn = nfulnl_log_packet, 865 .me = THIS_MODULE, 866 }; 867 868 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = { 869 [NFULA_CFG_CMD] = { .len = sizeof(struct nfulnl_msg_config_cmd) }, 870 [NFULA_CFG_MODE] = { .len = sizeof(struct nfulnl_msg_config_mode) }, 871 [NFULA_CFG_TIMEOUT] = { .type = NLA_U32 }, 872 [NFULA_CFG_QTHRESH] = { .type = NLA_U32 }, 873 [NFULA_CFG_NLBUFSIZ] = { .type = NLA_U32 }, 874 [NFULA_CFG_FLAGS] = { .type = NLA_U16 }, 875 }; 876 877 static int nfulnl_recv_config(struct net *net, struct sock *ctnl, 878 struct sk_buff *skb, const struct nlmsghdr *nlh, 879 const struct nlattr * const nfula[], 880 struct netlink_ext_ack *extack) 881 { 882 struct nfgenmsg *nfmsg = nlmsg_data(nlh); 883 u_int16_t group_num = ntohs(nfmsg->res_id); 884 struct nfulnl_instance *inst; 885 struct nfulnl_msg_config_cmd *cmd = NULL; 886 struct nfnl_log_net *log = nfnl_log_pernet(net); 887 int ret = 0; 888 u16 flags = 0; 889 890 if (nfula[NFULA_CFG_CMD]) { 891 u_int8_t pf = nfmsg->nfgen_family; 892 cmd = nla_data(nfula[NFULA_CFG_CMD]); 893 894 /* Commands without queue context */ 895 switch (cmd->command) { 896 case NFULNL_CFG_CMD_PF_BIND: 897 return nf_log_bind_pf(net, pf, &nfulnl_logger); 898 case NFULNL_CFG_CMD_PF_UNBIND: 899 nf_log_unbind_pf(net, pf); 900 return 0; 901 } 902 } 903 904 inst = instance_lookup_get(log, group_num); 905 if (inst && inst->peer_portid != NETLINK_CB(skb).portid) { 906 ret = -EPERM; 907 goto out_put; 908 } 909 910 /* Check if we support these flags in first place, dependencies should 911 * be there too not to break atomicity. 912 */ 913 if (nfula[NFULA_CFG_FLAGS]) { 914 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS])); 915 916 if ((flags & NFULNL_CFG_F_CONNTRACK) && 917 !rcu_access_pointer(nfnl_ct_hook)) { 918 #ifdef CONFIG_MODULES 919 nfnl_unlock(NFNL_SUBSYS_ULOG); 920 request_module("ip_conntrack_netlink"); 921 nfnl_lock(NFNL_SUBSYS_ULOG); 922 if (rcu_access_pointer(nfnl_ct_hook)) { 923 ret = -EAGAIN; 924 goto out_put; 925 } 926 #endif 927 ret = -EOPNOTSUPP; 928 goto out_put; 929 } 930 } 931 932 if (cmd != NULL) { 933 switch (cmd->command) { 934 case NFULNL_CFG_CMD_BIND: 935 if (inst) { 936 ret = -EBUSY; 937 goto out_put; 938 } 939 940 inst = instance_create(net, group_num, 941 NETLINK_CB(skb).portid, 942 sk_user_ns(NETLINK_CB(skb).sk)); 943 if (IS_ERR(inst)) { 944 ret = PTR_ERR(inst); 945 goto out; 946 } 947 break; 948 case NFULNL_CFG_CMD_UNBIND: 949 if (!inst) { 950 ret = -ENODEV; 951 goto out; 952 } 953 954 instance_destroy(log, inst); 955 goto out_put; 956 default: 957 ret = -ENOTSUPP; 958 goto out_put; 959 } 960 } else if (!inst) { 961 ret = -ENODEV; 962 goto out; 963 } 964 965 if (nfula[NFULA_CFG_MODE]) { 966 struct nfulnl_msg_config_mode *params = 967 nla_data(nfula[NFULA_CFG_MODE]); 968 969 nfulnl_set_mode(inst, params->copy_mode, 970 ntohl(params->copy_range)); 971 } 972 973 if (nfula[NFULA_CFG_TIMEOUT]) { 974 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]); 975 976 nfulnl_set_timeout(inst, ntohl(timeout)); 977 } 978 979 if (nfula[NFULA_CFG_NLBUFSIZ]) { 980 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]); 981 982 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz)); 983 } 984 985 if (nfula[NFULA_CFG_QTHRESH]) { 986 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]); 987 988 nfulnl_set_qthresh(inst, ntohl(qthresh)); 989 } 990 991 if (nfula[NFULA_CFG_FLAGS]) 992 nfulnl_set_flags(inst, flags); 993 994 out_put: 995 instance_put(inst); 996 out: 997 return ret; 998 } 999 1000 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = { 1001 [NFULNL_MSG_PACKET] = { .call = nfulnl_recv_unsupp, 1002 .attr_count = NFULA_MAX, }, 1003 [NFULNL_MSG_CONFIG] = { .call = nfulnl_recv_config, 1004 .attr_count = NFULA_CFG_MAX, 1005 .policy = nfula_cfg_policy }, 1006 }; 1007 1008 static const struct nfnetlink_subsystem nfulnl_subsys = { 1009 .name = "log", 1010 .subsys_id = NFNL_SUBSYS_ULOG, 1011 .cb_count = NFULNL_MSG_MAX, 1012 .cb = nfulnl_cb, 1013 }; 1014 1015 #ifdef CONFIG_PROC_FS 1016 struct iter_state { 1017 struct seq_net_private p; 1018 unsigned int bucket; 1019 }; 1020 1021 static struct hlist_node *get_first(struct net *net, struct iter_state *st) 1022 { 1023 struct nfnl_log_net *log; 1024 if (!st) 1025 return NULL; 1026 1027 log = nfnl_log_pernet(net); 1028 1029 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) { 1030 struct hlist_head *head = &log->instance_table[st->bucket]; 1031 1032 if (!hlist_empty(head)) 1033 return rcu_dereference_bh(hlist_first_rcu(head)); 1034 } 1035 return NULL; 1036 } 1037 1038 static struct hlist_node *get_next(struct net *net, struct iter_state *st, 1039 struct hlist_node *h) 1040 { 1041 h = rcu_dereference_bh(hlist_next_rcu(h)); 1042 while (!h) { 1043 struct nfnl_log_net *log; 1044 struct hlist_head *head; 1045 1046 if (++st->bucket >= INSTANCE_BUCKETS) 1047 return NULL; 1048 1049 log = nfnl_log_pernet(net); 1050 head = &log->instance_table[st->bucket]; 1051 h = rcu_dereference_bh(hlist_first_rcu(head)); 1052 } 1053 return h; 1054 } 1055 1056 static struct hlist_node *get_idx(struct net *net, struct iter_state *st, 1057 loff_t pos) 1058 { 1059 struct hlist_node *head; 1060 head = get_first(net, st); 1061 1062 if (head) 1063 while (pos && (head = get_next(net, st, head))) 1064 pos--; 1065 return pos ? NULL : head; 1066 } 1067 1068 static void *seq_start(struct seq_file *s, loff_t *pos) 1069 __acquires(rcu_bh) 1070 { 1071 rcu_read_lock_bh(); 1072 return get_idx(seq_file_net(s), s->private, *pos); 1073 } 1074 1075 static void *seq_next(struct seq_file *s, void *v, loff_t *pos) 1076 { 1077 (*pos)++; 1078 return get_next(seq_file_net(s), s->private, v); 1079 } 1080 1081 static void seq_stop(struct seq_file *s, void *v) 1082 __releases(rcu_bh) 1083 { 1084 rcu_read_unlock_bh(); 1085 } 1086 1087 static int seq_show(struct seq_file *s, void *v) 1088 { 1089 const struct nfulnl_instance *inst = v; 1090 1091 seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n", 1092 inst->group_num, 1093 inst->peer_portid, inst->qlen, 1094 inst->copy_mode, inst->copy_range, 1095 inst->flushtimeout, refcount_read(&inst->use)); 1096 1097 return 0; 1098 } 1099 1100 static const struct seq_operations nful_seq_ops = { 1101 .start = seq_start, 1102 .next = seq_next, 1103 .stop = seq_stop, 1104 .show = seq_show, 1105 }; 1106 #endif /* PROC_FS */ 1107 1108 static int __net_init nfnl_log_net_init(struct net *net) 1109 { 1110 unsigned int i; 1111 struct nfnl_log_net *log = nfnl_log_pernet(net); 1112 #ifdef CONFIG_PROC_FS 1113 struct proc_dir_entry *proc; 1114 kuid_t root_uid; 1115 kgid_t root_gid; 1116 #endif 1117 1118 for (i = 0; i < INSTANCE_BUCKETS; i++) 1119 INIT_HLIST_HEAD(&log->instance_table[i]); 1120 spin_lock_init(&log->instances_lock); 1121 1122 #ifdef CONFIG_PROC_FS 1123 proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter, 1124 &nful_seq_ops, sizeof(struct iter_state)); 1125 if (!proc) 1126 return -ENOMEM; 1127 1128 root_uid = make_kuid(net->user_ns, 0); 1129 root_gid = make_kgid(net->user_ns, 0); 1130 if (uid_valid(root_uid) && gid_valid(root_gid)) 1131 proc_set_user(proc, root_uid, root_gid); 1132 #endif 1133 return 0; 1134 } 1135 1136 static void __net_exit nfnl_log_net_exit(struct net *net) 1137 { 1138 struct nfnl_log_net *log = nfnl_log_pernet(net); 1139 unsigned int i; 1140 1141 #ifdef CONFIG_PROC_FS 1142 remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter); 1143 #endif 1144 nf_log_unset(net, &nfulnl_logger); 1145 for (i = 0; i < INSTANCE_BUCKETS; i++) 1146 WARN_ON_ONCE(!hlist_empty(&log->instance_table[i])); 1147 } 1148 1149 static struct pernet_operations nfnl_log_net_ops = { 1150 .init = nfnl_log_net_init, 1151 .exit = nfnl_log_net_exit, 1152 .id = &nfnl_log_net_id, 1153 .size = sizeof(struct nfnl_log_net), 1154 }; 1155 1156 static int __init nfnetlink_log_init(void) 1157 { 1158 int status; 1159 1160 status = register_pernet_subsys(&nfnl_log_net_ops); 1161 if (status < 0) { 1162 pr_err("failed to register pernet ops\n"); 1163 goto out; 1164 } 1165 1166 netlink_register_notifier(&nfulnl_rtnl_notifier); 1167 status = nfnetlink_subsys_register(&nfulnl_subsys); 1168 if (status < 0) { 1169 pr_err("failed to create netlink socket\n"); 1170 goto cleanup_netlink_notifier; 1171 } 1172 1173 status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger); 1174 if (status < 0) { 1175 pr_err("failed to register logger\n"); 1176 goto cleanup_subsys; 1177 } 1178 1179 return status; 1180 1181 cleanup_subsys: 1182 nfnetlink_subsys_unregister(&nfulnl_subsys); 1183 cleanup_netlink_notifier: 1184 netlink_unregister_notifier(&nfulnl_rtnl_notifier); 1185 unregister_pernet_subsys(&nfnl_log_net_ops); 1186 out: 1187 return status; 1188 } 1189 1190 static void __exit nfnetlink_log_fini(void) 1191 { 1192 nfnetlink_subsys_unregister(&nfulnl_subsys); 1193 netlink_unregister_notifier(&nfulnl_rtnl_notifier); 1194 unregister_pernet_subsys(&nfnl_log_net_ops); 1195 nf_log_unregister(&nfulnl_logger); 1196 } 1197 1198 MODULE_DESCRIPTION("netfilter userspace logging"); 1199 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); 1200 MODULE_LICENSE("GPL"); 1201 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG); 1202 MODULE_ALIAS_NF_LOGGER(AF_INET, 1); 1203 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1); 1204 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1); 1205 MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */ 1206 MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */ 1207 1208 module_init(nfnetlink_log_init); 1209 module_exit(nfnetlink_log_fini); 1210