| 1 | /* |
| 2 | * This is a module which is used for logging packets to userspace via |
| 3 | * nfetlink. |
| 4 | * |
| 5 | * (C) 2005 by Harald Welte <laforge@netfilter.org> |
| 6 | * |
| 7 | * Based on the old ipv4-only ipt_ULOG.c: |
| 8 | * (C) 2000-2004 by Harald Welte <laforge@netfilter.org> |
| 9 | * |
| 10 | * This program is free software; you can redistribute it and/or modify |
| 11 | * it under the terms of the GNU General Public License version 2 as |
| 12 | * published by the Free Software Foundation. |
| 13 | */ |
| 14 | #include <linux/module.h> |
| 15 | #include <linux/skbuff.h> |
| 16 | #include <linux/if_arp.h> |
| 17 | #include <linux/init.h> |
| 18 | #include <linux/ip.h> |
| 19 | #include <linux/ipv6.h> |
| 20 | #include <linux/netdevice.h> |
| 21 | #include <linux/netfilter.h> |
| 22 | #include <linux/netlink.h> |
| 23 | #include <linux/netfilter/nfnetlink.h> |
| 24 | #include <linux/netfilter/nfnetlink_log.h> |
| 25 | #include <linux/spinlock.h> |
| 26 | #include <linux/sysctl.h> |
| 27 | #include <linux/proc_fs.h> |
| 28 | #include <linux/security.h> |
| 29 | #include <linux/list.h> |
| 30 | #include <linux/jhash.h> |
| 31 | #include <linux/random.h> |
| 32 | #include <linux/slab.h> |
| 33 | #include <net/sock.h> |
| 34 | #include <net/netfilter/nf_log.h> |
| 35 | #include <net/netfilter/nfnetlink_log.h> |
| 36 | |
| 37 | #include <linux/atomic.h> |
| 38 | |
| 39 | #ifdef CONFIG_BRIDGE_NETFILTER |
| 40 | #include "../bridge/br_private.h" |
| 41 | #endif |
| 42 | |
| 43 | #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE |
| 44 | #define NFULNL_TIMEOUT_DEFAULT 100 /* every second */ |
| 45 | #define NFULNL_QTHRESH_DEFAULT 100 /* 100 packets */ |
| 46 | #define NFULNL_COPY_RANGE_MAX 0xFFFF /* max packet size is limited by 16-bit struct nfattr nfa_len field */ |
| 47 | |
| 48 | #define PRINTR(x, args...) do { if (net_ratelimit()) \ |
| 49 | printk(x, ## args); } while (0); |
| 50 | |
| 51 | struct nfulnl_instance { |
| 52 | struct hlist_node hlist; /* global list of instances */ |
| 53 | spinlock_t lock; |
| 54 | atomic_t use; /* use count */ |
| 55 | |
| 56 | unsigned int qlen; /* number of nlmsgs in skb */ |
| 57 | struct sk_buff *skb; /* pre-allocatd skb */ |
| 58 | struct timer_list timer; |
| 59 | struct user_namespace *peer_user_ns; /* User namespace of the peer process */ |
| 60 | int peer_portid; /* PORTID of the peer process */ |
| 61 | |
| 62 | /* configurable parameters */ |
| 63 | unsigned int flushtimeout; /* timeout until queue flush */ |
| 64 | unsigned int nlbufsiz; /* netlink buffer allocation size */ |
| 65 | unsigned int qthreshold; /* threshold of the queue */ |
| 66 | u_int32_t copy_range; |
| 67 | u_int32_t seq; /* instance-local sequential counter */ |
| 68 | u_int16_t group_num; /* number of this queue */ |
| 69 | u_int16_t flags; |
| 70 | u_int8_t copy_mode; |
| 71 | struct rcu_head rcu; |
| 72 | }; |
| 73 | |
| 74 | static DEFINE_SPINLOCK(instances_lock); |
| 75 | static atomic_t global_seq; |
| 76 | |
| 77 | #define INSTANCE_BUCKETS 16 |
| 78 | static struct hlist_head instance_table[INSTANCE_BUCKETS]; |
| 79 | static unsigned int hash_init; |
| 80 | |
| 81 | static inline u_int8_t instance_hashfn(u_int16_t group_num) |
| 82 | { |
| 83 | return ((group_num & 0xff) % INSTANCE_BUCKETS); |
| 84 | } |
| 85 | |
| 86 | static struct nfulnl_instance * |
| 87 | __instance_lookup(u_int16_t group_num) |
| 88 | { |
| 89 | struct hlist_head *head; |
| 90 | struct hlist_node *pos; |
| 91 | struct nfulnl_instance *inst; |
| 92 | |
| 93 | head = &instance_table[instance_hashfn(group_num)]; |
| 94 | hlist_for_each_entry_rcu(inst, pos, head, hlist) { |
| 95 | if (inst->group_num == group_num) |
| 96 | return inst; |
| 97 | } |
| 98 | return NULL; |
| 99 | } |
| 100 | |
| 101 | static inline void |
| 102 | instance_get(struct nfulnl_instance *inst) |
| 103 | { |
| 104 | atomic_inc(&inst->use); |
| 105 | } |
| 106 | |
| 107 | static struct nfulnl_instance * |
| 108 | instance_lookup_get(u_int16_t group_num) |
| 109 | { |
| 110 | struct nfulnl_instance *inst; |
| 111 | |
| 112 | rcu_read_lock_bh(); |
| 113 | inst = __instance_lookup(group_num); |
| 114 | if (inst && !atomic_inc_not_zero(&inst->use)) |
| 115 | inst = NULL; |
| 116 | rcu_read_unlock_bh(); |
| 117 | |
| 118 | return inst; |
| 119 | } |
| 120 | |
| 121 | static void nfulnl_instance_free_rcu(struct rcu_head *head) |
| 122 | { |
| 123 | kfree(container_of(head, struct nfulnl_instance, rcu)); |
| 124 | module_put(THIS_MODULE); |
| 125 | } |
| 126 | |
| 127 | static void |
| 128 | instance_put(struct nfulnl_instance *inst) |
| 129 | { |
| 130 | if (inst && atomic_dec_and_test(&inst->use)) |
| 131 | call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu); |
| 132 | } |
| 133 | |
| 134 | static void nfulnl_timer(unsigned long data); |
| 135 | |
| 136 | static struct nfulnl_instance * |
| 137 | instance_create(u_int16_t group_num, int portid, struct user_namespace *user_ns) |
| 138 | { |
| 139 | struct nfulnl_instance *inst; |
| 140 | int err; |
| 141 | |
| 142 | spin_lock_bh(&instances_lock); |
| 143 | if (__instance_lookup(group_num)) { |
| 144 | err = -EEXIST; |
| 145 | goto out_unlock; |
| 146 | } |
| 147 | |
| 148 | inst = kzalloc(sizeof(*inst), GFP_ATOMIC); |
| 149 | if (!inst) { |
| 150 | err = -ENOMEM; |
| 151 | goto out_unlock; |
| 152 | } |
| 153 | |
| 154 | if (!try_module_get(THIS_MODULE)) { |
| 155 | kfree(inst); |
| 156 | err = -EAGAIN; |
| 157 | goto out_unlock; |
| 158 | } |
| 159 | |
| 160 | INIT_HLIST_NODE(&inst->hlist); |
| 161 | spin_lock_init(&inst->lock); |
| 162 | /* needs to be two, since we _put() after creation */ |
| 163 | atomic_set(&inst->use, 2); |
| 164 | |
| 165 | setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst); |
| 166 | |
| 167 | inst->peer_user_ns = user_ns; |
| 168 | inst->peer_portid = portid; |
| 169 | inst->group_num = group_num; |
| 170 | |
| 171 | inst->qthreshold = NFULNL_QTHRESH_DEFAULT; |
| 172 | inst->flushtimeout = NFULNL_TIMEOUT_DEFAULT; |
| 173 | inst->nlbufsiz = NFULNL_NLBUFSIZ_DEFAULT; |
| 174 | inst->copy_mode = NFULNL_COPY_PACKET; |
| 175 | inst->copy_range = NFULNL_COPY_RANGE_MAX; |
| 176 | |
| 177 | hlist_add_head_rcu(&inst->hlist, |
| 178 | &instance_table[instance_hashfn(group_num)]); |
| 179 | |
| 180 | spin_unlock_bh(&instances_lock); |
| 181 | |
| 182 | return inst; |
| 183 | |
| 184 | out_unlock: |
| 185 | spin_unlock_bh(&instances_lock); |
| 186 | return ERR_PTR(err); |
| 187 | } |
| 188 | |
| 189 | static void __nfulnl_flush(struct nfulnl_instance *inst); |
| 190 | |
| 191 | /* called with BH disabled */ |
| 192 | static void |
| 193 | __instance_destroy(struct nfulnl_instance *inst) |
| 194 | { |
| 195 | /* first pull it out of the global list */ |
| 196 | hlist_del_rcu(&inst->hlist); |
| 197 | |
| 198 | /* then flush all pending packets from skb */ |
| 199 | |
| 200 | spin_lock(&inst->lock); |
| 201 | |
| 202 | /* lockless readers wont be able to use us */ |
| 203 | inst->copy_mode = NFULNL_COPY_DISABLED; |
| 204 | |
| 205 | if (inst->skb) |
| 206 | __nfulnl_flush(inst); |
| 207 | spin_unlock(&inst->lock); |
| 208 | |
| 209 | /* and finally put the refcount */ |
| 210 | instance_put(inst); |
| 211 | } |
| 212 | |
| 213 | static inline void |
| 214 | instance_destroy(struct nfulnl_instance *inst) |
| 215 | { |
| 216 | spin_lock_bh(&instances_lock); |
| 217 | __instance_destroy(inst); |
| 218 | spin_unlock_bh(&instances_lock); |
| 219 | } |
| 220 | |
| 221 | static int |
| 222 | nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode, |
| 223 | unsigned int range) |
| 224 | { |
| 225 | int status = 0; |
| 226 | |
| 227 | spin_lock_bh(&inst->lock); |
| 228 | |
| 229 | switch (mode) { |
| 230 | case NFULNL_COPY_NONE: |
| 231 | case NFULNL_COPY_META: |
| 232 | inst->copy_mode = mode; |
| 233 | inst->copy_range = 0; |
| 234 | break; |
| 235 | |
| 236 | case NFULNL_COPY_PACKET: |
| 237 | inst->copy_mode = mode; |
| 238 | inst->copy_range = min_t(unsigned int, |
| 239 | range, NFULNL_COPY_RANGE_MAX); |
| 240 | break; |
| 241 | |
| 242 | default: |
| 243 | status = -EINVAL; |
| 244 | break; |
| 245 | } |
| 246 | |
| 247 | spin_unlock_bh(&inst->lock); |
| 248 | |
| 249 | return status; |
| 250 | } |
| 251 | |
| 252 | static int |
| 253 | nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz) |
| 254 | { |
| 255 | int status; |
| 256 | |
| 257 | spin_lock_bh(&inst->lock); |
| 258 | if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT) |
| 259 | status = -ERANGE; |
| 260 | else if (nlbufsiz > 131072) |
| 261 | status = -ERANGE; |
| 262 | else { |
| 263 | inst->nlbufsiz = nlbufsiz; |
| 264 | status = 0; |
| 265 | } |
| 266 | spin_unlock_bh(&inst->lock); |
| 267 | |
| 268 | return status; |
| 269 | } |
| 270 | |
| 271 | static int |
| 272 | nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout) |
| 273 | { |
| 274 | spin_lock_bh(&inst->lock); |
| 275 | inst->flushtimeout = timeout; |
| 276 | spin_unlock_bh(&inst->lock); |
| 277 | |
| 278 | return 0; |
| 279 | } |
| 280 | |
| 281 | static int |
| 282 | nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh) |
| 283 | { |
| 284 | spin_lock_bh(&inst->lock); |
| 285 | inst->qthreshold = qthresh; |
| 286 | spin_unlock_bh(&inst->lock); |
| 287 | |
| 288 | return 0; |
| 289 | } |
| 290 | |
| 291 | static int |
| 292 | nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags) |
| 293 | { |
| 294 | spin_lock_bh(&inst->lock); |
| 295 | inst->flags = flags; |
| 296 | spin_unlock_bh(&inst->lock); |
| 297 | |
| 298 | return 0; |
| 299 | } |
| 300 | |
| 301 | static struct sk_buff * |
| 302 | nfulnl_alloc_skb(unsigned int inst_size, unsigned int pkt_size) |
| 303 | { |
| 304 | struct sk_buff *skb; |
| 305 | unsigned int n; |
| 306 | |
| 307 | /* alloc skb which should be big enough for a whole multipart |
| 308 | * message. WARNING: has to be <= 128k due to slab restrictions */ |
| 309 | |
| 310 | n = max(inst_size, pkt_size); |
| 311 | skb = alloc_skb(n, GFP_ATOMIC); |
| 312 | if (!skb) { |
| 313 | if (n > pkt_size) { |
| 314 | /* try to allocate only as much as we need for current |
| 315 | * packet */ |
| 316 | |
| 317 | skb = alloc_skb(pkt_size, GFP_ATOMIC); |
| 318 | if (!skb) |
| 319 | pr_err("nfnetlink_log: can't even alloc %u bytes\n", |
| 320 | pkt_size); |
| 321 | } |
| 322 | } |
| 323 | |
| 324 | return skb; |
| 325 | } |
| 326 | |
| 327 | static int |
| 328 | __nfulnl_send(struct nfulnl_instance *inst) |
| 329 | { |
| 330 | int status = -1; |
| 331 | |
| 332 | if (inst->qlen > 1) { |
| 333 | struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0, |
| 334 | NLMSG_DONE, |
| 335 | sizeof(struct nfgenmsg), |
| 336 | 0); |
| 337 | if (!nlh) |
| 338 | goto out; |
| 339 | } |
| 340 | status = nfnetlink_unicast(inst->skb, &init_net, inst->peer_portid, |
| 341 | MSG_DONTWAIT); |
| 342 | |
| 343 | inst->qlen = 0; |
| 344 | inst->skb = NULL; |
| 345 | out: |
| 346 | return status; |
| 347 | } |
| 348 | |
| 349 | static void |
| 350 | __nfulnl_flush(struct nfulnl_instance *inst) |
| 351 | { |
| 352 | /* timer holds a reference */ |
| 353 | if (del_timer(&inst->timer)) |
| 354 | instance_put(inst); |
| 355 | if (inst->skb) |
| 356 | __nfulnl_send(inst); |
| 357 | } |
| 358 | |
| 359 | static void |
| 360 | nfulnl_timer(unsigned long data) |
| 361 | { |
| 362 | struct nfulnl_instance *inst = (struct nfulnl_instance *)data; |
| 363 | |
| 364 | spin_lock_bh(&inst->lock); |
| 365 | if (inst->skb) |
| 366 | __nfulnl_send(inst); |
| 367 | spin_unlock_bh(&inst->lock); |
| 368 | instance_put(inst); |
| 369 | } |
| 370 | |
| 371 | /* This is an inline function, we don't really care about a long |
| 372 | * list of arguments */ |
| 373 | static inline int |
| 374 | __build_packet_message(struct nfulnl_instance *inst, |
| 375 | const struct sk_buff *skb, |
| 376 | unsigned int data_len, |
| 377 | u_int8_t pf, |
| 378 | unsigned int hooknum, |
| 379 | const struct net_device *indev, |
| 380 | const struct net_device *outdev, |
| 381 | const char *prefix, unsigned int plen) |
| 382 | { |
| 383 | struct nfulnl_msg_packet_hdr pmsg; |
| 384 | struct nlmsghdr *nlh; |
| 385 | struct nfgenmsg *nfmsg; |
| 386 | sk_buff_data_t old_tail = inst->skb->tail; |
| 387 | struct sock *sk; |
| 388 | const unsigned char *hwhdrp; |
| 389 | |
| 390 | nlh = nlmsg_put(inst->skb, 0, 0, |
| 391 | NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET, |
| 392 | sizeof(struct nfgenmsg), 0); |
| 393 | if (!nlh) |
| 394 | return -1; |
| 395 | nfmsg = nlmsg_data(nlh); |
| 396 | nfmsg->nfgen_family = pf; |
| 397 | nfmsg->version = NFNETLINK_V0; |
| 398 | nfmsg->res_id = htons(inst->group_num); |
| 399 | |
| 400 | pmsg.hw_protocol = skb->protocol; |
| 401 | pmsg.hook = hooknum; |
| 402 | |
| 403 | if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg)) |
| 404 | goto nla_put_failure; |
| 405 | |
| 406 | if (prefix && |
| 407 | nla_put(inst->skb, NFULA_PREFIX, plen, prefix)) |
| 408 | goto nla_put_failure; |
| 409 | |
| 410 | if (indev) { |
| 411 | #ifndef CONFIG_BRIDGE_NETFILTER |
| 412 | if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV, |
| 413 | htonl(indev->ifindex))) |
| 414 | goto nla_put_failure; |
| 415 | #else |
| 416 | if (pf == PF_BRIDGE) { |
| 417 | /* Case 1: outdev is physical input device, we need to |
| 418 | * look for bridge group (when called from |
| 419 | * netfilter_bridge) */ |
| 420 | if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV, |
| 421 | htonl(indev->ifindex)) || |
| 422 | /* this is the bridge group "brX" */ |
| 423 | /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */ |
| 424 | nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV, |
| 425 | htonl(br_port_get_rcu(indev)->br->dev->ifindex))) |
| 426 | goto nla_put_failure; |
| 427 | } else { |
| 428 | /* Case 2: indev is bridge group, we need to look for |
| 429 | * physical device (when called from ipv4) */ |
| 430 | if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV, |
| 431 | htonl(indev->ifindex))) |
| 432 | goto nla_put_failure; |
| 433 | if (skb->nf_bridge && skb->nf_bridge->physindev && |
| 434 | nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV, |
| 435 | htonl(skb->nf_bridge->physindev->ifindex))) |
| 436 | goto nla_put_failure; |
| 437 | } |
| 438 | #endif |
| 439 | } |
| 440 | |
| 441 | if (outdev) { |
| 442 | #ifndef CONFIG_BRIDGE_NETFILTER |
| 443 | if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV, |
| 444 | htonl(outdev->ifindex))) |
| 445 | goto nla_put_failure; |
| 446 | #else |
| 447 | if (pf == PF_BRIDGE) { |
| 448 | /* Case 1: outdev is physical output device, we need to |
| 449 | * look for bridge group (when called from |
| 450 | * netfilter_bridge) */ |
| 451 | if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV, |
| 452 | htonl(outdev->ifindex)) || |
| 453 | /* this is the bridge group "brX" */ |
| 454 | /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */ |
| 455 | nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV, |
| 456 | htonl(br_port_get_rcu(outdev)->br->dev->ifindex))) |
| 457 | goto nla_put_failure; |
| 458 | } else { |
| 459 | /* Case 2: indev is a bridge group, we need to look |
| 460 | * for physical device (when called from ipv4) */ |
| 461 | if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV, |
| 462 | htonl(outdev->ifindex))) |
| 463 | goto nla_put_failure; |
| 464 | if (skb->nf_bridge && skb->nf_bridge->physoutdev && |
| 465 | nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV, |
| 466 | htonl(skb->nf_bridge->physoutdev->ifindex))) |
| 467 | goto nla_put_failure; |
| 468 | } |
| 469 | #endif |
| 470 | } |
| 471 | |
| 472 | if (skb->mark && |
| 473 | nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark))) |
| 474 | goto nla_put_failure; |
| 475 | |
| 476 | if (indev && skb->dev && |
| 477 | skb->mac_header != skb->network_header) { |
| 478 | struct nfulnl_msg_packet_hw phw; |
| 479 | int len = dev_parse_header(skb, phw.hw_addr); |
| 480 | if (len > 0) { |
| 481 | phw.hw_addrlen = htons(len); |
| 482 | if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw)) |
| 483 | goto nla_put_failure; |
| 484 | } |
| 485 | } |
| 486 | |
| 487 | if (indev && skb_mac_header_was_set(skb)) { |
| 488 | if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) || |
| 489 | nla_put_be16(inst->skb, NFULA_HWLEN, |
| 490 | htons(skb->dev->hard_header_len))) |
| 491 | goto nla_put_failure; |
| 492 | |
| 493 | hwhdrp = skb_mac_header(skb); |
| 494 | |
| 495 | if (skb->dev->type == ARPHRD_SIT) |
| 496 | hwhdrp -= ETH_HLEN; |
| 497 | |
| 498 | if (hwhdrp >= skb->head && |
| 499 | nla_put(inst->skb, NFULA_HWHEADER, |
| 500 | skb->dev->hard_header_len, hwhdrp)) |
| 501 | goto nla_put_failure; |
| 502 | } |
| 503 | |
| 504 | if (skb->tstamp.tv64) { |
| 505 | struct nfulnl_msg_packet_timestamp ts; |
| 506 | struct timeval tv = ktime_to_timeval(skb->tstamp); |
| 507 | ts.sec = cpu_to_be64(tv.tv_sec); |
| 508 | ts.usec = cpu_to_be64(tv.tv_usec); |
| 509 | |
| 510 | if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts)) |
| 511 | goto nla_put_failure; |
| 512 | } |
| 513 | |
| 514 | /* UID */ |
| 515 | sk = skb->sk; |
| 516 | if (sk && sk->sk_state != TCP_TIME_WAIT) { |
| 517 | read_lock_bh(&sk->sk_callback_lock); |
| 518 | if (sk->sk_socket && sk->sk_socket->file) { |
| 519 | struct file *file = sk->sk_socket->file; |
| 520 | const struct cred *cred = file->f_cred; |
| 521 | struct user_namespace *user_ns = inst->peer_user_ns; |
| 522 | __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid)); |
| 523 | __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid)); |
| 524 | read_unlock_bh(&sk->sk_callback_lock); |
| 525 | if (nla_put_be32(inst->skb, NFULA_UID, uid) || |
| 526 | nla_put_be32(inst->skb, NFULA_GID, gid)) |
| 527 | goto nla_put_failure; |
| 528 | } else |
| 529 | read_unlock_bh(&sk->sk_callback_lock); |
| 530 | } |
| 531 | |
| 532 | /* local sequence number */ |
| 533 | if ((inst->flags & NFULNL_CFG_F_SEQ) && |
| 534 | nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++))) |
| 535 | goto nla_put_failure; |
| 536 | |
| 537 | /* global sequence number */ |
| 538 | if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) && |
| 539 | nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL, |
| 540 | htonl(atomic_inc_return(&global_seq)))) |
| 541 | goto nla_put_failure; |
| 542 | |
| 543 | if (data_len) { |
| 544 | struct nlattr *nla; |
| 545 | int size = nla_attr_size(data_len); |
| 546 | |
| 547 | if (skb_tailroom(inst->skb) < nla_total_size(data_len)) { |
| 548 | printk(KERN_WARNING "nfnetlink_log: no tailroom!\n"); |
| 549 | return -1; |
| 550 | } |
| 551 | |
| 552 | nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len)); |
| 553 | nla->nla_type = NFULA_PAYLOAD; |
| 554 | nla->nla_len = size; |
| 555 | |
| 556 | if (skb_copy_bits(skb, 0, nla_data(nla), data_len)) |
| 557 | BUG(); |
| 558 | } |
| 559 | |
| 560 | nlh->nlmsg_len = inst->skb->tail - old_tail; |
| 561 | return 0; |
| 562 | |
| 563 | nla_put_failure: |
| 564 | PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n"); |
| 565 | return -1; |
| 566 | } |
| 567 | |
| 568 | #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0) |
| 569 | |
| 570 | static struct nf_loginfo default_loginfo = { |
| 571 | .type = NF_LOG_TYPE_ULOG, |
| 572 | .u = { |
| 573 | .ulog = { |
| 574 | .copy_len = 0xffff, |
| 575 | .group = 0, |
| 576 | .qthreshold = 1, |
| 577 | }, |
| 578 | }, |
| 579 | }; |
| 580 | |
| 581 | /* log handler for internal netfilter logging api */ |
| 582 | void |
| 583 | nfulnl_log_packet(u_int8_t pf, |
| 584 | unsigned int hooknum, |
| 585 | const struct sk_buff *skb, |
| 586 | const struct net_device *in, |
| 587 | const struct net_device *out, |
| 588 | const struct nf_loginfo *li_user, |
| 589 | const char *prefix) |
| 590 | { |
| 591 | unsigned int size, data_len; |
| 592 | struct nfulnl_instance *inst; |
| 593 | const struct nf_loginfo *li; |
| 594 | unsigned int qthreshold; |
| 595 | unsigned int plen; |
| 596 | |
| 597 | if (li_user && li_user->type == NF_LOG_TYPE_ULOG) |
| 598 | li = li_user; |
| 599 | else |
| 600 | li = &default_loginfo; |
| 601 | |
| 602 | inst = instance_lookup_get(li->u.ulog.group); |
| 603 | if (!inst) |
| 604 | return; |
| 605 | |
| 606 | plen = 0; |
| 607 | if (prefix) |
| 608 | plen = strlen(prefix) + 1; |
| 609 | |
| 610 | /* FIXME: do we want to make the size calculation conditional based on |
| 611 | * what is actually present? way more branches and checks, but more |
| 612 | * memory efficient... */ |
| 613 | size = NLMSG_SPACE(sizeof(struct nfgenmsg)) |
| 614 | + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr)) |
| 615 | + nla_total_size(sizeof(u_int32_t)) /* ifindex */ |
| 616 | + nla_total_size(sizeof(u_int32_t)) /* ifindex */ |
| 617 | #ifdef CONFIG_BRIDGE_NETFILTER |
| 618 | + nla_total_size(sizeof(u_int32_t)) /* ifindex */ |
| 619 | + nla_total_size(sizeof(u_int32_t)) /* ifindex */ |
| 620 | #endif |
| 621 | + nla_total_size(sizeof(u_int32_t)) /* mark */ |
| 622 | + nla_total_size(sizeof(u_int32_t)) /* uid */ |
| 623 | + nla_total_size(sizeof(u_int32_t)) /* gid */ |
| 624 | + nla_total_size(plen) /* prefix */ |
| 625 | + nla_total_size(sizeof(struct nfulnl_msg_packet_hw)) |
| 626 | + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp)); |
| 627 | |
| 628 | if (in && skb_mac_header_was_set(skb)) { |
| 629 | size += nla_total_size(skb->dev->hard_header_len) |
| 630 | + nla_total_size(sizeof(u_int16_t)) /* hwtype */ |
| 631 | + nla_total_size(sizeof(u_int16_t)); /* hwlen */ |
| 632 | } |
| 633 | |
| 634 | spin_lock_bh(&inst->lock); |
| 635 | |
| 636 | if (inst->flags & NFULNL_CFG_F_SEQ) |
| 637 | size += nla_total_size(sizeof(u_int32_t)); |
| 638 | if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) |
| 639 | size += nla_total_size(sizeof(u_int32_t)); |
| 640 | |
| 641 | qthreshold = inst->qthreshold; |
| 642 | /* per-rule qthreshold overrides per-instance */ |
| 643 | if (li->u.ulog.qthreshold) |
| 644 | if (qthreshold > li->u.ulog.qthreshold) |
| 645 | qthreshold = li->u.ulog.qthreshold; |
| 646 | |
| 647 | |
| 648 | switch (inst->copy_mode) { |
| 649 | case NFULNL_COPY_META: |
| 650 | case NFULNL_COPY_NONE: |
| 651 | data_len = 0; |
| 652 | break; |
| 653 | |
| 654 | case NFULNL_COPY_PACKET: |
| 655 | if (inst->copy_range == 0 |
| 656 | || inst->copy_range > skb->len) |
| 657 | data_len = skb->len; |
| 658 | else |
| 659 | data_len = inst->copy_range; |
| 660 | |
| 661 | size += nla_total_size(data_len); |
| 662 | break; |
| 663 | |
| 664 | case NFULNL_COPY_DISABLED: |
| 665 | default: |
| 666 | goto unlock_and_release; |
| 667 | } |
| 668 | |
| 669 | if (inst->skb && |
| 670 | size > skb_tailroom(inst->skb) - sizeof(struct nfgenmsg)) { |
| 671 | /* either the queue len is too high or we don't have |
| 672 | * enough room in the skb left. flush to userspace. */ |
| 673 | __nfulnl_flush(inst); |
| 674 | } |
| 675 | |
| 676 | if (!inst->skb) { |
| 677 | inst->skb = nfulnl_alloc_skb(inst->nlbufsiz, size); |
| 678 | if (!inst->skb) |
| 679 | goto alloc_failure; |
| 680 | } |
| 681 | |
| 682 | inst->qlen++; |
| 683 | |
| 684 | __build_packet_message(inst, skb, data_len, pf, |
| 685 | hooknum, in, out, prefix, plen); |
| 686 | |
| 687 | if (inst->qlen >= qthreshold) |
| 688 | __nfulnl_flush(inst); |
| 689 | /* timer_pending always called within inst->lock, so there |
| 690 | * is no chance of a race here */ |
| 691 | else if (!timer_pending(&inst->timer)) { |
| 692 | instance_get(inst); |
| 693 | inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100); |
| 694 | add_timer(&inst->timer); |
| 695 | } |
| 696 | |
| 697 | unlock_and_release: |
| 698 | spin_unlock_bh(&inst->lock); |
| 699 | instance_put(inst); |
| 700 | return; |
| 701 | |
| 702 | alloc_failure: |
| 703 | /* FIXME: statistics */ |
| 704 | goto unlock_and_release; |
| 705 | } |
| 706 | EXPORT_SYMBOL_GPL(nfulnl_log_packet); |
| 707 | |
| 708 | static int |
| 709 | nfulnl_rcv_nl_event(struct notifier_block *this, |
| 710 | unsigned long event, void *ptr) |
| 711 | { |
| 712 | struct netlink_notify *n = ptr; |
| 713 | |
| 714 | if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) { |
| 715 | int i; |
| 716 | |
| 717 | /* destroy all instances for this portid */ |
| 718 | spin_lock_bh(&instances_lock); |
| 719 | for (i = 0; i < INSTANCE_BUCKETS; i++) { |
| 720 | struct hlist_node *tmp, *t2; |
| 721 | struct nfulnl_instance *inst; |
| 722 | struct hlist_head *head = &instance_table[i]; |
| 723 | |
| 724 | hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) { |
| 725 | if ((net_eq(n->net, &init_net)) && |
| 726 | (n->portid == inst->peer_portid)) |
| 727 | __instance_destroy(inst); |
| 728 | } |
| 729 | } |
| 730 | spin_unlock_bh(&instances_lock); |
| 731 | } |
| 732 | return NOTIFY_DONE; |
| 733 | } |
| 734 | |
| 735 | static struct notifier_block nfulnl_rtnl_notifier = { |
| 736 | .notifier_call = nfulnl_rcv_nl_event, |
| 737 | }; |
| 738 | |
| 739 | static int |
| 740 | nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb, |
| 741 | const struct nlmsghdr *nlh, |
| 742 | const struct nlattr * const nfqa[]) |
| 743 | { |
| 744 | return -ENOTSUPP; |
| 745 | } |
| 746 | |
| 747 | static struct nf_logger nfulnl_logger __read_mostly = { |
| 748 | .name = "nfnetlink_log", |
| 749 | .logfn = &nfulnl_log_packet, |
| 750 | .me = THIS_MODULE, |
| 751 | }; |
| 752 | |
| 753 | static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = { |
| 754 | [NFULA_CFG_CMD] = { .len = sizeof(struct nfulnl_msg_config_cmd) }, |
| 755 | [NFULA_CFG_MODE] = { .len = sizeof(struct nfulnl_msg_config_mode) }, |
| 756 | [NFULA_CFG_TIMEOUT] = { .type = NLA_U32 }, |
| 757 | [NFULA_CFG_QTHRESH] = { .type = NLA_U32 }, |
| 758 | [NFULA_CFG_NLBUFSIZ] = { .type = NLA_U32 }, |
| 759 | [NFULA_CFG_FLAGS] = { .type = NLA_U16 }, |
| 760 | }; |
| 761 | |
| 762 | static int |
| 763 | nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb, |
| 764 | const struct nlmsghdr *nlh, |
| 765 | const struct nlattr * const nfula[]) |
| 766 | { |
| 767 | struct nfgenmsg *nfmsg = nlmsg_data(nlh); |
| 768 | u_int16_t group_num = ntohs(nfmsg->res_id); |
| 769 | struct nfulnl_instance *inst; |
| 770 | struct nfulnl_msg_config_cmd *cmd = NULL; |
| 771 | int ret = 0; |
| 772 | |
| 773 | if (nfula[NFULA_CFG_CMD]) { |
| 774 | u_int8_t pf = nfmsg->nfgen_family; |
| 775 | cmd = nla_data(nfula[NFULA_CFG_CMD]); |
| 776 | |
| 777 | /* Commands without queue context */ |
| 778 | switch (cmd->command) { |
| 779 | case NFULNL_CFG_CMD_PF_BIND: |
| 780 | return nf_log_bind_pf(pf, &nfulnl_logger); |
| 781 | case NFULNL_CFG_CMD_PF_UNBIND: |
| 782 | nf_log_unbind_pf(pf); |
| 783 | return 0; |
| 784 | } |
| 785 | } |
| 786 | |
| 787 | inst = instance_lookup_get(group_num); |
| 788 | if (inst && inst->peer_portid != NETLINK_CB(skb).portid) { |
| 789 | ret = -EPERM; |
| 790 | goto out_put; |
| 791 | } |
| 792 | |
| 793 | if (cmd != NULL) { |
| 794 | switch (cmd->command) { |
| 795 | case NFULNL_CFG_CMD_BIND: |
| 796 | if (inst) { |
| 797 | ret = -EBUSY; |
| 798 | goto out_put; |
| 799 | } |
| 800 | |
| 801 | inst = instance_create(group_num, |
| 802 | NETLINK_CB(skb).portid, |
| 803 | sk_user_ns(NETLINK_CB(skb).ssk)); |
| 804 | if (IS_ERR(inst)) { |
| 805 | ret = PTR_ERR(inst); |
| 806 | goto out; |
| 807 | } |
| 808 | break; |
| 809 | case NFULNL_CFG_CMD_UNBIND: |
| 810 | if (!inst) { |
| 811 | ret = -ENODEV; |
| 812 | goto out; |
| 813 | } |
| 814 | |
| 815 | instance_destroy(inst); |
| 816 | goto out_put; |
| 817 | default: |
| 818 | ret = -ENOTSUPP; |
| 819 | break; |
| 820 | } |
| 821 | } |
| 822 | |
| 823 | if (nfula[NFULA_CFG_MODE]) { |
| 824 | struct nfulnl_msg_config_mode *params; |
| 825 | params = nla_data(nfula[NFULA_CFG_MODE]); |
| 826 | |
| 827 | if (!inst) { |
| 828 | ret = -ENODEV; |
| 829 | goto out; |
| 830 | } |
| 831 | nfulnl_set_mode(inst, params->copy_mode, |
| 832 | ntohl(params->copy_range)); |
| 833 | } |
| 834 | |
| 835 | if (nfula[NFULA_CFG_TIMEOUT]) { |
| 836 | __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]); |
| 837 | |
| 838 | if (!inst) { |
| 839 | ret = -ENODEV; |
| 840 | goto out; |
| 841 | } |
| 842 | nfulnl_set_timeout(inst, ntohl(timeout)); |
| 843 | } |
| 844 | |
| 845 | if (nfula[NFULA_CFG_NLBUFSIZ]) { |
| 846 | __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]); |
| 847 | |
| 848 | if (!inst) { |
| 849 | ret = -ENODEV; |
| 850 | goto out; |
| 851 | } |
| 852 | nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz)); |
| 853 | } |
| 854 | |
| 855 | if (nfula[NFULA_CFG_QTHRESH]) { |
| 856 | __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]); |
| 857 | |
| 858 | if (!inst) { |
| 859 | ret = -ENODEV; |
| 860 | goto out; |
| 861 | } |
| 862 | nfulnl_set_qthresh(inst, ntohl(qthresh)); |
| 863 | } |
| 864 | |
| 865 | if (nfula[NFULA_CFG_FLAGS]) { |
| 866 | __be16 flags = nla_get_be16(nfula[NFULA_CFG_FLAGS]); |
| 867 | |
| 868 | if (!inst) { |
| 869 | ret = -ENODEV; |
| 870 | goto out; |
| 871 | } |
| 872 | nfulnl_set_flags(inst, ntohs(flags)); |
| 873 | } |
| 874 | |
| 875 | out_put: |
| 876 | instance_put(inst); |
| 877 | out: |
| 878 | return ret; |
| 879 | } |
| 880 | |
| 881 | static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = { |
| 882 | [NFULNL_MSG_PACKET] = { .call = nfulnl_recv_unsupp, |
| 883 | .attr_count = NFULA_MAX, }, |
| 884 | [NFULNL_MSG_CONFIG] = { .call = nfulnl_recv_config, |
| 885 | .attr_count = NFULA_CFG_MAX, |
| 886 | .policy = nfula_cfg_policy }, |
| 887 | }; |
| 888 | |
| 889 | static const struct nfnetlink_subsystem nfulnl_subsys = { |
| 890 | .name = "log", |
| 891 | .subsys_id = NFNL_SUBSYS_ULOG, |
| 892 | .cb_count = NFULNL_MSG_MAX, |
| 893 | .cb = nfulnl_cb, |
| 894 | }; |
| 895 | |
| 896 | #ifdef CONFIG_PROC_FS |
| 897 | struct iter_state { |
| 898 | unsigned int bucket; |
| 899 | }; |
| 900 | |
| 901 | static struct hlist_node *get_first(struct iter_state *st) |
| 902 | { |
| 903 | if (!st) |
| 904 | return NULL; |
| 905 | |
| 906 | for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) { |
| 907 | if (!hlist_empty(&instance_table[st->bucket])) |
| 908 | return rcu_dereference_bh(hlist_first_rcu(&instance_table[st->bucket])); |
| 909 | } |
| 910 | return NULL; |
| 911 | } |
| 912 | |
| 913 | static struct hlist_node *get_next(struct iter_state *st, struct hlist_node *h) |
| 914 | { |
| 915 | h = rcu_dereference_bh(hlist_next_rcu(h)); |
| 916 | while (!h) { |
| 917 | if (++st->bucket >= INSTANCE_BUCKETS) |
| 918 | return NULL; |
| 919 | |
| 920 | h = rcu_dereference_bh(hlist_first_rcu(&instance_table[st->bucket])); |
| 921 | } |
| 922 | return h; |
| 923 | } |
| 924 | |
| 925 | static struct hlist_node *get_idx(struct iter_state *st, loff_t pos) |
| 926 | { |
| 927 | struct hlist_node *head; |
| 928 | head = get_first(st); |
| 929 | |
| 930 | if (head) |
| 931 | while (pos && (head = get_next(st, head))) |
| 932 | pos--; |
| 933 | return pos ? NULL : head; |
| 934 | } |
| 935 | |
| 936 | static void *seq_start(struct seq_file *seq, loff_t *pos) |
| 937 | __acquires(rcu_bh) |
| 938 | { |
| 939 | rcu_read_lock_bh(); |
| 940 | return get_idx(seq->private, *pos); |
| 941 | } |
| 942 | |
| 943 | static void *seq_next(struct seq_file *s, void *v, loff_t *pos) |
| 944 | { |
| 945 | (*pos)++; |
| 946 | return get_next(s->private, v); |
| 947 | } |
| 948 | |
| 949 | static void seq_stop(struct seq_file *s, void *v) |
| 950 | __releases(rcu_bh) |
| 951 | { |
| 952 | rcu_read_unlock_bh(); |
| 953 | } |
| 954 | |
| 955 | static int seq_show(struct seq_file *s, void *v) |
| 956 | { |
| 957 | const struct nfulnl_instance *inst = v; |
| 958 | |
| 959 | return seq_printf(s, "%5d %6d %5d %1d %5d %6d %2d\n", |
| 960 | inst->group_num, |
| 961 | inst->peer_portid, inst->qlen, |
| 962 | inst->copy_mode, inst->copy_range, |
| 963 | inst->flushtimeout, atomic_read(&inst->use)); |
| 964 | } |
| 965 | |
| 966 | static const struct seq_operations nful_seq_ops = { |
| 967 | .start = seq_start, |
| 968 | .next = seq_next, |
| 969 | .stop = seq_stop, |
| 970 | .show = seq_show, |
| 971 | }; |
| 972 | |
| 973 | static int nful_open(struct inode *inode, struct file *file) |
| 974 | { |
| 975 | return seq_open_private(file, &nful_seq_ops, |
| 976 | sizeof(struct iter_state)); |
| 977 | } |
| 978 | |
| 979 | static const struct file_operations nful_file_ops = { |
| 980 | .owner = THIS_MODULE, |
| 981 | .open = nful_open, |
| 982 | .read = seq_read, |
| 983 | .llseek = seq_lseek, |
| 984 | .release = seq_release_private, |
| 985 | }; |
| 986 | |
| 987 | #endif /* PROC_FS */ |
| 988 | |
| 989 | static int __init nfnetlink_log_init(void) |
| 990 | { |
| 991 | int i, status = -ENOMEM; |
| 992 | |
| 993 | for (i = 0; i < INSTANCE_BUCKETS; i++) |
| 994 | INIT_HLIST_HEAD(&instance_table[i]); |
| 995 | |
| 996 | /* it's not really all that important to have a random value, so |
| 997 | * we can do this from the init function, even if there hasn't |
| 998 | * been that much entropy yet */ |
| 999 | get_random_bytes(&hash_init, sizeof(hash_init)); |
| 1000 | |
| 1001 | netlink_register_notifier(&nfulnl_rtnl_notifier); |
| 1002 | status = nfnetlink_subsys_register(&nfulnl_subsys); |
| 1003 | if (status < 0) { |
| 1004 | printk(KERN_ERR "log: failed to create netlink socket\n"); |
| 1005 | goto cleanup_netlink_notifier; |
| 1006 | } |
| 1007 | |
| 1008 | status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger); |
| 1009 | if (status < 0) { |
| 1010 | printk(KERN_ERR "log: failed to register logger\n"); |
| 1011 | goto cleanup_subsys; |
| 1012 | } |
| 1013 | |
| 1014 | #ifdef CONFIG_PROC_FS |
| 1015 | if (!proc_create("nfnetlink_log", 0440, |
| 1016 | proc_net_netfilter, &nful_file_ops)) { |
| 1017 | status = -ENOMEM; |
| 1018 | goto cleanup_logger; |
| 1019 | } |
| 1020 | #endif |
| 1021 | return status; |
| 1022 | |
| 1023 | #ifdef CONFIG_PROC_FS |
| 1024 | cleanup_logger: |
| 1025 | nf_log_unregister(&nfulnl_logger); |
| 1026 | #endif |
| 1027 | cleanup_subsys: |
| 1028 | nfnetlink_subsys_unregister(&nfulnl_subsys); |
| 1029 | cleanup_netlink_notifier: |
| 1030 | netlink_unregister_notifier(&nfulnl_rtnl_notifier); |
| 1031 | return status; |
| 1032 | } |
| 1033 | |
| 1034 | static void __exit nfnetlink_log_fini(void) |
| 1035 | { |
| 1036 | nf_log_unregister(&nfulnl_logger); |
| 1037 | #ifdef CONFIG_PROC_FS |
| 1038 | remove_proc_entry("nfnetlink_log", proc_net_netfilter); |
| 1039 | #endif |
| 1040 | nfnetlink_subsys_unregister(&nfulnl_subsys); |
| 1041 | netlink_unregister_notifier(&nfulnl_rtnl_notifier); |
| 1042 | } |
| 1043 | |
| 1044 | MODULE_DESCRIPTION("netfilter userspace logging"); |
| 1045 | MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); |
| 1046 | MODULE_LICENSE("GPL"); |
| 1047 | MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG); |
| 1048 | |
| 1049 | module_init(nfnetlink_log_init); |
| 1050 | module_exit(nfnetlink_log_fini); |