]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - net/netfilter/nfnetlink_log.c
netfilter: add my copyright statements
[mirror_ubuntu-hirsute-kernel.git] / net / netfilter / nfnetlink_log.c
CommitLineData
0597f268
HW
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>
f229f6ce 6 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
0597f268
HW
7 *
8 * Based on the old ipv4-only ipt_ULOG.c:
9 * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
10 *
11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License version 2 as
13 * published by the Free Software Foundation.
0597f268
HW
14 */
15#include <linux/module.h>
16#include <linux/skbuff.h>
e035edd1 17#include <linux/if_arp.h>
0597f268
HW
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>
573ce260 23#include <net/netlink.h>
0597f268
HW
24#include <linux/netfilter/nfnetlink.h>
25#include <linux/netfilter/nfnetlink_log.h>
26#include <linux/spinlock.h>
27#include <linux/sysctl.h>
28#include <linux/proc_fs.h>
29#include <linux/security.h>
30#include <linux/list.h>
31#include <linux/jhash.h>
32#include <linux/random.h>
5a0e3ad6 33#include <linux/slab.h>
0597f268 34#include <net/sock.h>
f01ffbd6 35#include <net/netfilter/nf_log.h>
9368a53c 36#include <net/netns/generic.h>
d9e15007 37#include <net/netfilter/nfnetlink_log.h>
0597f268 38
60063497 39#include <linux/atomic.h>
0597f268 40
fbcd923c
HW
41#ifdef CONFIG_BRIDGE_NETFILTER
42#include "../bridge/br_private.h"
43#endif
44
c2db2924 45#define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
2c6764b7 46#define NFULNL_TIMEOUT_DEFAULT 100 /* every second */
0597f268 47#define NFULNL_QTHRESH_DEFAULT 100 /* 100 packets */
6b6ec99a 48#define NFULNL_COPY_RANGE_MAX 0xFFFF /* max packet size is limited by 16-bit struct nfattr nfa_len field */
0597f268
HW
49
50#define PRINTR(x, args...) do { if (net_ratelimit()) \
51 printk(x, ## args); } while (0);
52
0597f268
HW
53struct nfulnl_instance {
54 struct hlist_node hlist; /* global list of instances */
55 spinlock_t lock;
56 atomic_t use; /* use count */
57
58 unsigned int qlen; /* number of nlmsgs in skb */
59 struct sk_buff *skb; /* pre-allocatd skb */
0597f268 60 struct timer_list timer;
9368a53c 61 struct net *net;
9eea9515 62 struct user_namespace *peer_user_ns; /* User namespace of the peer process */
15e47304 63 int peer_portid; /* PORTID of the peer process */
0597f268
HW
64
65 /* configurable parameters */
66 unsigned int flushtimeout; /* timeout until queue flush */
67 unsigned int nlbufsiz; /* netlink buffer allocation size */
68 unsigned int qthreshold; /* threshold of the queue */
69 u_int32_t copy_range;
0af5f6c1 70 u_int32_t seq; /* instance-local sequential counter */
0597f268 71 u_int16_t group_num; /* number of this queue */
0af5f6c1 72 u_int16_t flags;
601e68e1 73 u_int8_t copy_mode;
bed1be20 74 struct rcu_head rcu;
0597f268
HW
75};
76
0597f268 77#define INSTANCE_BUCKETS 16
0597f268
HW
78static unsigned int hash_init;
79
9368a53c
G
80static int nfnl_log_net_id __read_mostly;
81
82struct nfnl_log_net {
83 spinlock_t instances_lock;
84 struct hlist_head instance_table[INSTANCE_BUCKETS];
85 atomic_t global_seq;
86};
87
88static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
89{
90 return net_generic(net, nfnl_log_net_id);
91}
92
0597f268
HW
93static inline u_int8_t instance_hashfn(u_int16_t group_num)
94{
95 return ((group_num & 0xff) % INSTANCE_BUCKETS);
96}
97
98static struct nfulnl_instance *
9368a53c 99__instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
0597f268
HW
100{
101 struct hlist_head *head;
0597f268
HW
102 struct nfulnl_instance *inst;
103
9368a53c 104 head = &log->instance_table[instance_hashfn(group_num)];
b67bfe0d 105 hlist_for_each_entry_rcu(inst, head, hlist) {
0597f268
HW
106 if (inst->group_num == group_num)
107 return inst;
108 }
109 return NULL;
110}
111
112static inline void
113instance_get(struct nfulnl_instance *inst)
114{
115 atomic_inc(&inst->use);
116}
117
118static struct nfulnl_instance *
9368a53c 119instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
0597f268
HW
120{
121 struct nfulnl_instance *inst;
122
bed1be20 123 rcu_read_lock_bh();
9368a53c 124 inst = __instance_lookup(log, group_num);
f5c5440d
ED
125 if (inst && !atomic_inc_not_zero(&inst->use))
126 inst = NULL;
bed1be20 127 rcu_read_unlock_bh();
0597f268
HW
128
129 return inst;
130}
131
bed1be20
ED
132static void nfulnl_instance_free_rcu(struct rcu_head *head)
133{
9368a53c
G
134 struct nfulnl_instance *inst =
135 container_of(head, struct nfulnl_instance, rcu);
136
137 put_net(inst->net);
138 kfree(inst);
bed1be20
ED
139 module_put(THIS_MODULE);
140}
141
0597f268
HW
142static void
143instance_put(struct nfulnl_instance *inst)
144{
bed1be20
ED
145 if (inst && atomic_dec_and_test(&inst->use))
146 call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
0597f268
HW
147}
148
149static void nfulnl_timer(unsigned long data);
150
151static struct nfulnl_instance *
9368a53c
G
152instance_create(struct net *net, u_int16_t group_num,
153 int portid, struct user_namespace *user_ns)
0597f268
HW
154{
155 struct nfulnl_instance *inst;
9368a53c 156 struct nfnl_log_net *log = nfnl_log_pernet(net);
baab2ce7 157 int err;
0597f268 158
9368a53c
G
159 spin_lock_bh(&log->instances_lock);
160 if (__instance_lookup(log, group_num)) {
baab2ce7 161 err = -EEXIST;
0597f268
HW
162 goto out_unlock;
163 }
164
10dfdc69 165 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
baab2ce7
PM
166 if (!inst) {
167 err = -ENOMEM;
0597f268 168 goto out_unlock;
baab2ce7 169 }
0597f268 170
aace57e0
MM
171 if (!try_module_get(THIS_MODULE)) {
172 kfree(inst);
baab2ce7 173 err = -EAGAIN;
aace57e0
MM
174 goto out_unlock;
175 }
176
0597f268 177 INIT_HLIST_NODE(&inst->hlist);
181a46a5 178 spin_lock_init(&inst->lock);
0597f268
HW
179 /* needs to be two, since we _put() after creation */
180 atomic_set(&inst->use, 2);
181
e6f689db 182 setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
0597f268 183
9368a53c 184 inst->net = get_net(net);
9eea9515 185 inst->peer_user_ns = user_ns;
15e47304 186 inst->peer_portid = portid;
0597f268
HW
187 inst->group_num = group_num;
188
189 inst->qthreshold = NFULNL_QTHRESH_DEFAULT;
190 inst->flushtimeout = NFULNL_TIMEOUT_DEFAULT;
191 inst->nlbufsiz = NFULNL_NLBUFSIZ_DEFAULT;
192 inst->copy_mode = NFULNL_COPY_PACKET;
6b6ec99a 193 inst->copy_range = NFULNL_COPY_RANGE_MAX;
0597f268 194
f5c5440d 195 hlist_add_head_rcu(&inst->hlist,
9368a53c
G
196 &log->instance_table[instance_hashfn(group_num)]);
197
0597f268 198
9368a53c 199 spin_unlock_bh(&log->instances_lock);
0597f268
HW
200
201 return inst;
202
0597f268 203out_unlock:
9368a53c 204 spin_unlock_bh(&log->instances_lock);
baab2ce7 205 return ERR_PTR(err);
0597f268
HW
206}
207
e3567061 208static void __nfulnl_flush(struct nfulnl_instance *inst);
0597f268 209
f5c5440d 210/* called with BH disabled */
0597f268 211static void
9afdb00c 212__instance_destroy(struct nfulnl_instance *inst)
0597f268
HW
213{
214 /* first pull it out of the global list */
f5c5440d 215 hlist_del_rcu(&inst->hlist);
0597f268 216
0597f268
HW
217 /* then flush all pending packets from skb */
218
f5c5440d
ED
219 spin_lock(&inst->lock);
220
221 /* lockless readers wont be able to use us */
222 inst->copy_mode = NFULNL_COPY_DISABLED;
223
e3567061
MM
224 if (inst->skb)
225 __nfulnl_flush(inst);
f5c5440d 226 spin_unlock(&inst->lock);
0597f268
HW
227
228 /* and finally put the refcount */
229 instance_put(inst);
0597f268
HW
230}
231
0597f268 232static inline void
9368a53c
G
233instance_destroy(struct nfnl_log_net *log,
234 struct nfulnl_instance *inst)
0597f268 235{
9368a53c 236 spin_lock_bh(&log->instances_lock);
9afdb00c 237 __instance_destroy(inst);
9368a53c 238 spin_unlock_bh(&log->instances_lock);
0597f268
HW
239}
240
241static int
242nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
243 unsigned int range)
244{
245 int status = 0;
246
247 spin_lock_bh(&inst->lock);
601e68e1 248
0597f268
HW
249 switch (mode) {
250 case NFULNL_COPY_NONE:
251 case NFULNL_COPY_META:
252 inst->copy_mode = mode;
253 inst->copy_range = 0;
254 break;
601e68e1 255
0597f268
HW
256 case NFULNL_COPY_PACKET:
257 inst->copy_mode = mode;
6b6ec99a
MM
258 inst->copy_range = min_t(unsigned int,
259 range, NFULNL_COPY_RANGE_MAX);
0597f268 260 break;
601e68e1 261
0597f268
HW
262 default:
263 status = -EINVAL;
264 break;
265 }
266
267 spin_unlock_bh(&inst->lock);
268
269 return status;
270}
271
272static int
273nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
274{
275 int status;
276
277 spin_lock_bh(&inst->lock);
278 if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
279 status = -ERANGE;
280 else if (nlbufsiz > 131072)
281 status = -ERANGE;
282 else {
283 inst->nlbufsiz = nlbufsiz;
284 status = 0;
285 }
286 spin_unlock_bh(&inst->lock);
287
288 return status;
289}
290
291static int
292nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
293{
294 spin_lock_bh(&inst->lock);
295 inst->flushtimeout = timeout;
296 spin_unlock_bh(&inst->lock);
297
298 return 0;
299}
300
301static int
302nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
303{
304 spin_lock_bh(&inst->lock);
305 inst->qthreshold = qthresh;
306 spin_unlock_bh(&inst->lock);
307
308 return 0;
309}
310
0af5f6c1
HW
311static int
312nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
313{
314 spin_lock_bh(&inst->lock);
ee433530 315 inst->flags = flags;
0af5f6c1
HW
316 spin_unlock_bh(&inst->lock);
317
318 return 0;
319}
320
c6a8f648
MM
321static struct sk_buff *
322nfulnl_alloc_skb(unsigned int inst_size, unsigned int pkt_size)
0597f268
HW
323{
324 struct sk_buff *skb;
ad2ad0f9 325 unsigned int n;
0597f268 326
0597f268
HW
327 /* alloc skb which should be big enough for a whole multipart
328 * message. WARNING: has to be <= 128k due to slab restrictions */
329
ad2ad0f9
PM
330 n = max(inst_size, pkt_size);
331 skb = alloc_skb(n, GFP_ATOMIC);
0597f268 332 if (!skb) {
ad2ad0f9
PM
333 if (n > pkt_size) {
334 /* try to allocate only as much as we need for current
335 * packet */
0597f268 336
ad2ad0f9
PM
337 skb = alloc_skb(pkt_size, GFP_ATOMIC);
338 if (!skb)
0a9ee813
JP
339 pr_err("nfnetlink_log: can't even alloc %u bytes\n",
340 pkt_size);
ad2ad0f9 341 }
0597f268
HW
342 }
343
344 return skb;
345}
346
347static int
348__nfulnl_send(struct nfulnl_instance *inst)
349{
29c5d4af 350 int status = -1;
0597f268 351
d550d095
DM
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 (!nlh)
358 goto out;
359 }
9368a53c 360 status = nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
cd8c20b6 361 MSG_DONTWAIT);
0597f268
HW
362
363 inst->qlen = 0;
364 inst->skb = NULL;
d550d095 365out:
0597f268
HW
366 return status;
367}
368
e3567061
MM
369static 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
c6a8f648
MM
379static void
380nfulnl_timer(unsigned long data)
0597f268 381{
601e68e1 382 struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
0597f268 383
0597f268 384 spin_lock_bh(&inst->lock);
370e6a87
MM
385 if (inst->skb)
386 __nfulnl_send(inst);
0597f268 387 spin_unlock_bh(&inst->lock);
05f7b7b3 388 instance_put(inst);
0597f268
HW
389}
390
0af5f6c1
HW
391/* This is an inline function, we don't really care about a long
392 * list of arguments */
601e68e1 393static inline int
9368a53c
G
394__build_packet_message(struct nfnl_log_net *log,
395 struct nfulnl_instance *inst,
601e68e1 396 const struct sk_buff *skb,
0597f268 397 unsigned int data_len,
76108cea 398 u_int8_t pf,
0597f268
HW
399 unsigned int hooknum,
400 const struct net_device *indev,
401 const struct net_device *outdev,
d7a5c324 402 const char *prefix, unsigned int plen)
0597f268 403{
0597f268
HW
404 struct nfulnl_msg_packet_hdr pmsg;
405 struct nlmsghdr *nlh;
406 struct nfgenmsg *nfmsg;
27a884dc 407 sk_buff_data_t old_tail = inst->skb->tail;
0626af31 408 struct sock *sk;
0c36b48b 409 const unsigned char *hwhdrp;
0597f268 410
d550d095 411 nlh = nlmsg_put(inst->skb, 0, 0,
0597f268 412 NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
d550d095
DM
413 sizeof(struct nfgenmsg), 0);
414 if (!nlh)
415 return -1;
416 nfmsg = nlmsg_data(nlh);
0597f268
HW
417 nfmsg->nfgen_family = pf;
418 nfmsg->version = NFNETLINK_V0;
419 nfmsg->res_id = htons(inst->group_num);
420
febf0a43 421 pmsg.hw_protocol = skb->protocol;
0597f268
HW
422 pmsg.hook = hooknum;
423
1db20a52
DM
424 if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
425 goto nla_put_failure;
0597f268 426
1db20a52
DM
427 if (prefix &&
428 nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
429 goto nla_put_failure;
0597f268
HW
430
431 if (indev) {
fbcd923c 432#ifndef CONFIG_BRIDGE_NETFILTER
1db20a52
DM
433 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
434 htonl(indev->ifindex)))
435 goto nla_put_failure;
fbcd923c
HW
436#else
437 if (pf == PF_BRIDGE) {
438 /* Case 1: outdev is physical input device, we need to
439 * look for bridge group (when called from
440 * netfilter_bridge) */
1db20a52
DM
441 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
442 htonl(indev->ifindex)) ||
fbcd923c 443 /* this is the bridge group "brX" */
f350a0a8 444 /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
1db20a52
DM
445 nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
446 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
447 goto nla_put_failure;
fbcd923c
HW
448 } else {
449 /* Case 2: indev is bridge group, we need to look for
450 * physical device (when called from ipv4) */
1db20a52
DM
451 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
452 htonl(indev->ifindex)))
453 goto nla_put_failure;
454 if (skb->nf_bridge && skb->nf_bridge->physindev &&
455 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
456 htonl(skb->nf_bridge->physindev->ifindex)))
457 goto nla_put_failure;
fbcd923c
HW
458 }
459#endif
0597f268
HW
460 }
461
462 if (outdev) {
fbcd923c 463#ifndef CONFIG_BRIDGE_NETFILTER
1db20a52
DM
464 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
465 htonl(outdev->ifindex)))
466 goto nla_put_failure;
fbcd923c
HW
467#else
468 if (pf == PF_BRIDGE) {
469 /* Case 1: outdev is physical output device, we need to
470 * look for bridge group (when called from
471 * netfilter_bridge) */
1db20a52
DM
472 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
473 htonl(outdev->ifindex)) ||
fbcd923c 474 /* this is the bridge group "brX" */
f350a0a8 475 /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
1db20a52
DM
476 nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
477 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
478 goto nla_put_failure;
fbcd923c
HW
479 } else {
480 /* Case 2: indev is a bridge group, we need to look
481 * for physical device (when called from ipv4) */
1db20a52
DM
482 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
483 htonl(outdev->ifindex)))
484 goto nla_put_failure;
485 if (skb->nf_bridge && skb->nf_bridge->physoutdev &&
486 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
487 htonl(skb->nf_bridge->physoutdev->ifindex)))
488 goto nla_put_failure;
fbcd923c
HW
489 }
490#endif
0597f268
HW
491 }
492
1db20a52
DM
493 if (skb->mark &&
494 nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
495 goto nla_put_failure;
0597f268 496
2c38de4c
NC
497 if (indev && skb->dev &&
498 skb->mac_header != skb->network_header) {
0597f268 499 struct nfulnl_msg_packet_hw phw;
b95cce35
SH
500 int len = dev_parse_header(skb, phw.hw_addr);
501 if (len > 0) {
502 phw.hw_addrlen = htons(len);
1db20a52
DM
503 if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
504 goto nla_put_failure;
b95cce35 505 }
0597f268
HW
506 }
507
72961ecf 508 if (indev && skb_mac_header_was_set(skb)) {
2dba62c3 509 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
1db20a52 510 nla_put_be16(inst->skb, NFULA_HWLEN,
0c36b48b
BH
511 htons(skb->dev->hard_header_len)))
512 goto nla_put_failure;
513
514 hwhdrp = skb_mac_header(skb);
515
516 if (skb->dev->type == ARPHRD_SIT)
517 hwhdrp -= ETH_HLEN;
518
519 if (hwhdrp >= skb->head &&
520 nla_put(inst->skb, NFULA_HWHEADER,
521 skb->dev->hard_header_len, hwhdrp))
1db20a52 522 goto nla_put_failure;
72961ecf
EL
523 }
524
b7aa0bf7 525 if (skb->tstamp.tv64) {
0597f268 526 struct nfulnl_msg_packet_timestamp ts;
b7aa0bf7
ED
527 struct timeval tv = ktime_to_timeval(skb->tstamp);
528 ts.sec = cpu_to_be64(tv.tv_sec);
529 ts.usec = cpu_to_be64(tv.tv_usec);
0597f268 530
1db20a52
DM
531 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
532 goto nla_put_failure;
0597f268
HW
533 }
534
535 /* UID */
0626af31
ED
536 sk = skb->sk;
537 if (sk && sk->sk_state != TCP_TIME_WAIT) {
538 read_lock_bh(&sk->sk_callback_lock);
539 if (sk->sk_socket && sk->sk_socket->file) {
540 struct file *file = sk->sk_socket->file;
437589a7
LT
541 const struct cred *cred = file->f_cred;
542 struct user_namespace *user_ns = inst->peer_user_ns;
543 __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
544 __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
0626af31 545 read_unlock_bh(&sk->sk_callback_lock);
1db20a52
DM
546 if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
547 nla_put_be32(inst->skb, NFULA_GID, gid))
548 goto nla_put_failure;
0597f268 549 } else
0626af31 550 read_unlock_bh(&sk->sk_callback_lock);
0597f268
HW
551 }
552
0af5f6c1 553 /* local sequence number */
1db20a52
DM
554 if ((inst->flags & NFULNL_CFG_F_SEQ) &&
555 nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
556 goto nla_put_failure;
0dfedd28 557
0af5f6c1 558 /* global sequence number */
1db20a52
DM
559 if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
560 nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
9368a53c 561 htonl(atomic_inc_return(&log->global_seq))))
1db20a52 562 goto nla_put_failure;
0af5f6c1 563
0597f268 564 if (data_len) {
df6fb868
PM
565 struct nlattr *nla;
566 int size = nla_attr_size(data_len);
0597f268 567
df6fb868 568 if (skb_tailroom(inst->skb) < nla_total_size(data_len)) {
0597f268 569 printk(KERN_WARNING "nfnetlink_log: no tailroom!\n");
d550d095 570 return -1;
0597f268
HW
571 }
572
df6fb868
PM
573 nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
574 nla->nla_type = NFULA_PAYLOAD;
575 nla->nla_len = size;
0597f268 576
df6fb868 577 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
0597f268
HW
578 BUG();
579 }
601e68e1 580
0597f268
HW
581 nlh->nlmsg_len = inst->skb->tail - old_tail;
582 return 0;
583
df6fb868 584nla_put_failure:
0597f268
HW
585 PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
586 return -1;
587}
588
589#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
590
591static struct nf_loginfo default_loginfo = {
592 .type = NF_LOG_TYPE_ULOG,
593 .u = {
594 .ulog = {
595 .copy_len = 0xffff,
596 .group = 0,
597 .qthreshold = 1,
598 },
599 },
600};
601
602/* log handler for internal netfilter logging api */
5f7340ef 603void
76108cea 604nfulnl_log_packet(u_int8_t pf,
0597f268
HW
605 unsigned int hooknum,
606 const struct sk_buff *skb,
607 const struct net_device *in,
608 const struct net_device *out,
609 const struct nf_loginfo *li_user,
610 const char *prefix)
611{
612 unsigned int size, data_len;
613 struct nfulnl_instance *inst;
614 const struct nf_loginfo *li;
615 unsigned int qthreshold;
d7a5c324 616 unsigned int plen;
9368a53c
G
617 struct net *net = dev_net(in ? in : out);
618 struct nfnl_log_net *log = nfnl_log_pernet(net);
0597f268 619
601e68e1 620 if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
0597f268
HW
621 li = li_user;
622 else
623 li = &default_loginfo;
624
9368a53c 625 inst = instance_lookup_get(log, li->u.ulog.group);
0597f268 626 if (!inst)
0597f268 627 return;
0597f268 628
d7a5c324
PM
629 plen = 0;
630 if (prefix)
881dbfe8 631 plen = strlen(prefix) + 1;
d7a5c324 632
0597f268
HW
633 /* FIXME: do we want to make the size calculation conditional based on
634 * what is actually present? way more branches and checks, but more
635 * memory efficient... */
573ce260 636 size = nlmsg_total_size(sizeof(struct nfgenmsg))
df6fb868
PM
637 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
638 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
639 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
fbcd923c 640#ifdef CONFIG_BRIDGE_NETFILTER
df6fb868
PM
641 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
642 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
fbcd923c 643#endif
df6fb868
PM
644 + nla_total_size(sizeof(u_int32_t)) /* mark */
645 + nla_total_size(sizeof(u_int32_t)) /* uid */
76aa1ce1 646 + nla_total_size(sizeof(u_int32_t)) /* gid */
df6fb868
PM
647 + nla_total_size(plen) /* prefix */
648 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
649 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp));
0597f268 650
eeff9bee
PNA
651 if (in && skb_mac_header_was_set(skb)) {
652 size += nla_total_size(skb->dev->hard_header_len)
653 + nla_total_size(sizeof(u_int16_t)) /* hwtype */
654 + nla_total_size(sizeof(u_int16_t)); /* hwlen */
655 }
656
0597f268
HW
657 spin_lock_bh(&inst->lock);
658
0af5f6c1 659 if (inst->flags & NFULNL_CFG_F_SEQ)
df6fb868 660 size += nla_total_size(sizeof(u_int32_t));
0af5f6c1 661 if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
df6fb868 662 size += nla_total_size(sizeof(u_int32_t));
0af5f6c1 663
0597f268
HW
664 qthreshold = inst->qthreshold;
665 /* per-rule qthreshold overrides per-instance */
5ca431f9
EL
666 if (li->u.ulog.qthreshold)
667 if (qthreshold > li->u.ulog.qthreshold)
668 qthreshold = li->u.ulog.qthreshold;
669
601e68e1 670
0597f268
HW
671 switch (inst->copy_mode) {
672 case NFULNL_COPY_META:
673 case NFULNL_COPY_NONE:
674 data_len = 0;
675 break;
601e68e1 676
0597f268 677 case NFULNL_COPY_PACKET:
601e68e1 678 if (inst->copy_range == 0
0597f268
HW
679 || inst->copy_range > skb->len)
680 data_len = skb->len;
681 else
682 data_len = inst->copy_range;
601e68e1 683
df6fb868 684 size += nla_total_size(data_len);
0597f268 685 break;
601e68e1 686
f5c5440d 687 case NFULNL_COPY_DISABLED:
0597f268 688 default:
55b5a91e 689 goto unlock_and_release;
0597f268
HW
690 }
691
d63b043d
MM
692 if (inst->skb &&
693 size > skb_tailroom(inst->skb) - sizeof(struct nfgenmsg)) {
0597f268
HW
694 /* either the queue len is too high or we don't have
695 * enough room in the skb left. flush to userspace. */
e3567061 696 __nfulnl_flush(inst);
55b5a91e 697 }
0597f268 698
55b5a91e
MM
699 if (!inst->skb) {
700 inst->skb = nfulnl_alloc_skb(inst->nlbufsiz, size);
701 if (!inst->skb)
0597f268 702 goto alloc_failure;
0597f268
HW
703 }
704
0597f268
HW
705 inst->qlen++;
706
9368a53c 707 __build_packet_message(log, inst, skb, data_len, pf,
8248779b 708 hooknum, in, out, prefix, plen);
0597f268 709
d63b043d
MM
710 if (inst->qlen >= qthreshold)
711 __nfulnl_flush(inst);
0597f268
HW
712 /* timer_pending always called within inst->lock, so there
713 * is no chance of a race here */
d63b043d 714 else if (!timer_pending(&inst->timer)) {
0597f268
HW
715 instance_get(inst);
716 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
717 add_timer(&inst->timer);
718 }
0597f268 719
ed32abea
MM
720unlock_and_release:
721 spin_unlock_bh(&inst->lock);
722 instance_put(inst);
0597f268
HW
723 return;
724
725alloc_failure:
0597f268 726 /* FIXME: statistics */
ed32abea 727 goto unlock_and_release;
0597f268 728}
5f7340ef 729EXPORT_SYMBOL_GPL(nfulnl_log_packet);
0597f268
HW
730
731static int
732nfulnl_rcv_nl_event(struct notifier_block *this,
733 unsigned long event, void *ptr)
734{
735 struct netlink_notify *n = ptr;
9368a53c 736 struct nfnl_log_net *log = nfnl_log_pernet(n->net);
0597f268 737
dee5817e 738 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
0597f268
HW
739 int i;
740
15e47304 741 /* destroy all instances for this portid */
9368a53c 742 spin_lock_bh(&log->instances_lock);
0597f268 743 for (i = 0; i < INSTANCE_BUCKETS; i++) {
b67bfe0d 744 struct hlist_node *t2;
0597f268 745 struct nfulnl_instance *inst;
9368a53c 746 struct hlist_head *head = &log->instance_table[i];
0597f268 747
b67bfe0d 748 hlist_for_each_entry_safe(inst, t2, head, hlist) {
9368a53c 749 if (n->portid == inst->peer_portid)
0597f268
HW
750 __instance_destroy(inst);
751 }
752 }
9368a53c 753 spin_unlock_bh(&log->instances_lock);
0597f268
HW
754 }
755 return NOTIFY_DONE;
756}
757
758static struct notifier_block nfulnl_rtnl_notifier = {
759 .notifier_call = nfulnl_rcv_nl_event,
760};
761
762static int
763nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
39938324
PM
764 const struct nlmsghdr *nlh,
765 const struct nlattr * const nfqa[])
0597f268
HW
766{
767 return -ENOTSUPP;
768}
769
ca735b3a 770static struct nf_logger nfulnl_logger __read_mostly = {
0597f268
HW
771 .name = "nfnetlink_log",
772 .logfn = &nfulnl_log_packet,
773 .me = THIS_MODULE,
774};
775
fd8281ad
PM
776static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
777 [NFULA_CFG_CMD] = { .len = sizeof(struct nfulnl_msg_config_cmd) },
778 [NFULA_CFG_MODE] = { .len = sizeof(struct nfulnl_msg_config_mode) },
779 [NFULA_CFG_TIMEOUT] = { .type = NLA_U32 },
780 [NFULA_CFG_QTHRESH] = { .type = NLA_U32 },
781 [NFULA_CFG_NLBUFSIZ] = { .type = NLA_U32 },
782 [NFULA_CFG_FLAGS] = { .type = NLA_U16 },
0597f268
HW
783};
784
785static int
786nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
39938324
PM
787 const struct nlmsghdr *nlh,
788 const struct nlattr * const nfula[])
0597f268 789{
d550d095 790 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
0597f268
HW
791 u_int16_t group_num = ntohs(nfmsg->res_id);
792 struct nfulnl_instance *inst;
b7047a1c 793 struct nfulnl_msg_config_cmd *cmd = NULL;
30e0c6a6 794 struct net *net = sock_net(ctnl);
9368a53c 795 struct nfnl_log_net *log = nfnl_log_pernet(net);
0597f268
HW
796 int ret = 0;
797
b7047a1c
PM
798 if (nfula[NFULA_CFG_CMD]) {
799 u_int8_t pf = nfmsg->nfgen_family;
800 cmd = nla_data(nfula[NFULA_CFG_CMD]);
801
802 /* Commands without queue context */
803 switch (cmd->command) {
804 case NFULNL_CFG_CMD_PF_BIND:
30e0c6a6 805 return nf_log_bind_pf(net, pf, &nfulnl_logger);
b7047a1c 806 case NFULNL_CFG_CMD_PF_UNBIND:
30e0c6a6 807 nf_log_unbind_pf(net, pf);
b7047a1c
PM
808 return 0;
809 }
810 }
811
9368a53c 812 inst = instance_lookup_get(log, group_num);
15e47304 813 if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
c0506365
PM
814 ret = -EPERM;
815 goto out_put;
816 }
817
b7047a1c 818 if (cmd != NULL) {
0597f268
HW
819 switch (cmd->command) {
820 case NFULNL_CFG_CMD_BIND:
821 if (inst) {
822 ret = -EBUSY;
823 goto out_put;
824 }
825
9368a53c 826 inst = instance_create(net, group_num,
15e47304 827 NETLINK_CB(skb).portid,
9eea9515 828 sk_user_ns(NETLINK_CB(skb).ssk));
baab2ce7
PM
829 if (IS_ERR(inst)) {
830 ret = PTR_ERR(inst);
f414c16c 831 goto out;
0597f268
HW
832 }
833 break;
834 case NFULNL_CFG_CMD_UNBIND:
835 if (!inst) {
836 ret = -ENODEV;
f414c16c 837 goto out;
0597f268
HW
838 }
839
9368a53c 840 instance_destroy(log, inst);
a49c6503 841 goto out_put;
0597f268 842 default:
cd21f0ac 843 ret = -ENOTSUPP;
0597f268
HW
844 break;
845 }
0597f268
HW
846 }
847
df6fb868 848 if (nfula[NFULA_CFG_MODE]) {
0597f268 849 struct nfulnl_msg_config_mode *params;
df6fb868 850 params = nla_data(nfula[NFULA_CFG_MODE]);
0597f268 851
c0506365
PM
852 if (!inst) {
853 ret = -ENODEV;
854 goto out;
855 }
0597f268 856 nfulnl_set_mode(inst, params->copy_mode,
d1208b99 857 ntohl(params->copy_range));
0597f268
HW
858 }
859
df6fb868 860 if (nfula[NFULA_CFG_TIMEOUT]) {
0dfedd28 861 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
0597f268 862
c0506365
PM
863 if (!inst) {
864 ret = -ENODEV;
865 goto out;
866 }
0597f268
HW
867 nfulnl_set_timeout(inst, ntohl(timeout));
868 }
869
df6fb868 870 if (nfula[NFULA_CFG_NLBUFSIZ]) {
0dfedd28 871 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
0597f268 872
c0506365
PM
873 if (!inst) {
874 ret = -ENODEV;
875 goto out;
876 }
0597f268
HW
877 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
878 }
879
df6fb868 880 if (nfula[NFULA_CFG_QTHRESH]) {
0dfedd28 881 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
0597f268 882
c0506365
PM
883 if (!inst) {
884 ret = -ENODEV;
885 goto out;
886 }
0597f268
HW
887 nfulnl_set_qthresh(inst, ntohl(qthresh));
888 }
889
df6fb868 890 if (nfula[NFULA_CFG_FLAGS]) {
0dfedd28 891 __be16 flags = nla_get_be16(nfula[NFULA_CFG_FLAGS]);
c0506365
PM
892
893 if (!inst) {
894 ret = -ENODEV;
895 goto out;
896 }
ee433530 897 nfulnl_set_flags(inst, ntohs(flags));
0af5f6c1
HW
898 }
899
0597f268
HW
900out_put:
901 instance_put(inst);
dd16704e 902out:
0597f268
HW
903 return ret;
904}
905
7c8d4cb4 906static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
0597f268 907 [NFULNL_MSG_PACKET] = { .call = nfulnl_recv_unsupp,
37d2e7a2 908 .attr_count = NFULA_MAX, },
0597f268 909 [NFULNL_MSG_CONFIG] = { .call = nfulnl_recv_config,
fd8281ad
PM
910 .attr_count = NFULA_CFG_MAX,
911 .policy = nfula_cfg_policy },
0597f268
HW
912};
913
7c8d4cb4 914static const struct nfnetlink_subsystem nfulnl_subsys = {
0597f268
HW
915 .name = "log",
916 .subsys_id = NFNL_SUBSYS_ULOG,
917 .cb_count = NFULNL_MSG_MAX,
0597f268
HW
918 .cb = nfulnl_cb,
919};
920
921#ifdef CONFIG_PROC_FS
922struct iter_state {
9368a53c 923 struct seq_net_private p;
0597f268
HW
924 unsigned int bucket;
925};
926
9368a53c 927static struct hlist_node *get_first(struct net *net, struct iter_state *st)
0597f268 928{
9368a53c 929 struct nfnl_log_net *log;
0597f268
HW
930 if (!st)
931 return NULL;
932
9368a53c
G
933 log = nfnl_log_pernet(net);
934
0597f268 935 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
9368a53c
G
936 struct hlist_head *head = &log->instance_table[st->bucket];
937
938 if (!hlist_empty(head))
939 return rcu_dereference_bh(hlist_first_rcu(head));
0597f268
HW
940 }
941 return NULL;
942}
943
9368a53c
G
944static struct hlist_node *get_next(struct net *net, struct iter_state *st,
945 struct hlist_node *h)
0597f268 946{
0e60ebe0 947 h = rcu_dereference_bh(hlist_next_rcu(h));
0597f268 948 while (!h) {
9368a53c
G
949 struct nfnl_log_net *log;
950 struct hlist_head *head;
951
0597f268
HW
952 if (++st->bucket >= INSTANCE_BUCKETS)
953 return NULL;
954
9368a53c
G
955 log = nfnl_log_pernet(net);
956 head = &log->instance_table[st->bucket];
957 h = rcu_dereference_bh(hlist_first_rcu(head));
0597f268
HW
958 }
959 return h;
960}
961
9368a53c
G
962static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
963 loff_t pos)
0597f268
HW
964{
965 struct hlist_node *head;
9368a53c 966 head = get_first(net, st);
0597f268
HW
967
968 if (head)
9368a53c 969 while (pos && (head = get_next(net, st, head)))
0597f268
HW
970 pos--;
971 return pos ? NULL : head;
972}
973
9368a53c 974static void *seq_start(struct seq_file *s, loff_t *pos)
bed1be20 975 __acquires(rcu_bh)
0597f268 976{
bed1be20 977 rcu_read_lock_bh();
9368a53c 978 return get_idx(seq_file_net(s), s->private, *pos);
0597f268
HW
979}
980
981static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
982{
983 (*pos)++;
9368a53c 984 return get_next(seq_file_net(s), s->private, v);
0597f268
HW
985}
986
987static void seq_stop(struct seq_file *s, void *v)
bed1be20 988 __releases(rcu_bh)
0597f268 989{
bed1be20 990 rcu_read_unlock_bh();
0597f268
HW
991}
992
993static int seq_show(struct seq_file *s, void *v)
994{
995 const struct nfulnl_instance *inst = v;
996
601e68e1 997 return seq_printf(s, "%5d %6d %5d %1d %5d %6d %2d\n",
0597f268 998 inst->group_num,
15e47304 999 inst->peer_portid, inst->qlen,
0597f268
HW
1000 inst->copy_mode, inst->copy_range,
1001 inst->flushtimeout, atomic_read(&inst->use));
1002}
1003
56b3d975 1004static const struct seq_operations nful_seq_ops = {
0597f268
HW
1005 .start = seq_start,
1006 .next = seq_next,
1007 .stop = seq_stop,
1008 .show = seq_show,
1009};
1010
1011static int nful_open(struct inode *inode, struct file *file)
1012{
9368a53c
G
1013 return seq_open_net(inode, file, &nful_seq_ops,
1014 sizeof(struct iter_state));
0597f268
HW
1015}
1016
da7071d7 1017static const struct file_operations nful_file_ops = {
0597f268
HW
1018 .owner = THIS_MODULE,
1019 .open = nful_open,
1020 .read = seq_read,
1021 .llseek = seq_lseek,
9368a53c 1022 .release = seq_release_net,
0597f268
HW
1023};
1024
1025#endif /* PROC_FS */
1026
9368a53c 1027static int __net_init nfnl_log_net_init(struct net *net)
0597f268 1028{
9368a53c
G
1029 unsigned int i;
1030 struct nfnl_log_net *log = nfnl_log_pernet(net);
601e68e1 1031
0597f268 1032 for (i = 0; i < INSTANCE_BUCKETS; i++)
9368a53c
G
1033 INIT_HLIST_HEAD(&log->instance_table[i]);
1034 spin_lock_init(&log->instances_lock);
1035
1036#ifdef CONFIG_PROC_FS
1037 if (!proc_create("nfnetlink_log", 0440,
1038 net->nf.proc_netfilter, &nful_file_ops))
1039 return -ENOMEM;
1040#endif
1041 return 0;
1042}
1043
1044static void __net_exit nfnl_log_net_exit(struct net *net)
1045{
1046 remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1047}
1048
1049static struct pernet_operations nfnl_log_net_ops = {
1050 .init = nfnl_log_net_init,
1051 .exit = nfnl_log_net_exit,
1052 .id = &nfnl_log_net_id,
1053 .size = sizeof(struct nfnl_log_net),
1054};
1055
1056static int __init nfnetlink_log_init(void)
1057{
1058 int status = -ENOMEM;
601e68e1 1059
0597f268
HW
1060 /* it's not really all that important to have a random value, so
1061 * we can do this from the init function, even if there hasn't
1062 * been that much entropy yet */
1063 get_random_bytes(&hash_init, sizeof(hash_init));
1064
1065 netlink_register_notifier(&nfulnl_rtnl_notifier);
1066 status = nfnetlink_subsys_register(&nfulnl_subsys);
1067 if (status < 0) {
9368a53c 1068 pr_err("log: failed to create netlink socket\n");
0597f268
HW
1069 goto cleanup_netlink_notifier;
1070 }
1071
ca735b3a
EL
1072 status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1073 if (status < 0) {
9368a53c 1074 pr_err("log: failed to register logger\n");
ca735b3a
EL
1075 goto cleanup_subsys;
1076 }
1077
9368a53c
G
1078 status = register_pernet_subsys(&nfnl_log_net_ops);
1079 if (status < 0) {
1080 pr_err("log: failed to register pernet ops\n");
ca735b3a 1081 goto cleanup_logger;
6fc09f10 1082 }
0597f268
HW
1083 return status;
1084
ca735b3a
EL
1085cleanup_logger:
1086 nf_log_unregister(&nfulnl_logger);
0597f268 1087cleanup_subsys:
0597f268
HW
1088 nfnetlink_subsys_unregister(&nfulnl_subsys);
1089cleanup_netlink_notifier:
1090 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1091 return status;
1092}
1093
65b4b4e8 1094static void __exit nfnetlink_log_fini(void)
0597f268 1095{
9368a53c 1096 unregister_pernet_subsys(&nfnl_log_net_ops);
e92ad99c 1097 nf_log_unregister(&nfulnl_logger);
32292a7f
PM
1098 nfnetlink_subsys_unregister(&nfulnl_subsys);
1099 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
0597f268
HW
1100}
1101
1102MODULE_DESCRIPTION("netfilter userspace logging");
1103MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1104MODULE_LICENSE("GPL");
f682faef 1105MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
0597f268 1106
65b4b4e8
AM
1107module_init(nfnetlink_log_init);
1108module_exit(nfnetlink_log_fini);