?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
perf tools: Don't clone maps from parent when synthesizing forks
[linux/fpc-iii.git] / net / netfilter / nfnetlink_log.c
blob332c69d27b4782698f6730befc7bc26d431ff735
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 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
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.
14 */
15
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17
18 #include <linux/module.h>
19 #include <linux/skbuff.h>
20 #include <linux/if_arp.h>
21 #include <linux/init.h>
22 #include <linux/ip.h>
23 #include <linux/ipv6.h>
24 #include <linux/netdevice.h>
25 #include <linux/netfilter.h>
26 #include <linux/netfilter_bridge.h>
27 #include <net/netlink.h>
28 #include <linux/netfilter/nfnetlink.h>
29 #include <linux/netfilter/nfnetlink_log.h>
30 #include <linux/netfilter/nf_conntrack_common.h>
31 #include <linux/spinlock.h>
32 #include <linux/sysctl.h>
33 #include <linux/proc_fs.h>
34 #include <linux/security.h>
35 #include <linux/list.h>
36 #include <linux/slab.h>
37 #include <net/sock.h>
38 #include <net/netfilter/nf_log.h>
39 #include <net/netns/generic.h>
40
41 #include <linux/atomic.h>
42 #include <linux/refcount.h>
43
44
45 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
46 #include "../bridge/br_private.h"
47 #endif
48
49 #define NFULNL_COPY_DISABLED 0xff
50 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
51 #define NFULNL_TIMEOUT_DEFAULT 100 /* every second */
52 #define NFULNL_QTHRESH_DEFAULT 100 /* 100 packets */
53 /* max packet size is limited by 16-bit struct nfattr nfa_len field */
54 #define NFULNL_COPY_RANGE_MAX (0xFFFF - NLA_HDRLEN)
55
56 #define PRINTR(x, args...) do { if (net_ratelimit()) \
57 printk(x, ## args); } while (0);
58
59 struct nfulnl_instance {
60 struct hlist_node hlist; /* global list of instances */
61 spinlock_t lock;
62 refcount_t use; /* use count */
63
64 unsigned int qlen; /* number of nlmsgs in skb */
65 struct sk_buff *skb; /* pre-allocatd skb */
66 struct timer_list timer;
67 struct net *net;
68 struct user_namespace *peer_user_ns; /* User namespace of the peer process */
69 u32 peer_portid; /* PORTID of the peer process */
70
71 /* configurable parameters */
72 unsigned int flushtimeout; /* timeout until queue flush */
73 unsigned int nlbufsiz; /* netlink buffer allocation size */
74 unsigned int qthreshold; /* threshold of the queue */
75 u_int32_t copy_range;
76 u_int32_t seq; /* instance-local sequential counter */
77 u_int16_t group_num; /* number of this queue */
78 u_int16_t flags;
79 u_int8_t copy_mode;
80 struct rcu_head rcu;
81 };
82
83 #define INSTANCE_BUCKETS 16
84
85 static unsigned int nfnl_log_net_id __read_mostly;
86
87 struct nfnl_log_net {
88 spinlock_t instances_lock;
89 struct hlist_head instance_table[INSTANCE_BUCKETS];
90 atomic_t global_seq;
91 };
92
93 static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
94 {
95 return net_generic(net, nfnl_log_net_id);
96 }
97
98 static inline u_int8_t instance_hashfn(u_int16_t group_num)
99 {
100 return ((group_num & 0xff) % INSTANCE_BUCKETS);
101 }
102
103 static struct nfulnl_instance *
104 __instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
105 {
106 struct hlist_head *head;
107 struct nfulnl_instance *inst;
108
109 head = &log->instance_table[instance_hashfn(group_num)];
110 hlist_for_each_entry_rcu(inst, head, hlist) {
111 if (inst->group_num == group_num)
112 return inst;
113 }
114 return NULL;
115 }
116
117 static inline void
118 instance_get(struct nfulnl_instance *inst)
119 {
120 refcount_inc(&inst->use);
121 }
122
123 static struct nfulnl_instance *
124 instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
125 {
126 struct nfulnl_instance *inst;
127
128 rcu_read_lock_bh();
129 inst = __instance_lookup(log, group_num);
130 if (inst && !refcount_inc_not_zero(&inst->use))
131 inst = NULL;
132 rcu_read_unlock_bh();
133
134 return inst;
135 }
136
137 static void nfulnl_instance_free_rcu(struct rcu_head *head)
138 {
139 struct nfulnl_instance *inst =
140 container_of(head, struct nfulnl_instance, rcu);
141
142 put_net(inst->net);
143 kfree(inst);
144 module_put(THIS_MODULE);
145 }
146
147 static void
148 instance_put(struct nfulnl_instance *inst)
149 {
150 if (inst && refcount_dec_and_test(&inst->use))
151 call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
152 }
153
154 static void nfulnl_timer(struct timer_list *t);
155
156 static struct nfulnl_instance *
157 instance_create(struct net *net, u_int16_t group_num,
158 u32 portid, struct user_namespace *user_ns)
159 {
160 struct nfulnl_instance *inst;
161 struct nfnl_log_net *log = nfnl_log_pernet(net);
162 int err;
163
164 spin_lock_bh(&log->instances_lock);
165 if (__instance_lookup(log, group_num)) {
166 err = -EEXIST;
167 goto out_unlock;
168 }
169
170 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
171 if (!inst) {
172 err = -ENOMEM;
173 goto out_unlock;
174 }
175
176 if (!try_module_get(THIS_MODULE)) {
177 kfree(inst);
178 err = -EAGAIN;
179 goto out_unlock;
180 }
181
182 INIT_HLIST_NODE(&inst->hlist);
183 spin_lock_init(&inst->lock);
184 /* needs to be two, since we _put() after creation */
185 refcount_set(&inst->use, 2);
186
187 timer_setup(&inst->timer, nfulnl_timer, 0);
188
189 inst->net = get_net(net);
190 inst->peer_user_ns = user_ns;
191 inst->peer_portid = portid;
192 inst->group_num = group_num;
193
194 inst->qthreshold = NFULNL_QTHRESH_DEFAULT;
195 inst->flushtimeout = NFULNL_TIMEOUT_DEFAULT;
196 inst->nlbufsiz = NFULNL_NLBUFSIZ_DEFAULT;
197 inst->copy_mode = NFULNL_COPY_PACKET;
198 inst->copy_range = NFULNL_COPY_RANGE_MAX;
199
200 hlist_add_head_rcu(&inst->hlist,
201 &log->instance_table[instance_hashfn(group_num)]);
202
203
204 spin_unlock_bh(&log->instances_lock);
205
206 return inst;
207
208 out_unlock:
209 spin_unlock_bh(&log->instances_lock);
210 return ERR_PTR(err);
211 }
212
213 static void __nfulnl_flush(struct nfulnl_instance *inst);
214
215 /* called with BH disabled */
216 static void
217 __instance_destroy(struct nfulnl_instance *inst)
218 {
219 /* first pull it out of the global list */
220 hlist_del_rcu(&inst->hlist);
221
222 /* then flush all pending packets from skb */
223
224 spin_lock(&inst->lock);
225
226 /* lockless readers wont be able to use us */
227 inst->copy_mode = NFULNL_COPY_DISABLED;
228
229 if (inst->skb)
230 __nfulnl_flush(inst);
231 spin_unlock(&inst->lock);
232
233 /* and finally put the refcount */
234 instance_put(inst);
235 }
236
237 static inline void
238 instance_destroy(struct nfnl_log_net *log,
239 struct nfulnl_instance *inst)
240 {
241 spin_lock_bh(&log->instances_lock);
242 __instance_destroy(inst);
243 spin_unlock_bh(&log->instances_lock);
244 }
245
246 static int
247 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
248 unsigned int range)
249 {
250 int status = 0;
251
252 spin_lock_bh(&inst->lock);
253
254 switch (mode) {
255 case NFULNL_COPY_NONE:
256 case NFULNL_COPY_META:
257 inst->copy_mode = mode;
258 inst->copy_range = 0;
259 break;
260
261 case NFULNL_COPY_PACKET:
262 inst->copy_mode = mode;
263 if (range == 0)
264 range = NFULNL_COPY_RANGE_MAX;
265 inst->copy_range = min_t(unsigned int,
266 range, NFULNL_COPY_RANGE_MAX);
267 break;
268
269 default:
270 status = -EINVAL;
271 break;
272 }
273
274 spin_unlock_bh(&inst->lock);
275
276 return status;
277 }
278
279 static int
280 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
281 {
282 int status;
283
284 spin_lock_bh(&inst->lock);
285 if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
286 status = -ERANGE;
287 else if (nlbufsiz > 131072)
288 status = -ERANGE;
289 else {
290 inst->nlbufsiz = nlbufsiz;
291 status = 0;
292 }
293 spin_unlock_bh(&inst->lock);
294
295 return status;
296 }
297
298 static void
299 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
300 {
301 spin_lock_bh(&inst->lock);
302 inst->flushtimeout = timeout;
303 spin_unlock_bh(&inst->lock);
304 }
305
306 static void
307 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
308 {
309 spin_lock_bh(&inst->lock);
310 inst->qthreshold = qthresh;
311 spin_unlock_bh(&inst->lock);
312 }
313
314 static int
315 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
316 {
317 spin_lock_bh(&inst->lock);
318 inst->flags = flags;
319 spin_unlock_bh(&inst->lock);
320
321 return 0;
322 }
323
324 static struct sk_buff *
325 nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
326 unsigned int pkt_size)
327 {
328 struct sk_buff *skb;
329 unsigned int n;
330
331 /* alloc skb which should be big enough for a whole multipart
332 * message. WARNING: has to be <= 128k due to slab restrictions */
333
334 n = max(inst_size, pkt_size);
335 skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
336 if (!skb) {
337 if (n > pkt_size) {
338 /* try to allocate only as much as we need for current
339 * packet */
340
341 skb = alloc_skb(pkt_size, GFP_ATOMIC);
342 }
343 }
344
345 return skb;
346 }
347
348 static void
349 __nfulnl_send(struct nfulnl_instance *inst)
350 {
351 if (inst->qlen > 1) {
352 struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
353 NLMSG_DONE,
354 sizeof(struct nfgenmsg),
355 0);
356 if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
357 inst->skb->len, skb_tailroom(inst->skb))) {
358 kfree_skb(inst->skb);
359 goto out;
360 }
361 }
362 nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
363 MSG_DONTWAIT);
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 /* This is an inline function, we don't really care about a long
392 * list of arguments */
393 static inline int
394 __build_packet_message(struct nfnl_log_net *log,
395 struct nfulnl_instance *inst,
396 const struct sk_buff *skb,
397 unsigned int data_len,
398 u_int8_t pf,
399 unsigned int hooknum,
400 const struct net_device *indev,
401 const struct net_device *outdev,
402 const char *prefix, unsigned int plen,
403 const struct nfnl_ct_hook *nfnl_ct,
404 struct nf_conn *ct, enum ip_conntrack_info ctinfo)
405 {
406 struct nfulnl_msg_packet_hdr pmsg;
407 struct nlmsghdr *nlh;
408 struct nfgenmsg *nfmsg;
409 sk_buff_data_t old_tail = inst->skb->tail;
410 struct sock *sk;
411 const unsigned char *hwhdrp;
412
413 nlh = nlmsg_put(inst->skb, 0, 0,
414 nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
415 sizeof(struct nfgenmsg), 0);
416 if (!nlh)
417 return -1;
418 nfmsg = nlmsg_data(nlh);
419 nfmsg->nfgen_family = pf;
420 nfmsg->version = NFNETLINK_V0;
421 nfmsg->res_id = htons(inst->group_num);
422
423 memset(&pmsg, 0, sizeof(pmsg));
424 pmsg.hw_protocol = skb->protocol;
425 pmsg.hook = hooknum;
426
427 if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
428 goto nla_put_failure;
429
430 if (prefix &&
431 nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
432 goto nla_put_failure;
433
434 if (indev) {
435 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
436 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
437 htonl(indev->ifindex)))
438 goto nla_put_failure;
439 #else
440 if (pf == PF_BRIDGE) {
441 /* Case 1: outdev is physical input device, we need to
442 * look for bridge group (when called from
443 * netfilter_bridge) */
444 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
445 htonl(indev->ifindex)) ||
446 /* this is the bridge group "brX" */
447 /* rcu_read_lock()ed by nf_hook_thresh or
448 * nf_log_packet.
449 */
450 nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
451 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
452 goto nla_put_failure;
453 } else {
454 struct net_device *physindev;
455
456 /* Case 2: indev is bridge group, we need to look for
457 * physical device (when called from ipv4) */
458 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
459 htonl(indev->ifindex)))
460 goto nla_put_failure;
461
462 physindev = nf_bridge_get_physindev(skb);
463 if (physindev &&
464 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
465 htonl(physindev->ifindex)))
466 goto nla_put_failure;
467 }
468 #endif
469 }
470
471 if (outdev) {
472 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
473 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
474 htonl(outdev->ifindex)))
475 goto nla_put_failure;
476 #else
477 if (pf == PF_BRIDGE) {
478 /* Case 1: outdev is physical output device, we need to
479 * look for bridge group (when called from
480 * netfilter_bridge) */
481 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
482 htonl(outdev->ifindex)) ||
483 /* this is the bridge group "brX" */
484 /* rcu_read_lock()ed by nf_hook_thresh or
485 * nf_log_packet.
486 */
487 nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
488 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
489 goto nla_put_failure;
490 } else {
491 struct net_device *physoutdev;
492
493 /* Case 2: indev is a bridge group, we need to look
494 * for physical device (when called from ipv4) */
495 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
496 htonl(outdev->ifindex)))
497 goto nla_put_failure;
498
499 physoutdev = nf_bridge_get_physoutdev(skb);
500 if (physoutdev &&
501 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
502 htonl(physoutdev->ifindex)))
503 goto nla_put_failure;
504 }
505 #endif
506 }
507
508 if (skb->mark &&
509 nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
510 goto nla_put_failure;
511
512 if (indev && skb->dev &&
513 skb->mac_header != skb->network_header) {
514 struct nfulnl_msg_packet_hw phw;
515 int len;
516
517 memset(&phw, 0, sizeof(phw));
518 len = dev_parse_header(skb, phw.hw_addr);
519 if (len > 0) {
520 phw.hw_addrlen = htons(len);
521 if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
522 goto nla_put_failure;
523 }
524 }
525
526 if (indev && skb_mac_header_was_set(skb)) {
527 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
528 nla_put_be16(inst->skb, NFULA_HWLEN,
529 htons(skb->dev->hard_header_len)))
530 goto nla_put_failure;
531
532 hwhdrp = skb_mac_header(skb);
533
534 if (skb->dev->type == ARPHRD_SIT)
535 hwhdrp -= ETH_HLEN;
536
537 if (hwhdrp >= skb->head &&
538 nla_put(inst->skb, NFULA_HWHEADER,
539 skb->dev->hard_header_len, hwhdrp))
540 goto nla_put_failure;
541 }
542
543 if (skb->tstamp) {
544 struct nfulnl_msg_packet_timestamp ts;
545 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
546 ts.sec = cpu_to_be64(kts.tv_sec);
547 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
548
549 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
550 goto nla_put_failure;
551 }
552
553 /* UID */
554 sk = skb->sk;
555 if (sk && sk_fullsock(sk)) {
556 read_lock_bh(&sk->sk_callback_lock);
557 if (sk->sk_socket && sk->sk_socket->file) {
558 struct file *file = sk->sk_socket->file;
559 const struct cred *cred = file->f_cred;
560 struct user_namespace *user_ns = inst->peer_user_ns;
561 __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
562 __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
563 read_unlock_bh(&sk->sk_callback_lock);
564 if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
565 nla_put_be32(inst->skb, NFULA_GID, gid))
566 goto nla_put_failure;
567 } else
568 read_unlock_bh(&sk->sk_callback_lock);
569 }
570
571 /* local sequence number */
572 if ((inst->flags & NFULNL_CFG_F_SEQ) &&
573 nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
574 goto nla_put_failure;
575
576 /* global sequence number */
577 if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
578 nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
579 htonl(atomic_inc_return(&log->global_seq))))
580 goto nla_put_failure;
581
582 if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
583 NFULA_CT, NFULA_CT_INFO) < 0)
584 goto nla_put_failure;
585
586 if (data_len) {
587 struct nlattr *nla;
588 int size = nla_attr_size(data_len);
589
590 if (skb_tailroom(inst->skb) < nla_total_size(data_len))
591 goto nla_put_failure;
592
593 nla = skb_put(inst->skb, nla_total_size(data_len));
594 nla->nla_type = NFULA_PAYLOAD;
595 nla->nla_len = size;
596
597 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
598 BUG();
599 }
600
601 nlh->nlmsg_len = inst->skb->tail - old_tail;
602 return 0;
603
604 nla_put_failure:
605 PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
606 return -1;
607 }
608
609 static const struct nf_loginfo default_loginfo = {
610 .type = NF_LOG_TYPE_ULOG,
611 .u = {
612 .ulog = {
613 .copy_len = 0xffff,
614 .group = 0,
615 .qthreshold = 1,
616 },
617 },
618 };
619
620 /* log handler for internal netfilter logging api */
621 static void
622 nfulnl_log_packet(struct net *net,
623 u_int8_t pf,
624 unsigned int hooknum,
625 const struct sk_buff *skb,
626 const struct net_device *in,
627 const struct net_device *out,
628 const struct nf_loginfo *li_user,
629 const char *prefix)
630 {
631 size_t size;
632 unsigned int data_len;
633 struct nfulnl_instance *inst;
634 const struct nf_loginfo *li;
635 unsigned int qthreshold;
636 unsigned int plen = 0;
637 struct nfnl_log_net *log = nfnl_log_pernet(net);
638 const struct nfnl_ct_hook *nfnl_ct = NULL;
639 struct nf_conn *ct = NULL;
640 enum ip_conntrack_info uninitialized_var(ctinfo);
641
642 if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
643 li = li_user;
644 else
645 li = &default_loginfo;
646
647 inst = instance_lookup_get(log, li->u.ulog.group);
648 if (!inst)
649 return;
650
651 if (prefix)
652 plen = strlen(prefix) + 1;
653
654 /* FIXME: do we want to make the size calculation conditional based on
655 * what is actually present? way more branches and checks, but more
656 * memory efficient... */
657 size = nlmsg_total_size(sizeof(struct nfgenmsg))
658 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
659 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
660 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
661 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
662 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
663 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
664 #endif
665 + nla_total_size(sizeof(u_int32_t)) /* mark */
666 + nla_total_size(sizeof(u_int32_t)) /* uid */
667 + nla_total_size(sizeof(u_int32_t)) /* gid */
668 + nla_total_size(plen) /* prefix */
669 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
670 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
671 + nla_total_size(sizeof(struct nfgenmsg)); /* NLMSG_DONE */
672
673 if (in && skb_mac_header_was_set(skb)) {
674 size += nla_total_size(skb->dev->hard_header_len)
675 + nla_total_size(sizeof(u_int16_t)) /* hwtype */
676 + nla_total_size(sizeof(u_int16_t)); /* hwlen */
677 }
678
679 spin_lock_bh(&inst->lock);
680
681 if (inst->flags & NFULNL_CFG_F_SEQ)
682 size += nla_total_size(sizeof(u_int32_t));
683 if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
684 size += nla_total_size(sizeof(u_int32_t));
685 if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
686 nfnl_ct = rcu_dereference(nfnl_ct_hook);
687 if (nfnl_ct != NULL) {
688 ct = nfnl_ct->get_ct(skb, &ctinfo);
689 if (ct != NULL)
690 size += nfnl_ct->build_size(ct);
691 }
692 }
693
694 qthreshold = inst->qthreshold;
695 /* per-rule qthreshold overrides per-instance */
696 if (li->u.ulog.qthreshold)
697 if (qthreshold > li->u.ulog.qthreshold)
698 qthreshold = li->u.ulog.qthreshold;
699
700
701 switch (inst->copy_mode) {
702 case NFULNL_COPY_META:
703 case NFULNL_COPY_NONE:
704 data_len = 0;
705 break;
706
707 case NFULNL_COPY_PACKET:
708 data_len = inst->copy_range;
709 if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
710 (li->u.ulog.copy_len < data_len))
711 data_len = li->u.ulog.copy_len;
712
713 if (data_len > skb->len)
714 data_len = skb->len;
715
716 size += nla_total_size(data_len);
717 break;
718
719 case NFULNL_COPY_DISABLED:
720 default:
721 goto unlock_and_release;
722 }
723
724 if (inst->skb && size > skb_tailroom(inst->skb)) {
725 /* either the queue len is too high or we don't have
726 * enough room in the skb left. flush to userspace. */
727 __nfulnl_flush(inst);
728 }
729
730 if (!inst->skb) {
731 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
732 inst->nlbufsiz, size);
733 if (!inst->skb)
734 goto alloc_failure;
735 }
736
737 inst->qlen++;
738
739 __build_packet_message(log, inst, skb, data_len, pf,
740 hooknum, in, out, prefix, plen,
741 nfnl_ct, ct, ctinfo);
742
743 if (inst->qlen >= qthreshold)
744 __nfulnl_flush(inst);
745 /* timer_pending always called within inst->lock, so there
746 * is no chance of a race here */
747 else if (!timer_pending(&inst->timer)) {
748 instance_get(inst);
749 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
750 add_timer(&inst->timer);
751 }
752
753 unlock_and_release:
754 spin_unlock_bh(&inst->lock);
755 instance_put(inst);
756 return;
757
758 alloc_failure:
759 /* FIXME: statistics */
760 goto unlock_and_release;
761 }
762
763 static int
764 nfulnl_rcv_nl_event(struct notifier_block *this,
765 unsigned long event, void *ptr)
766 {
767 struct netlink_notify *n = ptr;
768 struct nfnl_log_net *log = nfnl_log_pernet(n->net);
769
770 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
771 int i;
772
773 /* destroy all instances for this portid */
774 spin_lock_bh(&log->instances_lock);
775 for (i = 0; i < INSTANCE_BUCKETS; i++) {
776 struct hlist_node *t2;
777 struct nfulnl_instance *inst;
778 struct hlist_head *head = &log->instance_table[i];
779
780 hlist_for_each_entry_safe(inst, t2, head, hlist) {
781 if (n->portid == inst->peer_portid)
782 __instance_destroy(inst);
783 }
784 }
785 spin_unlock_bh(&log->instances_lock);
786 }
787 return NOTIFY_DONE;
788 }
789
790 static struct notifier_block nfulnl_rtnl_notifier = {
791 .notifier_call = nfulnl_rcv_nl_event,
792 };
793
794 static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl,
795 struct sk_buff *skb, const struct nlmsghdr *nlh,
796 const struct nlattr * const nfqa[],
797 struct netlink_ext_ack *extack)
798 {
799 return -ENOTSUPP;
800 }
801
802 static struct nf_logger nfulnl_logger __read_mostly = {
803 .name = "nfnetlink_log",
804 .type = NF_LOG_TYPE_ULOG,
805 .logfn = nfulnl_log_packet,
806 .me = THIS_MODULE,
807 };
808
809 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
810 [NFULA_CFG_CMD] = { .len = sizeof(struct nfulnl_msg_config_cmd) },
811 [NFULA_CFG_MODE] = { .len = sizeof(struct nfulnl_msg_config_mode) },
812 [NFULA_CFG_TIMEOUT] = { .type = NLA_U32 },
813 [NFULA_CFG_QTHRESH] = { .type = NLA_U32 },
814 [NFULA_CFG_NLBUFSIZ] = { .type = NLA_U32 },
815 [NFULA_CFG_FLAGS] = { .type = NLA_U16 },
816 };
817
818 static int nfulnl_recv_config(struct net *net, struct sock *ctnl,
819 struct sk_buff *skb, const struct nlmsghdr *nlh,
820 const struct nlattr * const nfula[],
821 struct netlink_ext_ack *extack)
822 {
823 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
824 u_int16_t group_num = ntohs(nfmsg->res_id);
825 struct nfulnl_instance *inst;
826 struct nfulnl_msg_config_cmd *cmd = NULL;
827 struct nfnl_log_net *log = nfnl_log_pernet(net);
828 int ret = 0;
829 u16 flags = 0;
830
831 if (nfula[NFULA_CFG_CMD]) {
832 u_int8_t pf = nfmsg->nfgen_family;
833 cmd = nla_data(nfula[NFULA_CFG_CMD]);
834
835 /* Commands without queue context */
836 switch (cmd->command) {
837 case NFULNL_CFG_CMD_PF_BIND:
838 return nf_log_bind_pf(net, pf, &nfulnl_logger);
839 case NFULNL_CFG_CMD_PF_UNBIND:
840 nf_log_unbind_pf(net, pf);
841 return 0;
842 }
843 }
844
845 inst = instance_lookup_get(log, group_num);
846 if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
847 ret = -EPERM;
848 goto out_put;
849 }
850
851 /* Check if we support these flags in first place, dependencies should
852 * be there too not to break atomicity.
853 */
854 if (nfula[NFULA_CFG_FLAGS]) {
855 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
856
857 if ((flags & NFULNL_CFG_F_CONNTRACK) &&
858 !rcu_access_pointer(nfnl_ct_hook)) {
859 #ifdef CONFIG_MODULES
860 nfnl_unlock(NFNL_SUBSYS_ULOG);
861 request_module("ip_conntrack_netlink");
862 nfnl_lock(NFNL_SUBSYS_ULOG);
863 if (rcu_access_pointer(nfnl_ct_hook)) {
864 ret = -EAGAIN;
865 goto out_put;
866 }
867 #endif
868 ret = -EOPNOTSUPP;
869 goto out_put;
870 }
871 }
872
873 if (cmd != NULL) {
874 switch (cmd->command) {
875 case NFULNL_CFG_CMD_BIND:
876 if (inst) {
877 ret = -EBUSY;
878 goto out_put;
879 }
880
881 inst = instance_create(net, group_num,
882 NETLINK_CB(skb).portid,
883 sk_user_ns(NETLINK_CB(skb).sk));
884 if (IS_ERR(inst)) {
885 ret = PTR_ERR(inst);
886 goto out;
887 }
888 break;
889 case NFULNL_CFG_CMD_UNBIND:
890 if (!inst) {
891 ret = -ENODEV;
892 goto out;
893 }
894
895 instance_destroy(log, inst);
896 goto out_put;
897 default:
898 ret = -ENOTSUPP;
899 goto out_put;
900 }
901 } else if (!inst) {
902 ret = -ENODEV;
903 goto out;
904 }
905
906 if (nfula[NFULA_CFG_MODE]) {
907 struct nfulnl_msg_config_mode *params =
908 nla_data(nfula[NFULA_CFG_MODE]);
909
910 nfulnl_set_mode(inst, params->copy_mode,
911 ntohl(params->copy_range));
912 }
913
914 if (nfula[NFULA_CFG_TIMEOUT]) {
915 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
916
917 nfulnl_set_timeout(inst, ntohl(timeout));
918 }
919
920 if (nfula[NFULA_CFG_NLBUFSIZ]) {
921 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
922
923 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
924 }
925
926 if (nfula[NFULA_CFG_QTHRESH]) {
927 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
928
929 nfulnl_set_qthresh(inst, ntohl(qthresh));
930 }
931
932 if (nfula[NFULA_CFG_FLAGS])
933 nfulnl_set_flags(inst, flags);
934
935 out_put:
936 instance_put(inst);
937 out:
938 return ret;
939 }
940
941 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
942 [NFULNL_MSG_PACKET] = { .call = nfulnl_recv_unsupp,
943 .attr_count = NFULA_MAX, },
944 [NFULNL_MSG_CONFIG] = { .call = nfulnl_recv_config,
945 .attr_count = NFULA_CFG_MAX,
946 .policy = nfula_cfg_policy },
947 };
948
949 static const struct nfnetlink_subsystem nfulnl_subsys = {
950 .name = "log",
951 .subsys_id = NFNL_SUBSYS_ULOG,
952 .cb_count = NFULNL_MSG_MAX,
953 .cb = nfulnl_cb,
954 };
955
956 #ifdef CONFIG_PROC_FS
957 struct iter_state {
958 struct seq_net_private p;
959 unsigned int bucket;
960 };
961
962 static struct hlist_node *get_first(struct net *net, struct iter_state *st)
963 {
964 struct nfnl_log_net *log;
965 if (!st)
966 return NULL;
967
968 log = nfnl_log_pernet(net);
969
970 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
971 struct hlist_head *head = &log->instance_table[st->bucket];
972
973 if (!hlist_empty(head))
974 return rcu_dereference_bh(hlist_first_rcu(head));
975 }
976 return NULL;
977 }
978
979 static struct hlist_node *get_next(struct net *net, struct iter_state *st,
980 struct hlist_node *h)
981 {
982 h = rcu_dereference_bh(hlist_next_rcu(h));
983 while (!h) {
984 struct nfnl_log_net *log;
985 struct hlist_head *head;
986
987 if (++st->bucket >= INSTANCE_BUCKETS)
988 return NULL;
989
990 log = nfnl_log_pernet(net);
991 head = &log->instance_table[st->bucket];
992 h = rcu_dereference_bh(hlist_first_rcu(head));
993 }
994 return h;
995 }
996
997 static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
998 loff_t pos)
999 {
1000 struct hlist_node *head;
1001 head = get_first(net, st);
1002
1003 if (head)
1004 while (pos && (head = get_next(net, st, head)))
1005 pos--;
1006 return pos ? NULL : head;
1007 }
1008
1009 static void *seq_start(struct seq_file *s, loff_t *pos)
1010 __acquires(rcu_bh)
1011 {
1012 rcu_read_lock_bh();
1013 return get_idx(seq_file_net(s), s->private, *pos);
1014 }
1015
1016 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1017 {
1018 (*pos)++;
1019 return get_next(seq_file_net(s), s->private, v);
1020 }
1021
1022 static void seq_stop(struct seq_file *s, void *v)
1023 __releases(rcu_bh)
1024 {
1025 rcu_read_unlock_bh();
1026 }
1027
1028 static int seq_show(struct seq_file *s, void *v)
1029 {
1030 const struct nfulnl_instance *inst = v;
1031
1032 seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1033 inst->group_num,
1034 inst->peer_portid, inst->qlen,
1035 inst->copy_mode, inst->copy_range,
1036 inst->flushtimeout, refcount_read(&inst->use));
1037
1038 return 0;
1039 }
1040
1041 static const struct seq_operations nful_seq_ops = {
1042 .start = seq_start,
1043 .next = seq_next,
1044 .stop = seq_stop,
1045 .show = seq_show,
1046 };
1047 #endif /* PROC_FS */
1048
1049 static int __net_init nfnl_log_net_init(struct net *net)
1050 {
1051 unsigned int i;
1052 struct nfnl_log_net *log = nfnl_log_pernet(net);
1053 #ifdef CONFIG_PROC_FS
1054 struct proc_dir_entry *proc;
1055 kuid_t root_uid;
1056 kgid_t root_gid;
1057 #endif
1058
1059 for (i = 0; i < INSTANCE_BUCKETS; i++)
1060 INIT_HLIST_HEAD(&log->instance_table[i]);
1061 spin_lock_init(&log->instances_lock);
1062
1063 #ifdef CONFIG_PROC_FS
1064 proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1065 &nful_seq_ops, sizeof(struct iter_state));
1066 if (!proc)
1067 return -ENOMEM;
1068
1069 root_uid = make_kuid(net->user_ns, 0);
1070 root_gid = make_kgid(net->user_ns, 0);
1071 if (uid_valid(root_uid) && gid_valid(root_gid))
1072 proc_set_user(proc, root_uid, root_gid);
1073 #endif
1074 return 0;
1075 }
1076
1077 static void __net_exit nfnl_log_net_exit(struct net *net)
1078 {
1079 struct nfnl_log_net *log = nfnl_log_pernet(net);
1080 unsigned int i;
1081
1082 #ifdef CONFIG_PROC_FS
1083 remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1084 #endif
1085 nf_log_unset(net, &nfulnl_logger);
1086 for (i = 0; i < INSTANCE_BUCKETS; i++)
1087 WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1088 }
1089
1090 static struct pernet_operations nfnl_log_net_ops = {
1091 .init = nfnl_log_net_init,
1092 .exit = nfnl_log_net_exit,
1093 .id = &nfnl_log_net_id,
1094 .size = sizeof(struct nfnl_log_net),
1095 };
1096
1097 static int __init nfnetlink_log_init(void)
1098 {
1099 int status;
1100
1101 status = register_pernet_subsys(&nfnl_log_net_ops);
1102 if (status < 0) {
1103 pr_err("failed to register pernet ops\n");
1104 goto out;
1105 }
1106
1107 netlink_register_notifier(&nfulnl_rtnl_notifier);
1108 status = nfnetlink_subsys_register(&nfulnl_subsys);
1109 if (status < 0) {
1110 pr_err("failed to create netlink socket\n");
1111 goto cleanup_netlink_notifier;
1112 }
1113
1114 status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1115 if (status < 0) {
1116 pr_err("failed to register logger\n");
1117 goto cleanup_subsys;
1118 }
1119
1120 return status;
1121
1122 cleanup_subsys:
1123 nfnetlink_subsys_unregister(&nfulnl_subsys);
1124 cleanup_netlink_notifier:
1125 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1126 unregister_pernet_subsys(&nfnl_log_net_ops);
1127 out:
1128 return status;
1129 }
1130
1131 static void __exit nfnetlink_log_fini(void)
1132 {
1133 nfnetlink_subsys_unregister(&nfulnl_subsys);
1134 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1135 unregister_pernet_subsys(&nfnl_log_net_ops);
1136 nf_log_unregister(&nfulnl_logger);
1137 }
1138
1139 MODULE_DESCRIPTION("netfilter userspace logging");
1140 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1141 MODULE_LICENSE("GPL");
1142 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1143 MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1144 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1145 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1146 MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1147 MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1148
1149 module_init(nfnetlink_log_init);
1150 module_exit(nfnetlink_log_fini);
Linux with added FPC-III support