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