Merge tag 'dmaengine-6.6-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul...
[platform/kernel/linux-starfive.git] / net / netfilter / nfnetlink_log.c
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
95 static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
96 {
97         return net_generic(net, nfnl_log_net_id);
98 }
99
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 *
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
120 instance_get(struct nfulnl_instance *inst)
121 {
122         refcount_inc(&inst->use);
123 }
124
125 static struct nfulnl_instance *
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 *
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
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
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 *
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(sizeof(*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
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
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
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
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
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
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
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 *
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
361 __nfulnl_send(struct nfulnl_instance *inst)
362 {
363         if (inst->qlen > 1) {
364                 struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
365                                                  NLMSG_DONE,
366                                                  sizeof(struct nfgenmsg),
367                                                  0);
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
381 __nfulnl_flush(struct nfulnl_instance *inst)
382 {
383         /* timer holds a reference */
384         if (del_timer(&inst->timer))
385                 instance_put(inst);
386         if (inst->skb)
387                 __nfulnl_send(inst);
388 }
389
390 static void
391 nfulnl_timer(struct timer_list *t)
392 {
393         struct nfulnl_instance *inst = from_timer(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
402 static u32 nfulnl_get_bridge_size(const struct sk_buff *skb)
403 {
404         u32 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         if (skb->network_header > skb->mac_header)
416                 size += nla_total_size(skb->network_header - skb->mac_header);
417
418         return size;
419 }
420
421 static int nfulnl_put_bridge(struct nfulnl_instance *inst, const struct sk_buff *skb)
422 {
423         if (!skb_mac_header_was_set(skb))
424                 return 0;
425
426         if (skb_vlan_tag_present(skb)) {
427                 struct nlattr *nest;
428
429                 nest = nla_nest_start(inst->skb, NFULA_VLAN);
430                 if (!nest)
431                         goto nla_put_failure;
432
433                 if (nla_put_be16(inst->skb, NFULA_VLAN_TCI, htons(skb->vlan_tci)) ||
434                     nla_put_be16(inst->skb, NFULA_VLAN_PROTO, skb->vlan_proto))
435                         goto nla_put_failure;
436
437                 nla_nest_end(inst->skb, nest);
438         }
439
440         if (skb->mac_header < skb->network_header) {
441                 int len = (int)(skb->network_header - skb->mac_header);
442
443                 if (nla_put(inst->skb, NFULA_L2HDR, len, skb_mac_header(skb)))
444                         goto nla_put_failure;
445         }
446
447         return 0;
448
449 nla_put_failure:
450         return -1;
451 }
452
453 /* This is an inline function, we don't really care about a long
454  * list of arguments */
455 static inline int
456 __build_packet_message(struct nfnl_log_net *log,
457                         struct nfulnl_instance *inst,
458                         const struct sk_buff *skb,
459                         unsigned int data_len,
460                         u_int8_t pf,
461                         unsigned int hooknum,
462                         const struct net_device *indev,
463                         const struct net_device *outdev,
464                         const char *prefix, unsigned int plen,
465                         const struct nfnl_ct_hook *nfnl_ct,
466                         struct nf_conn *ct, enum ip_conntrack_info ctinfo)
467 {
468         struct nfulnl_msg_packet_hdr pmsg;
469         struct nlmsghdr *nlh;
470         sk_buff_data_t old_tail = inst->skb->tail;
471         struct sock *sk;
472         const unsigned char *hwhdrp;
473
474         nlh = nfnl_msg_put(inst->skb, 0, 0,
475                            nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
476                            0, pf, NFNETLINK_V0, htons(inst->group_num));
477         if (!nlh)
478                 return -1;
479
480         memset(&pmsg, 0, sizeof(pmsg));
481         pmsg.hw_protocol        = skb->protocol;
482         pmsg.hook               = hooknum;
483
484         if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
485                 goto nla_put_failure;
486
487         if (prefix &&
488             nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
489                 goto nla_put_failure;
490
491         if (indev) {
492 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
493                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
494                                  htonl(indev->ifindex)))
495                         goto nla_put_failure;
496 #else
497                 if (pf == PF_BRIDGE) {
498                         /* Case 1: outdev is physical input device, we need to
499                          * look for bridge group (when called from
500                          * netfilter_bridge) */
501                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
502                                          htonl(indev->ifindex)) ||
503                         /* this is the bridge group "brX" */
504                         /* rcu_read_lock()ed by nf_hook_thresh or
505                          * nf_log_packet.
506                          */
507                             nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
508                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
509                                 goto nla_put_failure;
510                 } else {
511                         struct net_device *physindev;
512
513                         /* Case 2: indev is bridge group, we need to look for
514                          * physical device (when called from ipv4) */
515                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
516                                          htonl(indev->ifindex)))
517                                 goto nla_put_failure;
518
519                         physindev = nf_bridge_get_physindev(skb);
520                         if (physindev &&
521                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
522                                          htonl(physindev->ifindex)))
523                                 goto nla_put_failure;
524                 }
525 #endif
526         }
527
528         if (outdev) {
529 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
530                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
531                                  htonl(outdev->ifindex)))
532                         goto nla_put_failure;
533 #else
534                 if (pf == PF_BRIDGE) {
535                         /* Case 1: outdev is physical output device, we need to
536                          * look for bridge group (when called from
537                          * netfilter_bridge) */
538                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
539                                          htonl(outdev->ifindex)) ||
540                         /* this is the bridge group "brX" */
541                         /* rcu_read_lock()ed by nf_hook_thresh or
542                          * nf_log_packet.
543                          */
544                             nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
545                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
546                                 goto nla_put_failure;
547                 } else {
548                         struct net_device *physoutdev;
549
550                         /* Case 2: indev is a bridge group, we need to look
551                          * for physical device (when called from ipv4) */
552                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
553                                          htonl(outdev->ifindex)))
554                                 goto nla_put_failure;
555
556                         physoutdev = nf_bridge_get_physoutdev(skb);
557                         if (physoutdev &&
558                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
559                                          htonl(physoutdev->ifindex)))
560                                 goto nla_put_failure;
561                 }
562 #endif
563         }
564
565         if (skb->mark &&
566             nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
567                 goto nla_put_failure;
568
569         if (indev && skb->dev &&
570             skb_mac_header_was_set(skb) &&
571             skb_mac_header_len(skb) != 0) {
572                 struct nfulnl_msg_packet_hw phw;
573                 int len;
574
575                 memset(&phw, 0, sizeof(phw));
576                 len = dev_parse_header(skb, phw.hw_addr);
577                 if (len > 0) {
578                         phw.hw_addrlen = htons(len);
579                         if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
580                                 goto nla_put_failure;
581                 }
582         }
583
584         if (indev && skb_mac_header_was_set(skb)) {
585                 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
586                     nla_put_be16(inst->skb, NFULA_HWLEN,
587                                  htons(skb->dev->hard_header_len)))
588                         goto nla_put_failure;
589
590                 hwhdrp = skb_mac_header(skb);
591
592                 if (skb->dev->type == ARPHRD_SIT)
593                         hwhdrp -= ETH_HLEN;
594
595                 if (hwhdrp >= skb->head &&
596                     nla_put(inst->skb, NFULA_HWHEADER,
597                             skb->dev->hard_header_len, hwhdrp))
598                         goto nla_put_failure;
599         }
600
601         if (hooknum <= NF_INET_FORWARD) {
602                 struct timespec64 kts = ktime_to_timespec64(skb_tstamp_cond(skb, true));
603                 struct nfulnl_msg_packet_timestamp ts;
604                 ts.sec = cpu_to_be64(kts.tv_sec);
605                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
606
607                 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
608                         goto nla_put_failure;
609         }
610
611         /* UID */
612         sk = skb->sk;
613         if (sk && sk_fullsock(sk)) {
614                 read_lock_bh(&sk->sk_callback_lock);
615                 if (sk->sk_socket && sk->sk_socket->file) {
616                         struct file *file = sk->sk_socket->file;
617                         const struct cred *cred = file->f_cred;
618                         struct user_namespace *user_ns = inst->peer_user_ns;
619                         __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
620                         __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
621                         read_unlock_bh(&sk->sk_callback_lock);
622                         if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
623                             nla_put_be32(inst->skb, NFULA_GID, gid))
624                                 goto nla_put_failure;
625                 } else
626                         read_unlock_bh(&sk->sk_callback_lock);
627         }
628
629         /* local sequence number */
630         if ((inst->flags & NFULNL_CFG_F_SEQ) &&
631             nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
632                 goto nla_put_failure;
633
634         /* global sequence number */
635         if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
636             nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
637                          htonl(atomic_inc_return(&log->global_seq))))
638                 goto nla_put_failure;
639
640         if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
641                                  NFULA_CT, NFULA_CT_INFO) < 0)
642                 goto nla_put_failure;
643
644         if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) &&
645             nfulnl_put_bridge(inst, skb) < 0)
646                 goto nla_put_failure;
647
648         if (data_len) {
649                 struct nlattr *nla;
650                 int size = nla_attr_size(data_len);
651
652                 if (skb_tailroom(inst->skb) < nla_total_size(data_len))
653                         goto nla_put_failure;
654
655                 nla = skb_put(inst->skb, nla_total_size(data_len));
656                 nla->nla_type = NFULA_PAYLOAD;
657                 nla->nla_len = size;
658
659                 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
660                         BUG();
661         }
662
663         nlh->nlmsg_len = inst->skb->tail - old_tail;
664         return 0;
665
666 nla_put_failure:
667         PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
668         return -1;
669 }
670
671 static const struct nf_loginfo default_loginfo = {
672         .type =         NF_LOG_TYPE_ULOG,
673         .u = {
674                 .ulog = {
675                         .copy_len       = 0xffff,
676                         .group          = 0,
677                         .qthreshold     = 1,
678                 },
679         },
680 };
681
682 /* log handler for internal netfilter logging api */
683 static void
684 nfulnl_log_packet(struct net *net,
685                   u_int8_t pf,
686                   unsigned int hooknum,
687                   const struct sk_buff *skb,
688                   const struct net_device *in,
689                   const struct net_device *out,
690                   const struct nf_loginfo *li_user,
691                   const char *prefix)
692 {
693         size_t size;
694         unsigned int data_len;
695         struct nfulnl_instance *inst;
696         const struct nf_loginfo *li;
697         unsigned int qthreshold;
698         unsigned int plen = 0;
699         struct nfnl_log_net *log = nfnl_log_pernet(net);
700         const struct nfnl_ct_hook *nfnl_ct = NULL;
701         struct nf_conn *ct = NULL;
702         enum ip_conntrack_info ctinfo;
703
704         if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
705                 li = li_user;
706         else
707                 li = &default_loginfo;
708
709         inst = instance_lookup_get_rcu(log, li->u.ulog.group);
710         if (!inst)
711                 return;
712
713         if (prefix)
714                 plen = strlen(prefix) + 1;
715
716         /* FIXME: do we want to make the size calculation conditional based on
717          * what is actually present?  way more branches and checks, but more
718          * memory efficient... */
719         size = nlmsg_total_size(sizeof(struct nfgenmsg))
720                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
721                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
722                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
723 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
724                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
725                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
726 #endif
727                 + nla_total_size(sizeof(u_int32_t))     /* mark */
728                 + nla_total_size(sizeof(u_int32_t))     /* uid */
729                 + nla_total_size(sizeof(u_int32_t))     /* gid */
730                 + nla_total_size(plen)                  /* prefix */
731                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
732                 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
733                 + nla_total_size(sizeof(struct nfgenmsg));      /* NLMSG_DONE */
734
735         if (in && skb_mac_header_was_set(skb)) {
736                 size += nla_total_size(skb->dev->hard_header_len)
737                         + nla_total_size(sizeof(u_int16_t))     /* hwtype */
738                         + nla_total_size(sizeof(u_int16_t));    /* hwlen */
739         }
740
741         spin_lock_bh(&inst->lock);
742
743         if (inst->flags & NFULNL_CFG_F_SEQ)
744                 size += nla_total_size(sizeof(u_int32_t));
745         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
746                 size += nla_total_size(sizeof(u_int32_t));
747 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
748         if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
749                 nfnl_ct = rcu_dereference(nfnl_ct_hook);
750                 if (nfnl_ct != NULL) {
751                         ct = nf_ct_get(skb, &ctinfo);
752                         if (ct != NULL)
753                                 size += nfnl_ct->build_size(ct);
754                 }
755         }
756 #endif
757         if (pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE)
758                 size += nfulnl_get_bridge_size(skb);
759
760         qthreshold = inst->qthreshold;
761         /* per-rule qthreshold overrides per-instance */
762         if (li->u.ulog.qthreshold)
763                 if (qthreshold > li->u.ulog.qthreshold)
764                         qthreshold = li->u.ulog.qthreshold;
765
766
767         switch (inst->copy_mode) {
768         case NFULNL_COPY_META:
769         case NFULNL_COPY_NONE:
770                 data_len = 0;
771                 break;
772
773         case NFULNL_COPY_PACKET:
774                 data_len = inst->copy_range;
775                 if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
776                     (li->u.ulog.copy_len < data_len))
777                         data_len = li->u.ulog.copy_len;
778
779                 if (data_len > skb->len)
780                         data_len = skb->len;
781
782                 size += nla_total_size(data_len);
783                 break;
784
785         case NFULNL_COPY_DISABLED:
786         default:
787                 goto unlock_and_release;
788         }
789
790         if (inst->skb && size > skb_tailroom(inst->skb)) {
791                 /* either the queue len is too high or we don't have
792                  * enough room in the skb left. flush to userspace. */
793                 __nfulnl_flush(inst);
794         }
795
796         if (!inst->skb) {
797                 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
798                                              inst->nlbufsiz, size);
799                 if (!inst->skb)
800                         goto alloc_failure;
801         }
802
803         inst->qlen++;
804
805         __build_packet_message(log, inst, skb, data_len, pf,
806                                 hooknum, in, out, prefix, plen,
807                                 nfnl_ct, ct, ctinfo);
808
809         if (inst->qlen >= qthreshold)
810                 __nfulnl_flush(inst);
811         /* timer_pending always called within inst->lock, so there
812          * is no chance of a race here */
813         else if (!timer_pending(&inst->timer)) {
814                 instance_get(inst);
815                 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
816                 add_timer(&inst->timer);
817         }
818
819 unlock_and_release:
820         spin_unlock_bh(&inst->lock);
821         instance_put(inst);
822         return;
823
824 alloc_failure:
825         /* FIXME: statistics */
826         goto unlock_and_release;
827 }
828
829 static int
830 nfulnl_rcv_nl_event(struct notifier_block *this,
831                    unsigned long event, void *ptr)
832 {
833         struct netlink_notify *n = ptr;
834         struct nfnl_log_net *log = nfnl_log_pernet(n->net);
835
836         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
837                 int i;
838
839                 /* destroy all instances for this portid */
840                 spin_lock_bh(&log->instances_lock);
841                 for  (i = 0; i < INSTANCE_BUCKETS; i++) {
842                         struct hlist_node *t2;
843                         struct nfulnl_instance *inst;
844                         struct hlist_head *head = &log->instance_table[i];
845
846                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
847                                 if (n->portid == inst->peer_portid)
848                                         __instance_destroy(inst);
849                         }
850                 }
851                 spin_unlock_bh(&log->instances_lock);
852         }
853         return NOTIFY_DONE;
854 }
855
856 static struct notifier_block nfulnl_rtnl_notifier = {
857         .notifier_call  = nfulnl_rcv_nl_event,
858 };
859
860 static int nfulnl_recv_unsupp(struct sk_buff *skb, const struct nfnl_info *info,
861                               const struct nlattr * const nfula[])
862 {
863         return -ENOTSUPP;
864 }
865
866 static struct nf_logger nfulnl_logger __read_mostly = {
867         .name   = "nfnetlink_log",
868         .type   = NF_LOG_TYPE_ULOG,
869         .logfn  = nfulnl_log_packet,
870         .me     = THIS_MODULE,
871 };
872
873 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
874         [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
875         [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
876         [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
877         [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
878         [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
879         [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
880 };
881
882 static int nfulnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info,
883                               const struct nlattr * const nfula[])
884 {
885         struct nfnl_log_net *log = nfnl_log_pernet(info->net);
886         u_int16_t group_num = ntohs(info->nfmsg->res_id);
887         struct nfulnl_msg_config_cmd *cmd = NULL;
888         struct nfulnl_instance *inst;
889         u16 flags = 0;
890         int ret = 0;
891
892         if (nfula[NFULA_CFG_CMD]) {
893                 u_int8_t pf = info->nfmsg->nfgen_family;
894                 cmd = nla_data(nfula[NFULA_CFG_CMD]);
895
896                 /* Commands without queue context */
897                 switch (cmd->command) {
898                 case NFULNL_CFG_CMD_PF_BIND:
899                         return nf_log_bind_pf(info->net, pf, &nfulnl_logger);
900                 case NFULNL_CFG_CMD_PF_UNBIND:
901                         nf_log_unbind_pf(info->net, pf);
902                         return 0;
903                 }
904         }
905
906         inst = instance_lookup_get(log, group_num);
907         if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
908                 ret = -EPERM;
909                 goto out_put;
910         }
911
912         /* Check if we support these flags in first place, dependencies should
913          * be there too not to break atomicity.
914          */
915         if (nfula[NFULA_CFG_FLAGS]) {
916                 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
917
918                 if ((flags & NFULNL_CFG_F_CONNTRACK) &&
919                     !rcu_access_pointer(nfnl_ct_hook)) {
920 #ifdef CONFIG_MODULES
921                         nfnl_unlock(NFNL_SUBSYS_ULOG);
922                         request_module("ip_conntrack_netlink");
923                         nfnl_lock(NFNL_SUBSYS_ULOG);
924                         if (rcu_access_pointer(nfnl_ct_hook)) {
925                                 ret = -EAGAIN;
926                                 goto out_put;
927                         }
928 #endif
929                         ret = -EOPNOTSUPP;
930                         goto out_put;
931                 }
932         }
933
934         if (cmd != NULL) {
935                 switch (cmd->command) {
936                 case NFULNL_CFG_CMD_BIND:
937                         if (inst) {
938                                 ret = -EBUSY;
939                                 goto out_put;
940                         }
941
942                         inst = instance_create(info->net, group_num,
943                                                NETLINK_CB(skb).portid,
944                                                sk_user_ns(NETLINK_CB(skb).sk));
945                         if (IS_ERR(inst)) {
946                                 ret = PTR_ERR(inst);
947                                 goto out;
948                         }
949                         break;
950                 case NFULNL_CFG_CMD_UNBIND:
951                         if (!inst) {
952                                 ret = -ENODEV;
953                                 goto out;
954                         }
955
956                         instance_destroy(log, inst);
957                         goto out_put;
958                 default:
959                         ret = -ENOTSUPP;
960                         goto out_put;
961                 }
962         } else if (!inst) {
963                 ret = -ENODEV;
964                 goto out;
965         }
966
967         if (nfula[NFULA_CFG_MODE]) {
968                 struct nfulnl_msg_config_mode *params =
969                         nla_data(nfula[NFULA_CFG_MODE]);
970
971                 nfulnl_set_mode(inst, params->copy_mode,
972                                 ntohl(params->copy_range));
973         }
974
975         if (nfula[NFULA_CFG_TIMEOUT]) {
976                 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
977
978                 nfulnl_set_timeout(inst, ntohl(timeout));
979         }
980
981         if (nfula[NFULA_CFG_NLBUFSIZ]) {
982                 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
983
984                 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
985         }
986
987         if (nfula[NFULA_CFG_QTHRESH]) {
988                 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
989
990                 nfulnl_set_qthresh(inst, ntohl(qthresh));
991         }
992
993         if (nfula[NFULA_CFG_FLAGS])
994                 nfulnl_set_flags(inst, flags);
995
996 out_put:
997         instance_put(inst);
998 out:
999         return ret;
1000 }
1001
1002 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
1003         [NFULNL_MSG_PACKET]     = {
1004                 .call           = nfulnl_recv_unsupp,
1005                 .type           = NFNL_CB_MUTEX,
1006                 .attr_count     = NFULA_MAX,
1007         },
1008         [NFULNL_MSG_CONFIG]     = {
1009                 .call           = nfulnl_recv_config,
1010                 .type           = NFNL_CB_MUTEX,
1011                 .attr_count     = NFULA_CFG_MAX,
1012                 .policy         = nfula_cfg_policy
1013         },
1014 };
1015
1016 static const struct nfnetlink_subsystem nfulnl_subsys = {
1017         .name           = "log",
1018         .subsys_id      = NFNL_SUBSYS_ULOG,
1019         .cb_count       = NFULNL_MSG_MAX,
1020         .cb             = nfulnl_cb,
1021 };
1022
1023 #ifdef CONFIG_PROC_FS
1024 struct iter_state {
1025         struct seq_net_private p;
1026         unsigned int bucket;
1027 };
1028
1029 static struct hlist_node *get_first(struct net *net, struct iter_state *st)
1030 {
1031         struct nfnl_log_net *log;
1032         if (!st)
1033                 return NULL;
1034
1035         log = nfnl_log_pernet(net);
1036
1037         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1038                 struct hlist_head *head = &log->instance_table[st->bucket];
1039
1040                 if (!hlist_empty(head))
1041                         return rcu_dereference(hlist_first_rcu(head));
1042         }
1043         return NULL;
1044 }
1045
1046 static struct hlist_node *get_next(struct net *net, struct iter_state *st,
1047                                    struct hlist_node *h)
1048 {
1049         h = rcu_dereference(hlist_next_rcu(h));
1050         while (!h) {
1051                 struct nfnl_log_net *log;
1052                 struct hlist_head *head;
1053
1054                 if (++st->bucket >= INSTANCE_BUCKETS)
1055                         return NULL;
1056
1057                 log = nfnl_log_pernet(net);
1058                 head = &log->instance_table[st->bucket];
1059                 h = rcu_dereference(hlist_first_rcu(head));
1060         }
1061         return h;
1062 }
1063
1064 static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1065                                   loff_t pos)
1066 {
1067         struct hlist_node *head;
1068         head = get_first(net, st);
1069
1070         if (head)
1071                 while (pos && (head = get_next(net, st, head)))
1072                         pos--;
1073         return pos ? NULL : head;
1074 }
1075
1076 static void *seq_start(struct seq_file *s, loff_t *pos)
1077         __acquires(rcu)
1078 {
1079         rcu_read_lock();
1080         return get_idx(seq_file_net(s), s->private, *pos);
1081 }
1082
1083 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1084 {
1085         (*pos)++;
1086         return get_next(seq_file_net(s), s->private, v);
1087 }
1088
1089 static void seq_stop(struct seq_file *s, void *v)
1090         __releases(rcu)
1091 {
1092         rcu_read_unlock();
1093 }
1094
1095 static int seq_show(struct seq_file *s, void *v)
1096 {
1097         const struct nfulnl_instance *inst = v;
1098
1099         seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1100                    inst->group_num,
1101                    inst->peer_portid, inst->qlen,
1102                    inst->copy_mode, inst->copy_range,
1103                    inst->flushtimeout, refcount_read(&inst->use));
1104
1105         return 0;
1106 }
1107
1108 static const struct seq_operations nful_seq_ops = {
1109         .start  = seq_start,
1110         .next   = seq_next,
1111         .stop   = seq_stop,
1112         .show   = seq_show,
1113 };
1114 #endif /* PROC_FS */
1115
1116 static int __net_init nfnl_log_net_init(struct net *net)
1117 {
1118         unsigned int i;
1119         struct nfnl_log_net *log = nfnl_log_pernet(net);
1120 #ifdef CONFIG_PROC_FS
1121         struct proc_dir_entry *proc;
1122         kuid_t root_uid;
1123         kgid_t root_gid;
1124 #endif
1125
1126         for (i = 0; i < INSTANCE_BUCKETS; i++)
1127                 INIT_HLIST_HEAD(&log->instance_table[i]);
1128         spin_lock_init(&log->instances_lock);
1129
1130 #ifdef CONFIG_PROC_FS
1131         proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1132                         &nful_seq_ops, sizeof(struct iter_state));
1133         if (!proc)
1134                 return -ENOMEM;
1135
1136         root_uid = make_kuid(net->user_ns, 0);
1137         root_gid = make_kgid(net->user_ns, 0);
1138         if (uid_valid(root_uid) && gid_valid(root_gid))
1139                 proc_set_user(proc, root_uid, root_gid);
1140 #endif
1141         return 0;
1142 }
1143
1144 static void __net_exit nfnl_log_net_exit(struct net *net)
1145 {
1146         struct nfnl_log_net *log = nfnl_log_pernet(net);
1147         unsigned int i;
1148
1149 #ifdef CONFIG_PROC_FS
1150         remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1151 #endif
1152         nf_log_unset(net, &nfulnl_logger);
1153         for (i = 0; i < INSTANCE_BUCKETS; i++)
1154                 WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1155 }
1156
1157 static struct pernet_operations nfnl_log_net_ops = {
1158         .init   = nfnl_log_net_init,
1159         .exit   = nfnl_log_net_exit,
1160         .id     = &nfnl_log_net_id,
1161         .size   = sizeof(struct nfnl_log_net),
1162 };
1163
1164 static int __init nfnetlink_log_init(void)
1165 {
1166         int status;
1167
1168         status = register_pernet_subsys(&nfnl_log_net_ops);
1169         if (status < 0) {
1170                 pr_err("failed to register pernet ops\n");
1171                 goto out;
1172         }
1173
1174         netlink_register_notifier(&nfulnl_rtnl_notifier);
1175         status = nfnetlink_subsys_register(&nfulnl_subsys);
1176         if (status < 0) {
1177                 pr_err("failed to create netlink socket\n");
1178                 goto cleanup_netlink_notifier;
1179         }
1180
1181         status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1182         if (status < 0) {
1183                 pr_err("failed to register logger\n");
1184                 goto cleanup_subsys;
1185         }
1186
1187         return status;
1188
1189 cleanup_subsys:
1190         nfnetlink_subsys_unregister(&nfulnl_subsys);
1191 cleanup_netlink_notifier:
1192         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1193         unregister_pernet_subsys(&nfnl_log_net_ops);
1194 out:
1195         return status;
1196 }
1197
1198 static void __exit nfnetlink_log_fini(void)
1199 {
1200         nfnetlink_subsys_unregister(&nfulnl_subsys);
1201         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1202         unregister_pernet_subsys(&nfnl_log_net_ops);
1203         nf_log_unregister(&nfulnl_logger);
1204 }
1205
1206 MODULE_DESCRIPTION("netfilter userspace logging");
1207 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1208 MODULE_LICENSE("GPL");
1209 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1210 MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1211 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1212 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1213 MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1214 MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1215
1216 module_init(nfnetlink_log_init);
1217 module_exit(nfnetlink_log_fini);