]> git.kernelconcepts.de Git - karo-tx-linux.git/blob - net/netfilter/nfnetlink_queue.c
Merge branch 'for-linus' of git://git.kernel.dk/linux-block
[karo-tx-linux.git] / net / netfilter / nfnetlink_queue.c
1 /*
2  * This is a module which is used for queueing packets and communicating with
3  * userspace via nfnetlink.
4  *
5  * (C) 2005 by Harald Welte <laforge@netfilter.org>
6  * (C) 2007 by Patrick McHardy <kaber@trash.net>
7  *
8  * Based on the old ipv4-only ip_queue.c:
9  * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
10  * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
11  *
12  * This program is free software; you can redistribute it and/or modify
13  * it under the terms of the GNU General Public License version 2 as
14  * published by the Free Software Foundation.
15  *
16  */
17 #include <linux/module.h>
18 #include <linux/skbuff.h>
19 #include <linux/init.h>
20 #include <linux/spinlock.h>
21 #include <linux/slab.h>
22 #include <linux/notifier.h>
23 #include <linux/netdevice.h>
24 #include <linux/netfilter.h>
25 #include <linux/proc_fs.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/netfilter_ipv6.h>
28 #include <linux/netfilter_bridge.h>
29 #include <linux/netfilter/nfnetlink.h>
30 #include <linux/netfilter/nfnetlink_queue.h>
31 #include <linux/netfilter/nf_conntrack_common.h>
32 #include <linux/list.h>
33 #include <net/sock.h>
34 #include <net/tcp_states.h>
35 #include <net/netfilter/nf_queue.h>
36 #include <net/netns/generic.h>
37
38 #include <linux/atomic.h>
39
40 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
41 #include "../bridge/br_private.h"
42 #endif
43
44 #define NFQNL_QMAX_DEFAULT 1024
45
46 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
47  * includes the header length. Thus, the maximum packet length that we
48  * support is 65531 bytes. We send truncated packets if the specified length
49  * is larger than that.  Userspace can check for presence of NFQA_CAP_LEN
50  * attribute to detect truncation.
51  */
52 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
53
54 struct nfqnl_instance {
55         struct hlist_node hlist;                /* global list of queues */
56         struct rcu_head rcu;
57
58         u32 peer_portid;
59         unsigned int queue_maxlen;
60         unsigned int copy_range;
61         unsigned int queue_dropped;
62         unsigned int queue_user_dropped;
63
64
65         u_int16_t queue_num;                    /* number of this queue */
66         u_int8_t copy_mode;
67         u_int32_t flags;                        /* Set using NFQA_CFG_FLAGS */
68 /*
69  * Following fields are dirtied for each queued packet,
70  * keep them in same cache line if possible.
71  */
72         spinlock_t      lock    ____cacheline_aligned_in_smp;
73         unsigned int    queue_total;
74         unsigned int    id_sequence;            /* 'sequence' of pkt ids */
75         struct list_head queue_list;            /* packets in queue */
76 };
77
78 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
79
80 static unsigned int nfnl_queue_net_id __read_mostly;
81
82 #define INSTANCE_BUCKETS        16
83 struct nfnl_queue_net {
84         spinlock_t instances_lock;
85         struct hlist_head instance_table[INSTANCE_BUCKETS];
86 };
87
88 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
89 {
90         return net_generic(net, nfnl_queue_net_id);
91 }
92
93 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
94 {
95         return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
96 }
97
98 static struct nfqnl_instance *
99 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
100 {
101         struct hlist_head *head;
102         struct nfqnl_instance *inst;
103
104         head = &q->instance_table[instance_hashfn(queue_num)];
105         hlist_for_each_entry_rcu(inst, head, hlist) {
106                 if (inst->queue_num == queue_num)
107                         return inst;
108         }
109         return NULL;
110 }
111
112 static struct nfqnl_instance *
113 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
114 {
115         struct nfqnl_instance *inst;
116         unsigned int h;
117         int err;
118
119         spin_lock(&q->instances_lock);
120         if (instance_lookup(q, queue_num)) {
121                 err = -EEXIST;
122                 goto out_unlock;
123         }
124
125         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
126         if (!inst) {
127                 err = -ENOMEM;
128                 goto out_unlock;
129         }
130
131         inst->queue_num = queue_num;
132         inst->peer_portid = portid;
133         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
134         inst->copy_range = NFQNL_MAX_COPY_RANGE;
135         inst->copy_mode = NFQNL_COPY_NONE;
136         spin_lock_init(&inst->lock);
137         INIT_LIST_HEAD(&inst->queue_list);
138
139         if (!try_module_get(THIS_MODULE)) {
140                 err = -EAGAIN;
141                 goto out_free;
142         }
143
144         h = instance_hashfn(queue_num);
145         hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
146
147         spin_unlock(&q->instances_lock);
148
149         return inst;
150
151 out_free:
152         kfree(inst);
153 out_unlock:
154         spin_unlock(&q->instances_lock);
155         return ERR_PTR(err);
156 }
157
158 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
159                         unsigned long data);
160
161 static void
162 instance_destroy_rcu(struct rcu_head *head)
163 {
164         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
165                                                    rcu);
166
167         nfqnl_flush(inst, NULL, 0);
168         kfree(inst);
169         module_put(THIS_MODULE);
170 }
171
172 static void
173 __instance_destroy(struct nfqnl_instance *inst)
174 {
175         hlist_del_rcu(&inst->hlist);
176         call_rcu(&inst->rcu, instance_destroy_rcu);
177 }
178
179 static void
180 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
181 {
182         spin_lock(&q->instances_lock);
183         __instance_destroy(inst);
184         spin_unlock(&q->instances_lock);
185 }
186
187 static inline void
188 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
189 {
190        list_add_tail(&entry->list, &queue->queue_list);
191        queue->queue_total++;
192 }
193
194 static void
195 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
196 {
197         list_del(&entry->list);
198         queue->queue_total--;
199 }
200
201 static struct nf_queue_entry *
202 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
203 {
204         struct nf_queue_entry *entry = NULL, *i;
205
206         spin_lock_bh(&queue->lock);
207
208         list_for_each_entry(i, &queue->queue_list, list) {
209                 if (i->id == id) {
210                         entry = i;
211                         break;
212                 }
213         }
214
215         if (entry)
216                 __dequeue_entry(queue, entry);
217
218         spin_unlock_bh(&queue->lock);
219
220         return entry;
221 }
222
223 static void
224 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
225 {
226         struct nf_queue_entry *entry, *next;
227
228         spin_lock_bh(&queue->lock);
229         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
230                 if (!cmpfn || cmpfn(entry, data)) {
231                         list_del(&entry->list);
232                         queue->queue_total--;
233                         nf_reinject(entry, NF_DROP);
234                 }
235         }
236         spin_unlock_bh(&queue->lock);
237 }
238
239 static int
240 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
241                       bool csum_verify)
242 {
243         __u32 flags = 0;
244
245         if (packet->ip_summed == CHECKSUM_PARTIAL)
246                 flags = NFQA_SKB_CSUMNOTREADY;
247         else if (csum_verify)
248                 flags = NFQA_SKB_CSUM_NOTVERIFIED;
249
250         if (skb_is_gso(packet))
251                 flags |= NFQA_SKB_GSO;
252
253         return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
254 }
255
256 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
257 {
258         const struct cred *cred;
259
260         if (!sk_fullsock(sk))
261                 return 0;
262
263         read_lock_bh(&sk->sk_callback_lock);
264         if (sk->sk_socket && sk->sk_socket->file) {
265                 cred = sk->sk_socket->file->f_cred;
266                 if (nla_put_be32(skb, NFQA_UID,
267                     htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
268                         goto nla_put_failure;
269                 if (nla_put_be32(skb, NFQA_GID,
270                     htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
271                         goto nla_put_failure;
272         }
273         read_unlock_bh(&sk->sk_callback_lock);
274         return 0;
275
276 nla_put_failure:
277         read_unlock_bh(&sk->sk_callback_lock);
278         return -1;
279 }
280
281 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
282 {
283         u32 seclen = 0;
284 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
285         if (!skb || !sk_fullsock(skb->sk))
286                 return 0;
287
288         read_lock_bh(&skb->sk->sk_callback_lock);
289
290         if (skb->secmark)
291                 security_secid_to_secctx(skb->secmark, secdata, &seclen);
292
293         read_unlock_bh(&skb->sk->sk_callback_lock);
294 #endif
295         return seclen;
296 }
297
298 static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
299 {
300         struct sk_buff *entskb = entry->skb;
301         u32 nlalen = 0;
302
303         if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
304                 return 0;
305
306         if (skb_vlan_tag_present(entskb))
307                 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
308                                          nla_total_size(sizeof(__be16)));
309
310         if (entskb->network_header > entskb->mac_header)
311                 nlalen += nla_total_size((entskb->network_header -
312                                           entskb->mac_header));
313
314         return nlalen;
315 }
316
317 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
318 {
319         struct sk_buff *entskb = entry->skb;
320
321         if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
322                 return 0;
323
324         if (skb_vlan_tag_present(entskb)) {
325                 struct nlattr *nest;
326
327                 nest = nla_nest_start(skb, NFQA_VLAN | NLA_F_NESTED);
328                 if (!nest)
329                         goto nla_put_failure;
330
331                 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
332                     nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
333                         goto nla_put_failure;
334
335                 nla_nest_end(skb, nest);
336         }
337
338         if (entskb->mac_header < entskb->network_header) {
339                 int len = (int)(entskb->network_header - entskb->mac_header);
340
341                 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
342                         goto nla_put_failure;
343         }
344
345         return 0;
346
347 nla_put_failure:
348         return -1;
349 }
350
351 static struct sk_buff *
352 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
353                            struct nf_queue_entry *entry,
354                            __be32 **packet_id_ptr)
355 {
356         size_t size;
357         size_t data_len = 0, cap_len = 0;
358         unsigned int hlen = 0;
359         struct sk_buff *skb;
360         struct nlattr *nla;
361         struct nfqnl_msg_packet_hdr *pmsg;
362         struct nlmsghdr *nlh;
363         struct nfgenmsg *nfmsg;
364         struct sk_buff *entskb = entry->skb;
365         struct net_device *indev;
366         struct net_device *outdev;
367         struct nf_conn *ct = NULL;
368         enum ip_conntrack_info uninitialized_var(ctinfo);
369         struct nfnl_ct_hook *nfnl_ct;
370         bool csum_verify;
371         char *secdata = NULL;
372         u32 seclen = 0;
373
374         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
375                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
376                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
377                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
378 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
379                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
380                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
381 #endif
382                 + nla_total_size(sizeof(u_int32_t))     /* mark */
383                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
384                 + nla_total_size(sizeof(u_int32_t))     /* skbinfo */
385                 + nla_total_size(sizeof(u_int32_t));    /* cap_len */
386
387         if (entskb->tstamp)
388                 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
389
390         size += nfqnl_get_bridge_size(entry);
391
392         if (entry->state.hook <= NF_INET_FORWARD ||
393            (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
394                 csum_verify = !skb_csum_unnecessary(entskb);
395         else
396                 csum_verify = false;
397
398         outdev = entry->state.out;
399
400         switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
401         case NFQNL_COPY_META:
402         case NFQNL_COPY_NONE:
403                 break;
404
405         case NFQNL_COPY_PACKET:
406                 if (!(queue->flags & NFQA_CFG_F_GSO) &&
407                     entskb->ip_summed == CHECKSUM_PARTIAL &&
408                     skb_checksum_help(entskb))
409                         return NULL;
410
411                 data_len = ACCESS_ONCE(queue->copy_range);
412                 if (data_len > entskb->len)
413                         data_len = entskb->len;
414
415                 hlen = skb_zerocopy_headlen(entskb);
416                 hlen = min_t(unsigned int, hlen, data_len);
417                 size += sizeof(struct nlattr) + hlen;
418                 cap_len = entskb->len;
419                 break;
420         }
421
422         nfnl_ct = rcu_dereference(nfnl_ct_hook);
423
424         if (queue->flags & NFQA_CFG_F_CONNTRACK) {
425                 if (nfnl_ct != NULL) {
426                         ct = nfnl_ct->get_ct(entskb, &ctinfo);
427                         if (ct != NULL)
428                                 size += nfnl_ct->build_size(ct);
429                 }
430         }
431
432         if (queue->flags & NFQA_CFG_F_UID_GID) {
433                 size +=  (nla_total_size(sizeof(u_int32_t))     /* uid */
434                         + nla_total_size(sizeof(u_int32_t)));   /* gid */
435         }
436
437         if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
438                 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
439                 if (seclen)
440                         size += nla_total_size(seclen);
441         }
442
443         skb = alloc_skb(size, GFP_ATOMIC);
444         if (!skb) {
445                 skb_tx_error(entskb);
446                 goto nlmsg_failure;
447         }
448
449         nlh = nlmsg_put(skb, 0, 0,
450                         nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
451                         sizeof(struct nfgenmsg), 0);
452         if (!nlh) {
453                 skb_tx_error(entskb);
454                 kfree_skb(skb);
455                 goto nlmsg_failure;
456         }
457         nfmsg = nlmsg_data(nlh);
458         nfmsg->nfgen_family = entry->state.pf;
459         nfmsg->version = NFNETLINK_V0;
460         nfmsg->res_id = htons(queue->queue_num);
461
462         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
463         pmsg = nla_data(nla);
464         pmsg->hw_protocol       = entskb->protocol;
465         pmsg->hook              = entry->state.hook;
466         *packet_id_ptr          = &pmsg->packet_id;
467
468         indev = entry->state.in;
469         if (indev) {
470 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
471                 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
472                         goto nla_put_failure;
473 #else
474                 if (entry->state.pf == PF_BRIDGE) {
475                         /* Case 1: indev is physical input device, we need to
476                          * look for bridge group (when called from
477                          * netfilter_bridge) */
478                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
479                                          htonl(indev->ifindex)) ||
480                         /* this is the bridge group "brX" */
481                         /* rcu_read_lock()ed by __nf_queue */
482                             nla_put_be32(skb, NFQA_IFINDEX_INDEV,
483                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
484                                 goto nla_put_failure;
485                 } else {
486                         int physinif;
487
488                         /* Case 2: indev is bridge group, we need to look for
489                          * physical device (when called from ipv4) */
490                         if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
491                                          htonl(indev->ifindex)))
492                                 goto nla_put_failure;
493
494                         physinif = nf_bridge_get_physinif(entskb);
495                         if (physinif &&
496                             nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
497                                          htonl(physinif)))
498                                 goto nla_put_failure;
499                 }
500 #endif
501         }
502
503         if (outdev) {
504 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
505                 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
506                         goto nla_put_failure;
507 #else
508                 if (entry->state.pf == PF_BRIDGE) {
509                         /* Case 1: outdev is physical output device, we need to
510                          * look for bridge group (when called from
511                          * netfilter_bridge) */
512                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
513                                          htonl(outdev->ifindex)) ||
514                         /* this is the bridge group "brX" */
515                         /* rcu_read_lock()ed by __nf_queue */
516                             nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
517                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
518                                 goto nla_put_failure;
519                 } else {
520                         int physoutif;
521
522                         /* Case 2: outdev is bridge group, we need to look for
523                          * physical output device (when called from ipv4) */
524                         if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
525                                          htonl(outdev->ifindex)))
526                                 goto nla_put_failure;
527
528                         physoutif = nf_bridge_get_physoutif(entskb);
529                         if (physoutif &&
530                             nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
531                                          htonl(physoutif)))
532                                 goto nla_put_failure;
533                 }
534 #endif
535         }
536
537         if (entskb->mark &&
538             nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
539                 goto nla_put_failure;
540
541         if (indev && entskb->dev &&
542             entskb->mac_header != entskb->network_header) {
543                 struct nfqnl_msg_packet_hw phw;
544                 int len;
545
546                 memset(&phw, 0, sizeof(phw));
547                 len = dev_parse_header(entskb, phw.hw_addr);
548                 if (len) {
549                         phw.hw_addrlen = htons(len);
550                         if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
551                                 goto nla_put_failure;
552                 }
553         }
554
555         if (nfqnl_put_bridge(entry, skb) < 0)
556                 goto nla_put_failure;
557
558         if (entskb->tstamp) {
559                 struct nfqnl_msg_packet_timestamp ts;
560                 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
561
562                 ts.sec = cpu_to_be64(kts.tv_sec);
563                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
564
565                 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
566                         goto nla_put_failure;
567         }
568
569         if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
570             nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
571                 goto nla_put_failure;
572
573         if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
574                 goto nla_put_failure;
575
576         if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
577                 goto nla_put_failure;
578
579         if (cap_len > data_len &&
580             nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
581                 goto nla_put_failure;
582
583         if (nfqnl_put_packet_info(skb, entskb, csum_verify))
584                 goto nla_put_failure;
585
586         if (data_len) {
587                 struct nlattr *nla;
588
589                 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
590                         goto nla_put_failure;
591
592                 nla = skb_put(skb, sizeof(*nla));
593                 nla->nla_type = NFQA_PAYLOAD;
594                 nla->nla_len = nla_attr_size(data_len);
595
596                 if (skb_zerocopy(skb, entskb, data_len, hlen))
597                         goto nla_put_failure;
598         }
599
600         nlh->nlmsg_len = skb->len;
601         if (seclen)
602                 security_release_secctx(secdata, seclen);
603         return skb;
604
605 nla_put_failure:
606         skb_tx_error(entskb);
607         kfree_skb(skb);
608         net_err_ratelimited("nf_queue: error creating packet message\n");
609 nlmsg_failure:
610         if (seclen)
611                 security_release_secctx(secdata, seclen);
612         return NULL;
613 }
614
615 static int
616 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
617                         struct nf_queue_entry *entry)
618 {
619         struct sk_buff *nskb;
620         int err = -ENOBUFS;
621         __be32 *packet_id_ptr;
622         int failopen = 0;
623
624         nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
625         if (nskb == NULL) {
626                 err = -ENOMEM;
627                 goto err_out;
628         }
629         spin_lock_bh(&queue->lock);
630
631         if (queue->queue_total >= queue->queue_maxlen) {
632                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
633                         failopen = 1;
634                         err = 0;
635                 } else {
636                         queue->queue_dropped++;
637                         net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
638                                              queue->queue_total);
639                 }
640                 goto err_out_free_nskb;
641         }
642         entry->id = ++queue->id_sequence;
643         *packet_id_ptr = htonl(entry->id);
644
645         /* nfnetlink_unicast will either free the nskb or add it to a socket */
646         err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
647         if (err < 0) {
648                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
649                         failopen = 1;
650                         err = 0;
651                 } else {
652                         queue->queue_user_dropped++;
653                 }
654                 goto err_out_unlock;
655         }
656
657         __enqueue_entry(queue, entry);
658
659         spin_unlock_bh(&queue->lock);
660         return 0;
661
662 err_out_free_nskb:
663         kfree_skb(nskb);
664 err_out_unlock:
665         spin_unlock_bh(&queue->lock);
666         if (failopen)
667                 nf_reinject(entry, NF_ACCEPT);
668 err_out:
669         return err;
670 }
671
672 static struct nf_queue_entry *
673 nf_queue_entry_dup(struct nf_queue_entry *e)
674 {
675         struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
676         if (entry)
677                 nf_queue_entry_get_refs(entry);
678         return entry;
679 }
680
681 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
682 /* When called from bridge netfilter, skb->data must point to MAC header
683  * before calling skb_gso_segment(). Else, original MAC header is lost
684  * and segmented skbs will be sent to wrong destination.
685  */
686 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
687 {
688         if (skb->nf_bridge)
689                 __skb_push(skb, skb->network_header - skb->mac_header);
690 }
691
692 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
693 {
694         if (skb->nf_bridge)
695                 __skb_pull(skb, skb->network_header - skb->mac_header);
696 }
697 #else
698 #define nf_bridge_adjust_skb_data(s) do {} while (0)
699 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
700 #endif
701
702 static void free_entry(struct nf_queue_entry *entry)
703 {
704         nf_queue_entry_release_refs(entry);
705         kfree(entry);
706 }
707
708 static int
709 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
710                            struct sk_buff *skb, struct nf_queue_entry *entry)
711 {
712         int ret = -ENOMEM;
713         struct nf_queue_entry *entry_seg;
714
715         nf_bridge_adjust_segmented_data(skb);
716
717         if (skb->next == NULL) { /* last packet, no need to copy entry */
718                 struct sk_buff *gso_skb = entry->skb;
719                 entry->skb = skb;
720                 ret = __nfqnl_enqueue_packet(net, queue, entry);
721                 if (ret)
722                         entry->skb = gso_skb;
723                 return ret;
724         }
725
726         skb->next = NULL;
727
728         entry_seg = nf_queue_entry_dup(entry);
729         if (entry_seg) {
730                 entry_seg->skb = skb;
731                 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
732                 if (ret)
733                         free_entry(entry_seg);
734         }
735         return ret;
736 }
737
738 static int
739 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
740 {
741         unsigned int queued;
742         struct nfqnl_instance *queue;
743         struct sk_buff *skb, *segs;
744         int err = -ENOBUFS;
745         struct net *net = entry->state.net;
746         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
747
748         /* rcu_read_lock()ed by nf_hook_thresh */
749         queue = instance_lookup(q, queuenum);
750         if (!queue)
751                 return -ESRCH;
752
753         if (queue->copy_mode == NFQNL_COPY_NONE)
754                 return -EINVAL;
755
756         skb = entry->skb;
757
758         switch (entry->state.pf) {
759         case NFPROTO_IPV4:
760                 skb->protocol = htons(ETH_P_IP);
761                 break;
762         case NFPROTO_IPV6:
763                 skb->protocol = htons(ETH_P_IPV6);
764                 break;
765         }
766
767         if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
768                 return __nfqnl_enqueue_packet(net, queue, entry);
769
770         nf_bridge_adjust_skb_data(skb);
771         segs = skb_gso_segment(skb, 0);
772         /* Does not use PTR_ERR to limit the number of error codes that can be
773          * returned by nf_queue.  For instance, callers rely on -ESRCH to
774          * mean 'ignore this hook'.
775          */
776         if (IS_ERR_OR_NULL(segs))
777                 goto out_err;
778         queued = 0;
779         err = 0;
780         do {
781                 struct sk_buff *nskb = segs->next;
782                 if (err == 0)
783                         err = __nfqnl_enqueue_packet_gso(net, queue,
784                                                         segs, entry);
785                 if (err == 0)
786                         queued++;
787                 else
788                         kfree_skb(segs);
789                 segs = nskb;
790         } while (segs);
791
792         if (queued) {
793                 if (err) /* some segments are already queued */
794                         free_entry(entry);
795                 kfree_skb(skb);
796                 return 0;
797         }
798  out_err:
799         nf_bridge_adjust_segmented_data(skb);
800         return err;
801 }
802
803 static int
804 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
805 {
806         struct sk_buff *nskb;
807
808         if (diff < 0) {
809                 if (pskb_trim(e->skb, data_len))
810                         return -ENOMEM;
811         } else if (diff > 0) {
812                 if (data_len > 0xFFFF)
813                         return -EINVAL;
814                 if (diff > skb_tailroom(e->skb)) {
815                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
816                                                diff, GFP_ATOMIC);
817                         if (!nskb) {
818                                 printk(KERN_WARNING "nf_queue: OOM "
819                                       "in mangle, dropping packet\n");
820                                 return -ENOMEM;
821                         }
822                         kfree_skb(e->skb);
823                         e->skb = nskb;
824                 }
825                 skb_put(e->skb, diff);
826         }
827         if (!skb_make_writable(e->skb, data_len))
828                 return -ENOMEM;
829         skb_copy_to_linear_data(e->skb, data, data_len);
830         e->skb->ip_summed = CHECKSUM_NONE;
831         return 0;
832 }
833
834 static int
835 nfqnl_set_mode(struct nfqnl_instance *queue,
836                unsigned char mode, unsigned int range)
837 {
838         int status = 0;
839
840         spin_lock_bh(&queue->lock);
841         switch (mode) {
842         case NFQNL_COPY_NONE:
843         case NFQNL_COPY_META:
844                 queue->copy_mode = mode;
845                 queue->copy_range = 0;
846                 break;
847
848         case NFQNL_COPY_PACKET:
849                 queue->copy_mode = mode;
850                 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
851                         queue->copy_range = NFQNL_MAX_COPY_RANGE;
852                 else
853                         queue->copy_range = range;
854                 break;
855
856         default:
857                 status = -EINVAL;
858
859         }
860         spin_unlock_bh(&queue->lock);
861
862         return status;
863 }
864
865 static int
866 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
867 {
868         if (entry->state.in)
869                 if (entry->state.in->ifindex == ifindex)
870                         return 1;
871         if (entry->state.out)
872                 if (entry->state.out->ifindex == ifindex)
873                         return 1;
874 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
875         if (entry->skb->nf_bridge) {
876                 int physinif, physoutif;
877
878                 physinif = nf_bridge_get_physinif(entry->skb);
879                 physoutif = nf_bridge_get_physoutif(entry->skb);
880
881                 if (physinif == ifindex || physoutif == ifindex)
882                         return 1;
883         }
884 #endif
885         return 0;
886 }
887
888 /* drop all packets with either indev or outdev == ifindex from all queue
889  * instances */
890 static void
891 nfqnl_dev_drop(struct net *net, int ifindex)
892 {
893         int i;
894         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
895
896         rcu_read_lock();
897
898         for (i = 0; i < INSTANCE_BUCKETS; i++) {
899                 struct nfqnl_instance *inst;
900                 struct hlist_head *head = &q->instance_table[i];
901
902                 hlist_for_each_entry_rcu(inst, head, hlist)
903                         nfqnl_flush(inst, dev_cmp, ifindex);
904         }
905
906         rcu_read_unlock();
907 }
908
909 static int
910 nfqnl_rcv_dev_event(struct notifier_block *this,
911                     unsigned long event, void *ptr)
912 {
913         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
914
915         /* Drop any packets associated with the downed device */
916         if (event == NETDEV_DOWN)
917                 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
918         return NOTIFY_DONE;
919 }
920
921 static struct notifier_block nfqnl_dev_notifier = {
922         .notifier_call  = nfqnl_rcv_dev_event,
923 };
924
925 static unsigned int nfqnl_nf_hook_drop(struct net *net)
926 {
927         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
928         unsigned int instances = 0;
929         int i;
930
931         rcu_read_lock();
932         for (i = 0; i < INSTANCE_BUCKETS; i++) {
933                 struct nfqnl_instance *inst;
934                 struct hlist_head *head = &q->instance_table[i];
935
936                 hlist_for_each_entry_rcu(inst, head, hlist) {
937                         nfqnl_flush(inst, NULL, 0);
938                         instances++;
939                 }
940         }
941         rcu_read_unlock();
942
943         return instances;
944 }
945
946 static int
947 nfqnl_rcv_nl_event(struct notifier_block *this,
948                    unsigned long event, void *ptr)
949 {
950         struct netlink_notify *n = ptr;
951         struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
952
953         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
954                 int i;
955
956                 /* destroy all instances for this portid */
957                 spin_lock(&q->instances_lock);
958                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
959                         struct hlist_node *t2;
960                         struct nfqnl_instance *inst;
961                         struct hlist_head *head = &q->instance_table[i];
962
963                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
964                                 if (n->portid == inst->peer_portid)
965                                         __instance_destroy(inst);
966                         }
967                 }
968                 spin_unlock(&q->instances_lock);
969         }
970         return NOTIFY_DONE;
971 }
972
973 static struct notifier_block nfqnl_rtnl_notifier = {
974         .notifier_call  = nfqnl_rcv_nl_event,
975 };
976
977 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
978         [NFQA_VLAN_TCI]         = { .type = NLA_U16},
979         [NFQA_VLAN_PROTO]       = { .type = NLA_U16},
980 };
981
982 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
983         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
984         [NFQA_MARK]             = { .type = NLA_U32 },
985         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
986         [NFQA_CT]               = { .type = NLA_UNSPEC },
987         [NFQA_EXP]              = { .type = NLA_UNSPEC },
988         [NFQA_VLAN]             = { .type = NLA_NESTED },
989 };
990
991 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
992         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
993         [NFQA_MARK]             = { .type = NLA_U32 },
994 };
995
996 static struct nfqnl_instance *
997 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
998 {
999         struct nfqnl_instance *queue;
1000
1001         queue = instance_lookup(q, queue_num);
1002         if (!queue)
1003                 return ERR_PTR(-ENODEV);
1004
1005         if (queue->peer_portid != nlportid)
1006                 return ERR_PTR(-EPERM);
1007
1008         return queue;
1009 }
1010
1011 static struct nfqnl_msg_verdict_hdr*
1012 verdicthdr_get(const struct nlattr * const nfqa[])
1013 {
1014         struct nfqnl_msg_verdict_hdr *vhdr;
1015         unsigned int verdict;
1016
1017         if (!nfqa[NFQA_VERDICT_HDR])
1018                 return NULL;
1019
1020         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1021         verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1022         if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1023                 return NULL;
1024         return vhdr;
1025 }
1026
1027 static int nfq_id_after(unsigned int id, unsigned int max)
1028 {
1029         return (int)(id - max) > 0;
1030 }
1031
1032 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1033                                     struct sk_buff *skb,
1034                                     const struct nlmsghdr *nlh,
1035                                     const struct nlattr * const nfqa[],
1036                                     struct netlink_ext_ack *extack)
1037 {
1038         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1039         struct nf_queue_entry *entry, *tmp;
1040         unsigned int verdict, maxid;
1041         struct nfqnl_msg_verdict_hdr *vhdr;
1042         struct nfqnl_instance *queue;
1043         LIST_HEAD(batch_list);
1044         u16 queue_num = ntohs(nfmsg->res_id);
1045         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1046
1047         queue = verdict_instance_lookup(q, queue_num,
1048                                         NETLINK_CB(skb).portid);
1049         if (IS_ERR(queue))
1050                 return PTR_ERR(queue);
1051
1052         vhdr = verdicthdr_get(nfqa);
1053         if (!vhdr)
1054                 return -EINVAL;
1055
1056         verdict = ntohl(vhdr->verdict);
1057         maxid = ntohl(vhdr->id);
1058
1059         spin_lock_bh(&queue->lock);
1060
1061         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1062                 if (nfq_id_after(entry->id, maxid))
1063                         break;
1064                 __dequeue_entry(queue, entry);
1065                 list_add_tail(&entry->list, &batch_list);
1066         }
1067
1068         spin_unlock_bh(&queue->lock);
1069
1070         if (list_empty(&batch_list))
1071                 return -ENOENT;
1072
1073         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1074                 if (nfqa[NFQA_MARK])
1075                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1076                 nf_reinject(entry, verdict);
1077         }
1078         return 0;
1079 }
1080
1081 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1082                                       const struct nlmsghdr *nlh,
1083                                       const struct nlattr * const nfqa[],
1084                                       struct nf_queue_entry *entry,
1085                                       enum ip_conntrack_info *ctinfo)
1086 {
1087         struct nf_conn *ct;
1088
1089         ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1090         if (ct == NULL)
1091                 return NULL;
1092
1093         if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1094                 return NULL;
1095
1096         if (nfqa[NFQA_EXP])
1097                 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1098                                       NETLINK_CB(entry->skb).portid,
1099                                       nlmsg_report(nlh));
1100         return ct;
1101 }
1102
1103 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1104                              const struct nlattr * const nfqa[])
1105 {
1106         if (nfqa[NFQA_VLAN]) {
1107                 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1108                 int err;
1109
1110                 err = nla_parse_nested(tb, NFQA_VLAN_MAX, nfqa[NFQA_VLAN],
1111                                        nfqa_vlan_policy, NULL);
1112                 if (err < 0)
1113                         return err;
1114
1115                 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1116                         return -EINVAL;
1117
1118                 entry->skb->vlan_tci = ntohs(nla_get_be16(tb[NFQA_VLAN_TCI]));
1119                 entry->skb->vlan_proto = nla_get_be16(tb[NFQA_VLAN_PROTO]);
1120         }
1121
1122         if (nfqa[NFQA_L2HDR]) {
1123                 int mac_header_len = entry->skb->network_header -
1124                         entry->skb->mac_header;
1125
1126                 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1127                         return -EINVAL;
1128                 else if (mac_header_len > 0)
1129                         memcpy(skb_mac_header(entry->skb),
1130                                nla_data(nfqa[NFQA_L2HDR]),
1131                                mac_header_len);
1132         }
1133
1134         return 0;
1135 }
1136
1137 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1138                               struct sk_buff *skb,
1139                               const struct nlmsghdr *nlh,
1140                               const struct nlattr * const nfqa[],
1141                               struct netlink_ext_ack *extack)
1142 {
1143         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1144         u_int16_t queue_num = ntohs(nfmsg->res_id);
1145         struct nfqnl_msg_verdict_hdr *vhdr;
1146         struct nfqnl_instance *queue;
1147         unsigned int verdict;
1148         struct nf_queue_entry *entry;
1149         enum ip_conntrack_info uninitialized_var(ctinfo);
1150         struct nfnl_ct_hook *nfnl_ct;
1151         struct nf_conn *ct = NULL;
1152         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1153         int err;
1154
1155         queue = verdict_instance_lookup(q, queue_num,
1156                                         NETLINK_CB(skb).portid);
1157         if (IS_ERR(queue))
1158                 return PTR_ERR(queue);
1159
1160         vhdr = verdicthdr_get(nfqa);
1161         if (!vhdr)
1162                 return -EINVAL;
1163
1164         verdict = ntohl(vhdr->verdict);
1165
1166         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1167         if (entry == NULL)
1168                 return -ENOENT;
1169
1170         /* rcu lock already held from nfnl->call_rcu. */
1171         nfnl_ct = rcu_dereference(nfnl_ct_hook);
1172
1173         if (nfqa[NFQA_CT]) {
1174                 if (nfnl_ct != NULL)
1175                         ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1176         }
1177
1178         if (entry->state.pf == PF_BRIDGE) {
1179                 err = nfqa_parse_bridge(entry, nfqa);
1180                 if (err < 0)
1181                         return err;
1182         }
1183
1184         if (nfqa[NFQA_PAYLOAD]) {
1185                 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1186                 int diff = payload_len - entry->skb->len;
1187
1188                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1189                                  payload_len, entry, diff) < 0)
1190                         verdict = NF_DROP;
1191
1192                 if (ct && diff)
1193                         nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1194         }
1195
1196         if (nfqa[NFQA_MARK])
1197                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1198
1199         nf_reinject(entry, verdict);
1200         return 0;
1201 }
1202
1203 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1204                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1205                              const struct nlattr * const nfqa[],
1206                              struct netlink_ext_ack *extack)
1207 {
1208         return -ENOTSUPP;
1209 }
1210
1211 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1212         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1213         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
1214 };
1215
1216 static const struct nf_queue_handler nfqh = {
1217         .outfn          = nfqnl_enqueue_packet,
1218         .nf_hook_drop   = nfqnl_nf_hook_drop,
1219 };
1220
1221 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1222                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1223                              const struct nlattr * const nfqa[],
1224                              struct netlink_ext_ack *extack)
1225 {
1226         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1227         u_int16_t queue_num = ntohs(nfmsg->res_id);
1228         struct nfqnl_instance *queue;
1229         struct nfqnl_msg_config_cmd *cmd = NULL;
1230         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1231         __u32 flags = 0, mask = 0;
1232         int ret = 0;
1233
1234         if (nfqa[NFQA_CFG_CMD]) {
1235                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1236
1237                 /* Obsolete commands without queue context */
1238                 switch (cmd->command) {
1239                 case NFQNL_CFG_CMD_PF_BIND: return 0;
1240                 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1241                 }
1242         }
1243
1244         /* Check if we support these flags in first place, dependencies should
1245          * be there too not to break atomicity.
1246          */
1247         if (nfqa[NFQA_CFG_FLAGS]) {
1248                 if (!nfqa[NFQA_CFG_MASK]) {
1249                         /* A mask is needed to specify which flags are being
1250                          * changed.
1251                          */
1252                         return -EINVAL;
1253                 }
1254
1255                 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1256                 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1257
1258                 if (flags >= NFQA_CFG_F_MAX)
1259                         return -EOPNOTSUPP;
1260
1261 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1262                 if (flags & mask & NFQA_CFG_F_SECCTX)
1263                         return -EOPNOTSUPP;
1264 #endif
1265                 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1266                     !rcu_access_pointer(nfnl_ct_hook)) {
1267 #ifdef CONFIG_MODULES
1268                         nfnl_unlock(NFNL_SUBSYS_QUEUE);
1269                         request_module("ip_conntrack_netlink");
1270                         nfnl_lock(NFNL_SUBSYS_QUEUE);
1271                         if (rcu_access_pointer(nfnl_ct_hook))
1272                                 return -EAGAIN;
1273 #endif
1274                         return -EOPNOTSUPP;
1275                 }
1276         }
1277
1278         rcu_read_lock();
1279         queue = instance_lookup(q, queue_num);
1280         if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1281                 ret = -EPERM;
1282                 goto err_out_unlock;
1283         }
1284
1285         if (cmd != NULL) {
1286                 switch (cmd->command) {
1287                 case NFQNL_CFG_CMD_BIND:
1288                         if (queue) {
1289                                 ret = -EBUSY;
1290                                 goto err_out_unlock;
1291                         }
1292                         queue = instance_create(q, queue_num,
1293                                                 NETLINK_CB(skb).portid);
1294                         if (IS_ERR(queue)) {
1295                                 ret = PTR_ERR(queue);
1296                                 goto err_out_unlock;
1297                         }
1298                         break;
1299                 case NFQNL_CFG_CMD_UNBIND:
1300                         if (!queue) {
1301                                 ret = -ENODEV;
1302                                 goto err_out_unlock;
1303                         }
1304                         instance_destroy(q, queue);
1305                         goto err_out_unlock;
1306                 case NFQNL_CFG_CMD_PF_BIND:
1307                 case NFQNL_CFG_CMD_PF_UNBIND:
1308                         break;
1309                 default:
1310                         ret = -ENOTSUPP;
1311                         goto err_out_unlock;
1312                 }
1313         }
1314
1315         if (!queue) {
1316                 ret = -ENODEV;
1317                 goto err_out_unlock;
1318         }
1319
1320         if (nfqa[NFQA_CFG_PARAMS]) {
1321                 struct nfqnl_msg_config_params *params =
1322                         nla_data(nfqa[NFQA_CFG_PARAMS]);
1323
1324                 nfqnl_set_mode(queue, params->copy_mode,
1325                                 ntohl(params->copy_range));
1326         }
1327
1328         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1329                 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1330
1331                 spin_lock_bh(&queue->lock);
1332                 queue->queue_maxlen = ntohl(*queue_maxlen);
1333                 spin_unlock_bh(&queue->lock);
1334         }
1335
1336         if (nfqa[NFQA_CFG_FLAGS]) {
1337                 spin_lock_bh(&queue->lock);
1338                 queue->flags &= ~mask;
1339                 queue->flags |= flags & mask;
1340                 spin_unlock_bh(&queue->lock);
1341         }
1342
1343 err_out_unlock:
1344         rcu_read_unlock();
1345         return ret;
1346 }
1347
1348 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1349         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
1350                                     .attr_count = NFQA_MAX, },
1351         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
1352                                     .attr_count = NFQA_MAX,
1353                                     .policy = nfqa_verdict_policy },
1354         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
1355                                     .attr_count = NFQA_CFG_MAX,
1356                                     .policy = nfqa_cfg_policy },
1357         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1358                                     .attr_count = NFQA_MAX,
1359                                     .policy = nfqa_verdict_batch_policy },
1360 };
1361
1362 static const struct nfnetlink_subsystem nfqnl_subsys = {
1363         .name           = "nf_queue",
1364         .subsys_id      = NFNL_SUBSYS_QUEUE,
1365         .cb_count       = NFQNL_MSG_MAX,
1366         .cb             = nfqnl_cb,
1367 };
1368
1369 #ifdef CONFIG_PROC_FS
1370 struct iter_state {
1371         struct seq_net_private p;
1372         unsigned int bucket;
1373 };
1374
1375 static struct hlist_node *get_first(struct seq_file *seq)
1376 {
1377         struct iter_state *st = seq->private;
1378         struct net *net;
1379         struct nfnl_queue_net *q;
1380
1381         if (!st)
1382                 return NULL;
1383
1384         net = seq_file_net(seq);
1385         q = nfnl_queue_pernet(net);
1386         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1387                 if (!hlist_empty(&q->instance_table[st->bucket]))
1388                         return q->instance_table[st->bucket].first;
1389         }
1390         return NULL;
1391 }
1392
1393 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1394 {
1395         struct iter_state *st = seq->private;
1396         struct net *net = seq_file_net(seq);
1397
1398         h = h->next;
1399         while (!h) {
1400                 struct nfnl_queue_net *q;
1401
1402                 if (++st->bucket >= INSTANCE_BUCKETS)
1403                         return NULL;
1404
1405                 q = nfnl_queue_pernet(net);
1406                 h = q->instance_table[st->bucket].first;
1407         }
1408         return h;
1409 }
1410
1411 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1412 {
1413         struct hlist_node *head;
1414         head = get_first(seq);
1415
1416         if (head)
1417                 while (pos && (head = get_next(seq, head)))
1418                         pos--;
1419         return pos ? NULL : head;
1420 }
1421
1422 static void *seq_start(struct seq_file *s, loff_t *pos)
1423         __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1424 {
1425         spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1426         return get_idx(s, *pos);
1427 }
1428
1429 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1430 {
1431         (*pos)++;
1432         return get_next(s, v);
1433 }
1434
1435 static void seq_stop(struct seq_file *s, void *v)
1436         __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1437 {
1438         spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1439 }
1440
1441 static int seq_show(struct seq_file *s, void *v)
1442 {
1443         const struct nfqnl_instance *inst = v;
1444
1445         seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1446                    inst->queue_num,
1447                    inst->peer_portid, inst->queue_total,
1448                    inst->copy_mode, inst->copy_range,
1449                    inst->queue_dropped, inst->queue_user_dropped,
1450                    inst->id_sequence, 1);
1451         return 0;
1452 }
1453
1454 static const struct seq_operations nfqnl_seq_ops = {
1455         .start  = seq_start,
1456         .next   = seq_next,
1457         .stop   = seq_stop,
1458         .show   = seq_show,
1459 };
1460
1461 static int nfqnl_open(struct inode *inode, struct file *file)
1462 {
1463         return seq_open_net(inode, file, &nfqnl_seq_ops,
1464                         sizeof(struct iter_state));
1465 }
1466
1467 static const struct file_operations nfqnl_file_ops = {
1468         .owner   = THIS_MODULE,
1469         .open    = nfqnl_open,
1470         .read    = seq_read,
1471         .llseek  = seq_lseek,
1472         .release = seq_release_net,
1473 };
1474
1475 #endif /* PROC_FS */
1476
1477 static int __net_init nfnl_queue_net_init(struct net *net)
1478 {
1479         unsigned int i;
1480         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1481
1482         for (i = 0; i < INSTANCE_BUCKETS; i++)
1483                 INIT_HLIST_HEAD(&q->instance_table[i]);
1484
1485         spin_lock_init(&q->instances_lock);
1486
1487 #ifdef CONFIG_PROC_FS
1488         if (!proc_create("nfnetlink_queue", 0440,
1489                          net->nf.proc_netfilter, &nfqnl_file_ops))
1490                 return -ENOMEM;
1491 #endif
1492         nf_register_queue_handler(net, &nfqh);
1493         return 0;
1494 }
1495
1496 static void __net_exit nfnl_queue_net_exit(struct net *net)
1497 {
1498         nf_unregister_queue_handler(net);
1499 #ifdef CONFIG_PROC_FS
1500         remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1501 #endif
1502 }
1503
1504 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1505 {
1506         synchronize_rcu();
1507 }
1508
1509 static struct pernet_operations nfnl_queue_net_ops = {
1510         .init           = nfnl_queue_net_init,
1511         .exit           = nfnl_queue_net_exit,
1512         .exit_batch     = nfnl_queue_net_exit_batch,
1513         .id             = &nfnl_queue_net_id,
1514         .size           = sizeof(struct nfnl_queue_net),
1515 };
1516
1517 static int __init nfnetlink_queue_init(void)
1518 {
1519         int status;
1520
1521         status = register_pernet_subsys(&nfnl_queue_net_ops);
1522         if (status < 0) {
1523                 pr_err("nf_queue: failed to register pernet ops\n");
1524                 goto out;
1525         }
1526
1527         netlink_register_notifier(&nfqnl_rtnl_notifier);
1528         status = nfnetlink_subsys_register(&nfqnl_subsys);
1529         if (status < 0) {
1530                 pr_err("nf_queue: failed to create netlink socket\n");
1531                 goto cleanup_netlink_notifier;
1532         }
1533
1534         status = register_netdevice_notifier(&nfqnl_dev_notifier);
1535         if (status < 0) {
1536                 pr_err("nf_queue: failed to register netdevice notifier\n");
1537                 goto cleanup_netlink_subsys;
1538         }
1539
1540         return status;
1541
1542 cleanup_netlink_subsys:
1543         nfnetlink_subsys_unregister(&nfqnl_subsys);
1544 cleanup_netlink_notifier:
1545         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1546         unregister_pernet_subsys(&nfnl_queue_net_ops);
1547 out:
1548         return status;
1549 }
1550
1551 static void __exit nfnetlink_queue_fini(void)
1552 {
1553         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1554         nfnetlink_subsys_unregister(&nfqnl_subsys);
1555         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1556         unregister_pernet_subsys(&nfnl_queue_net_ops);
1557
1558         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1559 }
1560
1561 MODULE_DESCRIPTION("netfilter packet queue handler");
1562 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1563 MODULE_LICENSE("GPL");
1564 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1565
1566 module_init(nfnetlink_queue_init);
1567 module_exit(nfnetlink_queue_fini);