~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/netfilter/nfnetlink_queue.c

Version: ~ [ linux-6.1-rc7 ] ~ [ linux-6.0.10 ] ~ [ linux-5.19.17 ] ~ [ linux-5.18.19 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.80 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.156 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.225 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.267 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.300 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.334 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.302 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.9 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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;
 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 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 struct sk_buff *
299 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
300                            struct nf_queue_entry *entry,
301                            __be32 **packet_id_ptr)
302 {
303         size_t size;
304         size_t data_len = 0, cap_len = 0;
305         unsigned int hlen = 0;
306         struct sk_buff *skb;
307         struct nlattr *nla;
308         struct nfqnl_msg_packet_hdr *pmsg;
309         struct nlmsghdr *nlh;
310         struct nfgenmsg *nfmsg;
311         struct sk_buff *entskb = entry->skb;
312         struct net_device *indev;
313         struct net_device *outdev;
314         struct nf_conn *ct = NULL;
315         enum ip_conntrack_info uninitialized_var(ctinfo);
316         struct nfnl_ct_hook *nfnl_ct;
317         bool csum_verify;
318         char *secdata = NULL;
319         u32 seclen = 0;
320 
321         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
322                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
323                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
324                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
325 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
326                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
327                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
328 #endif
329                 + nla_total_size(sizeof(u_int32_t))     /* mark */
330                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
331                 + nla_total_size(sizeof(u_int32_t))     /* skbinfo */
332                 + nla_total_size(sizeof(u_int32_t));    /* cap_len */
333 
334         if (entskb->tstamp.tv64)
335                 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
336 
337         if (entry->state.hook <= NF_INET_FORWARD ||
338            (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
339                 csum_verify = !skb_csum_unnecessary(entskb);
340         else
341                 csum_verify = false;
342 
343         outdev = entry->state.out;
344 
345         switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
346         case NFQNL_COPY_META:
347         case NFQNL_COPY_NONE:
348                 break;
349 
350         case NFQNL_COPY_PACKET:
351                 if (!(queue->flags & NFQA_CFG_F_GSO) &&
352                     entskb->ip_summed == CHECKSUM_PARTIAL &&
353                     skb_checksum_help(entskb))
354                         return NULL;
355 
356                 data_len = ACCESS_ONCE(queue->copy_range);
357                 if (data_len > entskb->len)
358                         data_len = entskb->len;
359 
360                 hlen = skb_zerocopy_headlen(entskb);
361                 hlen = min_t(unsigned int, hlen, data_len);
362                 size += sizeof(struct nlattr) + hlen;
363                 cap_len = entskb->len;
364                 break;
365         }
366 
367         nfnl_ct = rcu_dereference(nfnl_ct_hook);
368 
369         if (queue->flags & NFQA_CFG_F_CONNTRACK) {
370                 if (nfnl_ct != NULL) {
371                         ct = nfnl_ct->get_ct(entskb, &ctinfo);
372                         if (ct != NULL)
373                                 size += nfnl_ct->build_size(ct);
374                 }
375         }
376 
377         if (queue->flags & NFQA_CFG_F_UID_GID) {
378                 size +=  (nla_total_size(sizeof(u_int32_t))     /* uid */
379                         + nla_total_size(sizeof(u_int32_t)));   /* gid */
380         }
381 
382         if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
383                 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
384                 if (seclen)
385                         size += nla_total_size(seclen);
386         }
387 
388         skb = alloc_skb(size, GFP_ATOMIC);
389         if (!skb) {
390                 skb_tx_error(entskb);
391                 return NULL;
392         }
393 
394         nlh = nlmsg_put(skb, 0, 0,
395                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
396                         sizeof(struct nfgenmsg), 0);
397         if (!nlh) {
398                 skb_tx_error(entskb);
399                 kfree_skb(skb);
400                 return NULL;
401         }
402         nfmsg = nlmsg_data(nlh);
403         nfmsg->nfgen_family = entry->state.pf;
404         nfmsg->version = NFNETLINK_V0;
405         nfmsg->res_id = htons(queue->queue_num);
406 
407         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
408         pmsg = nla_data(nla);
409         pmsg->hw_protocol       = entskb->protocol;
410         pmsg->hook              = entry->state.hook;
411         *packet_id_ptr          = &pmsg->packet_id;
412 
413         indev = entry->state.in;
414         if (indev) {
415 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
416                 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
417                         goto nla_put_failure;
418 #else
419                 if (entry->state.pf == PF_BRIDGE) {
420                         /* Case 1: indev is physical input device, we need to
421                          * look for bridge group (when called from
422                          * netfilter_bridge) */
423                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
424                                          htonl(indev->ifindex)) ||
425                         /* this is the bridge group "brX" */
426                         /* rcu_read_lock()ed by __nf_queue */
427                             nla_put_be32(skb, NFQA_IFINDEX_INDEV,
428                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
429                                 goto nla_put_failure;
430                 } else {
431                         int physinif;
432 
433                         /* Case 2: indev is bridge group, we need to look for
434                          * physical device (when called from ipv4) */
435                         if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
436                                          htonl(indev->ifindex)))
437                                 goto nla_put_failure;
438 
439                         physinif = nf_bridge_get_physinif(entskb);
440                         if (physinif &&
441                             nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
442                                          htonl(physinif)))
443                                 goto nla_put_failure;
444                 }
445 #endif
446         }
447 
448         if (outdev) {
449 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
450                 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
451                         goto nla_put_failure;
452 #else
453                 if (entry->state.pf == PF_BRIDGE) {
454                         /* Case 1: outdev is physical output device, we need to
455                          * look for bridge group (when called from
456                          * netfilter_bridge) */
457                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
458                                          htonl(outdev->ifindex)) ||
459                         /* this is the bridge group "brX" */
460                         /* rcu_read_lock()ed by __nf_queue */
461                             nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
462                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
463                                 goto nla_put_failure;
464                 } else {
465                         int physoutif;
466 
467                         /* Case 2: outdev is bridge group, we need to look for
468                          * physical output device (when called from ipv4) */
469                         if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
470                                          htonl(outdev->ifindex)))
471                                 goto nla_put_failure;
472 
473                         physoutif = nf_bridge_get_physoutif(entskb);
474                         if (physoutif &&
475                             nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
476                                          htonl(physoutif)))
477                                 goto nla_put_failure;
478                 }
479 #endif
480         }
481 
482         if (entskb->mark &&
483             nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
484                 goto nla_put_failure;
485 
486         if (indev && entskb->dev &&
487             entskb->mac_header != entskb->network_header) {
488                 struct nfqnl_msg_packet_hw phw;
489                 int len;
490 
491                 memset(&phw, 0, sizeof(phw));
492                 len = dev_parse_header(entskb, phw.hw_addr);
493                 if (len) {
494                         phw.hw_addrlen = htons(len);
495                         if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
496                                 goto nla_put_failure;
497                 }
498         }
499 
500         if (entskb->tstamp.tv64) {
501                 struct nfqnl_msg_packet_timestamp ts;
502                 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
503 
504                 ts.sec = cpu_to_be64(kts.tv_sec);
505                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
506 
507                 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
508                         goto nla_put_failure;
509         }
510 
511         if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
512             nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
513                 goto nla_put_failure;
514 
515         if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
516                 goto nla_put_failure;
517 
518         if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
519                 goto nla_put_failure;
520 
521         if (cap_len > data_len &&
522             nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
523                 goto nla_put_failure;
524 
525         if (nfqnl_put_packet_info(skb, entskb, csum_verify))
526                 goto nla_put_failure;
527 
528         if (data_len) {
529                 struct nlattr *nla;
530 
531                 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
532                         goto nla_put_failure;
533 
534                 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
535                 nla->nla_type = NFQA_PAYLOAD;
536                 nla->nla_len = nla_attr_size(data_len);
537 
538                 if (skb_zerocopy(skb, entskb, data_len, hlen))
539                         goto nla_put_failure;
540         }
541 
542         nlh->nlmsg_len = skb->len;
543         return skb;
544 
545 nla_put_failure:
546         skb_tx_error(entskb);
547         kfree_skb(skb);
548         net_err_ratelimited("nf_queue: error creating packet message\n");
549         return NULL;
550 }
551 
552 static int
553 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
554                         struct nf_queue_entry *entry)
555 {
556         struct sk_buff *nskb;
557         int err = -ENOBUFS;
558         __be32 *packet_id_ptr;
559         int failopen = 0;
560 
561         nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
562         if (nskb == NULL) {
563                 err = -ENOMEM;
564                 goto err_out;
565         }
566         spin_lock_bh(&queue->lock);
567 
568         if (queue->queue_total >= queue->queue_maxlen) {
569                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
570                         failopen = 1;
571                         err = 0;
572                 } else {
573                         queue->queue_dropped++;
574                         net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
575                                              queue->queue_total);
576                 }
577                 goto err_out_free_nskb;
578         }
579         entry->id = ++queue->id_sequence;
580         *packet_id_ptr = htonl(entry->id);
581 
582         /* nfnetlink_unicast will either free the nskb or add it to a socket */
583         err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
584         if (err < 0) {
585                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
586                         failopen = 1;
587                         err = 0;
588                 } else {
589                         queue->queue_user_dropped++;
590                 }
591                 goto err_out_unlock;
592         }
593 
594         __enqueue_entry(queue, entry);
595 
596         spin_unlock_bh(&queue->lock);
597         return 0;
598 
599 err_out_free_nskb:
600         kfree_skb(nskb);
601 err_out_unlock:
602         spin_unlock_bh(&queue->lock);
603         if (failopen)
604                 nf_reinject(entry, NF_ACCEPT);
605 err_out:
606         return err;
607 }
608 
609 static struct nf_queue_entry *
610 nf_queue_entry_dup(struct nf_queue_entry *e)
611 {
612         struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
613         if (entry)
614                 nf_queue_entry_get_refs(entry);
615         return entry;
616 }
617 
618 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
619 /* When called from bridge netfilter, skb->data must point to MAC header
620  * before calling skb_gso_segment(). Else, original MAC header is lost
621  * and segmented skbs will be sent to wrong destination.
622  */
623 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
624 {
625         if (skb->nf_bridge)
626                 __skb_push(skb, skb->network_header - skb->mac_header);
627 }
628 
629 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
630 {
631         if (skb->nf_bridge)
632                 __skb_pull(skb, skb->network_header - skb->mac_header);
633 }
634 #else
635 #define nf_bridge_adjust_skb_data(s) do {} while (0)
636 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
637 #endif
638 
639 static void free_entry(struct nf_queue_entry *entry)
640 {
641         nf_queue_entry_release_refs(entry);
642         kfree(entry);
643 }
644 
645 static int
646 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
647                            struct sk_buff *skb, struct nf_queue_entry *entry)
648 {
649         int ret = -ENOMEM;
650         struct nf_queue_entry *entry_seg;
651 
652         nf_bridge_adjust_segmented_data(skb);
653 
654         if (skb->next == NULL) { /* last packet, no need to copy entry */
655                 struct sk_buff *gso_skb = entry->skb;
656                 entry->skb = skb;
657                 ret = __nfqnl_enqueue_packet(net, queue, entry);
658                 if (ret)
659                         entry->skb = gso_skb;
660                 return ret;
661         }
662 
663         skb->next = NULL;
664 
665         entry_seg = nf_queue_entry_dup(entry);
666         if (entry_seg) {
667                 entry_seg->skb = skb;
668                 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
669                 if (ret)
670                         free_entry(entry_seg);
671         }
672         return ret;
673 }
674 
675 static int
676 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
677 {
678         unsigned int queued;
679         struct nfqnl_instance *queue;
680         struct sk_buff *skb, *segs;
681         int err = -ENOBUFS;
682         struct net *net = entry->state.net;
683         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
684 
685         /* rcu_read_lock()ed by nf_hook_slow() */
686         queue = instance_lookup(q, queuenum);
687         if (!queue)
688                 return -ESRCH;
689 
690         if (queue->copy_mode == NFQNL_COPY_NONE)
691                 return -EINVAL;
692 
693         skb = entry->skb;
694 
695         switch (entry->state.pf) {
696         case NFPROTO_IPV4:
697                 skb->protocol = htons(ETH_P_IP);
698                 break;
699         case NFPROTO_IPV6:
700                 skb->protocol = htons(ETH_P_IPV6);
701                 break;
702         }
703 
704         if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
705                 return __nfqnl_enqueue_packet(net, queue, entry);
706 
707         nf_bridge_adjust_skb_data(skb);
708         segs = skb_gso_segment(skb, 0);
709         /* Does not use PTR_ERR to limit the number of error codes that can be
710          * returned by nf_queue.  For instance, callers rely on -ESRCH to
711          * mean 'ignore this hook'.
712          */
713         if (IS_ERR_OR_NULL(segs))
714                 goto out_err;
715         queued = 0;
716         err = 0;
717         do {
718                 struct sk_buff *nskb = segs->next;
719                 if (err == 0)
720                         err = __nfqnl_enqueue_packet_gso(net, queue,
721                                                         segs, entry);
722                 if (err == 0)
723                         queued++;
724                 else
725                         kfree_skb(segs);
726                 segs = nskb;
727         } while (segs);
728 
729         if (queued) {
730                 if (err) /* some segments are already queued */
731                         free_entry(entry);
732                 kfree_skb(skb);
733                 return 0;
734         }
735  out_err:
736         nf_bridge_adjust_segmented_data(skb);
737         return err;
738 }
739 
740 static int
741 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
742 {
743         struct sk_buff *nskb;
744 
745         if (diff < 0) {
746                 if (pskb_trim(e->skb, data_len))
747                         return -ENOMEM;
748         } else if (diff > 0) {
749                 if (data_len > 0xFFFF)
750                         return -EINVAL;
751                 if (diff > skb_tailroom(e->skb)) {
752                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
753                                                diff, GFP_ATOMIC);
754                         if (!nskb) {
755                                 printk(KERN_WARNING "nf_queue: OOM "
756                                       "in mangle, dropping packet\n");
757                                 return -ENOMEM;
758                         }
759                         kfree_skb(e->skb);
760                         e->skb = nskb;
761                 }
762                 skb_put(e->skb, diff);
763         }
764         if (!skb_make_writable(e->skb, data_len))
765                 return -ENOMEM;
766         skb_copy_to_linear_data(e->skb, data, data_len);
767         e->skb->ip_summed = CHECKSUM_NONE;
768         return 0;
769 }
770 
771 static int
772 nfqnl_set_mode(struct nfqnl_instance *queue,
773                unsigned char mode, unsigned int range)
774 {
775         int status = 0;
776 
777         spin_lock_bh(&queue->lock);
778         switch (mode) {
779         case NFQNL_COPY_NONE:
780         case NFQNL_COPY_META:
781                 queue->copy_mode = mode;
782                 queue->copy_range = 0;
783                 break;
784 
785         case NFQNL_COPY_PACKET:
786                 queue->copy_mode = mode;
787                 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
788                         queue->copy_range = NFQNL_MAX_COPY_RANGE;
789                 else
790                         queue->copy_range = range;
791                 break;
792 
793         default:
794                 status = -EINVAL;
795 
796         }
797         spin_unlock_bh(&queue->lock);
798 
799         return status;
800 }
801 
802 static int
803 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
804 {
805         if (entry->state.in)
806                 if (entry->state.in->ifindex == ifindex)
807                         return 1;
808         if (entry->state.out)
809                 if (entry->state.out->ifindex == ifindex)
810                         return 1;
811 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
812         if (entry->skb->nf_bridge) {
813                 int physinif, physoutif;
814 
815                 physinif = nf_bridge_get_physinif(entry->skb);
816                 physoutif = nf_bridge_get_physoutif(entry->skb);
817 
818                 if (physinif == ifindex || physoutif == ifindex)
819                         return 1;
820         }
821 #endif
822         return 0;
823 }
824 
825 /* drop all packets with either indev or outdev == ifindex from all queue
826  * instances */
827 static void
828 nfqnl_dev_drop(struct net *net, int ifindex)
829 {
830         int i;
831         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
832 
833         rcu_read_lock();
834 
835         for (i = 0; i < INSTANCE_BUCKETS; i++) {
836                 struct nfqnl_instance *inst;
837                 struct hlist_head *head = &q->instance_table[i];
838 
839                 hlist_for_each_entry_rcu(inst, head, hlist)
840                         nfqnl_flush(inst, dev_cmp, ifindex);
841         }
842 
843         rcu_read_unlock();
844 }
845 
846 static int
847 nfqnl_rcv_dev_event(struct notifier_block *this,
848                     unsigned long event, void *ptr)
849 {
850         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
851 
852         /* Drop any packets associated with the downed device */
853         if (event == NETDEV_DOWN)
854                 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
855         return NOTIFY_DONE;
856 }
857 
858 static struct notifier_block nfqnl_dev_notifier = {
859         .notifier_call  = nfqnl_rcv_dev_event,
860 };
861 
862 static int nf_hook_cmp(struct nf_queue_entry *entry, unsigned long ops_ptr)
863 {
864         return entry->elem == (struct nf_hook_ops *)ops_ptr;
865 }
866 
867 static void nfqnl_nf_hook_drop(struct net *net, struct nf_hook_ops *hook)
868 {
869         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
870         int i;
871 
872         rcu_read_lock();
873         for (i = 0; i < INSTANCE_BUCKETS; i++) {
874                 struct nfqnl_instance *inst;
875                 struct hlist_head *head = &q->instance_table[i];
876 
877                 hlist_for_each_entry_rcu(inst, head, hlist)
878                         nfqnl_flush(inst, nf_hook_cmp, (unsigned long)hook);
879         }
880         rcu_read_unlock();
881 }
882 
883 static int
884 nfqnl_rcv_nl_event(struct notifier_block *this,
885                    unsigned long event, void *ptr)
886 {
887         struct netlink_notify *n = ptr;
888         struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
889 
890         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
891                 int i;
892 
893                 /* destroy all instances for this portid */
894                 spin_lock(&q->instances_lock);
895                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
896                         struct hlist_node *t2;
897                         struct nfqnl_instance *inst;
898                         struct hlist_head *head = &q->instance_table[i];
899 
900                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
901                                 if (n->portid == inst->peer_portid)
902                                         __instance_destroy(inst);
903                         }
904                 }
905                 spin_unlock(&q->instances_lock);
906         }
907         return NOTIFY_DONE;
908 }
909 
910 static struct notifier_block nfqnl_rtnl_notifier = {
911         .notifier_call  = nfqnl_rcv_nl_event,
912 };
913 
914 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
915         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
916         [NFQA_MARK]             = { .type = NLA_U32 },
917         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
918         [NFQA_CT]               = { .type = NLA_UNSPEC },
919         [NFQA_EXP]              = { .type = NLA_UNSPEC },
920 };
921 
922 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
923         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
924         [NFQA_MARK]             = { .type = NLA_U32 },
925 };
926 
927 static struct nfqnl_instance *
928 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
929 {
930         struct nfqnl_instance *queue;
931 
932         queue = instance_lookup(q, queue_num);
933         if (!queue)
934                 return ERR_PTR(-ENODEV);
935 
936         if (queue->peer_portid != nlportid)
937                 return ERR_PTR(-EPERM);
938 
939         return queue;
940 }
941 
942 static struct nfqnl_msg_verdict_hdr*
943 verdicthdr_get(const struct nlattr * const nfqa[])
944 {
945         struct nfqnl_msg_verdict_hdr *vhdr;
946         unsigned int verdict;
947 
948         if (!nfqa[NFQA_VERDICT_HDR])
949                 return NULL;
950 
951         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
952         verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
953         if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
954                 return NULL;
955         return vhdr;
956 }
957 
958 static int nfq_id_after(unsigned int id, unsigned int max)
959 {
960         return (int)(id - max) > 0;
961 }
962 
963 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
964                                     struct sk_buff *skb,
965                                     const struct nlmsghdr *nlh,
966                                     const struct nlattr * const nfqa[])
967 {
968         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
969         struct nf_queue_entry *entry, *tmp;
970         unsigned int verdict, maxid;
971         struct nfqnl_msg_verdict_hdr *vhdr;
972         struct nfqnl_instance *queue;
973         LIST_HEAD(batch_list);
974         u16 queue_num = ntohs(nfmsg->res_id);
975         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
976 
977         queue = verdict_instance_lookup(q, queue_num,
978                                         NETLINK_CB(skb).portid);
979         if (IS_ERR(queue))
980                 return PTR_ERR(queue);
981 
982         vhdr = verdicthdr_get(nfqa);
983         if (!vhdr)
984                 return -EINVAL;
985 
986         verdict = ntohl(vhdr->verdict);
987         maxid = ntohl(vhdr->id);
988 
989         spin_lock_bh(&queue->lock);
990 
991         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
992                 if (nfq_id_after(entry->id, maxid))
993                         break;
994                 __dequeue_entry(queue, entry);
995                 list_add_tail(&entry->list, &batch_list);
996         }
997 
998         spin_unlock_bh(&queue->lock);
999 
1000         if (list_empty(&batch_list))
1001                 return -ENOENT;
1002 
1003         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1004                 if (nfqa[NFQA_MARK])
1005                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1006                 nf_reinject(entry, verdict);
1007         }
1008         return 0;
1009 }
1010 
1011 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1012                                       const struct nlmsghdr *nlh,
1013                                       const struct nlattr * const nfqa[],
1014                                       struct nf_queue_entry *entry,
1015                                       enum ip_conntrack_info *ctinfo)
1016 {
1017         struct nf_conn *ct;
1018 
1019         ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1020         if (ct == NULL)
1021                 return NULL;
1022 
1023         if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1024                 return NULL;
1025 
1026         if (nfqa[NFQA_EXP])
1027                 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1028                                       NETLINK_CB(entry->skb).portid,
1029                                       nlmsg_report(nlh));
1030         return ct;
1031 }
1032 
1033 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1034                               struct sk_buff *skb,
1035                               const struct nlmsghdr *nlh,
1036                               const struct nlattr * const nfqa[])
1037 {
1038         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1039         u_int16_t queue_num = ntohs(nfmsg->res_id);
1040         struct nfqnl_msg_verdict_hdr *vhdr;
1041         struct nfqnl_instance *queue;
1042         unsigned int verdict;
1043         struct nf_queue_entry *entry;
1044         enum ip_conntrack_info uninitialized_var(ctinfo);
1045         struct nfnl_ct_hook *nfnl_ct;
1046         struct nf_conn *ct = NULL;
1047         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1048 
1049         queue = instance_lookup(q, queue_num);
1050         if (!queue)
1051                 queue = verdict_instance_lookup(q, queue_num,
1052                                                 NETLINK_CB(skb).portid);
1053         if (IS_ERR(queue))
1054                 return PTR_ERR(queue);
1055 
1056         vhdr = verdicthdr_get(nfqa);
1057         if (!vhdr)
1058                 return -EINVAL;
1059 
1060         verdict = ntohl(vhdr->verdict);
1061 
1062         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1063         if (entry == NULL)
1064                 return -ENOENT;
1065 
1066         /* rcu lock already held from nfnl->call_rcu. */
1067         nfnl_ct = rcu_dereference(nfnl_ct_hook);
1068 
1069         if (nfqa[NFQA_CT]) {
1070                 if (nfnl_ct != NULL)
1071                         ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1072         }
1073 
1074         if (nfqa[NFQA_PAYLOAD]) {
1075                 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1076                 int diff = payload_len - entry->skb->len;
1077 
1078                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1079                                  payload_len, entry, diff) < 0)
1080                         verdict = NF_DROP;
1081 
1082                 if (ct && diff)
1083                         nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1084         }
1085 
1086         if (nfqa[NFQA_MARK])
1087                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1088 
1089         nf_reinject(entry, verdict);
1090         return 0;
1091 }
1092 
1093 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1094                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1095                              const struct nlattr * const nfqa[])
1096 {
1097         return -ENOTSUPP;
1098 }
1099 
1100 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1101         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1102         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
1103 };
1104 
1105 static const struct nf_queue_handler nfqh = {
1106         .outfn          = &nfqnl_enqueue_packet,
1107         .nf_hook_drop   = &nfqnl_nf_hook_drop,
1108 };
1109 
1110 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1111                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1112                              const struct nlattr * const nfqa[])
1113 {
1114         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1115         u_int16_t queue_num = ntohs(nfmsg->res_id);
1116         struct nfqnl_instance *queue;
1117         struct nfqnl_msg_config_cmd *cmd = NULL;
1118         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1119         __u32 flags = 0, mask = 0;
1120         int ret = 0;
1121 
1122         if (nfqa[NFQA_CFG_CMD]) {
1123                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1124 
1125                 /* Obsolete commands without queue context */
1126                 switch (cmd->command) {
1127                 case NFQNL_CFG_CMD_PF_BIND: return 0;
1128                 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1129                 }
1130         }
1131 
1132         /* Check if we support these flags in first place, dependencies should
1133          * be there too not to break atomicity.
1134          */
1135         if (nfqa[NFQA_CFG_FLAGS]) {
1136                 if (!nfqa[NFQA_CFG_MASK]) {
1137                         /* A mask is needed to specify which flags are being
1138                          * changed.
1139                          */
1140                         return -EINVAL;
1141                 }
1142 
1143                 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1144                 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1145 
1146                 if (flags >= NFQA_CFG_F_MAX)
1147                         return -EOPNOTSUPP;
1148 
1149 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1150                 if (flags & mask & NFQA_CFG_F_SECCTX)
1151                         return -EOPNOTSUPP;
1152 #endif
1153                 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1154                     !rcu_access_pointer(nfnl_ct_hook)) {
1155 #ifdef CONFIG_MODULES
1156                         nfnl_unlock(NFNL_SUBSYS_QUEUE);
1157                         request_module("ip_conntrack_netlink");
1158                         nfnl_lock(NFNL_SUBSYS_QUEUE);
1159                         if (rcu_access_pointer(nfnl_ct_hook))
1160                                 return -EAGAIN;
1161 #endif
1162                         return -EOPNOTSUPP;
1163                 }
1164         }
1165 
1166         rcu_read_lock();
1167         queue = instance_lookup(q, queue_num);
1168         if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1169                 ret = -EPERM;
1170                 goto err_out_unlock;
1171         }
1172 
1173         if (cmd != NULL) {
1174                 switch (cmd->command) {
1175                 case NFQNL_CFG_CMD_BIND:
1176                         if (queue) {
1177                                 ret = -EBUSY;
1178                                 goto err_out_unlock;
1179                         }
1180                         queue = instance_create(q, queue_num,
1181                                                 NETLINK_CB(skb).portid);
1182                         if (IS_ERR(queue)) {
1183                                 ret = PTR_ERR(queue);
1184                                 goto err_out_unlock;
1185                         }
1186                         break;
1187                 case NFQNL_CFG_CMD_UNBIND:
1188                         if (!queue) {
1189                                 ret = -ENODEV;
1190                                 goto err_out_unlock;
1191                         }
1192                         instance_destroy(q, queue);
1193                         goto err_out_unlock;
1194                 case NFQNL_CFG_CMD_PF_BIND:
1195                 case NFQNL_CFG_CMD_PF_UNBIND:
1196                         break;
1197                 default:
1198                         ret = -ENOTSUPP;
1199                         goto err_out_unlock;
1200                 }
1201         }
1202 
1203         if (!queue) {
1204                 ret = -ENODEV;
1205                 goto err_out_unlock;
1206         }
1207 
1208         if (nfqa[NFQA_CFG_PARAMS]) {
1209                 struct nfqnl_msg_config_params *params =
1210                         nla_data(nfqa[NFQA_CFG_PARAMS]);
1211 
1212                 nfqnl_set_mode(queue, params->copy_mode,
1213                                 ntohl(params->copy_range));
1214         }
1215 
1216         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1217                 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1218 
1219                 spin_lock_bh(&queue->lock);
1220                 queue->queue_maxlen = ntohl(*queue_maxlen);
1221                 spin_unlock_bh(&queue->lock);
1222         }
1223 
1224         if (nfqa[NFQA_CFG_FLAGS]) {
1225                 spin_lock_bh(&queue->lock);
1226                 queue->flags &= ~mask;
1227                 queue->flags |= flags & mask;
1228                 spin_unlock_bh(&queue->lock);
1229         }
1230 
1231 err_out_unlock:
1232         rcu_read_unlock();
1233         return ret;
1234 }
1235 
1236 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1237         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
1238                                     .attr_count = NFQA_MAX, },
1239         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
1240                                     .attr_count = NFQA_MAX,
1241                                     .policy = nfqa_verdict_policy },
1242         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
1243                                     .attr_count = NFQA_CFG_MAX,
1244                                     .policy = nfqa_cfg_policy },
1245         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1246                                     .attr_count = NFQA_MAX,
1247                                     .policy = nfqa_verdict_batch_policy },
1248 };
1249 
1250 static const struct nfnetlink_subsystem nfqnl_subsys = {
1251         .name           = "nf_queue",
1252         .subsys_id      = NFNL_SUBSYS_QUEUE,
1253         .cb_count       = NFQNL_MSG_MAX,
1254         .cb             = nfqnl_cb,
1255 };
1256 
1257 #ifdef CONFIG_PROC_FS
1258 struct iter_state {
1259         struct seq_net_private p;
1260         unsigned int bucket;
1261 };
1262 
1263 static struct hlist_node *get_first(struct seq_file *seq)
1264 {
1265         struct iter_state *st = seq->private;
1266         struct net *net;
1267         struct nfnl_queue_net *q;
1268 
1269         if (!st)
1270                 return NULL;
1271 
1272         net = seq_file_net(seq);
1273         q = nfnl_queue_pernet(net);
1274         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1275                 if (!hlist_empty(&q->instance_table[st->bucket]))
1276                         return q->instance_table[st->bucket].first;
1277         }
1278         return NULL;
1279 }
1280 
1281 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1282 {
1283         struct iter_state *st = seq->private;
1284         struct net *net = seq_file_net(seq);
1285 
1286         h = h->next;
1287         while (!h) {
1288                 struct nfnl_queue_net *q;
1289 
1290                 if (++st->bucket >= INSTANCE_BUCKETS)
1291                         return NULL;
1292 
1293                 q = nfnl_queue_pernet(net);
1294                 h = q->instance_table[st->bucket].first;
1295         }
1296         return h;
1297 }
1298 
1299 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1300 {
1301         struct hlist_node *head;
1302         head = get_first(seq);
1303 
1304         if (head)
1305                 while (pos && (head = get_next(seq, head)))
1306                         pos--;
1307         return pos ? NULL : head;
1308 }
1309 
1310 static void *seq_start(struct seq_file *s, loff_t *pos)
1311         __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1312 {
1313         spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1314         return get_idx(s, *pos);
1315 }
1316 
1317 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1318 {
1319         (*pos)++;
1320         return get_next(s, v);
1321 }
1322 
1323 static void seq_stop(struct seq_file *s, void *v)
1324         __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1325 {
1326         spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1327 }
1328 
1329 static int seq_show(struct seq_file *s, void *v)
1330 {
1331         const struct nfqnl_instance *inst = v;
1332 
1333         seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1334                    inst->queue_num,
1335                    inst->peer_portid, inst->queue_total,
1336                    inst->copy_mode, inst->copy_range,
1337                    inst->queue_dropped, inst->queue_user_dropped,
1338                    inst->id_sequence, 1);
1339         return 0;
1340 }
1341 
1342 static const struct seq_operations nfqnl_seq_ops = {
1343         .start  = seq_start,
1344         .next   = seq_next,
1345         .stop   = seq_stop,
1346         .show   = seq_show,
1347 };
1348 
1349 static int nfqnl_open(struct inode *inode, struct file *file)
1350 {
1351         return seq_open_net(inode, file, &nfqnl_seq_ops,
1352                         sizeof(struct iter_state));
1353 }
1354 
1355 static const struct file_operations nfqnl_file_ops = {
1356         .owner   = THIS_MODULE,
1357         .open    = nfqnl_open,
1358         .read    = seq_read,
1359         .llseek  = seq_lseek,
1360         .release = seq_release_net,
1361 };
1362 
1363 #endif /* PROC_FS */
1364 
1365 static int __net_init nfnl_queue_net_init(struct net *net)
1366 {
1367         unsigned int i;
1368         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1369 
1370         for (i = 0; i < INSTANCE_BUCKETS; i++)
1371                 INIT_HLIST_HEAD(&q->instance_table[i]);
1372 
1373         spin_lock_init(&q->instances_lock);
1374 
1375 #ifdef CONFIG_PROC_FS
1376         if (!proc_create("nfnetlink_queue", 0440,
1377                          net->nf.proc_netfilter, &nfqnl_file_ops))
1378                 return -ENOMEM;
1379 #endif
1380         return 0;
1381 }
1382 
1383 static void __net_exit nfnl_queue_net_exit(struct net *net)
1384 {
1385 #ifdef CONFIG_PROC_FS
1386         remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1387 #endif
1388 }
1389 
1390 static struct pernet_operations nfnl_queue_net_ops = {
1391         .init   = nfnl_queue_net_init,
1392         .exit   = nfnl_queue_net_exit,
1393         .id     = &nfnl_queue_net_id,
1394         .size   = sizeof(struct nfnl_queue_net),
1395 };
1396 
1397 static int __init nfnetlink_queue_init(void)
1398 {
1399         int status;
1400 
1401         status = register_pernet_subsys(&nfnl_queue_net_ops);
1402         if (status < 0) {
1403                 pr_err("nf_queue: failed to register pernet ops\n");
1404                 goto out;
1405         }
1406 
1407         netlink_register_notifier(&nfqnl_rtnl_notifier);
1408         status = nfnetlink_subsys_register(&nfqnl_subsys);
1409         if (status < 0) {
1410                 pr_err("nf_queue: failed to create netlink socket\n");
1411                 goto cleanup_netlink_notifier;
1412         }
1413 
1414         register_netdevice_notifier(&nfqnl_dev_notifier);
1415         nf_register_queue_handler(&nfqh);
1416         return status;
1417 
1418 cleanup_netlink_notifier:
1419         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1420         unregister_pernet_subsys(&nfnl_queue_net_ops);
1421 out:
1422         return status;
1423 }
1424 
1425 static void __exit nfnetlink_queue_fini(void)
1426 {
1427         nf_unregister_queue_handler();
1428         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1429         nfnetlink_subsys_unregister(&nfqnl_subsys);
1430         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1431         unregister_pernet_subsys(&nfnl_queue_net_ops);
1432 
1433         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1434 }
1435 
1436 MODULE_DESCRIPTION("netfilter packet queue handler");
1437 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1438 MODULE_LICENSE("GPL");
1439 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1440 
1441 module_init(nfnetlink_queue_init);
1442 module_exit(nfnetlink_queue_fini);
1443 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp