2 * This is a module which is used for queueing packets and communicating with
3 * userspace via nfnetlink.
5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
6 * (C) 2007 by Patrick McHardy <kaber@trash.net>
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>
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.
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/list.h>
33 #include <net/tcp_states.h>
34 #include <net/netfilter/nf_queue.h>
35 #include <net/netns/generic.h>
36 #include <net/netfilter/nfnetlink_queue.h>
38 #include <linux/atomic.h>
40 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
41 #include "../bridge/br_private.h"
44 #define NFQNL_QMAX_DEFAULT 1024
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.
52 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
54 struct nfqnl_instance {
55 struct hlist_node hlist; /* global list of queues */
59 unsigned int queue_maxlen;
60 unsigned int copy_range;
61 unsigned int queue_dropped;
62 unsigned int queue_user_dropped;
65 u_int16_t queue_num; /* number of this queue */
67 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
69 * Following fields are dirtied for each queued packet,
70 * keep them in same cache line if possible.
73 unsigned int queue_total;
74 unsigned int id_sequence; /* 'sequence' of pkt ids */
75 struct list_head queue_list; /* packets in queue */
78 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
80 static int nfnl_queue_net_id __read_mostly;
82 #define INSTANCE_BUCKETS 16
83 struct nfnl_queue_net {
84 spinlock_t instances_lock;
85 struct hlist_head instance_table[INSTANCE_BUCKETS];
88 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
90 return net_generic(net, nfnl_queue_net_id);
93 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
95 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
98 static struct nfqnl_instance *
99 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
101 struct hlist_head *head;
102 struct nfqnl_instance *inst;
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)
112 static struct nfqnl_instance *
113 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
115 struct nfqnl_instance *inst;
119 spin_lock(&q->instances_lock);
120 if (instance_lookup(q, queue_num)) {
125 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
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);
139 if (!try_module_get(THIS_MODULE)) {
144 h = instance_hashfn(queue_num);
145 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
147 spin_unlock(&q->instances_lock);
154 spin_unlock(&q->instances_lock);
158 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
162 instance_destroy_rcu(struct rcu_head *head)
164 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
167 nfqnl_flush(inst, NULL, 0);
169 module_put(THIS_MODULE);
173 __instance_destroy(struct nfqnl_instance *inst)
175 hlist_del_rcu(&inst->hlist);
176 call_rcu(&inst->rcu, instance_destroy_rcu);
180 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
182 spin_lock(&q->instances_lock);
183 __instance_destroy(inst);
184 spin_unlock(&q->instances_lock);
188 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
190 list_add_tail(&entry->list, &queue->queue_list);
191 queue->queue_total++;
195 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
197 list_del(&entry->list);
198 queue->queue_total--;
201 static struct nf_queue_entry *
202 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
204 struct nf_queue_entry *entry = NULL, *i;
206 spin_lock_bh(&queue->lock);
208 list_for_each_entry(i, &queue->queue_list, list) {
216 __dequeue_entry(queue, entry);
218 spin_unlock_bh(&queue->lock);
224 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
226 struct nf_queue_entry *entry, *next;
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);
236 spin_unlock_bh(&queue->lock);
240 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
245 if (packet->ip_summed == CHECKSUM_PARTIAL)
246 flags = NFQA_SKB_CSUMNOTREADY;
247 else if (csum_verify)
248 flags = NFQA_SKB_CSUM_NOTVERIFIED;
250 if (skb_is_gso(packet))
251 flags |= NFQA_SKB_GSO;
253 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
256 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
258 const struct cred *cred;
260 if (!sk_fullsock(sk))
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;
273 read_unlock_bh(&sk->sk_callback_lock);
277 read_unlock_bh(&sk->sk_callback_lock);
281 static struct sk_buff *
282 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
283 struct nf_queue_entry *entry,
284 __be32 **packet_id_ptr)
287 size_t data_len = 0, cap_len = 0;
288 unsigned int hlen = 0;
291 struct nfqnl_msg_packet_hdr *pmsg;
292 struct nlmsghdr *nlh;
293 struct nfgenmsg *nfmsg;
294 struct sk_buff *entskb = entry->skb;
295 struct net_device *indev;
296 struct net_device *outdev;
297 struct nf_conn *ct = NULL;
298 enum ip_conntrack_info uninitialized_var(ctinfo);
301 size = nlmsg_total_size(sizeof(struct nfgenmsg))
302 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
303 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
304 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
305 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
306 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
307 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
309 + nla_total_size(sizeof(u_int32_t)) /* mark */
310 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
311 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
312 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
314 if (entskb->tstamp.tv64)
315 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
317 if (entry->state.hook <= NF_INET_FORWARD ||
318 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
319 csum_verify = !skb_csum_unnecessary(entskb);
323 outdev = entry->state.out;
325 switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
326 case NFQNL_COPY_META:
327 case NFQNL_COPY_NONE:
330 case NFQNL_COPY_PACKET:
331 if (!(queue->flags & NFQA_CFG_F_GSO) &&
332 entskb->ip_summed == CHECKSUM_PARTIAL &&
333 skb_checksum_help(entskb))
336 data_len = ACCESS_ONCE(queue->copy_range);
337 if (data_len > entskb->len)
338 data_len = entskb->len;
340 hlen = skb_zerocopy_headlen(entskb);
341 hlen = min_t(unsigned int, hlen, data_len);
342 size += sizeof(struct nlattr) + hlen;
343 cap_len = entskb->len;
347 if (queue->flags & NFQA_CFG_F_CONNTRACK)
348 ct = nfqnl_ct_get(entskb, &size, &ctinfo);
350 if (queue->flags & NFQA_CFG_F_UID_GID) {
351 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
352 + nla_total_size(sizeof(u_int32_t))); /* gid */
355 skb = nfnetlink_alloc_skb(net, size, queue->peer_portid,
358 skb_tx_error(entskb);
362 nlh = nlmsg_put(skb, 0, 0,
363 NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
364 sizeof(struct nfgenmsg), 0);
366 skb_tx_error(entskb);
370 nfmsg = nlmsg_data(nlh);
371 nfmsg->nfgen_family = entry->state.pf;
372 nfmsg->version = NFNETLINK_V0;
373 nfmsg->res_id = htons(queue->queue_num);
375 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
376 pmsg = nla_data(nla);
377 pmsg->hw_protocol = entskb->protocol;
378 pmsg->hook = entry->state.hook;
379 *packet_id_ptr = &pmsg->packet_id;
381 indev = entry->state.in;
383 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
384 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
385 goto nla_put_failure;
387 if (entry->state.pf == PF_BRIDGE) {
388 /* Case 1: indev is physical input device, we need to
389 * look for bridge group (when called from
390 * netfilter_bridge) */
391 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
392 htonl(indev->ifindex)) ||
393 /* this is the bridge group "brX" */
394 /* rcu_read_lock()ed by __nf_queue */
395 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
396 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
397 goto nla_put_failure;
401 /* Case 2: indev is bridge group, we need to look for
402 * physical device (when called from ipv4) */
403 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
404 htonl(indev->ifindex)))
405 goto nla_put_failure;
407 physinif = nf_bridge_get_physinif(entskb);
409 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
411 goto nla_put_failure;
417 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
418 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
419 goto nla_put_failure;
421 if (entry->state.pf == PF_BRIDGE) {
422 /* Case 1: outdev is physical output device, we need to
423 * look for bridge group (when called from
424 * netfilter_bridge) */
425 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
426 htonl(outdev->ifindex)) ||
427 /* this is the bridge group "brX" */
428 /* rcu_read_lock()ed by __nf_queue */
429 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
430 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
431 goto nla_put_failure;
435 /* Case 2: outdev is bridge group, we need to look for
436 * physical output device (when called from ipv4) */
437 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
438 htonl(outdev->ifindex)))
439 goto nla_put_failure;
441 physoutif = nf_bridge_get_physoutif(entskb);
443 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
445 goto nla_put_failure;
451 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
452 goto nla_put_failure;
454 if (indev && entskb->dev &&
455 entskb->mac_header != entskb->network_header) {
456 struct nfqnl_msg_packet_hw phw;
459 memset(&phw, 0, sizeof(phw));
460 len = dev_parse_header(entskb, phw.hw_addr);
462 phw.hw_addrlen = htons(len);
463 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
464 goto nla_put_failure;
468 if (entskb->tstamp.tv64) {
469 struct nfqnl_msg_packet_timestamp ts;
470 struct timeval tv = ktime_to_timeval(entskb->tstamp);
471 ts.sec = cpu_to_be64(tv.tv_sec);
472 ts.usec = cpu_to_be64(tv.tv_usec);
474 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
475 goto nla_put_failure;
478 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
479 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
480 goto nla_put_failure;
482 if (ct && nfqnl_ct_put(skb, ct, ctinfo) < 0)
483 goto nla_put_failure;
485 if (cap_len > data_len &&
486 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
487 goto nla_put_failure;
489 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
490 goto nla_put_failure;
495 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
496 goto nla_put_failure;
498 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
499 nla->nla_type = NFQA_PAYLOAD;
500 nla->nla_len = nla_attr_size(data_len);
502 if (skb_zerocopy(skb, entskb, data_len, hlen))
503 goto nla_put_failure;
506 nlh->nlmsg_len = skb->len;
510 skb_tx_error(entskb);
512 net_err_ratelimited("nf_queue: error creating packet message\n");
517 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
518 struct nf_queue_entry *entry)
520 struct sk_buff *nskb;
522 __be32 *packet_id_ptr;
525 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
530 spin_lock_bh(&queue->lock);
532 if (queue->queue_total >= queue->queue_maxlen) {
533 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
537 queue->queue_dropped++;
538 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
541 goto err_out_free_nskb;
543 entry->id = ++queue->id_sequence;
544 *packet_id_ptr = htonl(entry->id);
546 /* nfnetlink_unicast will either free the nskb or add it to a socket */
547 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
549 queue->queue_user_dropped++;
553 __enqueue_entry(queue, entry);
555 spin_unlock_bh(&queue->lock);
561 spin_unlock_bh(&queue->lock);
563 nf_reinject(entry, NF_ACCEPT);
568 static struct nf_queue_entry *
569 nf_queue_entry_dup(struct nf_queue_entry *e)
571 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
573 if (nf_queue_entry_get_refs(entry))
580 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
581 /* When called from bridge netfilter, skb->data must point to MAC header
582 * before calling skb_gso_segment(). Else, original MAC header is lost
583 * and segmented skbs will be sent to wrong destination.
585 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
588 __skb_push(skb, skb->network_header - skb->mac_header);
591 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
594 __skb_pull(skb, skb->network_header - skb->mac_header);
597 #define nf_bridge_adjust_skb_data(s) do {} while (0)
598 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
601 static void free_entry(struct nf_queue_entry *entry)
603 nf_queue_entry_release_refs(entry);
608 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
609 struct sk_buff *skb, struct nf_queue_entry *entry)
612 struct nf_queue_entry *entry_seg;
614 nf_bridge_adjust_segmented_data(skb);
616 if (skb->next == NULL) { /* last packet, no need to copy entry */
617 struct sk_buff *gso_skb = entry->skb;
619 ret = __nfqnl_enqueue_packet(net, queue, entry);
621 entry->skb = gso_skb;
627 entry_seg = nf_queue_entry_dup(entry);
629 entry_seg->skb = skb;
630 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
632 free_entry(entry_seg);
638 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
641 struct nfqnl_instance *queue;
642 struct sk_buff *skb, *segs;
644 struct net *net = dev_net(entry->state.in ?
645 entry->state.in : entry->state.out);
646 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
648 /* rcu_read_lock()ed by nf_hook_slow() */
649 queue = instance_lookup(q, queuenum);
653 if (queue->copy_mode == NFQNL_COPY_NONE)
658 switch (entry->state.pf) {
660 skb->protocol = htons(ETH_P_IP);
663 skb->protocol = htons(ETH_P_IPV6);
667 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
668 return __nfqnl_enqueue_packet(net, queue, entry);
670 nf_bridge_adjust_skb_data(skb);
671 segs = skb_gso_segment(skb, 0);
672 /* Does not use PTR_ERR to limit the number of error codes that can be
673 * returned by nf_queue. For instance, callers rely on -ECANCELED to
674 * mean 'ignore this hook'.
676 if (IS_ERR_OR_NULL(segs))
681 struct sk_buff *nskb = segs->next;
683 err = __nfqnl_enqueue_packet_gso(net, queue,
693 if (err) /* some segments are already queued */
699 nf_bridge_adjust_segmented_data(skb);
704 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
706 struct sk_buff *nskb;
709 if (pskb_trim(e->skb, data_len))
711 } else if (diff > 0) {
712 if (data_len > 0xFFFF)
714 if (diff > skb_tailroom(e->skb)) {
715 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
718 printk(KERN_WARNING "nf_queue: OOM "
719 "in mangle, dropping packet\n");
725 skb_put(e->skb, diff);
727 if (!skb_make_writable(e->skb, data_len))
729 skb_copy_to_linear_data(e->skb, data, data_len);
730 e->skb->ip_summed = CHECKSUM_NONE;
735 nfqnl_set_mode(struct nfqnl_instance *queue,
736 unsigned char mode, unsigned int range)
740 spin_lock_bh(&queue->lock);
742 case NFQNL_COPY_NONE:
743 case NFQNL_COPY_META:
744 queue->copy_mode = mode;
745 queue->copy_range = 0;
748 case NFQNL_COPY_PACKET:
749 queue->copy_mode = mode;
750 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
751 queue->copy_range = NFQNL_MAX_COPY_RANGE;
753 queue->copy_range = range;
760 spin_unlock_bh(&queue->lock);
766 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
769 if (entry->state.in->ifindex == ifindex)
771 if (entry->state.out)
772 if (entry->state.out->ifindex == ifindex)
774 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
775 if (entry->skb->nf_bridge) {
776 int physinif, physoutif;
778 physinif = nf_bridge_get_physinif(entry->skb);
779 physoutif = nf_bridge_get_physoutif(entry->skb);
781 if (physinif == ifindex || physoutif == ifindex)
788 /* drop all packets with either indev or outdev == ifindex from all queue
791 nfqnl_dev_drop(struct net *net, int ifindex)
794 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
798 for (i = 0; i < INSTANCE_BUCKETS; i++) {
799 struct nfqnl_instance *inst;
800 struct hlist_head *head = &q->instance_table[i];
802 hlist_for_each_entry_rcu(inst, head, hlist)
803 nfqnl_flush(inst, dev_cmp, ifindex);
809 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
812 nfqnl_rcv_dev_event(struct notifier_block *this,
813 unsigned long event, void *ptr)
815 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
817 /* Drop any packets associated with the downed device */
818 if (event == NETDEV_DOWN)
819 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
823 static struct notifier_block nfqnl_dev_notifier = {
824 .notifier_call = nfqnl_rcv_dev_event,
828 nfqnl_rcv_nl_event(struct notifier_block *this,
829 unsigned long event, void *ptr)
831 struct netlink_notify *n = ptr;
832 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
834 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
837 /* destroy all instances for this portid */
838 spin_lock(&q->instances_lock);
839 for (i = 0; i < INSTANCE_BUCKETS; i++) {
840 struct hlist_node *t2;
841 struct nfqnl_instance *inst;
842 struct hlist_head *head = &q->instance_table[i];
844 hlist_for_each_entry_safe(inst, t2, head, hlist) {
845 if (n->portid == inst->peer_portid)
846 __instance_destroy(inst);
849 spin_unlock(&q->instances_lock);
854 static struct notifier_block nfqnl_rtnl_notifier = {
855 .notifier_call = nfqnl_rcv_nl_event,
858 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
859 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
860 [NFQA_MARK] = { .type = NLA_U32 },
861 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
862 [NFQA_CT] = { .type = NLA_UNSPEC },
863 [NFQA_EXP] = { .type = NLA_UNSPEC },
866 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
867 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
868 [NFQA_MARK] = { .type = NLA_U32 },
871 static struct nfqnl_instance *
872 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
874 struct nfqnl_instance *queue;
876 queue = instance_lookup(q, queue_num);
878 return ERR_PTR(-ENODEV);
880 if (queue->peer_portid != nlportid)
881 return ERR_PTR(-EPERM);
886 static struct nfqnl_msg_verdict_hdr*
887 verdicthdr_get(const struct nlattr * const nfqa[])
889 struct nfqnl_msg_verdict_hdr *vhdr;
890 unsigned int verdict;
892 if (!nfqa[NFQA_VERDICT_HDR])
895 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
896 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
897 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
902 static int nfq_id_after(unsigned int id, unsigned int max)
904 return (int)(id - max) > 0;
908 nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
909 const struct nlmsghdr *nlh,
910 const struct nlattr * const nfqa[])
912 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
913 struct nf_queue_entry *entry, *tmp;
914 unsigned int verdict, maxid;
915 struct nfqnl_msg_verdict_hdr *vhdr;
916 struct nfqnl_instance *queue;
917 LIST_HEAD(batch_list);
918 u16 queue_num = ntohs(nfmsg->res_id);
920 struct net *net = sock_net(ctnl);
921 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
923 queue = verdict_instance_lookup(q, queue_num,
924 NETLINK_CB(skb).portid);
926 return PTR_ERR(queue);
928 vhdr = verdicthdr_get(nfqa);
932 verdict = ntohl(vhdr->verdict);
933 maxid = ntohl(vhdr->id);
935 spin_lock_bh(&queue->lock);
937 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
938 if (nfq_id_after(entry->id, maxid))
940 __dequeue_entry(queue, entry);
941 list_add_tail(&entry->list, &batch_list);
944 spin_unlock_bh(&queue->lock);
946 if (list_empty(&batch_list))
949 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
951 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
952 nf_reinject(entry, verdict);
958 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
959 const struct nlmsghdr *nlh,
960 const struct nlattr * const nfqa[])
962 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
963 u_int16_t queue_num = ntohs(nfmsg->res_id);
965 struct nfqnl_msg_verdict_hdr *vhdr;
966 struct nfqnl_instance *queue;
967 unsigned int verdict;
968 struct nf_queue_entry *entry;
969 enum ip_conntrack_info uninitialized_var(ctinfo);
970 struct nf_conn *ct = NULL;
972 struct net *net = sock_net(ctnl);
973 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
975 queue = instance_lookup(q, queue_num);
977 queue = verdict_instance_lookup(q, queue_num,
978 NETLINK_CB(skb).portid);
980 return PTR_ERR(queue);
982 vhdr = verdicthdr_get(nfqa);
986 verdict = ntohl(vhdr->verdict);
988 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
993 ct = nfqnl_ct_parse(entry->skb, nfqa[NFQA_CT], &ctinfo);
994 if (ct && nfqa[NFQA_EXP]) {
995 nfqnl_attach_expect(ct, nfqa[NFQA_EXP],
996 NETLINK_CB(skb).portid,
1001 if (nfqa[NFQA_PAYLOAD]) {
1002 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1003 int diff = payload_len - entry->skb->len;
1005 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1006 payload_len, entry, diff) < 0)
1010 nfqnl_ct_seq_adjust(entry->skb, ct, ctinfo, diff);
1013 if (nfqa[NFQA_MARK])
1014 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1016 nf_reinject(entry, verdict);
1021 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1022 const struct nlmsghdr *nlh,
1023 const struct nlattr * const nfqa[])
1028 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1029 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1030 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1033 static const struct nf_queue_handler nfqh = {
1034 .outfn = &nfqnl_enqueue_packet,
1038 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1039 const struct nlmsghdr *nlh,
1040 const struct nlattr * const nfqa[])
1042 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1043 u_int16_t queue_num = ntohs(nfmsg->res_id);
1044 struct nfqnl_instance *queue;
1045 struct nfqnl_msg_config_cmd *cmd = NULL;
1046 struct net *net = sock_net(ctnl);
1047 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1050 if (nfqa[NFQA_CFG_CMD]) {
1051 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1053 /* Obsolete commands without queue context */
1054 switch (cmd->command) {
1055 case NFQNL_CFG_CMD_PF_BIND: return 0;
1056 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1061 queue = instance_lookup(q, queue_num);
1062 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1064 goto err_out_unlock;
1068 switch (cmd->command) {
1069 case NFQNL_CFG_CMD_BIND:
1072 goto err_out_unlock;
1074 queue = instance_create(q, queue_num,
1075 NETLINK_CB(skb).portid);
1076 if (IS_ERR(queue)) {
1077 ret = PTR_ERR(queue);
1078 goto err_out_unlock;
1081 case NFQNL_CFG_CMD_UNBIND:
1084 goto err_out_unlock;
1086 instance_destroy(q, queue);
1088 case NFQNL_CFG_CMD_PF_BIND:
1089 case NFQNL_CFG_CMD_PF_UNBIND:
1097 if (nfqa[NFQA_CFG_PARAMS]) {
1098 struct nfqnl_msg_config_params *params;
1102 goto err_out_unlock;
1104 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1105 nfqnl_set_mode(queue, params->copy_mode,
1106 ntohl(params->copy_range));
1109 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1110 __be32 *queue_maxlen;
1114 goto err_out_unlock;
1116 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1117 spin_lock_bh(&queue->lock);
1118 queue->queue_maxlen = ntohl(*queue_maxlen);
1119 spin_unlock_bh(&queue->lock);
1122 if (nfqa[NFQA_CFG_FLAGS]) {
1127 goto err_out_unlock;
1130 if (!nfqa[NFQA_CFG_MASK]) {
1131 /* A mask is needed to specify which flags are being
1135 goto err_out_unlock;
1138 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1139 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1141 if (flags >= NFQA_CFG_F_MAX) {
1143 goto err_out_unlock;
1146 spin_lock_bh(&queue->lock);
1147 queue->flags &= ~mask;
1148 queue->flags |= flags & mask;
1149 spin_unlock_bh(&queue->lock);
1157 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1158 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1159 .attr_count = NFQA_MAX, },
1160 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1161 .attr_count = NFQA_MAX,
1162 .policy = nfqa_verdict_policy },
1163 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1164 .attr_count = NFQA_CFG_MAX,
1165 .policy = nfqa_cfg_policy },
1166 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1167 .attr_count = NFQA_MAX,
1168 .policy = nfqa_verdict_batch_policy },
1171 static const struct nfnetlink_subsystem nfqnl_subsys = {
1173 .subsys_id = NFNL_SUBSYS_QUEUE,
1174 .cb_count = NFQNL_MSG_MAX,
1178 #ifdef CONFIG_PROC_FS
1180 struct seq_net_private p;
1181 unsigned int bucket;
1184 static struct hlist_node *get_first(struct seq_file *seq)
1186 struct iter_state *st = seq->private;
1188 struct nfnl_queue_net *q;
1193 net = seq_file_net(seq);
1194 q = nfnl_queue_pernet(net);
1195 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1196 if (!hlist_empty(&q->instance_table[st->bucket]))
1197 return q->instance_table[st->bucket].first;
1202 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1204 struct iter_state *st = seq->private;
1205 struct net *net = seq_file_net(seq);
1209 struct nfnl_queue_net *q;
1211 if (++st->bucket >= INSTANCE_BUCKETS)
1214 q = nfnl_queue_pernet(net);
1215 h = q->instance_table[st->bucket].first;
1220 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1222 struct hlist_node *head;
1223 head = get_first(seq);
1226 while (pos && (head = get_next(seq, head)))
1228 return pos ? NULL : head;
1231 static void *seq_start(struct seq_file *s, loff_t *pos)
1232 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1234 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1235 return get_idx(s, *pos);
1238 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1241 return get_next(s, v);
1244 static void seq_stop(struct seq_file *s, void *v)
1245 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1247 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1250 static int seq_show(struct seq_file *s, void *v)
1252 const struct nfqnl_instance *inst = v;
1254 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1256 inst->peer_portid, inst->queue_total,
1257 inst->copy_mode, inst->copy_range,
1258 inst->queue_dropped, inst->queue_user_dropped,
1259 inst->id_sequence, 1);
1263 static const struct seq_operations nfqnl_seq_ops = {
1270 static int nfqnl_open(struct inode *inode, struct file *file)
1272 return seq_open_net(inode, file, &nfqnl_seq_ops,
1273 sizeof(struct iter_state));
1276 static const struct file_operations nfqnl_file_ops = {
1277 .owner = THIS_MODULE,
1280 .llseek = seq_lseek,
1281 .release = seq_release_net,
1284 #endif /* PROC_FS */
1286 static int __net_init nfnl_queue_net_init(struct net *net)
1289 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1291 for (i = 0; i < INSTANCE_BUCKETS; i++)
1292 INIT_HLIST_HEAD(&q->instance_table[i]);
1294 spin_lock_init(&q->instances_lock);
1296 #ifdef CONFIG_PROC_FS
1297 if (!proc_create("nfnetlink_queue", 0440,
1298 net->nf.proc_netfilter, &nfqnl_file_ops))
1304 static void __net_exit nfnl_queue_net_exit(struct net *net)
1306 #ifdef CONFIG_PROC_FS
1307 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1311 static struct pernet_operations nfnl_queue_net_ops = {
1312 .init = nfnl_queue_net_init,
1313 .exit = nfnl_queue_net_exit,
1314 .id = &nfnl_queue_net_id,
1315 .size = sizeof(struct nfnl_queue_net),
1318 static int __init nfnetlink_queue_init(void)
1320 int status = -ENOMEM;
1322 netlink_register_notifier(&nfqnl_rtnl_notifier);
1323 status = nfnetlink_subsys_register(&nfqnl_subsys);
1325 pr_err("nf_queue: failed to create netlink socket\n");
1326 goto cleanup_netlink_notifier;
1329 status = register_pernet_subsys(&nfnl_queue_net_ops);
1331 pr_err("nf_queue: failed to register pernet ops\n");
1332 goto cleanup_subsys;
1334 register_netdevice_notifier(&nfqnl_dev_notifier);
1335 nf_register_queue_handler(&nfqh);
1339 nfnetlink_subsys_unregister(&nfqnl_subsys);
1340 cleanup_netlink_notifier:
1341 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1345 static void __exit nfnetlink_queue_fini(void)
1347 nf_unregister_queue_handler();
1348 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1349 unregister_pernet_subsys(&nfnl_queue_net_ops);
1350 nfnetlink_subsys_unregister(&nfqnl_subsys);
1351 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1353 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1356 MODULE_DESCRIPTION("netfilter packet queue handler");
1357 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1358 MODULE_LICENSE("GPL");
1359 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1361 module_init(nfnetlink_queue_init);
1362 module_exit(nfnetlink_queue_fini);