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 u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
284 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
285 if (!skb || !sk_fullsock(skb->sk))
288 read_lock_bh(&skb->sk->sk_callback_lock);
291 security_secid_to_secctx(skb->secmark, secdata, &seclen);
293 read_unlock_bh(&skb->sk->sk_callback_lock);
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)
304 size_t data_len = 0, cap_len = 0, rem_len = 0;
305 unsigned int hlen = 0;
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);
317 char *secdata = NULL;
320 size = nlmsg_total_size(sizeof(struct nfgenmsg))
321 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
322 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
323 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
324 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
325 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
326 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
328 + nla_total_size(sizeof(u_int32_t)) /* mark */
329 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
330 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
331 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
333 if (entskb->tstamp.tv64)
334 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
336 if (entry->state.hook <= NF_INET_FORWARD ||
337 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
338 csum_verify = !skb_csum_unnecessary(entskb);
342 outdev = entry->state.out;
344 switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
345 case NFQNL_COPY_META:
346 case NFQNL_COPY_NONE:
349 case NFQNL_COPY_PACKET:
350 if (!(queue->flags & NFQA_CFG_F_GSO) &&
351 entskb->ip_summed == CHECKSUM_PARTIAL &&
352 skb_checksum_help(entskb))
355 data_len = ACCESS_ONCE(queue->copy_range);
356 if (data_len > entskb->len)
357 data_len = entskb->len;
359 hlen = skb_zerocopy_headlen(entskb);
360 hlen = min_t(unsigned int, hlen, data_len);
361 size += sizeof(struct nlattr) + hlen;
362 cap_len = entskb->len;
363 rem_len = data_len - hlen;
367 if (queue->flags & NFQA_CFG_F_CONNTRACK)
368 ct = nfqnl_ct_get(entskb, &size, &ctinfo);
370 if (queue->flags & NFQA_CFG_F_UID_GID) {
371 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
372 + nla_total_size(sizeof(u_int32_t))); /* gid */
375 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
376 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
378 size += nla_total_size(seclen);
381 skb = __netlink_alloc_skb(net->nfnl, size, rem_len, queue->peer_portid,
384 skb_tx_error(entskb);
388 nlh = nlmsg_put(skb, 0, 0,
389 NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
390 sizeof(struct nfgenmsg), 0);
392 skb_tx_error(entskb);
396 nfmsg = nlmsg_data(nlh);
397 nfmsg->nfgen_family = entry->state.pf;
398 nfmsg->version = NFNETLINK_V0;
399 nfmsg->res_id = htons(queue->queue_num);
401 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
402 pmsg = nla_data(nla);
403 pmsg->hw_protocol = entskb->protocol;
404 pmsg->hook = entry->state.hook;
405 *packet_id_ptr = &pmsg->packet_id;
407 indev = entry->state.in;
409 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
410 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
411 goto nla_put_failure;
413 if (entry->state.pf == PF_BRIDGE) {
414 /* Case 1: indev is physical input device, we need to
415 * look for bridge group (when called from
416 * netfilter_bridge) */
417 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
418 htonl(indev->ifindex)) ||
419 /* this is the bridge group "brX" */
420 /* rcu_read_lock()ed by __nf_queue */
421 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
422 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
423 goto nla_put_failure;
427 /* Case 2: indev is bridge group, we need to look for
428 * physical device (when called from ipv4) */
429 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
430 htonl(indev->ifindex)))
431 goto nla_put_failure;
433 physinif = nf_bridge_get_physinif(entskb);
435 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
437 goto nla_put_failure;
443 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
444 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
445 goto nla_put_failure;
447 if (entry->state.pf == PF_BRIDGE) {
448 /* Case 1: outdev is physical output device, we need to
449 * look for bridge group (when called from
450 * netfilter_bridge) */
451 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
452 htonl(outdev->ifindex)) ||
453 /* this is the bridge group "brX" */
454 /* rcu_read_lock()ed by __nf_queue */
455 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
456 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
457 goto nla_put_failure;
461 /* Case 2: outdev is bridge group, we need to look for
462 * physical output device (when called from ipv4) */
463 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
464 htonl(outdev->ifindex)))
465 goto nla_put_failure;
467 physoutif = nf_bridge_get_physoutif(entskb);
469 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
471 goto nla_put_failure;
477 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
478 goto nla_put_failure;
480 if (indev && entskb->dev &&
481 entskb->mac_header != entskb->network_header) {
482 struct nfqnl_msg_packet_hw phw;
485 memset(&phw, 0, sizeof(phw));
486 len = dev_parse_header(entskb, phw.hw_addr);
488 phw.hw_addrlen = htons(len);
489 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
490 goto nla_put_failure;
494 if (entskb->tstamp.tv64) {
495 struct nfqnl_msg_packet_timestamp ts;
496 struct timeval tv = ktime_to_timeval(entskb->tstamp);
497 ts.sec = cpu_to_be64(tv.tv_sec);
498 ts.usec = cpu_to_be64(tv.tv_usec);
500 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
501 goto nla_put_failure;
504 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
505 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
506 goto nla_put_failure;
508 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
509 goto nla_put_failure;
511 if (ct && nfqnl_ct_put(skb, ct, ctinfo) < 0)
512 goto nla_put_failure;
514 if (cap_len > data_len &&
515 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
516 goto nla_put_failure;
518 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
519 goto nla_put_failure;
524 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
525 goto nla_put_failure;
527 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
528 nla->nla_type = NFQA_PAYLOAD;
529 nla->nla_len = nla_attr_size(data_len);
531 if (skb_zerocopy(skb, entskb, data_len, hlen))
532 goto nla_put_failure;
535 nlh->nlmsg_len = skb->len;
539 skb_tx_error(entskb);
541 net_err_ratelimited("nf_queue: error creating packet message\n");
546 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
547 struct nf_queue_entry *entry)
549 struct sk_buff *nskb;
551 __be32 *packet_id_ptr;
554 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
559 spin_lock_bh(&queue->lock);
561 if (queue->queue_total >= queue->queue_maxlen) {
562 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
566 queue->queue_dropped++;
567 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
570 goto err_out_free_nskb;
572 entry->id = ++queue->id_sequence;
573 *packet_id_ptr = htonl(entry->id);
575 /* nfnetlink_unicast will either free the nskb or add it to a socket */
576 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
578 queue->queue_user_dropped++;
582 __enqueue_entry(queue, entry);
584 spin_unlock_bh(&queue->lock);
590 spin_unlock_bh(&queue->lock);
592 nf_reinject(entry, NF_ACCEPT);
597 static struct nf_queue_entry *
598 nf_queue_entry_dup(struct nf_queue_entry *e)
600 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
602 if (nf_queue_entry_get_refs(entry))
609 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
610 /* When called from bridge netfilter, skb->data must point to MAC header
611 * before calling skb_gso_segment(). Else, original MAC header is lost
612 * and segmented skbs will be sent to wrong destination.
614 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
617 __skb_push(skb, skb->network_header - skb->mac_header);
620 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
623 __skb_pull(skb, skb->network_header - skb->mac_header);
626 #define nf_bridge_adjust_skb_data(s) do {} while (0)
627 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
630 static void free_entry(struct nf_queue_entry *entry)
632 nf_queue_entry_release_refs(entry);
637 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
638 struct sk_buff *skb, struct nf_queue_entry *entry)
641 struct nf_queue_entry *entry_seg;
643 nf_bridge_adjust_segmented_data(skb);
645 if (skb->next == NULL) { /* last packet, no need to copy entry */
646 struct sk_buff *gso_skb = entry->skb;
648 ret = __nfqnl_enqueue_packet(net, queue, entry);
650 entry->skb = gso_skb;
656 entry_seg = nf_queue_entry_dup(entry);
658 entry_seg->skb = skb;
659 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
661 free_entry(entry_seg);
667 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
670 struct nfqnl_instance *queue;
671 struct sk_buff *skb, *segs;
673 struct net *net = dev_net(entry->state.in ?
674 entry->state.in : entry->state.out);
675 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
677 /* rcu_read_lock()ed by nf_hook_slow() */
678 queue = instance_lookup(q, queuenum);
682 if (queue->copy_mode == NFQNL_COPY_NONE)
687 switch (entry->state.pf) {
689 skb->protocol = htons(ETH_P_IP);
692 skb->protocol = htons(ETH_P_IPV6);
696 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
697 return __nfqnl_enqueue_packet(net, queue, entry);
699 nf_bridge_adjust_skb_data(skb);
700 segs = skb_gso_segment(skb, 0);
701 /* Does not use PTR_ERR to limit the number of error codes that can be
702 * returned by nf_queue. For instance, callers rely on -ECANCELED to
703 * mean 'ignore this hook'.
705 if (IS_ERR_OR_NULL(segs))
710 struct sk_buff *nskb = segs->next;
712 err = __nfqnl_enqueue_packet_gso(net, queue,
722 if (err) /* some segments are already queued */
728 nf_bridge_adjust_segmented_data(skb);
733 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
735 struct sk_buff *nskb;
738 if (pskb_trim(e->skb, data_len))
740 } else if (diff > 0) {
741 if (data_len > 0xFFFF)
743 if (diff > skb_tailroom(e->skb)) {
744 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
747 printk(KERN_WARNING "nf_queue: OOM "
748 "in mangle, dropping packet\n");
754 skb_put(e->skb, diff);
756 if (!skb_make_writable(e->skb, data_len))
758 skb_copy_to_linear_data(e->skb, data, data_len);
759 e->skb->ip_summed = CHECKSUM_NONE;
764 nfqnl_set_mode(struct nfqnl_instance *queue,
765 unsigned char mode, unsigned int range)
769 spin_lock_bh(&queue->lock);
771 case NFQNL_COPY_NONE:
772 case NFQNL_COPY_META:
773 queue->copy_mode = mode;
774 queue->copy_range = 0;
777 case NFQNL_COPY_PACKET:
778 queue->copy_mode = mode;
779 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
780 queue->copy_range = NFQNL_MAX_COPY_RANGE;
782 queue->copy_range = range;
789 spin_unlock_bh(&queue->lock);
795 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
798 if (entry->state.in->ifindex == ifindex)
800 if (entry->state.out)
801 if (entry->state.out->ifindex == ifindex)
803 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
804 if (entry->skb->nf_bridge) {
805 int physinif, physoutif;
807 physinif = nf_bridge_get_physinif(entry->skb);
808 physoutif = nf_bridge_get_physoutif(entry->skb);
810 if (physinif == ifindex || physoutif == ifindex)
817 /* drop all packets with either indev or outdev == ifindex from all queue
820 nfqnl_dev_drop(struct net *net, int ifindex)
823 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
827 for (i = 0; i < INSTANCE_BUCKETS; i++) {
828 struct nfqnl_instance *inst;
829 struct hlist_head *head = &q->instance_table[i];
831 hlist_for_each_entry_rcu(inst, head, hlist)
832 nfqnl_flush(inst, dev_cmp, ifindex);
839 nfqnl_rcv_dev_event(struct notifier_block *this,
840 unsigned long event, void *ptr)
842 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
844 /* Drop any packets associated with the downed device */
845 if (event == NETDEV_DOWN)
846 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
850 static struct notifier_block nfqnl_dev_notifier = {
851 .notifier_call = nfqnl_rcv_dev_event,
854 static int nf_hook_cmp(struct nf_queue_entry *entry, unsigned long ops_ptr)
856 return entry->elem == (struct nf_hook_ops *)ops_ptr;
859 static void nfqnl_nf_hook_drop(struct net *net, struct nf_hook_ops *hook)
861 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
865 for (i = 0; i < INSTANCE_BUCKETS; i++) {
866 struct nfqnl_instance *inst;
867 struct hlist_head *head = &q->instance_table[i];
869 hlist_for_each_entry_rcu(inst, head, hlist)
870 nfqnl_flush(inst, nf_hook_cmp, (unsigned long)hook);
876 nfqnl_rcv_nl_event(struct notifier_block *this,
877 unsigned long event, void *ptr)
879 struct netlink_notify *n = ptr;
880 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
882 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
885 /* destroy all instances for this portid */
886 spin_lock(&q->instances_lock);
887 for (i = 0; i < INSTANCE_BUCKETS; i++) {
888 struct hlist_node *t2;
889 struct nfqnl_instance *inst;
890 struct hlist_head *head = &q->instance_table[i];
892 hlist_for_each_entry_safe(inst, t2, head, hlist) {
893 if (n->portid == inst->peer_portid)
894 __instance_destroy(inst);
897 spin_unlock(&q->instances_lock);
902 static struct notifier_block nfqnl_rtnl_notifier = {
903 .notifier_call = nfqnl_rcv_nl_event,
906 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
907 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
908 [NFQA_MARK] = { .type = NLA_U32 },
909 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
910 [NFQA_CT] = { .type = NLA_UNSPEC },
911 [NFQA_EXP] = { .type = NLA_UNSPEC },
914 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
915 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
916 [NFQA_MARK] = { .type = NLA_U32 },
919 static struct nfqnl_instance *
920 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
922 struct nfqnl_instance *queue;
924 queue = instance_lookup(q, queue_num);
926 return ERR_PTR(-ENODEV);
928 if (queue->peer_portid != nlportid)
929 return ERR_PTR(-EPERM);
934 static struct nfqnl_msg_verdict_hdr*
935 verdicthdr_get(const struct nlattr * const nfqa[])
937 struct nfqnl_msg_verdict_hdr *vhdr;
938 unsigned int verdict;
940 if (!nfqa[NFQA_VERDICT_HDR])
943 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
944 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
945 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
950 static int nfq_id_after(unsigned int id, unsigned int max)
952 return (int)(id - max) > 0;
956 nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
957 const struct nlmsghdr *nlh,
958 const struct nlattr * const nfqa[])
960 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
961 struct nf_queue_entry *entry, *tmp;
962 unsigned int verdict, maxid;
963 struct nfqnl_msg_verdict_hdr *vhdr;
964 struct nfqnl_instance *queue;
965 LIST_HEAD(batch_list);
966 u16 queue_num = ntohs(nfmsg->res_id);
968 struct net *net = sock_net(ctnl);
969 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
971 queue = verdict_instance_lookup(q, queue_num,
972 NETLINK_CB(skb).portid);
974 return PTR_ERR(queue);
976 vhdr = verdicthdr_get(nfqa);
980 verdict = ntohl(vhdr->verdict);
981 maxid = ntohl(vhdr->id);
983 spin_lock_bh(&queue->lock);
985 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
986 if (nfq_id_after(entry->id, maxid))
988 __dequeue_entry(queue, entry);
989 list_add_tail(&entry->list, &batch_list);
992 spin_unlock_bh(&queue->lock);
994 if (list_empty(&batch_list))
997 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
999 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1000 nf_reinject(entry, verdict);
1006 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
1007 const struct nlmsghdr *nlh,
1008 const struct nlattr * const nfqa[])
1010 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1011 u_int16_t queue_num = ntohs(nfmsg->res_id);
1013 struct nfqnl_msg_verdict_hdr *vhdr;
1014 struct nfqnl_instance *queue;
1015 unsigned int verdict;
1016 struct nf_queue_entry *entry;
1017 enum ip_conntrack_info uninitialized_var(ctinfo);
1018 struct nf_conn *ct = NULL;
1020 struct net *net = sock_net(ctnl);
1021 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1023 queue = instance_lookup(q, queue_num);
1025 queue = verdict_instance_lookup(q, queue_num,
1026 NETLINK_CB(skb).portid);
1028 return PTR_ERR(queue);
1030 vhdr = verdicthdr_get(nfqa);
1034 verdict = ntohl(vhdr->verdict);
1036 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1040 if (nfqa[NFQA_CT]) {
1041 ct = nfqnl_ct_parse(entry->skb, nfqa[NFQA_CT], &ctinfo);
1042 if (ct && nfqa[NFQA_EXP]) {
1043 nfqnl_attach_expect(ct, nfqa[NFQA_EXP],
1044 NETLINK_CB(skb).portid,
1049 if (nfqa[NFQA_PAYLOAD]) {
1050 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1051 int diff = payload_len - entry->skb->len;
1053 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1054 payload_len, entry, diff) < 0)
1058 nfqnl_ct_seq_adjust(entry->skb, ct, ctinfo, diff);
1061 if (nfqa[NFQA_MARK])
1062 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1064 nf_reinject(entry, verdict);
1069 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1070 const struct nlmsghdr *nlh,
1071 const struct nlattr * const nfqa[])
1076 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1077 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1078 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1081 static const struct nf_queue_handler nfqh = {
1082 .outfn = &nfqnl_enqueue_packet,
1083 .nf_hook_drop = &nfqnl_nf_hook_drop,
1087 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1088 const struct nlmsghdr *nlh,
1089 const struct nlattr * const nfqa[])
1091 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1092 u_int16_t queue_num = ntohs(nfmsg->res_id);
1093 struct nfqnl_instance *queue;
1094 struct nfqnl_msg_config_cmd *cmd = NULL;
1095 struct net *net = sock_net(ctnl);
1096 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1099 if (nfqa[NFQA_CFG_CMD]) {
1100 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1102 /* Obsolete commands without queue context */
1103 switch (cmd->command) {
1104 case NFQNL_CFG_CMD_PF_BIND: return 0;
1105 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1110 queue = instance_lookup(q, queue_num);
1111 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1113 goto err_out_unlock;
1117 switch (cmd->command) {
1118 case NFQNL_CFG_CMD_BIND:
1121 goto err_out_unlock;
1123 queue = instance_create(q, queue_num,
1124 NETLINK_CB(skb).portid);
1125 if (IS_ERR(queue)) {
1126 ret = PTR_ERR(queue);
1127 goto err_out_unlock;
1130 case NFQNL_CFG_CMD_UNBIND:
1133 goto err_out_unlock;
1135 instance_destroy(q, queue);
1137 case NFQNL_CFG_CMD_PF_BIND:
1138 case NFQNL_CFG_CMD_PF_UNBIND:
1146 if (nfqa[NFQA_CFG_PARAMS]) {
1147 struct nfqnl_msg_config_params *params;
1151 goto err_out_unlock;
1153 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1154 nfqnl_set_mode(queue, params->copy_mode,
1155 ntohl(params->copy_range));
1158 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1159 __be32 *queue_maxlen;
1163 goto err_out_unlock;
1165 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1166 spin_lock_bh(&queue->lock);
1167 queue->queue_maxlen = ntohl(*queue_maxlen);
1168 spin_unlock_bh(&queue->lock);
1171 if (nfqa[NFQA_CFG_FLAGS]) {
1176 goto err_out_unlock;
1179 if (!nfqa[NFQA_CFG_MASK]) {
1180 /* A mask is needed to specify which flags are being
1184 goto err_out_unlock;
1187 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1188 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1190 if (flags >= NFQA_CFG_F_MAX) {
1192 goto err_out_unlock;
1194 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1195 if (flags & mask & NFQA_CFG_F_SECCTX) {
1197 goto err_out_unlock;
1200 spin_lock_bh(&queue->lock);
1201 queue->flags &= ~mask;
1202 queue->flags |= flags & mask;
1203 spin_unlock_bh(&queue->lock);
1211 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1212 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1213 .attr_count = NFQA_MAX, },
1214 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1215 .attr_count = NFQA_MAX,
1216 .policy = nfqa_verdict_policy },
1217 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1218 .attr_count = NFQA_CFG_MAX,
1219 .policy = nfqa_cfg_policy },
1220 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1221 .attr_count = NFQA_MAX,
1222 .policy = nfqa_verdict_batch_policy },
1225 static const struct nfnetlink_subsystem nfqnl_subsys = {
1227 .subsys_id = NFNL_SUBSYS_QUEUE,
1228 .cb_count = NFQNL_MSG_MAX,
1232 #ifdef CONFIG_PROC_FS
1234 struct seq_net_private p;
1235 unsigned int bucket;
1238 static struct hlist_node *get_first(struct seq_file *seq)
1240 struct iter_state *st = seq->private;
1242 struct nfnl_queue_net *q;
1247 net = seq_file_net(seq);
1248 q = nfnl_queue_pernet(net);
1249 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1250 if (!hlist_empty(&q->instance_table[st->bucket]))
1251 return q->instance_table[st->bucket].first;
1256 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1258 struct iter_state *st = seq->private;
1259 struct net *net = seq_file_net(seq);
1263 struct nfnl_queue_net *q;
1265 if (++st->bucket >= INSTANCE_BUCKETS)
1268 q = nfnl_queue_pernet(net);
1269 h = q->instance_table[st->bucket].first;
1274 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1276 struct hlist_node *head;
1277 head = get_first(seq);
1280 while (pos && (head = get_next(seq, head)))
1282 return pos ? NULL : head;
1285 static void *seq_start(struct seq_file *s, loff_t *pos)
1286 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1288 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1289 return get_idx(s, *pos);
1292 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1295 return get_next(s, v);
1298 static void seq_stop(struct seq_file *s, void *v)
1299 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1301 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1304 static int seq_show(struct seq_file *s, void *v)
1306 const struct nfqnl_instance *inst = v;
1308 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1310 inst->peer_portid, inst->queue_total,
1311 inst->copy_mode, inst->copy_range,
1312 inst->queue_dropped, inst->queue_user_dropped,
1313 inst->id_sequence, 1);
1317 static const struct seq_operations nfqnl_seq_ops = {
1324 static int nfqnl_open(struct inode *inode, struct file *file)
1326 return seq_open_net(inode, file, &nfqnl_seq_ops,
1327 sizeof(struct iter_state));
1330 static const struct file_operations nfqnl_file_ops = {
1331 .owner = THIS_MODULE,
1334 .llseek = seq_lseek,
1335 .release = seq_release_net,
1338 #endif /* PROC_FS */
1340 static int __net_init nfnl_queue_net_init(struct net *net)
1343 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1345 for (i = 0; i < INSTANCE_BUCKETS; i++)
1346 INIT_HLIST_HEAD(&q->instance_table[i]);
1348 spin_lock_init(&q->instances_lock);
1350 #ifdef CONFIG_PROC_FS
1351 if (!proc_create("nfnetlink_queue", 0440,
1352 net->nf.proc_netfilter, &nfqnl_file_ops))
1358 static void __net_exit nfnl_queue_net_exit(struct net *net)
1360 #ifdef CONFIG_PROC_FS
1361 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1365 static struct pernet_operations nfnl_queue_net_ops = {
1366 .init = nfnl_queue_net_init,
1367 .exit = nfnl_queue_net_exit,
1368 .id = &nfnl_queue_net_id,
1369 .size = sizeof(struct nfnl_queue_net),
1372 static int __init nfnetlink_queue_init(void)
1376 status = register_pernet_subsys(&nfnl_queue_net_ops);
1378 pr_err("nf_queue: failed to register pernet ops\n");
1382 netlink_register_notifier(&nfqnl_rtnl_notifier);
1383 status = nfnetlink_subsys_register(&nfqnl_subsys);
1385 pr_err("nf_queue: failed to create netlink socket\n");
1386 goto cleanup_netlink_notifier;
1389 register_netdevice_notifier(&nfqnl_dev_notifier);
1390 nf_register_queue_handler(&nfqh);
1393 cleanup_netlink_notifier:
1394 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1399 static void __exit nfnetlink_queue_fini(void)
1401 nf_unregister_queue_handler();
1402 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1403 nfnetlink_subsys_unregister(&nfqnl_subsys);
1404 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1405 unregister_pernet_subsys(&nfnl_queue_net_ops);
1407 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1410 MODULE_DESCRIPTION("netfilter packet queue handler");
1411 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1412 MODULE_LICENSE("GPL");
1413 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1415 module_init(nfnetlink_queue_init);
1416 module_exit(nfnetlink_queue_fini);