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/netfilter/nf_conntrack_common.h>
32 #include <linux/list.h>
34 #include <net/tcp_states.h>
35 #include <net/netfilter/nf_queue.h>
36 #include <net/netns/generic.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 u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
300 struct sk_buff *entskb = entry->skb;
303 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
306 if (skb_vlan_tag_present(entskb))
307 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
308 nla_total_size(sizeof(__be16)));
310 if (entskb->network_header > entskb->mac_header)
311 nlalen += nla_total_size((entskb->network_header -
312 entskb->mac_header));
317 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
319 struct sk_buff *entskb = entry->skb;
321 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
324 if (skb_vlan_tag_present(entskb)) {
327 nest = nla_nest_start(skb, NFQA_VLAN | NLA_F_NESTED);
329 goto nla_put_failure;
331 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
332 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
333 goto nla_put_failure;
335 nla_nest_end(skb, nest);
338 if (entskb->mac_header < entskb->network_header) {
339 int len = (int)(entskb->network_header - entskb->mac_header);
341 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
342 goto nla_put_failure;
351 static struct sk_buff *
352 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
353 struct nf_queue_entry *entry,
354 __be32 **packet_id_ptr)
357 size_t data_len = 0, cap_len = 0;
358 unsigned int hlen = 0;
361 struct nfqnl_msg_packet_hdr *pmsg;
362 struct nlmsghdr *nlh;
363 struct nfgenmsg *nfmsg;
364 struct sk_buff *entskb = entry->skb;
365 struct net_device *indev;
366 struct net_device *outdev;
367 struct nf_conn *ct = NULL;
368 enum ip_conntrack_info uninitialized_var(ctinfo);
369 struct nfnl_ct_hook *nfnl_ct;
371 char *secdata = NULL;
374 size = nlmsg_total_size(sizeof(struct nfgenmsg))
375 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
376 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
377 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
378 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
379 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
380 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
382 + nla_total_size(sizeof(u_int32_t)) /* mark */
383 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
384 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
385 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
387 if (entskb->tstamp.tv64)
388 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
390 size += nfqnl_get_bridge_size(entry);
392 if (entry->state.hook <= NF_INET_FORWARD ||
393 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
394 csum_verify = !skb_csum_unnecessary(entskb);
398 outdev = entry->state.out;
400 switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
401 case NFQNL_COPY_META:
402 case NFQNL_COPY_NONE:
405 case NFQNL_COPY_PACKET:
406 if (!(queue->flags & NFQA_CFG_F_GSO) &&
407 entskb->ip_summed == CHECKSUM_PARTIAL &&
408 skb_checksum_help(entskb))
411 data_len = ACCESS_ONCE(queue->copy_range);
412 if (data_len > entskb->len)
413 data_len = entskb->len;
415 hlen = skb_zerocopy_headlen(entskb);
416 hlen = min_t(unsigned int, hlen, data_len);
417 size += sizeof(struct nlattr) + hlen;
418 cap_len = entskb->len;
422 nfnl_ct = rcu_dereference(nfnl_ct_hook);
424 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
425 if (nfnl_ct != NULL) {
426 ct = nfnl_ct->get_ct(entskb, &ctinfo);
428 size += nfnl_ct->build_size(ct);
432 if (queue->flags & NFQA_CFG_F_UID_GID) {
433 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
434 + nla_total_size(sizeof(u_int32_t))); /* gid */
437 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
438 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
440 size += nla_total_size(seclen);
443 skb = alloc_skb(size, GFP_ATOMIC);
445 skb_tx_error(entskb);
449 nlh = nlmsg_put(skb, 0, 0,
450 NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
451 sizeof(struct nfgenmsg), 0);
453 skb_tx_error(entskb);
457 nfmsg = nlmsg_data(nlh);
458 nfmsg->nfgen_family = entry->state.pf;
459 nfmsg->version = NFNETLINK_V0;
460 nfmsg->res_id = htons(queue->queue_num);
462 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
463 pmsg = nla_data(nla);
464 pmsg->hw_protocol = entskb->protocol;
465 pmsg->hook = entry->state.hook;
466 *packet_id_ptr = &pmsg->packet_id;
468 indev = entry->state.in;
470 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
471 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
472 goto nla_put_failure;
474 if (entry->state.pf == PF_BRIDGE) {
475 /* Case 1: indev is physical input device, we need to
476 * look for bridge group (when called from
477 * netfilter_bridge) */
478 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
479 htonl(indev->ifindex)) ||
480 /* this is the bridge group "brX" */
481 /* rcu_read_lock()ed by __nf_queue */
482 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
483 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
484 goto nla_put_failure;
488 /* Case 2: indev is bridge group, we need to look for
489 * physical device (when called from ipv4) */
490 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
491 htonl(indev->ifindex)))
492 goto nla_put_failure;
494 physinif = nf_bridge_get_physinif(entskb);
496 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
498 goto nla_put_failure;
504 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
505 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
506 goto nla_put_failure;
508 if (entry->state.pf == PF_BRIDGE) {
509 /* Case 1: outdev is physical output device, we need to
510 * look for bridge group (when called from
511 * netfilter_bridge) */
512 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
513 htonl(outdev->ifindex)) ||
514 /* this is the bridge group "brX" */
515 /* rcu_read_lock()ed by __nf_queue */
516 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
517 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
518 goto nla_put_failure;
522 /* Case 2: outdev is bridge group, we need to look for
523 * physical output device (when called from ipv4) */
524 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
525 htonl(outdev->ifindex)))
526 goto nla_put_failure;
528 physoutif = nf_bridge_get_physoutif(entskb);
530 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
532 goto nla_put_failure;
538 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
539 goto nla_put_failure;
541 if (indev && entskb->dev &&
542 skb_mac_header_was_set(entskb)) {
543 struct nfqnl_msg_packet_hw phw;
546 memset(&phw, 0, sizeof(phw));
547 len = dev_parse_header(entskb, phw.hw_addr);
549 phw.hw_addrlen = htons(len);
550 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
551 goto nla_put_failure;
555 if (nfqnl_put_bridge(entry, skb) < 0)
556 goto nla_put_failure;
558 if (entskb->tstamp.tv64) {
559 struct nfqnl_msg_packet_timestamp ts;
560 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
562 ts.sec = cpu_to_be64(kts.tv_sec);
563 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
565 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
566 goto nla_put_failure;
569 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
570 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
571 goto nla_put_failure;
573 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
574 goto nla_put_failure;
576 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
577 goto nla_put_failure;
579 if (cap_len > data_len &&
580 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
581 goto nla_put_failure;
583 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
584 goto nla_put_failure;
589 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
590 goto nla_put_failure;
592 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
593 nla->nla_type = NFQA_PAYLOAD;
594 nla->nla_len = nla_attr_size(data_len);
596 if (skb_zerocopy(skb, entskb, data_len, hlen))
597 goto nla_put_failure;
600 nlh->nlmsg_len = skb->len;
602 security_release_secctx(secdata, seclen);
606 skb_tx_error(entskb);
608 net_err_ratelimited("nf_queue: error creating packet message\n");
611 security_release_secctx(secdata, seclen);
616 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
617 struct nf_queue_entry *entry)
619 struct sk_buff *nskb;
621 __be32 *packet_id_ptr;
624 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
629 spin_lock_bh(&queue->lock);
631 if (queue->queue_total >= queue->queue_maxlen) {
632 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
636 queue->queue_dropped++;
637 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
640 goto err_out_free_nskb;
642 entry->id = ++queue->id_sequence;
643 *packet_id_ptr = htonl(entry->id);
645 /* nfnetlink_unicast will either free the nskb or add it to a socket */
646 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
648 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
652 queue->queue_user_dropped++;
657 __enqueue_entry(queue, entry);
659 spin_unlock_bh(&queue->lock);
665 spin_unlock_bh(&queue->lock);
667 nf_reinject(entry, NF_ACCEPT);
672 static struct nf_queue_entry *
673 nf_queue_entry_dup(struct nf_queue_entry *e)
675 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
680 if (nf_queue_entry_get_refs(entry))
687 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
688 /* When called from bridge netfilter, skb->data must point to MAC header
689 * before calling skb_gso_segment(). Else, original MAC header is lost
690 * and segmented skbs will be sent to wrong destination.
692 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
695 __skb_push(skb, skb->network_header - skb->mac_header);
698 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
701 __skb_pull(skb, skb->network_header - skb->mac_header);
704 #define nf_bridge_adjust_skb_data(s) do {} while (0)
705 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
708 static void free_entry(struct nf_queue_entry *entry)
710 nf_queue_entry_release_refs(entry);
715 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
716 struct sk_buff *skb, struct nf_queue_entry *entry)
719 struct nf_queue_entry *entry_seg;
721 nf_bridge_adjust_segmented_data(skb);
723 if (skb->next == NULL) { /* last packet, no need to copy entry */
724 struct sk_buff *gso_skb = entry->skb;
726 ret = __nfqnl_enqueue_packet(net, queue, entry);
728 entry->skb = gso_skb;
734 entry_seg = nf_queue_entry_dup(entry);
736 entry_seg->skb = skb;
737 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
739 free_entry(entry_seg);
745 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
748 struct nfqnl_instance *queue;
749 struct sk_buff *skb, *segs;
751 struct net *net = entry->state.net;
752 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
754 /* rcu_read_lock()ed by nf_hook_thresh */
755 queue = instance_lookup(q, queuenum);
759 if (queue->copy_mode == NFQNL_COPY_NONE)
764 switch (entry->state.pf) {
766 skb->protocol = htons(ETH_P_IP);
769 skb->protocol = htons(ETH_P_IPV6);
773 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
774 return __nfqnl_enqueue_packet(net, queue, entry);
776 nf_bridge_adjust_skb_data(skb);
777 segs = skb_gso_segment(skb, 0);
778 /* Does not use PTR_ERR to limit the number of error codes that can be
779 * returned by nf_queue. For instance, callers rely on -ESRCH to
780 * mean 'ignore this hook'.
782 if (IS_ERR_OR_NULL(segs))
787 struct sk_buff *nskb = segs->next;
789 err = __nfqnl_enqueue_packet_gso(net, queue,
799 if (err) /* some segments are already queued */
805 nf_bridge_adjust_segmented_data(skb);
810 nfqnl_mangle(void *data, unsigned int data_len, struct nf_queue_entry *e, int diff)
812 struct sk_buff *nskb;
815 unsigned int min_len = skb_transport_offset(e->skb);
817 if (data_len < min_len)
820 if (pskb_trim(e->skb, data_len))
822 } else if (diff > 0) {
823 if (data_len > 0xFFFF)
825 if (diff > skb_tailroom(e->skb)) {
826 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
829 printk(KERN_WARNING "nf_queue: OOM "
830 "in mangle, dropping packet\n");
836 skb_put(e->skb, diff);
838 if (!skb_make_writable(e->skb, data_len))
840 skb_copy_to_linear_data(e->skb, data, data_len);
841 e->skb->ip_summed = CHECKSUM_NONE;
846 nfqnl_set_mode(struct nfqnl_instance *queue,
847 unsigned char mode, unsigned int range)
851 spin_lock_bh(&queue->lock);
853 case NFQNL_COPY_NONE:
854 case NFQNL_COPY_META:
855 queue->copy_mode = mode;
856 queue->copy_range = 0;
859 case NFQNL_COPY_PACKET:
860 queue->copy_mode = mode;
861 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
862 queue->copy_range = NFQNL_MAX_COPY_RANGE;
864 queue->copy_range = range;
871 spin_unlock_bh(&queue->lock);
877 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
880 if (entry->state.in->ifindex == ifindex)
882 if (entry->state.out)
883 if (entry->state.out->ifindex == ifindex)
885 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
886 if (entry->skb->nf_bridge) {
887 int physinif, physoutif;
889 physinif = nf_bridge_get_physinif(entry->skb);
890 physoutif = nf_bridge_get_physoutif(entry->skb);
892 if (physinif == ifindex || physoutif == ifindex)
899 /* drop all packets with either indev or outdev == ifindex from all queue
902 nfqnl_dev_drop(struct net *net, int ifindex)
905 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
909 for (i = 0; i < INSTANCE_BUCKETS; i++) {
910 struct nfqnl_instance *inst;
911 struct hlist_head *head = &q->instance_table[i];
913 hlist_for_each_entry_rcu(inst, head, hlist)
914 nfqnl_flush(inst, dev_cmp, ifindex);
921 nfqnl_rcv_dev_event(struct notifier_block *this,
922 unsigned long event, void *ptr)
924 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
926 /* Drop any packets associated with the downed device */
927 if (event == NETDEV_DOWN)
928 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
932 static struct notifier_block nfqnl_dev_notifier = {
933 .notifier_call = nfqnl_rcv_dev_event,
936 static int nf_hook_cmp(struct nf_queue_entry *entry, unsigned long entry_ptr)
938 return rcu_access_pointer(entry->state.hook_entries) ==
939 (struct nf_hook_entry *)entry_ptr;
942 static void nfqnl_nf_hook_drop(struct net *net,
943 const struct nf_hook_entry *hook)
945 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
949 for (i = 0; i < INSTANCE_BUCKETS; i++) {
950 struct nfqnl_instance *inst;
951 struct hlist_head *head = &q->instance_table[i];
953 hlist_for_each_entry_rcu(inst, head, hlist)
954 nfqnl_flush(inst, nf_hook_cmp, (unsigned long)hook);
960 nfqnl_rcv_nl_event(struct notifier_block *this,
961 unsigned long event, void *ptr)
963 struct netlink_notify *n = ptr;
964 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
966 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
969 /* destroy all instances for this portid */
970 spin_lock(&q->instances_lock);
971 for (i = 0; i < INSTANCE_BUCKETS; i++) {
972 struct hlist_node *t2;
973 struct nfqnl_instance *inst;
974 struct hlist_head *head = &q->instance_table[i];
976 hlist_for_each_entry_safe(inst, t2, head, hlist) {
977 if (n->portid == inst->peer_portid)
978 __instance_destroy(inst);
981 spin_unlock(&q->instances_lock);
986 static struct notifier_block nfqnl_rtnl_notifier = {
987 .notifier_call = nfqnl_rcv_nl_event,
990 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
991 [NFQA_VLAN_TCI] = { .type = NLA_U16},
992 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
995 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
996 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
997 [NFQA_MARK] = { .type = NLA_U32 },
998 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
999 [NFQA_CT] = { .type = NLA_UNSPEC },
1000 [NFQA_EXP] = { .type = NLA_UNSPEC },
1001 [NFQA_VLAN] = { .type = NLA_NESTED },
1004 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1005 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1006 [NFQA_MARK] = { .type = NLA_U32 },
1009 static struct nfqnl_instance *
1010 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1012 struct nfqnl_instance *queue;
1014 queue = instance_lookup(q, queue_num);
1016 return ERR_PTR(-ENODEV);
1018 if (queue->peer_portid != nlportid)
1019 return ERR_PTR(-EPERM);
1024 static struct nfqnl_msg_verdict_hdr*
1025 verdicthdr_get(const struct nlattr * const nfqa[])
1027 struct nfqnl_msg_verdict_hdr *vhdr;
1028 unsigned int verdict;
1030 if (!nfqa[NFQA_VERDICT_HDR])
1033 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1034 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1035 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1040 static int nfq_id_after(unsigned int id, unsigned int max)
1042 return (int)(id - max) > 0;
1045 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1046 struct sk_buff *skb,
1047 const struct nlmsghdr *nlh,
1048 const struct nlattr * const nfqa[])
1050 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1051 struct nf_queue_entry *entry, *tmp;
1052 unsigned int verdict, maxid;
1053 struct nfqnl_msg_verdict_hdr *vhdr;
1054 struct nfqnl_instance *queue;
1055 LIST_HEAD(batch_list);
1056 u16 queue_num = ntohs(nfmsg->res_id);
1057 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1059 queue = verdict_instance_lookup(q, queue_num,
1060 NETLINK_CB(skb).portid);
1062 return PTR_ERR(queue);
1064 vhdr = verdicthdr_get(nfqa);
1068 verdict = ntohl(vhdr->verdict);
1069 maxid = ntohl(vhdr->id);
1071 spin_lock_bh(&queue->lock);
1073 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1074 if (nfq_id_after(entry->id, maxid))
1076 __dequeue_entry(queue, entry);
1077 list_add_tail(&entry->list, &batch_list);
1080 spin_unlock_bh(&queue->lock);
1082 if (list_empty(&batch_list))
1085 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1086 if (nfqa[NFQA_MARK])
1087 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1088 nf_reinject(entry, verdict);
1093 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1094 const struct nlmsghdr *nlh,
1095 const struct nlattr * const nfqa[],
1096 struct nf_queue_entry *entry,
1097 enum ip_conntrack_info *ctinfo)
1101 ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1105 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1109 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1110 NETLINK_CB(entry->skb).portid,
1115 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1116 const struct nlattr * const nfqa[])
1118 if (nfqa[NFQA_VLAN]) {
1119 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1122 err = nla_parse_nested(tb, NFQA_VLAN_MAX, nfqa[NFQA_VLAN],
1127 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1130 entry->skb->vlan_tci = ntohs(nla_get_be16(tb[NFQA_VLAN_TCI]));
1131 entry->skb->vlan_proto = nla_get_be16(tb[NFQA_VLAN_PROTO]);
1134 if (nfqa[NFQA_L2HDR]) {
1135 int mac_header_len = entry->skb->network_header -
1136 entry->skb->mac_header;
1138 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1140 else if (mac_header_len > 0)
1141 memcpy(skb_mac_header(entry->skb),
1142 nla_data(nfqa[NFQA_L2HDR]),
1149 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1150 struct sk_buff *skb,
1151 const struct nlmsghdr *nlh,
1152 const struct nlattr * const nfqa[])
1154 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1155 u_int16_t queue_num = ntohs(nfmsg->res_id);
1156 struct nfqnl_msg_verdict_hdr *vhdr;
1157 struct nfqnl_instance *queue;
1158 unsigned int verdict;
1159 struct nf_queue_entry *entry;
1160 enum ip_conntrack_info uninitialized_var(ctinfo);
1161 struct nfnl_ct_hook *nfnl_ct;
1162 struct nf_conn *ct = NULL;
1163 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1166 queue = verdict_instance_lookup(q, queue_num,
1167 NETLINK_CB(skb).portid);
1169 return PTR_ERR(queue);
1171 vhdr = verdicthdr_get(nfqa);
1175 verdict = ntohl(vhdr->verdict);
1177 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1181 /* rcu lock already held from nfnl->call_rcu. */
1182 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1184 if (nfqa[NFQA_CT]) {
1185 if (nfnl_ct != NULL)
1186 ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1189 if (entry->state.pf == PF_BRIDGE) {
1190 err = nfqa_parse_bridge(entry, nfqa);
1195 if (nfqa[NFQA_PAYLOAD]) {
1196 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1197 int diff = payload_len - entry->skb->len;
1199 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1200 payload_len, entry, diff) < 0)
1204 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1207 if (nfqa[NFQA_MARK])
1208 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1210 nf_reinject(entry, verdict);
1214 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1215 struct sk_buff *skb, const struct nlmsghdr *nlh,
1216 const struct nlattr * const nfqa[])
1221 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1222 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1223 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1224 [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 },
1225 [NFQA_CFG_MASK] = { .type = NLA_U32 },
1226 [NFQA_CFG_FLAGS] = { .type = NLA_U32 },
1229 static const struct nf_queue_handler nfqh = {
1230 .outfn = &nfqnl_enqueue_packet,
1231 .nf_hook_drop = &nfqnl_nf_hook_drop,
1234 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1235 struct sk_buff *skb, const struct nlmsghdr *nlh,
1236 const struct nlattr * const nfqa[])
1238 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1239 u_int16_t queue_num = ntohs(nfmsg->res_id);
1240 struct nfqnl_instance *queue;
1241 struct nfqnl_msg_config_cmd *cmd = NULL;
1242 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1243 __u32 flags = 0, mask = 0;
1246 if (nfqa[NFQA_CFG_CMD]) {
1247 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1249 /* Obsolete commands without queue context */
1250 switch (cmd->command) {
1251 case NFQNL_CFG_CMD_PF_BIND: return 0;
1252 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1256 /* Check if we support these flags in first place, dependencies should
1257 * be there too not to break atomicity.
1259 if (nfqa[NFQA_CFG_FLAGS]) {
1260 if (!nfqa[NFQA_CFG_MASK]) {
1261 /* A mask is needed to specify which flags are being
1267 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1268 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1270 if (flags >= NFQA_CFG_F_MAX)
1273 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1274 if (flags & mask & NFQA_CFG_F_SECCTX)
1277 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1278 !rcu_access_pointer(nfnl_ct_hook)) {
1279 #ifdef CONFIG_MODULES
1280 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1281 request_module("ip_conntrack_netlink");
1282 nfnl_lock(NFNL_SUBSYS_QUEUE);
1283 if (rcu_access_pointer(nfnl_ct_hook))
1291 queue = instance_lookup(q, queue_num);
1292 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1294 goto err_out_unlock;
1298 switch (cmd->command) {
1299 case NFQNL_CFG_CMD_BIND:
1302 goto err_out_unlock;
1304 queue = instance_create(q, queue_num,
1305 NETLINK_CB(skb).portid);
1306 if (IS_ERR(queue)) {
1307 ret = PTR_ERR(queue);
1308 goto err_out_unlock;
1311 case NFQNL_CFG_CMD_UNBIND:
1314 goto err_out_unlock;
1316 instance_destroy(q, queue);
1317 goto err_out_unlock;
1318 case NFQNL_CFG_CMD_PF_BIND:
1319 case NFQNL_CFG_CMD_PF_UNBIND:
1323 goto err_out_unlock;
1329 goto err_out_unlock;
1332 if (nfqa[NFQA_CFG_PARAMS]) {
1333 struct nfqnl_msg_config_params *params =
1334 nla_data(nfqa[NFQA_CFG_PARAMS]);
1336 nfqnl_set_mode(queue, params->copy_mode,
1337 ntohl(params->copy_range));
1340 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1341 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1343 spin_lock_bh(&queue->lock);
1344 queue->queue_maxlen = ntohl(*queue_maxlen);
1345 spin_unlock_bh(&queue->lock);
1348 if (nfqa[NFQA_CFG_FLAGS]) {
1349 spin_lock_bh(&queue->lock);
1350 queue->flags &= ~mask;
1351 queue->flags |= flags & mask;
1352 spin_unlock_bh(&queue->lock);
1360 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1361 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1362 .attr_count = NFQA_MAX, },
1363 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1364 .attr_count = NFQA_MAX,
1365 .policy = nfqa_verdict_policy },
1366 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1367 .attr_count = NFQA_CFG_MAX,
1368 .policy = nfqa_cfg_policy },
1369 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1370 .attr_count = NFQA_MAX,
1371 .policy = nfqa_verdict_batch_policy },
1374 static const struct nfnetlink_subsystem nfqnl_subsys = {
1376 .subsys_id = NFNL_SUBSYS_QUEUE,
1377 .cb_count = NFQNL_MSG_MAX,
1381 #ifdef CONFIG_PROC_FS
1383 struct seq_net_private p;
1384 unsigned int bucket;
1387 static struct hlist_node *get_first(struct seq_file *seq)
1389 struct iter_state *st = seq->private;
1391 struct nfnl_queue_net *q;
1396 net = seq_file_net(seq);
1397 q = nfnl_queue_pernet(net);
1398 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1399 if (!hlist_empty(&q->instance_table[st->bucket]))
1400 return q->instance_table[st->bucket].first;
1405 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1407 struct iter_state *st = seq->private;
1408 struct net *net = seq_file_net(seq);
1412 struct nfnl_queue_net *q;
1414 if (++st->bucket >= INSTANCE_BUCKETS)
1417 q = nfnl_queue_pernet(net);
1418 h = q->instance_table[st->bucket].first;
1423 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1425 struct hlist_node *head;
1426 head = get_first(seq);
1429 while (pos && (head = get_next(seq, head)))
1431 return pos ? NULL : head;
1434 static void *seq_start(struct seq_file *s, loff_t *pos)
1435 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1437 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1438 return get_idx(s, *pos);
1441 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1444 return get_next(s, v);
1447 static void seq_stop(struct seq_file *s, void *v)
1448 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1450 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1453 static int seq_show(struct seq_file *s, void *v)
1455 const struct nfqnl_instance *inst = v;
1457 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1459 inst->peer_portid, inst->queue_total,
1460 inst->copy_mode, inst->copy_range,
1461 inst->queue_dropped, inst->queue_user_dropped,
1462 inst->id_sequence, 1);
1466 static const struct seq_operations nfqnl_seq_ops = {
1473 static int nfqnl_open(struct inode *inode, struct file *file)
1475 return seq_open_net(inode, file, &nfqnl_seq_ops,
1476 sizeof(struct iter_state));
1479 static const struct file_operations nfqnl_file_ops = {
1480 .owner = THIS_MODULE,
1483 .llseek = seq_lseek,
1484 .release = seq_release_net,
1487 #endif /* PROC_FS */
1489 static int __net_init nfnl_queue_net_init(struct net *net)
1492 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1494 for (i = 0; i < INSTANCE_BUCKETS; i++)
1495 INIT_HLIST_HEAD(&q->instance_table[i]);
1497 spin_lock_init(&q->instances_lock);
1499 #ifdef CONFIG_PROC_FS
1500 if (!proc_create("nfnetlink_queue", 0440,
1501 net->nf.proc_netfilter, &nfqnl_file_ops))
1504 nf_register_queue_handler(net, &nfqh);
1508 static void __net_exit nfnl_queue_net_exit(struct net *net)
1510 nf_unregister_queue_handler(net);
1511 #ifdef CONFIG_PROC_FS
1512 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1516 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1521 static struct pernet_operations nfnl_queue_net_ops = {
1522 .init = nfnl_queue_net_init,
1523 .exit = nfnl_queue_net_exit,
1524 .exit_batch = nfnl_queue_net_exit_batch,
1525 .id = &nfnl_queue_net_id,
1526 .size = sizeof(struct nfnl_queue_net),
1529 static int __init nfnetlink_queue_init(void)
1533 status = register_pernet_subsys(&nfnl_queue_net_ops);
1535 pr_err("nf_queue: failed to register pernet ops\n");
1539 netlink_register_notifier(&nfqnl_rtnl_notifier);
1540 status = nfnetlink_subsys_register(&nfqnl_subsys);
1542 pr_err("nf_queue: failed to create netlink socket\n");
1543 goto cleanup_netlink_notifier;
1546 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1548 pr_err("nf_queue: failed to register netdevice notifier\n");
1549 goto cleanup_netlink_subsys;
1554 cleanup_netlink_subsys:
1555 nfnetlink_subsys_unregister(&nfqnl_subsys);
1556 cleanup_netlink_notifier:
1557 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1558 unregister_pernet_subsys(&nfnl_queue_net_ops);
1563 static void __exit nfnetlink_queue_fini(void)
1565 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1566 nfnetlink_subsys_unregister(&nfqnl_subsys);
1567 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1568 unregister_pernet_subsys(&nfnl_queue_net_ops);
1570 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1573 MODULE_DESCRIPTION("netfilter packet queue handler");
1574 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1575 MODULE_LICENSE("GPL");
1576 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1578 module_init(nfnetlink_queue_init);
1579 module_exit(nfnetlink_queue_fini);