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.
18 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20 #include <linux/module.h>
21 #include <linux/skbuff.h>
22 #include <linux/init.h>
23 #include <linux/spinlock.h>
24 #include <linux/slab.h>
25 #include <linux/notifier.h>
26 #include <linux/netdevice.h>
27 #include <linux/netfilter.h>
28 #include <linux/proc_fs.h>
29 #include <linux/netfilter_ipv4.h>
30 #include <linux/netfilter_ipv6.h>
31 #include <linux/netfilter_bridge.h>
32 #include <linux/netfilter/nfnetlink.h>
33 #include <linux/netfilter/nfnetlink_queue.h>
34 #include <linux/netfilter/nf_conntrack_common.h>
35 #include <linux/list.h>
37 #include <net/tcp_states.h>
38 #include <net/netfilter/nf_queue.h>
39 #include <net/netns/generic.h>
41 #include <linux/atomic.h>
43 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
44 #include "../bridge/br_private.h"
47 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
48 #include <net/netfilter/nf_conntrack.h>
51 #define NFQNL_QMAX_DEFAULT 1024
53 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
54 * includes the header length. Thus, the maximum packet length that we
55 * support is 65531 bytes. We send truncated packets if the specified length
56 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
57 * attribute to detect truncation.
59 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
61 struct nfqnl_instance {
62 struct hlist_node hlist; /* global list of queues */
66 unsigned int queue_maxlen;
67 unsigned int copy_range;
68 unsigned int queue_dropped;
69 unsigned int queue_user_dropped;
72 u_int16_t queue_num; /* number of this queue */
74 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
76 * Following fields are dirtied for each queued packet,
77 * keep them in same cache line if possible.
79 spinlock_t lock ____cacheline_aligned_in_smp;
80 unsigned int queue_total;
81 unsigned int id_sequence; /* 'sequence' of pkt ids */
82 struct list_head queue_list; /* packets in queue */
85 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
87 static unsigned int nfnl_queue_net_id __read_mostly;
89 #define INSTANCE_BUCKETS 16
90 struct nfnl_queue_net {
91 spinlock_t instances_lock;
92 struct hlist_head instance_table[INSTANCE_BUCKETS];
95 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
97 return net_generic(net, nfnl_queue_net_id);
100 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
102 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
105 static struct nfqnl_instance *
106 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
108 struct hlist_head *head;
109 struct nfqnl_instance *inst;
111 head = &q->instance_table[instance_hashfn(queue_num)];
112 hlist_for_each_entry_rcu(inst, head, hlist) {
113 if (inst->queue_num == queue_num)
119 static struct nfqnl_instance *
120 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
122 struct nfqnl_instance *inst;
126 spin_lock(&q->instances_lock);
127 if (instance_lookup(q, queue_num)) {
132 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
138 inst->queue_num = queue_num;
139 inst->peer_portid = portid;
140 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
141 inst->copy_range = NFQNL_MAX_COPY_RANGE;
142 inst->copy_mode = NFQNL_COPY_NONE;
143 spin_lock_init(&inst->lock);
144 INIT_LIST_HEAD(&inst->queue_list);
146 if (!try_module_get(THIS_MODULE)) {
151 h = instance_hashfn(queue_num);
152 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
154 spin_unlock(&q->instances_lock);
161 spin_unlock(&q->instances_lock);
165 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
169 instance_destroy_rcu(struct rcu_head *head)
171 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
174 nfqnl_flush(inst, NULL, 0);
176 module_put(THIS_MODULE);
180 __instance_destroy(struct nfqnl_instance *inst)
182 hlist_del_rcu(&inst->hlist);
183 call_rcu(&inst->rcu, instance_destroy_rcu);
187 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
189 spin_lock(&q->instances_lock);
190 __instance_destroy(inst);
191 spin_unlock(&q->instances_lock);
195 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
197 list_add_tail(&entry->list, &queue->queue_list);
198 queue->queue_total++;
202 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
204 list_del(&entry->list);
205 queue->queue_total--;
208 static struct nf_queue_entry *
209 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
211 struct nf_queue_entry *entry = NULL, *i;
213 spin_lock_bh(&queue->lock);
215 list_for_each_entry(i, &queue->queue_list, list) {
223 __dequeue_entry(queue, entry);
225 spin_unlock_bh(&queue->lock);
230 static void nfqnl_reinject(struct nf_queue_entry *entry, unsigned int verdict)
232 struct nf_ct_hook *ct_hook;
235 if (verdict == NF_ACCEPT ||
236 verdict == NF_STOP) {
238 ct_hook = rcu_dereference(nf_ct_hook);
240 err = ct_hook->update(entry->state.net, entry->skb);
246 nf_reinject(entry, verdict);
250 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
252 struct nf_queue_entry *entry, *next;
254 spin_lock_bh(&queue->lock);
255 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
256 if (!cmpfn || cmpfn(entry, data)) {
257 list_del(&entry->list);
258 queue->queue_total--;
259 nfqnl_reinject(entry, NF_DROP);
262 spin_unlock_bh(&queue->lock);
266 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
271 if (packet->ip_summed == CHECKSUM_PARTIAL)
272 flags = NFQA_SKB_CSUMNOTREADY;
273 else if (csum_verify)
274 flags = NFQA_SKB_CSUM_NOTVERIFIED;
276 if (skb_is_gso(packet))
277 flags |= NFQA_SKB_GSO;
279 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
282 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
284 const struct cred *cred;
286 if (!sk_fullsock(sk))
289 read_lock_bh(&sk->sk_callback_lock);
290 if (sk->sk_socket && sk->sk_socket->file) {
291 cred = sk->sk_socket->file->f_cred;
292 if (nla_put_be32(skb, NFQA_UID,
293 htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
294 goto nla_put_failure;
295 if (nla_put_be32(skb, NFQA_GID,
296 htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
297 goto nla_put_failure;
299 read_unlock_bh(&sk->sk_callback_lock);
303 read_unlock_bh(&sk->sk_callback_lock);
307 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
310 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
311 if (!skb || !sk_fullsock(skb->sk))
314 read_lock_bh(&skb->sk->sk_callback_lock);
317 security_secid_to_secctx(skb->secmark, secdata, &seclen);
319 read_unlock_bh(&skb->sk->sk_callback_lock);
324 static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
326 struct sk_buff *entskb = entry->skb;
329 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
332 if (skb_vlan_tag_present(entskb))
333 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
334 nla_total_size(sizeof(__be16)));
336 if (entskb->network_header > entskb->mac_header)
337 nlalen += nla_total_size((entskb->network_header -
338 entskb->mac_header));
343 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
345 struct sk_buff *entskb = entry->skb;
347 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
350 if (skb_vlan_tag_present(entskb)) {
353 nest = nla_nest_start(skb, NFQA_VLAN | NLA_F_NESTED);
355 goto nla_put_failure;
357 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
358 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
359 goto nla_put_failure;
361 nla_nest_end(skb, nest);
364 if (entskb->mac_header < entskb->network_header) {
365 int len = (int)(entskb->network_header - entskb->mac_header);
367 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
368 goto nla_put_failure;
377 static struct sk_buff *
378 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
379 struct nf_queue_entry *entry,
380 __be32 **packet_id_ptr)
383 size_t data_len = 0, cap_len = 0;
384 unsigned int hlen = 0;
387 struct nfqnl_msg_packet_hdr *pmsg;
388 struct nlmsghdr *nlh;
389 struct nfgenmsg *nfmsg;
390 struct sk_buff *entskb = entry->skb;
391 struct net_device *indev;
392 struct net_device *outdev;
393 struct nf_conn *ct = NULL;
394 enum ip_conntrack_info uninitialized_var(ctinfo);
395 struct nfnl_ct_hook *nfnl_ct;
397 char *secdata = NULL;
400 size = nlmsg_total_size(sizeof(struct nfgenmsg))
401 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
402 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
403 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
404 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
405 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
406 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
408 + nla_total_size(sizeof(u_int32_t)) /* mark */
409 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
410 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
411 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
414 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
416 size += nfqnl_get_bridge_size(entry);
418 if (entry->state.hook <= NF_INET_FORWARD ||
419 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
420 csum_verify = !skb_csum_unnecessary(entskb);
424 outdev = entry->state.out;
426 switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) {
427 case NFQNL_COPY_META:
428 case NFQNL_COPY_NONE:
431 case NFQNL_COPY_PACKET:
432 if (!(queue->flags & NFQA_CFG_F_GSO) &&
433 entskb->ip_summed == CHECKSUM_PARTIAL &&
434 skb_checksum_help(entskb))
437 data_len = READ_ONCE(queue->copy_range);
438 if (data_len > entskb->len)
439 data_len = entskb->len;
441 hlen = skb_zerocopy_headlen(entskb);
442 hlen = min_t(unsigned int, hlen, data_len);
443 size += sizeof(struct nlattr) + hlen;
444 cap_len = entskb->len;
448 nfnl_ct = rcu_dereference(nfnl_ct_hook);
450 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
451 if (nfnl_ct != NULL) {
452 ct = nfnl_ct->get_ct(entskb, &ctinfo);
454 size += nfnl_ct->build_size(ct);
458 if (queue->flags & NFQA_CFG_F_UID_GID) {
459 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
460 + nla_total_size(sizeof(u_int32_t))); /* gid */
463 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
464 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
466 size += nla_total_size(seclen);
469 skb = alloc_skb(size, GFP_ATOMIC);
471 skb_tx_error(entskb);
475 nlh = nlmsg_put(skb, 0, 0,
476 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
477 sizeof(struct nfgenmsg), 0);
479 skb_tx_error(entskb);
483 nfmsg = nlmsg_data(nlh);
484 nfmsg->nfgen_family = entry->state.pf;
485 nfmsg->version = NFNETLINK_V0;
486 nfmsg->res_id = htons(queue->queue_num);
488 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
489 pmsg = nla_data(nla);
490 pmsg->hw_protocol = entskb->protocol;
491 pmsg->hook = entry->state.hook;
492 *packet_id_ptr = &pmsg->packet_id;
494 indev = entry->state.in;
496 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
497 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
498 goto nla_put_failure;
500 if (entry->state.pf == PF_BRIDGE) {
501 /* Case 1: indev is physical input device, we need to
502 * look for bridge group (when called from
503 * netfilter_bridge) */
504 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
505 htonl(indev->ifindex)) ||
506 /* this is the bridge group "brX" */
507 /* rcu_read_lock()ed by __nf_queue */
508 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
509 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
510 goto nla_put_failure;
514 /* Case 2: indev is bridge group, we need to look for
515 * physical device (when called from ipv4) */
516 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
517 htonl(indev->ifindex)))
518 goto nla_put_failure;
520 physinif = nf_bridge_get_physinif(entskb);
522 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
524 goto nla_put_failure;
530 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
531 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
532 goto nla_put_failure;
534 if (entry->state.pf == PF_BRIDGE) {
535 /* Case 1: outdev is physical output device, we need to
536 * look for bridge group (when called from
537 * netfilter_bridge) */
538 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
539 htonl(outdev->ifindex)) ||
540 /* this is the bridge group "brX" */
541 /* rcu_read_lock()ed by __nf_queue */
542 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
543 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
544 goto nla_put_failure;
548 /* Case 2: outdev is bridge group, we need to look for
549 * physical output device (when called from ipv4) */
550 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
551 htonl(outdev->ifindex)))
552 goto nla_put_failure;
554 physoutif = nf_bridge_get_physoutif(entskb);
556 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
558 goto nla_put_failure;
564 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
565 goto nla_put_failure;
567 if (indev && entskb->dev &&
568 entskb->mac_header != entskb->network_header) {
569 struct nfqnl_msg_packet_hw phw;
572 memset(&phw, 0, sizeof(phw));
573 len = dev_parse_header(entskb, phw.hw_addr);
575 phw.hw_addrlen = htons(len);
576 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
577 goto nla_put_failure;
581 if (nfqnl_put_bridge(entry, skb) < 0)
582 goto nla_put_failure;
584 if (entskb->tstamp) {
585 struct nfqnl_msg_packet_timestamp ts;
586 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
588 ts.sec = cpu_to_be64(kts.tv_sec);
589 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
591 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
592 goto nla_put_failure;
595 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
596 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
597 goto nla_put_failure;
599 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
600 goto nla_put_failure;
602 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
603 goto nla_put_failure;
605 if (cap_len > data_len &&
606 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
607 goto nla_put_failure;
609 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
610 goto nla_put_failure;
615 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
616 goto nla_put_failure;
618 nla = skb_put(skb, sizeof(*nla));
619 nla->nla_type = NFQA_PAYLOAD;
620 nla->nla_len = nla_attr_size(data_len);
622 if (skb_zerocopy(skb, entskb, data_len, hlen))
623 goto nla_put_failure;
626 nlh->nlmsg_len = skb->len;
628 security_release_secctx(secdata, seclen);
632 skb_tx_error(entskb);
634 net_err_ratelimited("nf_queue: error creating packet message\n");
637 security_release_secctx(secdata, seclen);
641 static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry)
643 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
644 static const unsigned long flags = IPS_CONFIRMED | IPS_DYING;
645 const struct nf_conn *ct = (void *)skb_nfct(entry->skb);
647 if (ct && ((ct->status & flags) == IPS_DYING))
654 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
655 struct nf_queue_entry *entry)
657 struct sk_buff *nskb;
659 __be32 *packet_id_ptr;
662 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
667 spin_lock_bh(&queue->lock);
669 if (nf_ct_drop_unconfirmed(entry))
670 goto err_out_free_nskb;
672 if (queue->queue_total >= queue->queue_maxlen) {
673 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
677 queue->queue_dropped++;
678 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
681 goto err_out_free_nskb;
683 entry->id = ++queue->id_sequence;
684 *packet_id_ptr = htonl(entry->id);
686 /* nfnetlink_unicast will either free the nskb or add it to a socket */
687 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
689 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
693 queue->queue_user_dropped++;
698 __enqueue_entry(queue, entry);
700 spin_unlock_bh(&queue->lock);
706 spin_unlock_bh(&queue->lock);
708 nfqnl_reinject(entry, NF_ACCEPT);
713 static struct nf_queue_entry *
714 nf_queue_entry_dup(struct nf_queue_entry *e)
716 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
718 nf_queue_entry_get_refs(entry);
722 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
723 /* When called from bridge netfilter, skb->data must point to MAC header
724 * before calling skb_gso_segment(). Else, original MAC header is lost
725 * and segmented skbs will be sent to wrong destination.
727 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
730 __skb_push(skb, skb->network_header - skb->mac_header);
733 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
736 __skb_pull(skb, skb->network_header - skb->mac_header);
739 #define nf_bridge_adjust_skb_data(s) do {} while (0)
740 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
743 static void free_entry(struct nf_queue_entry *entry)
745 nf_queue_entry_release_refs(entry);
750 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
751 struct sk_buff *skb, struct nf_queue_entry *entry)
754 struct nf_queue_entry *entry_seg;
756 nf_bridge_adjust_segmented_data(skb);
758 if (skb->next == NULL) { /* last packet, no need to copy entry */
759 struct sk_buff *gso_skb = entry->skb;
761 ret = __nfqnl_enqueue_packet(net, queue, entry);
763 entry->skb = gso_skb;
769 entry_seg = nf_queue_entry_dup(entry);
771 entry_seg->skb = skb;
772 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
774 free_entry(entry_seg);
780 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
783 struct nfqnl_instance *queue;
784 struct sk_buff *skb, *segs;
786 struct net *net = entry->state.net;
787 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
789 /* rcu_read_lock()ed by nf_hook_thresh */
790 queue = instance_lookup(q, queuenum);
794 if (queue->copy_mode == NFQNL_COPY_NONE)
799 switch (entry->state.pf) {
801 skb->protocol = htons(ETH_P_IP);
804 skb->protocol = htons(ETH_P_IPV6);
808 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
809 return __nfqnl_enqueue_packet(net, queue, entry);
811 nf_bridge_adjust_skb_data(skb);
812 segs = skb_gso_segment(skb, 0);
813 /* Does not use PTR_ERR to limit the number of error codes that can be
814 * returned by nf_queue. For instance, callers rely on -ESRCH to
815 * mean 'ignore this hook'.
817 if (IS_ERR_OR_NULL(segs))
822 struct sk_buff *nskb = segs->next;
824 err = __nfqnl_enqueue_packet_gso(net, queue,
834 if (err) /* some segments are already queued */
840 nf_bridge_adjust_segmented_data(skb);
845 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
847 struct sk_buff *nskb;
850 if (pskb_trim(e->skb, data_len))
852 } else if (diff > 0) {
853 if (data_len > 0xFFFF)
855 if (diff > skb_tailroom(e->skb)) {
856 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
863 skb_put(e->skb, diff);
865 if (!skb_make_writable(e->skb, data_len))
867 skb_copy_to_linear_data(e->skb, data, data_len);
868 e->skb->ip_summed = CHECKSUM_NONE;
873 nfqnl_set_mode(struct nfqnl_instance *queue,
874 unsigned char mode, unsigned int range)
878 spin_lock_bh(&queue->lock);
880 case NFQNL_COPY_NONE:
881 case NFQNL_COPY_META:
882 queue->copy_mode = mode;
883 queue->copy_range = 0;
886 case NFQNL_COPY_PACKET:
887 queue->copy_mode = mode;
888 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
889 queue->copy_range = NFQNL_MAX_COPY_RANGE;
891 queue->copy_range = range;
898 spin_unlock_bh(&queue->lock);
904 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
907 if (entry->state.in->ifindex == ifindex)
909 if (entry->state.out)
910 if (entry->state.out->ifindex == ifindex)
912 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
913 if (entry->skb->nf_bridge) {
914 int physinif, physoutif;
916 physinif = nf_bridge_get_physinif(entry->skb);
917 physoutif = nf_bridge_get_physoutif(entry->skb);
919 if (physinif == ifindex || physoutif == ifindex)
926 /* drop all packets with either indev or outdev == ifindex from all queue
929 nfqnl_dev_drop(struct net *net, int ifindex)
932 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
936 for (i = 0; i < INSTANCE_BUCKETS; i++) {
937 struct nfqnl_instance *inst;
938 struct hlist_head *head = &q->instance_table[i];
940 hlist_for_each_entry_rcu(inst, head, hlist)
941 nfqnl_flush(inst, dev_cmp, ifindex);
948 nfqnl_rcv_dev_event(struct notifier_block *this,
949 unsigned long event, void *ptr)
951 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
953 /* Drop any packets associated with the downed device */
954 if (event == NETDEV_DOWN)
955 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
959 static struct notifier_block nfqnl_dev_notifier = {
960 .notifier_call = nfqnl_rcv_dev_event,
963 static void nfqnl_nf_hook_drop(struct net *net)
965 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
968 for (i = 0; i < INSTANCE_BUCKETS; i++) {
969 struct nfqnl_instance *inst;
970 struct hlist_head *head = &q->instance_table[i];
972 hlist_for_each_entry_rcu(inst, head, hlist)
973 nfqnl_flush(inst, NULL, 0);
978 nfqnl_rcv_nl_event(struct notifier_block *this,
979 unsigned long event, void *ptr)
981 struct netlink_notify *n = ptr;
982 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
984 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
987 /* destroy all instances for this portid */
988 spin_lock(&q->instances_lock);
989 for (i = 0; i < INSTANCE_BUCKETS; i++) {
990 struct hlist_node *t2;
991 struct nfqnl_instance *inst;
992 struct hlist_head *head = &q->instance_table[i];
994 hlist_for_each_entry_safe(inst, t2, head, hlist) {
995 if (n->portid == inst->peer_portid)
996 __instance_destroy(inst);
999 spin_unlock(&q->instances_lock);
1004 static struct notifier_block nfqnl_rtnl_notifier = {
1005 .notifier_call = nfqnl_rcv_nl_event,
1008 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
1009 [NFQA_VLAN_TCI] = { .type = NLA_U16},
1010 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
1013 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
1014 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1015 [NFQA_MARK] = { .type = NLA_U32 },
1016 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
1017 [NFQA_CT] = { .type = NLA_UNSPEC },
1018 [NFQA_EXP] = { .type = NLA_UNSPEC },
1019 [NFQA_VLAN] = { .type = NLA_NESTED },
1022 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1023 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1024 [NFQA_MARK] = { .type = NLA_U32 },
1027 static struct nfqnl_instance *
1028 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1030 struct nfqnl_instance *queue;
1032 queue = instance_lookup(q, queue_num);
1034 return ERR_PTR(-ENODEV);
1036 if (queue->peer_portid != nlportid)
1037 return ERR_PTR(-EPERM);
1042 static struct nfqnl_msg_verdict_hdr*
1043 verdicthdr_get(const struct nlattr * const nfqa[])
1045 struct nfqnl_msg_verdict_hdr *vhdr;
1046 unsigned int verdict;
1048 if (!nfqa[NFQA_VERDICT_HDR])
1051 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1052 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1053 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1058 static int nfq_id_after(unsigned int id, unsigned int max)
1060 return (int)(id - max) > 0;
1063 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1064 struct sk_buff *skb,
1065 const struct nlmsghdr *nlh,
1066 const struct nlattr * const nfqa[],
1067 struct netlink_ext_ack *extack)
1069 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1070 struct nf_queue_entry *entry, *tmp;
1071 unsigned int verdict, maxid;
1072 struct nfqnl_msg_verdict_hdr *vhdr;
1073 struct nfqnl_instance *queue;
1074 LIST_HEAD(batch_list);
1075 u16 queue_num = ntohs(nfmsg->res_id);
1076 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1078 queue = verdict_instance_lookup(q, queue_num,
1079 NETLINK_CB(skb).portid);
1081 return PTR_ERR(queue);
1083 vhdr = verdicthdr_get(nfqa);
1087 verdict = ntohl(vhdr->verdict);
1088 maxid = ntohl(vhdr->id);
1090 spin_lock_bh(&queue->lock);
1092 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1093 if (nfq_id_after(entry->id, maxid))
1095 __dequeue_entry(queue, entry);
1096 list_add_tail(&entry->list, &batch_list);
1099 spin_unlock_bh(&queue->lock);
1101 if (list_empty(&batch_list))
1104 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1105 if (nfqa[NFQA_MARK])
1106 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1108 nfqnl_reinject(entry, verdict);
1113 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1114 const struct nlmsghdr *nlh,
1115 const struct nlattr * const nfqa[],
1116 struct nf_queue_entry *entry,
1117 enum ip_conntrack_info *ctinfo)
1121 ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1125 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1129 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1130 NETLINK_CB(entry->skb).portid,
1135 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1136 const struct nlattr * const nfqa[])
1138 if (nfqa[NFQA_VLAN]) {
1139 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1142 err = nla_parse_nested(tb, NFQA_VLAN_MAX, nfqa[NFQA_VLAN],
1143 nfqa_vlan_policy, NULL);
1147 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1150 entry->skb->vlan_tci = ntohs(nla_get_be16(tb[NFQA_VLAN_TCI]));
1151 entry->skb->vlan_proto = nla_get_be16(tb[NFQA_VLAN_PROTO]);
1154 if (nfqa[NFQA_L2HDR]) {
1155 int mac_header_len = entry->skb->network_header -
1156 entry->skb->mac_header;
1158 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1160 else if (mac_header_len > 0)
1161 memcpy(skb_mac_header(entry->skb),
1162 nla_data(nfqa[NFQA_L2HDR]),
1169 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1170 struct sk_buff *skb,
1171 const struct nlmsghdr *nlh,
1172 const struct nlattr * const nfqa[],
1173 struct netlink_ext_ack *extack)
1175 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1176 u_int16_t queue_num = ntohs(nfmsg->res_id);
1177 struct nfqnl_msg_verdict_hdr *vhdr;
1178 struct nfqnl_instance *queue;
1179 unsigned int verdict;
1180 struct nf_queue_entry *entry;
1181 enum ip_conntrack_info uninitialized_var(ctinfo);
1182 struct nfnl_ct_hook *nfnl_ct;
1183 struct nf_conn *ct = NULL;
1184 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1187 queue = verdict_instance_lookup(q, queue_num,
1188 NETLINK_CB(skb).portid);
1190 return PTR_ERR(queue);
1192 vhdr = verdicthdr_get(nfqa);
1196 verdict = ntohl(vhdr->verdict);
1198 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1202 /* rcu lock already held from nfnl->call_rcu. */
1203 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1205 if (nfqa[NFQA_CT]) {
1206 if (nfnl_ct != NULL)
1207 ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1210 if (entry->state.pf == PF_BRIDGE) {
1211 err = nfqa_parse_bridge(entry, nfqa);
1216 if (nfqa[NFQA_PAYLOAD]) {
1217 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1218 int diff = payload_len - entry->skb->len;
1220 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1221 payload_len, entry, diff) < 0)
1225 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1228 if (nfqa[NFQA_MARK])
1229 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1231 nfqnl_reinject(entry, verdict);
1235 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1236 struct sk_buff *skb, const struct nlmsghdr *nlh,
1237 const struct nlattr * const nfqa[],
1238 struct netlink_ext_ack *extack)
1243 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1244 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1245 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1248 static const struct nf_queue_handler nfqh = {
1249 .outfn = nfqnl_enqueue_packet,
1250 .nf_hook_drop = nfqnl_nf_hook_drop,
1253 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1254 struct sk_buff *skb, const struct nlmsghdr *nlh,
1255 const struct nlattr * const nfqa[],
1256 struct netlink_ext_ack *extack)
1258 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1259 u_int16_t queue_num = ntohs(nfmsg->res_id);
1260 struct nfqnl_instance *queue;
1261 struct nfqnl_msg_config_cmd *cmd = NULL;
1262 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1263 __u32 flags = 0, mask = 0;
1266 if (nfqa[NFQA_CFG_CMD]) {
1267 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1269 /* Obsolete commands without queue context */
1270 switch (cmd->command) {
1271 case NFQNL_CFG_CMD_PF_BIND: return 0;
1272 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1276 /* Check if we support these flags in first place, dependencies should
1277 * be there too not to break atomicity.
1279 if (nfqa[NFQA_CFG_FLAGS]) {
1280 if (!nfqa[NFQA_CFG_MASK]) {
1281 /* A mask is needed to specify which flags are being
1287 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1288 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1290 if (flags >= NFQA_CFG_F_MAX)
1293 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1294 if (flags & mask & NFQA_CFG_F_SECCTX)
1297 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1298 !rcu_access_pointer(nfnl_ct_hook)) {
1299 #ifdef CONFIG_MODULES
1300 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1301 request_module("ip_conntrack_netlink");
1302 nfnl_lock(NFNL_SUBSYS_QUEUE);
1303 if (rcu_access_pointer(nfnl_ct_hook))
1311 queue = instance_lookup(q, queue_num);
1312 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1314 goto err_out_unlock;
1318 switch (cmd->command) {
1319 case NFQNL_CFG_CMD_BIND:
1322 goto err_out_unlock;
1324 queue = instance_create(q, queue_num,
1325 NETLINK_CB(skb).portid);
1326 if (IS_ERR(queue)) {
1327 ret = PTR_ERR(queue);
1328 goto err_out_unlock;
1331 case NFQNL_CFG_CMD_UNBIND:
1334 goto err_out_unlock;
1336 instance_destroy(q, queue);
1337 goto err_out_unlock;
1338 case NFQNL_CFG_CMD_PF_BIND:
1339 case NFQNL_CFG_CMD_PF_UNBIND:
1343 goto err_out_unlock;
1349 goto err_out_unlock;
1352 if (nfqa[NFQA_CFG_PARAMS]) {
1353 struct nfqnl_msg_config_params *params =
1354 nla_data(nfqa[NFQA_CFG_PARAMS]);
1356 nfqnl_set_mode(queue, params->copy_mode,
1357 ntohl(params->copy_range));
1360 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1361 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1363 spin_lock_bh(&queue->lock);
1364 queue->queue_maxlen = ntohl(*queue_maxlen);
1365 spin_unlock_bh(&queue->lock);
1368 if (nfqa[NFQA_CFG_FLAGS]) {
1369 spin_lock_bh(&queue->lock);
1370 queue->flags &= ~mask;
1371 queue->flags |= flags & mask;
1372 spin_unlock_bh(&queue->lock);
1380 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1381 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1382 .attr_count = NFQA_MAX, },
1383 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1384 .attr_count = NFQA_MAX,
1385 .policy = nfqa_verdict_policy },
1386 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1387 .attr_count = NFQA_CFG_MAX,
1388 .policy = nfqa_cfg_policy },
1389 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1390 .attr_count = NFQA_MAX,
1391 .policy = nfqa_verdict_batch_policy },
1394 static const struct nfnetlink_subsystem nfqnl_subsys = {
1396 .subsys_id = NFNL_SUBSYS_QUEUE,
1397 .cb_count = NFQNL_MSG_MAX,
1401 #ifdef CONFIG_PROC_FS
1403 struct seq_net_private p;
1404 unsigned int bucket;
1407 static struct hlist_node *get_first(struct seq_file *seq)
1409 struct iter_state *st = seq->private;
1411 struct nfnl_queue_net *q;
1416 net = seq_file_net(seq);
1417 q = nfnl_queue_pernet(net);
1418 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1419 if (!hlist_empty(&q->instance_table[st->bucket]))
1420 return q->instance_table[st->bucket].first;
1425 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1427 struct iter_state *st = seq->private;
1428 struct net *net = seq_file_net(seq);
1432 struct nfnl_queue_net *q;
1434 if (++st->bucket >= INSTANCE_BUCKETS)
1437 q = nfnl_queue_pernet(net);
1438 h = q->instance_table[st->bucket].first;
1443 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1445 struct hlist_node *head;
1446 head = get_first(seq);
1449 while (pos && (head = get_next(seq, head)))
1451 return pos ? NULL : head;
1454 static void *seq_start(struct seq_file *s, loff_t *pos)
1455 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1457 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1458 return get_idx(s, *pos);
1461 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1464 return get_next(s, v);
1467 static void seq_stop(struct seq_file *s, void *v)
1468 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1470 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1473 static int seq_show(struct seq_file *s, void *v)
1475 const struct nfqnl_instance *inst = v;
1477 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1479 inst->peer_portid, inst->queue_total,
1480 inst->copy_mode, inst->copy_range,
1481 inst->queue_dropped, inst->queue_user_dropped,
1482 inst->id_sequence, 1);
1486 static const struct seq_operations nfqnl_seq_ops = {
1492 #endif /* PROC_FS */
1494 static int __net_init nfnl_queue_net_init(struct net *net)
1497 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1499 for (i = 0; i < INSTANCE_BUCKETS; i++)
1500 INIT_HLIST_HEAD(&q->instance_table[i]);
1502 spin_lock_init(&q->instances_lock);
1504 #ifdef CONFIG_PROC_FS
1505 if (!proc_create_net("nfnetlink_queue", 0440, net->nf.proc_netfilter,
1506 &nfqnl_seq_ops, sizeof(struct iter_state)))
1509 nf_register_queue_handler(net, &nfqh);
1513 static void __net_exit nfnl_queue_net_exit(struct net *net)
1515 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1518 nf_unregister_queue_handler(net);
1519 #ifdef CONFIG_PROC_FS
1520 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1522 for (i = 0; i < INSTANCE_BUCKETS; i++)
1523 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1526 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1531 static struct pernet_operations nfnl_queue_net_ops = {
1532 .init = nfnl_queue_net_init,
1533 .exit = nfnl_queue_net_exit,
1534 .exit_batch = nfnl_queue_net_exit_batch,
1535 .id = &nfnl_queue_net_id,
1536 .size = sizeof(struct nfnl_queue_net),
1539 static int __init nfnetlink_queue_init(void)
1543 status = register_pernet_subsys(&nfnl_queue_net_ops);
1545 pr_err("failed to register pernet ops\n");
1549 netlink_register_notifier(&nfqnl_rtnl_notifier);
1550 status = nfnetlink_subsys_register(&nfqnl_subsys);
1552 pr_err("failed to create netlink socket\n");
1553 goto cleanup_netlink_notifier;
1556 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1558 pr_err("failed to register netdevice notifier\n");
1559 goto cleanup_netlink_subsys;
1564 cleanup_netlink_subsys:
1565 nfnetlink_subsys_unregister(&nfqnl_subsys);
1566 cleanup_netlink_notifier:
1567 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1568 unregister_pernet_subsys(&nfnl_queue_net_ops);
1573 static void __exit nfnetlink_queue_fini(void)
1575 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1576 nfnetlink_subsys_unregister(&nfqnl_subsys);
1577 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1578 unregister_pernet_subsys(&nfnl_queue_net_ops);
1580 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1583 MODULE_DESCRIPTION("netfilter packet queue handler");
1584 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1585 MODULE_LICENSE("GPL");
1586 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1588 module_init(nfnetlink_queue_init);
1589 module_exit(nfnetlink_queue_fini);