1 // SPDX-License-Identifier: GPL-2.0-only
3 * This is a module which is used for queueing packets and communicating with
4 * userspace via nfnetlink.
6 * (C) 2005 by Harald Welte <laforge@netfilter.org>
7 * (C) 2007 by Patrick McHardy <kaber@trash.net>
9 * Based on the old ipv4-only ip_queue.c:
10 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
11 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
14 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/init.h>
19 #include <linux/spinlock.h>
20 #include <linux/slab.h>
21 #include <linux/notifier.h>
22 #include <linux/netdevice.h>
23 #include <linux/netfilter.h>
24 #include <linux/proc_fs.h>
25 #include <linux/netfilter_ipv4.h>
26 #include <linux/netfilter_ipv6.h>
27 #include <linux/netfilter_bridge.h>
28 #include <linux/netfilter/nfnetlink.h>
29 #include <linux/netfilter/nfnetlink_queue.h>
30 #include <linux/netfilter/nf_conntrack_common.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>
37 #include <linux/atomic.h>
39 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
40 #include "../bridge/br_private.h"
43 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
44 #include <net/netfilter/nf_conntrack.h>
47 #define NFQNL_QMAX_DEFAULT 1024
49 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
50 * includes the header length. Thus, the maximum packet length that we
51 * support is 65531 bytes. We send truncated packets if the specified length
52 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
53 * attribute to detect truncation.
55 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
57 struct nfqnl_instance {
58 struct hlist_node hlist; /* global list of queues */
62 unsigned int queue_maxlen;
63 unsigned int copy_range;
64 unsigned int queue_dropped;
65 unsigned int queue_user_dropped;
68 u_int16_t queue_num; /* number of this queue */
70 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
72 * Following fields are dirtied for each queued packet,
73 * keep them in same cache line if possible.
75 spinlock_t lock ____cacheline_aligned_in_smp;
76 unsigned int queue_total;
77 unsigned int id_sequence; /* 'sequence' of pkt ids */
78 struct list_head queue_list; /* packets in queue */
81 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
83 static unsigned int nfnl_queue_net_id __read_mostly;
85 #define INSTANCE_BUCKETS 16
86 struct nfnl_queue_net {
87 spinlock_t instances_lock;
88 struct hlist_head instance_table[INSTANCE_BUCKETS];
91 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
93 return net_generic(net, nfnl_queue_net_id);
96 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
98 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
101 static struct nfqnl_instance *
102 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
104 struct hlist_head *head;
105 struct nfqnl_instance *inst;
107 head = &q->instance_table[instance_hashfn(queue_num)];
108 hlist_for_each_entry_rcu(inst, head, hlist) {
109 if (inst->queue_num == queue_num)
115 static struct nfqnl_instance *
116 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
118 struct nfqnl_instance *inst;
122 spin_lock(&q->instances_lock);
123 if (instance_lookup(q, queue_num)) {
128 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
134 inst->queue_num = queue_num;
135 inst->peer_portid = portid;
136 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
137 inst->copy_range = NFQNL_MAX_COPY_RANGE;
138 inst->copy_mode = NFQNL_COPY_NONE;
139 spin_lock_init(&inst->lock);
140 INIT_LIST_HEAD(&inst->queue_list);
142 if (!try_module_get(THIS_MODULE)) {
147 h = instance_hashfn(queue_num);
148 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
150 spin_unlock(&q->instances_lock);
157 spin_unlock(&q->instances_lock);
161 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
165 instance_destroy_rcu(struct rcu_head *head)
167 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
170 nfqnl_flush(inst, NULL, 0);
172 module_put(THIS_MODULE);
176 __instance_destroy(struct nfqnl_instance *inst)
178 hlist_del_rcu(&inst->hlist);
179 call_rcu(&inst->rcu, instance_destroy_rcu);
183 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
185 spin_lock(&q->instances_lock);
186 __instance_destroy(inst);
187 spin_unlock(&q->instances_lock);
191 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
193 list_add_tail(&entry->list, &queue->queue_list);
194 queue->queue_total++;
198 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
200 list_del(&entry->list);
201 queue->queue_total--;
204 static struct nf_queue_entry *
205 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
207 struct nf_queue_entry *entry = NULL, *i;
209 spin_lock_bh(&queue->lock);
211 list_for_each_entry(i, &queue->queue_list, list) {
219 __dequeue_entry(queue, entry);
221 spin_unlock_bh(&queue->lock);
226 static void nfqnl_reinject(struct nf_queue_entry *entry, unsigned int verdict)
228 struct nf_ct_hook *ct_hook;
231 if (verdict == NF_ACCEPT ||
232 verdict == NF_REPEAT ||
233 verdict == NF_STOP) {
235 ct_hook = rcu_dereference(nf_ct_hook);
237 err = ct_hook->update(entry->state.net, entry->skb);
243 nf_reinject(entry, verdict);
247 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
249 struct nf_queue_entry *entry, *next;
251 spin_lock_bh(&queue->lock);
252 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
253 if (!cmpfn || cmpfn(entry, data)) {
254 list_del(&entry->list);
255 queue->queue_total--;
256 nfqnl_reinject(entry, NF_DROP);
259 spin_unlock_bh(&queue->lock);
263 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
268 if (packet->ip_summed == CHECKSUM_PARTIAL)
269 flags = NFQA_SKB_CSUMNOTREADY;
270 else if (csum_verify)
271 flags = NFQA_SKB_CSUM_NOTVERIFIED;
273 if (skb_is_gso(packet))
274 flags |= NFQA_SKB_GSO;
276 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
279 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
281 const struct cred *cred;
283 if (!sk_fullsock(sk))
286 read_lock_bh(&sk->sk_callback_lock);
287 if (sk->sk_socket && sk->sk_socket->file) {
288 cred = sk->sk_socket->file->f_cred;
289 if (nla_put_be32(skb, NFQA_UID,
290 htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
291 goto nla_put_failure;
292 if (nla_put_be32(skb, NFQA_GID,
293 htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
294 goto nla_put_failure;
296 read_unlock_bh(&sk->sk_callback_lock);
300 read_unlock_bh(&sk->sk_callback_lock);
304 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
307 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
308 if (!skb || !sk_fullsock(skb->sk))
311 read_lock_bh(&skb->sk->sk_callback_lock);
314 security_secid_to_secctx(skb->secmark, secdata, &seclen);
316 read_unlock_bh(&skb->sk->sk_callback_lock);
321 static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
323 struct sk_buff *entskb = entry->skb;
326 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
329 if (skb_vlan_tag_present(entskb))
330 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
331 nla_total_size(sizeof(__be16)));
333 if (entskb->network_header > entskb->mac_header)
334 nlalen += nla_total_size((entskb->network_header -
335 entskb->mac_header));
340 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
342 struct sk_buff *entskb = entry->skb;
344 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
347 if (skb_vlan_tag_present(entskb)) {
350 nest = nla_nest_start(skb, NFQA_VLAN);
352 goto nla_put_failure;
354 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
355 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
356 goto nla_put_failure;
358 nla_nest_end(skb, nest);
361 if (entskb->mac_header < entskb->network_header) {
362 int len = (int)(entskb->network_header - entskb->mac_header);
364 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
365 goto nla_put_failure;
374 static struct sk_buff *
375 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
376 struct nf_queue_entry *entry,
377 __be32 **packet_id_ptr)
380 size_t data_len = 0, cap_len = 0;
381 unsigned int hlen = 0;
384 struct nfqnl_msg_packet_hdr *pmsg;
385 struct nlmsghdr *nlh;
386 struct nfgenmsg *nfmsg;
387 struct sk_buff *entskb = entry->skb;
388 struct net_device *indev;
389 struct net_device *outdev;
390 struct nf_conn *ct = NULL;
391 enum ip_conntrack_info ctinfo;
392 struct nfnl_ct_hook *nfnl_ct;
394 char *secdata = NULL;
397 size = nlmsg_total_size(sizeof(struct nfgenmsg))
398 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
399 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
400 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
401 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
402 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
403 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
405 + nla_total_size(sizeof(u_int32_t)) /* mark */
406 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
407 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
408 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
411 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
413 size += nfqnl_get_bridge_size(entry);
415 if (entry->state.hook <= NF_INET_FORWARD ||
416 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
417 csum_verify = !skb_csum_unnecessary(entskb);
421 outdev = entry->state.out;
423 switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) {
424 case NFQNL_COPY_META:
425 case NFQNL_COPY_NONE:
428 case NFQNL_COPY_PACKET:
429 if (!(queue->flags & NFQA_CFG_F_GSO) &&
430 entskb->ip_summed == CHECKSUM_PARTIAL &&
431 skb_checksum_help(entskb))
434 data_len = READ_ONCE(queue->copy_range);
435 if (data_len > entskb->len)
436 data_len = entskb->len;
438 hlen = skb_zerocopy_headlen(entskb);
439 hlen = min_t(unsigned int, hlen, data_len);
440 size += sizeof(struct nlattr) + hlen;
441 cap_len = entskb->len;
445 nfnl_ct = rcu_dereference(nfnl_ct_hook);
447 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
448 if (nfnl_ct != NULL) {
449 ct = nfnl_ct->get_ct(entskb, &ctinfo);
451 size += nfnl_ct->build_size(ct);
455 if (queue->flags & NFQA_CFG_F_UID_GID) {
456 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
457 + nla_total_size(sizeof(u_int32_t))); /* gid */
460 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
461 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
463 size += nla_total_size(seclen);
466 skb = alloc_skb(size, GFP_ATOMIC);
468 skb_tx_error(entskb);
472 nlh = nlmsg_put(skb, 0, 0,
473 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
474 sizeof(struct nfgenmsg), 0);
476 skb_tx_error(entskb);
480 nfmsg = nlmsg_data(nlh);
481 nfmsg->nfgen_family = entry->state.pf;
482 nfmsg->version = NFNETLINK_V0;
483 nfmsg->res_id = htons(queue->queue_num);
485 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
486 pmsg = nla_data(nla);
487 pmsg->hw_protocol = entskb->protocol;
488 pmsg->hook = entry->state.hook;
489 *packet_id_ptr = &pmsg->packet_id;
491 indev = entry->state.in;
493 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
494 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
495 goto nla_put_failure;
497 if (entry->state.pf == PF_BRIDGE) {
498 /* Case 1: indev is physical input device, we need to
499 * look for bridge group (when called from
500 * netfilter_bridge) */
501 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
502 htonl(indev->ifindex)) ||
503 /* this is the bridge group "brX" */
504 /* rcu_read_lock()ed by __nf_queue */
505 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
506 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
507 goto nla_put_failure;
511 /* Case 2: indev is bridge group, we need to look for
512 * physical device (when called from ipv4) */
513 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
514 htonl(indev->ifindex)))
515 goto nla_put_failure;
517 physinif = nf_bridge_get_physinif(entskb);
519 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
521 goto nla_put_failure;
527 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
528 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
529 goto nla_put_failure;
531 if (entry->state.pf == PF_BRIDGE) {
532 /* Case 1: outdev is physical output device, we need to
533 * look for bridge group (when called from
534 * netfilter_bridge) */
535 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
536 htonl(outdev->ifindex)) ||
537 /* this is the bridge group "brX" */
538 /* rcu_read_lock()ed by __nf_queue */
539 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
540 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
541 goto nla_put_failure;
545 /* Case 2: outdev is bridge group, we need to look for
546 * physical output device (when called from ipv4) */
547 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
548 htonl(outdev->ifindex)))
549 goto nla_put_failure;
551 physoutif = nf_bridge_get_physoutif(entskb);
553 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
555 goto nla_put_failure;
561 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
562 goto nla_put_failure;
564 if (indev && entskb->dev &&
565 entskb->mac_header != entskb->network_header) {
566 struct nfqnl_msg_packet_hw phw;
569 memset(&phw, 0, sizeof(phw));
570 len = dev_parse_header(entskb, phw.hw_addr);
572 phw.hw_addrlen = htons(len);
573 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
574 goto nla_put_failure;
578 if (nfqnl_put_bridge(entry, skb) < 0)
579 goto nla_put_failure;
581 if (entry->state.hook <= NF_INET_FORWARD && entskb->tstamp) {
582 struct nfqnl_msg_packet_timestamp ts;
583 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
585 ts.sec = cpu_to_be64(kts.tv_sec);
586 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
588 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
589 goto nla_put_failure;
592 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
593 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
594 goto nla_put_failure;
596 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
597 goto nla_put_failure;
599 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
600 goto nla_put_failure;
602 if (cap_len > data_len &&
603 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
604 goto nla_put_failure;
606 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
607 goto nla_put_failure;
612 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
613 goto nla_put_failure;
615 nla = skb_put(skb, sizeof(*nla));
616 nla->nla_type = NFQA_PAYLOAD;
617 nla->nla_len = nla_attr_size(data_len);
619 if (skb_zerocopy(skb, entskb, data_len, hlen))
620 goto nla_put_failure;
623 nlh->nlmsg_len = skb->len;
625 security_release_secctx(secdata, seclen);
629 skb_tx_error(entskb);
631 net_err_ratelimited("nf_queue: error creating packet message\n");
634 security_release_secctx(secdata, seclen);
638 static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry)
640 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
641 static const unsigned long flags = IPS_CONFIRMED | IPS_DYING;
642 const struct nf_conn *ct = (void *)skb_nfct(entry->skb);
644 if (ct && ((ct->status & flags) == IPS_DYING))
651 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
652 struct nf_queue_entry *entry)
654 struct sk_buff *nskb;
656 __be32 *packet_id_ptr;
659 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
664 spin_lock_bh(&queue->lock);
666 if (nf_ct_drop_unconfirmed(entry))
667 goto err_out_free_nskb;
669 if (queue->queue_total >= queue->queue_maxlen) {
670 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
674 queue->queue_dropped++;
675 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
678 goto err_out_free_nskb;
680 entry->id = ++queue->id_sequence;
681 *packet_id_ptr = htonl(entry->id);
683 /* nfnetlink_unicast will either free the nskb or add it to a socket */
684 err = nfnetlink_unicast(nskb, net, queue->peer_portid);
686 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
690 queue->queue_user_dropped++;
695 __enqueue_entry(queue, entry);
697 spin_unlock_bh(&queue->lock);
703 spin_unlock_bh(&queue->lock);
705 nfqnl_reinject(entry, NF_ACCEPT);
710 static struct nf_queue_entry *
711 nf_queue_entry_dup(struct nf_queue_entry *e)
713 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
715 nf_queue_entry_get_refs(entry);
719 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
720 /* When called from bridge netfilter, skb->data must point to MAC header
721 * before calling skb_gso_segment(). Else, original MAC header is lost
722 * and segmented skbs will be sent to wrong destination.
724 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
726 if (nf_bridge_info_get(skb))
727 __skb_push(skb, skb->network_header - skb->mac_header);
730 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
732 if (nf_bridge_info_get(skb))
733 __skb_pull(skb, skb->network_header - skb->mac_header);
736 #define nf_bridge_adjust_skb_data(s) do {} while (0)
737 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
741 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
742 struct sk_buff *skb, struct nf_queue_entry *entry)
745 struct nf_queue_entry *entry_seg;
747 nf_bridge_adjust_segmented_data(skb);
749 if (skb->next == NULL) { /* last packet, no need to copy entry */
750 struct sk_buff *gso_skb = entry->skb;
752 ret = __nfqnl_enqueue_packet(net, queue, entry);
754 entry->skb = gso_skb;
758 skb_mark_not_on_list(skb);
760 entry_seg = nf_queue_entry_dup(entry);
762 entry_seg->skb = skb;
763 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
765 nf_queue_entry_free(entry_seg);
771 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
774 struct nfqnl_instance *queue;
775 struct sk_buff *skb, *segs, *nskb;
777 struct net *net = entry->state.net;
778 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
780 /* rcu_read_lock()ed by nf_hook_thresh */
781 queue = instance_lookup(q, queuenum);
785 if (queue->copy_mode == NFQNL_COPY_NONE)
790 switch (entry->state.pf) {
792 skb->protocol = htons(ETH_P_IP);
795 skb->protocol = htons(ETH_P_IPV6);
799 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
800 return __nfqnl_enqueue_packet(net, queue, entry);
802 nf_bridge_adjust_skb_data(skb);
803 segs = skb_gso_segment(skb, 0);
804 /* Does not use PTR_ERR to limit the number of error codes that can be
805 * returned by nf_queue. For instance, callers rely on -ESRCH to
806 * mean 'ignore this hook'.
808 if (IS_ERR_OR_NULL(segs))
812 skb_list_walk_safe(segs, segs, nskb) {
814 err = __nfqnl_enqueue_packet_gso(net, queue,
823 if (err) /* some segments are already queued */
824 nf_queue_entry_free(entry);
829 nf_bridge_adjust_segmented_data(skb);
834 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
836 struct sk_buff *nskb;
839 if (pskb_trim(e->skb, data_len))
841 } else if (diff > 0) {
842 if (data_len > 0xFFFF)
844 if (diff > skb_tailroom(e->skb)) {
845 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
852 skb_put(e->skb, diff);
854 if (skb_ensure_writable(e->skb, data_len))
856 skb_copy_to_linear_data(e->skb, data, data_len);
857 e->skb->ip_summed = CHECKSUM_NONE;
862 nfqnl_set_mode(struct nfqnl_instance *queue,
863 unsigned char mode, unsigned int range)
867 spin_lock_bh(&queue->lock);
869 case NFQNL_COPY_NONE:
870 case NFQNL_COPY_META:
871 queue->copy_mode = mode;
872 queue->copy_range = 0;
875 case NFQNL_COPY_PACKET:
876 queue->copy_mode = mode;
877 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
878 queue->copy_range = NFQNL_MAX_COPY_RANGE;
880 queue->copy_range = range;
887 spin_unlock_bh(&queue->lock);
893 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
895 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
896 int physinif, physoutif;
898 physinif = nf_bridge_get_physinif(entry->skb);
899 physoutif = nf_bridge_get_physoutif(entry->skb);
901 if (physinif == ifindex || physoutif == ifindex)
905 if (entry->state.in->ifindex == ifindex)
907 if (entry->state.out)
908 if (entry->state.out->ifindex == ifindex)
914 /* drop all packets with either indev or outdev == ifindex from all queue
917 nfqnl_dev_drop(struct net *net, int ifindex)
920 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
924 for (i = 0; i < INSTANCE_BUCKETS; i++) {
925 struct nfqnl_instance *inst;
926 struct hlist_head *head = &q->instance_table[i];
928 hlist_for_each_entry_rcu(inst, head, hlist)
929 nfqnl_flush(inst, dev_cmp, ifindex);
936 nfqnl_rcv_dev_event(struct notifier_block *this,
937 unsigned long event, void *ptr)
939 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
941 /* Drop any packets associated with the downed device */
942 if (event == NETDEV_DOWN)
943 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
947 static struct notifier_block nfqnl_dev_notifier = {
948 .notifier_call = nfqnl_rcv_dev_event,
951 static void nfqnl_nf_hook_drop(struct net *net)
953 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
956 for (i = 0; i < INSTANCE_BUCKETS; i++) {
957 struct nfqnl_instance *inst;
958 struct hlist_head *head = &q->instance_table[i];
960 hlist_for_each_entry_rcu(inst, head, hlist)
961 nfqnl_flush(inst, NULL, 0);
966 nfqnl_rcv_nl_event(struct notifier_block *this,
967 unsigned long event, void *ptr)
969 struct netlink_notify *n = ptr;
970 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
972 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
975 /* destroy all instances for this portid */
976 spin_lock(&q->instances_lock);
977 for (i = 0; i < INSTANCE_BUCKETS; i++) {
978 struct hlist_node *t2;
979 struct nfqnl_instance *inst;
980 struct hlist_head *head = &q->instance_table[i];
982 hlist_for_each_entry_safe(inst, t2, head, hlist) {
983 if (n->portid == inst->peer_portid)
984 __instance_destroy(inst);
987 spin_unlock(&q->instances_lock);
992 static struct notifier_block nfqnl_rtnl_notifier = {
993 .notifier_call = nfqnl_rcv_nl_event,
996 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
997 [NFQA_VLAN_TCI] = { .type = NLA_U16},
998 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
1001 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
1002 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1003 [NFQA_MARK] = { .type = NLA_U32 },
1004 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
1005 [NFQA_CT] = { .type = NLA_UNSPEC },
1006 [NFQA_EXP] = { .type = NLA_UNSPEC },
1007 [NFQA_VLAN] = { .type = NLA_NESTED },
1010 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1011 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1012 [NFQA_MARK] = { .type = NLA_U32 },
1015 static struct nfqnl_instance *
1016 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1018 struct nfqnl_instance *queue;
1020 queue = instance_lookup(q, queue_num);
1022 return ERR_PTR(-ENODEV);
1024 if (queue->peer_portid != nlportid)
1025 return ERR_PTR(-EPERM);
1030 static struct nfqnl_msg_verdict_hdr*
1031 verdicthdr_get(const struct nlattr * const nfqa[])
1033 struct nfqnl_msg_verdict_hdr *vhdr;
1034 unsigned int verdict;
1036 if (!nfqa[NFQA_VERDICT_HDR])
1039 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1040 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1041 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1046 static int nfq_id_after(unsigned int id, unsigned int max)
1048 return (int)(id - max) > 0;
1051 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1052 struct sk_buff *skb,
1053 const struct nlmsghdr *nlh,
1054 const struct nlattr * const nfqa[],
1055 struct netlink_ext_ack *extack)
1057 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1058 struct nf_queue_entry *entry, *tmp;
1059 unsigned int verdict, maxid;
1060 struct nfqnl_msg_verdict_hdr *vhdr;
1061 struct nfqnl_instance *queue;
1062 LIST_HEAD(batch_list);
1063 u16 queue_num = ntohs(nfmsg->res_id);
1064 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1066 queue = verdict_instance_lookup(q, queue_num,
1067 NETLINK_CB(skb).portid);
1069 return PTR_ERR(queue);
1071 vhdr = verdicthdr_get(nfqa);
1075 verdict = ntohl(vhdr->verdict);
1076 maxid = ntohl(vhdr->id);
1078 spin_lock_bh(&queue->lock);
1080 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1081 if (nfq_id_after(entry->id, maxid))
1083 __dequeue_entry(queue, entry);
1084 list_add_tail(&entry->list, &batch_list);
1087 spin_unlock_bh(&queue->lock);
1089 if (list_empty(&batch_list))
1092 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1093 if (nfqa[NFQA_MARK])
1094 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1096 nfqnl_reinject(entry, verdict);
1101 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1102 const struct nlmsghdr *nlh,
1103 const struct nlattr * const nfqa[],
1104 struct nf_queue_entry *entry,
1105 enum ip_conntrack_info *ctinfo)
1109 ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1113 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1117 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1118 NETLINK_CB(entry->skb).portid,
1123 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1124 const struct nlattr * const nfqa[])
1126 if (nfqa[NFQA_VLAN]) {
1127 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1130 err = nla_parse_nested_deprecated(tb, NFQA_VLAN_MAX,
1132 nfqa_vlan_policy, NULL);
1136 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1139 __vlan_hwaccel_put_tag(entry->skb,
1140 nla_get_be16(tb[NFQA_VLAN_PROTO]),
1141 ntohs(nla_get_be16(tb[NFQA_VLAN_TCI])));
1144 if (nfqa[NFQA_L2HDR]) {
1145 int mac_header_len = entry->skb->network_header -
1146 entry->skb->mac_header;
1148 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1150 else if (mac_header_len > 0)
1151 memcpy(skb_mac_header(entry->skb),
1152 nla_data(nfqa[NFQA_L2HDR]),
1159 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1160 struct sk_buff *skb,
1161 const struct nlmsghdr *nlh,
1162 const struct nlattr * const nfqa[],
1163 struct netlink_ext_ack *extack)
1165 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1166 u_int16_t queue_num = ntohs(nfmsg->res_id);
1167 struct nfqnl_msg_verdict_hdr *vhdr;
1168 struct nfqnl_instance *queue;
1169 unsigned int verdict;
1170 struct nf_queue_entry *entry;
1171 enum ip_conntrack_info ctinfo;
1172 struct nfnl_ct_hook *nfnl_ct;
1173 struct nf_conn *ct = NULL;
1174 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1177 queue = verdict_instance_lookup(q, queue_num,
1178 NETLINK_CB(skb).portid);
1180 return PTR_ERR(queue);
1182 vhdr = verdicthdr_get(nfqa);
1186 verdict = ntohl(vhdr->verdict);
1188 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1192 /* rcu lock already held from nfnl->call_rcu. */
1193 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1195 if (nfqa[NFQA_CT]) {
1196 if (nfnl_ct != NULL)
1197 ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1200 if (entry->state.pf == PF_BRIDGE) {
1201 err = nfqa_parse_bridge(entry, nfqa);
1206 if (nfqa[NFQA_PAYLOAD]) {
1207 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1208 int diff = payload_len - entry->skb->len;
1210 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1211 payload_len, entry, diff) < 0)
1215 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1218 if (nfqa[NFQA_MARK])
1219 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1221 nfqnl_reinject(entry, verdict);
1225 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1226 struct sk_buff *skb, const struct nlmsghdr *nlh,
1227 const struct nlattr * const nfqa[],
1228 struct netlink_ext_ack *extack)
1233 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1234 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1235 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1236 [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 },
1237 [NFQA_CFG_MASK] = { .type = NLA_U32 },
1238 [NFQA_CFG_FLAGS] = { .type = NLA_U32 },
1241 static const struct nf_queue_handler nfqh = {
1242 .outfn = nfqnl_enqueue_packet,
1243 .nf_hook_drop = nfqnl_nf_hook_drop,
1246 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1247 struct sk_buff *skb, const struct nlmsghdr *nlh,
1248 const struct nlattr * const nfqa[],
1249 struct netlink_ext_ack *extack)
1251 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1252 u_int16_t queue_num = ntohs(nfmsg->res_id);
1253 struct nfqnl_instance *queue;
1254 struct nfqnl_msg_config_cmd *cmd = NULL;
1255 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1256 __u32 flags = 0, mask = 0;
1259 if (nfqa[NFQA_CFG_CMD]) {
1260 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1262 /* Obsolete commands without queue context */
1263 switch (cmd->command) {
1264 case NFQNL_CFG_CMD_PF_BIND: return 0;
1265 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1269 /* Check if we support these flags in first place, dependencies should
1270 * be there too not to break atomicity.
1272 if (nfqa[NFQA_CFG_FLAGS]) {
1273 if (!nfqa[NFQA_CFG_MASK]) {
1274 /* A mask is needed to specify which flags are being
1280 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1281 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1283 if (flags >= NFQA_CFG_F_MAX)
1286 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1287 if (flags & mask & NFQA_CFG_F_SECCTX)
1290 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1291 !rcu_access_pointer(nfnl_ct_hook)) {
1292 #ifdef CONFIG_MODULES
1293 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1294 request_module("ip_conntrack_netlink");
1295 nfnl_lock(NFNL_SUBSYS_QUEUE);
1296 if (rcu_access_pointer(nfnl_ct_hook))
1304 queue = instance_lookup(q, queue_num);
1305 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1307 goto err_out_unlock;
1311 switch (cmd->command) {
1312 case NFQNL_CFG_CMD_BIND:
1315 goto err_out_unlock;
1317 queue = instance_create(q, queue_num,
1318 NETLINK_CB(skb).portid);
1319 if (IS_ERR(queue)) {
1320 ret = PTR_ERR(queue);
1321 goto err_out_unlock;
1324 case NFQNL_CFG_CMD_UNBIND:
1327 goto err_out_unlock;
1329 instance_destroy(q, queue);
1330 goto err_out_unlock;
1331 case NFQNL_CFG_CMD_PF_BIND:
1332 case NFQNL_CFG_CMD_PF_UNBIND:
1336 goto err_out_unlock;
1342 goto err_out_unlock;
1345 if (nfqa[NFQA_CFG_PARAMS]) {
1346 struct nfqnl_msg_config_params *params =
1347 nla_data(nfqa[NFQA_CFG_PARAMS]);
1349 nfqnl_set_mode(queue, params->copy_mode,
1350 ntohl(params->copy_range));
1353 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1354 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1356 spin_lock_bh(&queue->lock);
1357 queue->queue_maxlen = ntohl(*queue_maxlen);
1358 spin_unlock_bh(&queue->lock);
1361 if (nfqa[NFQA_CFG_FLAGS]) {
1362 spin_lock_bh(&queue->lock);
1363 queue->flags &= ~mask;
1364 queue->flags |= flags & mask;
1365 spin_unlock_bh(&queue->lock);
1373 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1374 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1375 .attr_count = NFQA_MAX, },
1376 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1377 .attr_count = NFQA_MAX,
1378 .policy = nfqa_verdict_policy },
1379 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1380 .attr_count = NFQA_CFG_MAX,
1381 .policy = nfqa_cfg_policy },
1382 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1383 .attr_count = NFQA_MAX,
1384 .policy = nfqa_verdict_batch_policy },
1387 static const struct nfnetlink_subsystem nfqnl_subsys = {
1389 .subsys_id = NFNL_SUBSYS_QUEUE,
1390 .cb_count = NFQNL_MSG_MAX,
1394 #ifdef CONFIG_PROC_FS
1396 struct seq_net_private p;
1397 unsigned int bucket;
1400 static struct hlist_node *get_first(struct seq_file *seq)
1402 struct iter_state *st = seq->private;
1404 struct nfnl_queue_net *q;
1409 net = seq_file_net(seq);
1410 q = nfnl_queue_pernet(net);
1411 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1412 if (!hlist_empty(&q->instance_table[st->bucket]))
1413 return q->instance_table[st->bucket].first;
1418 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1420 struct iter_state *st = seq->private;
1421 struct net *net = seq_file_net(seq);
1425 struct nfnl_queue_net *q;
1427 if (++st->bucket >= INSTANCE_BUCKETS)
1430 q = nfnl_queue_pernet(net);
1431 h = q->instance_table[st->bucket].first;
1436 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1438 struct hlist_node *head;
1439 head = get_first(seq);
1442 while (pos && (head = get_next(seq, head)))
1444 return pos ? NULL : head;
1447 static void *seq_start(struct seq_file *s, loff_t *pos)
1448 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1450 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1451 return get_idx(s, *pos);
1454 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1457 return get_next(s, v);
1460 static void seq_stop(struct seq_file *s, void *v)
1461 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1463 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1466 static int seq_show(struct seq_file *s, void *v)
1468 const struct nfqnl_instance *inst = v;
1470 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1472 inst->peer_portid, inst->queue_total,
1473 inst->copy_mode, inst->copy_range,
1474 inst->queue_dropped, inst->queue_user_dropped,
1475 inst->id_sequence, 1);
1479 static const struct seq_operations nfqnl_seq_ops = {
1485 #endif /* PROC_FS */
1487 static int __net_init nfnl_queue_net_init(struct net *net)
1490 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1492 for (i = 0; i < INSTANCE_BUCKETS; i++)
1493 INIT_HLIST_HEAD(&q->instance_table[i]);
1495 spin_lock_init(&q->instances_lock);
1497 #ifdef CONFIG_PROC_FS
1498 if (!proc_create_net("nfnetlink_queue", 0440, net->nf.proc_netfilter,
1499 &nfqnl_seq_ops, sizeof(struct iter_state)))
1502 nf_register_queue_handler(net, &nfqh);
1506 static void __net_exit nfnl_queue_net_exit(struct net *net)
1508 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1511 nf_unregister_queue_handler(net);
1512 #ifdef CONFIG_PROC_FS
1513 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1515 for (i = 0; i < INSTANCE_BUCKETS; i++)
1516 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1519 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1524 static struct pernet_operations nfnl_queue_net_ops = {
1525 .init = nfnl_queue_net_init,
1526 .exit = nfnl_queue_net_exit,
1527 .exit_batch = nfnl_queue_net_exit_batch,
1528 .id = &nfnl_queue_net_id,
1529 .size = sizeof(struct nfnl_queue_net),
1532 static int __init nfnetlink_queue_init(void)
1536 status = register_pernet_subsys(&nfnl_queue_net_ops);
1538 pr_err("failed to register pernet ops\n");
1542 netlink_register_notifier(&nfqnl_rtnl_notifier);
1543 status = nfnetlink_subsys_register(&nfqnl_subsys);
1545 pr_err("failed to create netlink socket\n");
1546 goto cleanup_netlink_notifier;
1549 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1551 pr_err("failed to register netdevice notifier\n");
1552 goto cleanup_netlink_subsys;
1557 cleanup_netlink_subsys:
1558 nfnetlink_subsys_unregister(&nfqnl_subsys);
1559 cleanup_netlink_notifier:
1560 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1561 unregister_pernet_subsys(&nfnl_queue_net_ops);
1566 static void __exit nfnetlink_queue_fini(void)
1568 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1569 nfnetlink_subsys_unregister(&nfqnl_subsys);
1570 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1571 unregister_pernet_subsys(&nfnl_queue_net_ops);
1573 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1576 MODULE_DESCRIPTION("netfilter packet queue handler");
1577 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1578 MODULE_LICENSE("GPL");
1579 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1581 module_init(nfnetlink_queue_init);
1582 module_exit(nfnetlink_queue_fini);