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 IS_ENABLED(CONFIG_NF_CONNTRACK)
448 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
449 if (nfnl_ct != NULL) {
450 ct = nf_ct_get(entskb, &ctinfo);
452 size += nfnl_ct->build_size(ct);
457 if (queue->flags & NFQA_CFG_F_UID_GID) {
458 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
459 + nla_total_size(sizeof(u_int32_t))); /* gid */
462 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
463 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
465 size += nla_total_size(seclen);
468 skb = alloc_skb(size, GFP_ATOMIC);
470 skb_tx_error(entskb);
474 nlh = nlmsg_put(skb, 0, 0,
475 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
476 sizeof(struct nfgenmsg), 0);
478 skb_tx_error(entskb);
482 nfmsg = nlmsg_data(nlh);
483 nfmsg->nfgen_family = entry->state.pf;
484 nfmsg->version = NFNETLINK_V0;
485 nfmsg->res_id = htons(queue->queue_num);
487 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
488 pmsg = nla_data(nla);
489 pmsg->hw_protocol = entskb->protocol;
490 pmsg->hook = entry->state.hook;
491 *packet_id_ptr = &pmsg->packet_id;
493 indev = entry->state.in;
495 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
496 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
497 goto nla_put_failure;
499 if (entry->state.pf == PF_BRIDGE) {
500 /* Case 1: indev is physical input device, we need to
501 * look for bridge group (when called from
502 * netfilter_bridge) */
503 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
504 htonl(indev->ifindex)) ||
505 /* this is the bridge group "brX" */
506 /* rcu_read_lock()ed by __nf_queue */
507 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
508 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
509 goto nla_put_failure;
513 /* Case 2: indev is bridge group, we need to look for
514 * physical device (when called from ipv4) */
515 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
516 htonl(indev->ifindex)))
517 goto nla_put_failure;
519 physinif = nf_bridge_get_physinif(entskb);
521 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
523 goto nla_put_failure;
529 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
530 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
531 goto nla_put_failure;
533 if (entry->state.pf == PF_BRIDGE) {
534 /* Case 1: outdev is physical output device, we need to
535 * look for bridge group (when called from
536 * netfilter_bridge) */
537 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
538 htonl(outdev->ifindex)) ||
539 /* this is the bridge group "brX" */
540 /* rcu_read_lock()ed by __nf_queue */
541 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
542 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
543 goto nla_put_failure;
547 /* Case 2: outdev is bridge group, we need to look for
548 * physical output device (when called from ipv4) */
549 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
550 htonl(outdev->ifindex)))
551 goto nla_put_failure;
553 physoutif = nf_bridge_get_physoutif(entskb);
555 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
557 goto nla_put_failure;
563 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
564 goto nla_put_failure;
566 if (indev && entskb->dev &&
567 entskb->mac_header != entskb->network_header) {
568 struct nfqnl_msg_packet_hw phw;
571 memset(&phw, 0, sizeof(phw));
572 len = dev_parse_header(entskb, phw.hw_addr);
574 phw.hw_addrlen = htons(len);
575 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
576 goto nla_put_failure;
580 if (nfqnl_put_bridge(entry, skb) < 0)
581 goto nla_put_failure;
583 if (entry->state.hook <= NF_INET_FORWARD && entskb->tstamp) {
584 struct nfqnl_msg_packet_timestamp ts;
585 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
587 ts.sec = cpu_to_be64(kts.tv_sec);
588 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
590 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
591 goto nla_put_failure;
594 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
595 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
596 goto nla_put_failure;
598 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
599 goto nla_put_failure;
601 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
602 goto nla_put_failure;
604 if (cap_len > data_len &&
605 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
606 goto nla_put_failure;
608 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
609 goto nla_put_failure;
614 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
615 goto nla_put_failure;
617 nla = skb_put(skb, sizeof(*nla));
618 nla->nla_type = NFQA_PAYLOAD;
619 nla->nla_len = nla_attr_size(data_len);
621 if (skb_zerocopy(skb, entskb, data_len, hlen))
622 goto nla_put_failure;
625 nlh->nlmsg_len = skb->len;
627 security_release_secctx(secdata, seclen);
631 skb_tx_error(entskb);
633 net_err_ratelimited("nf_queue: error creating packet message\n");
636 security_release_secctx(secdata, seclen);
640 static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry)
642 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
643 static const unsigned long flags = IPS_CONFIRMED | IPS_DYING;
644 const struct nf_conn *ct = (void *)skb_nfct(entry->skb);
646 if (ct && ((ct->status & flags) == IPS_DYING))
653 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
654 struct nf_queue_entry *entry)
656 struct sk_buff *nskb;
658 __be32 *packet_id_ptr;
661 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
666 spin_lock_bh(&queue->lock);
668 if (nf_ct_drop_unconfirmed(entry))
669 goto err_out_free_nskb;
671 if (queue->queue_total >= queue->queue_maxlen) {
672 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
676 queue->queue_dropped++;
677 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
680 goto err_out_free_nskb;
682 entry->id = ++queue->id_sequence;
683 *packet_id_ptr = htonl(entry->id);
685 /* nfnetlink_unicast will either free the nskb or add it to a socket */
686 err = nfnetlink_unicast(nskb, net, queue->peer_portid);
688 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
692 queue->queue_user_dropped++;
697 __enqueue_entry(queue, entry);
699 spin_unlock_bh(&queue->lock);
705 spin_unlock_bh(&queue->lock);
707 nfqnl_reinject(entry, NF_ACCEPT);
712 static struct nf_queue_entry *
713 nf_queue_entry_dup(struct nf_queue_entry *e)
715 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
717 nf_queue_entry_get_refs(entry);
721 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
722 /* When called from bridge netfilter, skb->data must point to MAC header
723 * before calling skb_gso_segment(). Else, original MAC header is lost
724 * and segmented skbs will be sent to wrong destination.
726 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
728 if (nf_bridge_info_get(skb))
729 __skb_push(skb, skb->network_header - skb->mac_header);
732 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
734 if (nf_bridge_info_get(skb))
735 __skb_pull(skb, skb->network_header - skb->mac_header);
738 #define nf_bridge_adjust_skb_data(s) do {} while (0)
739 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
743 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
744 struct sk_buff *skb, struct nf_queue_entry *entry)
747 struct nf_queue_entry *entry_seg;
749 nf_bridge_adjust_segmented_data(skb);
751 if (skb->next == NULL) { /* last packet, no need to copy entry */
752 struct sk_buff *gso_skb = entry->skb;
754 ret = __nfqnl_enqueue_packet(net, queue, entry);
756 entry->skb = gso_skb;
760 skb_mark_not_on_list(skb);
762 entry_seg = nf_queue_entry_dup(entry);
764 entry_seg->skb = skb;
765 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
767 nf_queue_entry_free(entry_seg);
773 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
776 struct nfqnl_instance *queue;
777 struct sk_buff *skb, *segs, *nskb;
779 struct net *net = entry->state.net;
780 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
782 /* rcu_read_lock()ed by nf_hook_thresh */
783 queue = instance_lookup(q, queuenum);
787 if (queue->copy_mode == NFQNL_COPY_NONE)
792 switch (entry->state.pf) {
794 skb->protocol = htons(ETH_P_IP);
797 skb->protocol = htons(ETH_P_IPV6);
801 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
802 return __nfqnl_enqueue_packet(net, queue, entry);
804 nf_bridge_adjust_skb_data(skb);
805 segs = skb_gso_segment(skb, 0);
806 /* Does not use PTR_ERR to limit the number of error codes that can be
807 * returned by nf_queue. For instance, callers rely on -ESRCH to
808 * mean 'ignore this hook'.
810 if (IS_ERR_OR_NULL(segs))
814 skb_list_walk_safe(segs, segs, nskb) {
816 err = __nfqnl_enqueue_packet_gso(net, queue,
825 if (err) /* some segments are already queued */
826 nf_queue_entry_free(entry);
831 nf_bridge_adjust_segmented_data(skb);
836 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
838 struct sk_buff *nskb;
841 if (pskb_trim(e->skb, data_len))
843 } else if (diff > 0) {
844 if (data_len > 0xFFFF)
846 if (diff > skb_tailroom(e->skb)) {
847 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
854 skb_put(e->skb, diff);
856 if (skb_ensure_writable(e->skb, data_len))
858 skb_copy_to_linear_data(e->skb, data, data_len);
859 e->skb->ip_summed = CHECKSUM_NONE;
864 nfqnl_set_mode(struct nfqnl_instance *queue,
865 unsigned char mode, unsigned int range)
869 spin_lock_bh(&queue->lock);
871 case NFQNL_COPY_NONE:
872 case NFQNL_COPY_META:
873 queue->copy_mode = mode;
874 queue->copy_range = 0;
877 case NFQNL_COPY_PACKET:
878 queue->copy_mode = mode;
879 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
880 queue->copy_range = NFQNL_MAX_COPY_RANGE;
882 queue->copy_range = range;
889 spin_unlock_bh(&queue->lock);
895 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
897 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
898 int physinif, physoutif;
900 physinif = nf_bridge_get_physinif(entry->skb);
901 physoutif = nf_bridge_get_physoutif(entry->skb);
903 if (physinif == ifindex || physoutif == ifindex)
907 if (entry->state.in->ifindex == ifindex)
909 if (entry->state.out)
910 if (entry->state.out->ifindex == ifindex)
916 /* drop all packets with either indev or outdev == ifindex from all queue
919 nfqnl_dev_drop(struct net *net, int ifindex)
922 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
926 for (i = 0; i < INSTANCE_BUCKETS; i++) {
927 struct nfqnl_instance *inst;
928 struct hlist_head *head = &q->instance_table[i];
930 hlist_for_each_entry_rcu(inst, head, hlist)
931 nfqnl_flush(inst, dev_cmp, ifindex);
938 nfqnl_rcv_dev_event(struct notifier_block *this,
939 unsigned long event, void *ptr)
941 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
943 /* Drop any packets associated with the downed device */
944 if (event == NETDEV_DOWN)
945 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
949 static struct notifier_block nfqnl_dev_notifier = {
950 .notifier_call = nfqnl_rcv_dev_event,
953 static void nfqnl_nf_hook_drop(struct net *net)
955 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
958 for (i = 0; i < INSTANCE_BUCKETS; i++) {
959 struct nfqnl_instance *inst;
960 struct hlist_head *head = &q->instance_table[i];
962 hlist_for_each_entry_rcu(inst, head, hlist)
963 nfqnl_flush(inst, NULL, 0);
968 nfqnl_rcv_nl_event(struct notifier_block *this,
969 unsigned long event, void *ptr)
971 struct netlink_notify *n = ptr;
972 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
974 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
977 /* destroy all instances for this portid */
978 spin_lock(&q->instances_lock);
979 for (i = 0; i < INSTANCE_BUCKETS; i++) {
980 struct hlist_node *t2;
981 struct nfqnl_instance *inst;
982 struct hlist_head *head = &q->instance_table[i];
984 hlist_for_each_entry_safe(inst, t2, head, hlist) {
985 if (n->portid == inst->peer_portid)
986 __instance_destroy(inst);
989 spin_unlock(&q->instances_lock);
994 static struct notifier_block nfqnl_rtnl_notifier = {
995 .notifier_call = nfqnl_rcv_nl_event,
998 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
999 [NFQA_VLAN_TCI] = { .type = NLA_U16},
1000 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
1003 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
1004 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1005 [NFQA_MARK] = { .type = NLA_U32 },
1006 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
1007 [NFQA_CT] = { .type = NLA_UNSPEC },
1008 [NFQA_EXP] = { .type = NLA_UNSPEC },
1009 [NFQA_VLAN] = { .type = NLA_NESTED },
1012 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1013 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1014 [NFQA_MARK] = { .type = NLA_U32 },
1017 static struct nfqnl_instance *
1018 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1020 struct nfqnl_instance *queue;
1022 queue = instance_lookup(q, queue_num);
1024 return ERR_PTR(-ENODEV);
1026 if (queue->peer_portid != nlportid)
1027 return ERR_PTR(-EPERM);
1032 static struct nfqnl_msg_verdict_hdr*
1033 verdicthdr_get(const struct nlattr * const nfqa[])
1035 struct nfqnl_msg_verdict_hdr *vhdr;
1036 unsigned int verdict;
1038 if (!nfqa[NFQA_VERDICT_HDR])
1041 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1042 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1043 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1048 static int nfq_id_after(unsigned int id, unsigned int max)
1050 return (int)(id - max) > 0;
1053 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1054 struct sk_buff *skb,
1055 const struct nlmsghdr *nlh,
1056 const struct nlattr * const nfqa[],
1057 struct netlink_ext_ack *extack)
1059 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1060 struct nf_queue_entry *entry, *tmp;
1061 unsigned int verdict, maxid;
1062 struct nfqnl_msg_verdict_hdr *vhdr;
1063 struct nfqnl_instance *queue;
1064 LIST_HEAD(batch_list);
1065 u16 queue_num = ntohs(nfmsg->res_id);
1066 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1068 queue = verdict_instance_lookup(q, queue_num,
1069 NETLINK_CB(skb).portid);
1071 return PTR_ERR(queue);
1073 vhdr = verdicthdr_get(nfqa);
1077 verdict = ntohl(vhdr->verdict);
1078 maxid = ntohl(vhdr->id);
1080 spin_lock_bh(&queue->lock);
1082 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1083 if (nfq_id_after(entry->id, maxid))
1085 __dequeue_entry(queue, entry);
1086 list_add_tail(&entry->list, &batch_list);
1089 spin_unlock_bh(&queue->lock);
1091 if (list_empty(&batch_list))
1094 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1095 if (nfqa[NFQA_MARK])
1096 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1098 nfqnl_reinject(entry, verdict);
1103 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1104 const struct nlmsghdr *nlh,
1105 const struct nlattr * const nfqa[],
1106 struct nf_queue_entry *entry,
1107 enum ip_conntrack_info *ctinfo)
1109 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1112 ct = nf_ct_get(entry->skb, ctinfo);
1116 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1120 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1121 NETLINK_CB(entry->skb).portid,
1129 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1130 const struct nlattr * const nfqa[])
1132 if (nfqa[NFQA_VLAN]) {
1133 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1136 err = nla_parse_nested_deprecated(tb, NFQA_VLAN_MAX,
1138 nfqa_vlan_policy, NULL);
1142 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1145 __vlan_hwaccel_put_tag(entry->skb,
1146 nla_get_be16(tb[NFQA_VLAN_PROTO]),
1147 ntohs(nla_get_be16(tb[NFQA_VLAN_TCI])));
1150 if (nfqa[NFQA_L2HDR]) {
1151 int mac_header_len = entry->skb->network_header -
1152 entry->skb->mac_header;
1154 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1156 else if (mac_header_len > 0)
1157 memcpy(skb_mac_header(entry->skb),
1158 nla_data(nfqa[NFQA_L2HDR]),
1165 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1166 struct sk_buff *skb,
1167 const struct nlmsghdr *nlh,
1168 const struct nlattr * const nfqa[],
1169 struct netlink_ext_ack *extack)
1171 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1172 u_int16_t queue_num = ntohs(nfmsg->res_id);
1173 struct nfqnl_msg_verdict_hdr *vhdr;
1174 struct nfqnl_instance *queue;
1175 unsigned int verdict;
1176 struct nf_queue_entry *entry;
1177 enum ip_conntrack_info ctinfo;
1178 struct nfnl_ct_hook *nfnl_ct;
1179 struct nf_conn *ct = NULL;
1180 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1183 queue = verdict_instance_lookup(q, queue_num,
1184 NETLINK_CB(skb).portid);
1186 return PTR_ERR(queue);
1188 vhdr = verdicthdr_get(nfqa);
1192 verdict = ntohl(vhdr->verdict);
1194 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1198 /* rcu lock already held from nfnl->call_rcu. */
1199 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1201 if (nfqa[NFQA_CT]) {
1202 if (nfnl_ct != NULL)
1203 ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1206 if (entry->state.pf == PF_BRIDGE) {
1207 err = nfqa_parse_bridge(entry, nfqa);
1212 if (nfqa[NFQA_PAYLOAD]) {
1213 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1214 int diff = payload_len - entry->skb->len;
1216 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1217 payload_len, entry, diff) < 0)
1221 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1224 if (nfqa[NFQA_MARK])
1225 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1227 nfqnl_reinject(entry, verdict);
1231 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1232 struct sk_buff *skb, const struct nlmsghdr *nlh,
1233 const struct nlattr * const nfqa[],
1234 struct netlink_ext_ack *extack)
1239 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1240 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1241 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1242 [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 },
1243 [NFQA_CFG_MASK] = { .type = NLA_U32 },
1244 [NFQA_CFG_FLAGS] = { .type = NLA_U32 },
1247 static const struct nf_queue_handler nfqh = {
1248 .outfn = nfqnl_enqueue_packet,
1249 .nf_hook_drop = nfqnl_nf_hook_drop,
1252 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1253 struct sk_buff *skb, const struct nlmsghdr *nlh,
1254 const struct nlattr * const nfqa[],
1255 struct netlink_ext_ack *extack)
1257 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1258 u_int16_t queue_num = ntohs(nfmsg->res_id);
1259 struct nfqnl_instance *queue;
1260 struct nfqnl_msg_config_cmd *cmd = NULL;
1261 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1262 __u32 flags = 0, mask = 0;
1265 if (nfqa[NFQA_CFG_CMD]) {
1266 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1268 /* Obsolete commands without queue context */
1269 switch (cmd->command) {
1270 case NFQNL_CFG_CMD_PF_BIND: return 0;
1271 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1275 /* Check if we support these flags in first place, dependencies should
1276 * be there too not to break atomicity.
1278 if (nfqa[NFQA_CFG_FLAGS]) {
1279 if (!nfqa[NFQA_CFG_MASK]) {
1280 /* A mask is needed to specify which flags are being
1286 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1287 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1289 if (flags >= NFQA_CFG_F_MAX)
1292 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1293 if (flags & mask & NFQA_CFG_F_SECCTX)
1296 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1297 !rcu_access_pointer(nfnl_ct_hook)) {
1298 #ifdef CONFIG_MODULES
1299 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1300 request_module("ip_conntrack_netlink");
1301 nfnl_lock(NFNL_SUBSYS_QUEUE);
1302 if (rcu_access_pointer(nfnl_ct_hook))
1310 queue = instance_lookup(q, queue_num);
1311 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1313 goto err_out_unlock;
1317 switch (cmd->command) {
1318 case NFQNL_CFG_CMD_BIND:
1321 goto err_out_unlock;
1323 queue = instance_create(q, queue_num,
1324 NETLINK_CB(skb).portid);
1325 if (IS_ERR(queue)) {
1326 ret = PTR_ERR(queue);
1327 goto err_out_unlock;
1330 case NFQNL_CFG_CMD_UNBIND:
1333 goto err_out_unlock;
1335 instance_destroy(q, queue);
1336 goto err_out_unlock;
1337 case NFQNL_CFG_CMD_PF_BIND:
1338 case NFQNL_CFG_CMD_PF_UNBIND:
1342 goto err_out_unlock;
1348 goto err_out_unlock;
1351 if (nfqa[NFQA_CFG_PARAMS]) {
1352 struct nfqnl_msg_config_params *params =
1353 nla_data(nfqa[NFQA_CFG_PARAMS]);
1355 nfqnl_set_mode(queue, params->copy_mode,
1356 ntohl(params->copy_range));
1359 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1360 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1362 spin_lock_bh(&queue->lock);
1363 queue->queue_maxlen = ntohl(*queue_maxlen);
1364 spin_unlock_bh(&queue->lock);
1367 if (nfqa[NFQA_CFG_FLAGS]) {
1368 spin_lock_bh(&queue->lock);
1369 queue->flags &= ~mask;
1370 queue->flags |= flags & mask;
1371 spin_unlock_bh(&queue->lock);
1379 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1380 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1381 .attr_count = NFQA_MAX, },
1382 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1383 .attr_count = NFQA_MAX,
1384 .policy = nfqa_verdict_policy },
1385 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1386 .attr_count = NFQA_CFG_MAX,
1387 .policy = nfqa_cfg_policy },
1388 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1389 .attr_count = NFQA_MAX,
1390 .policy = nfqa_verdict_batch_policy },
1393 static const struct nfnetlink_subsystem nfqnl_subsys = {
1395 .subsys_id = NFNL_SUBSYS_QUEUE,
1396 .cb_count = NFQNL_MSG_MAX,
1400 #ifdef CONFIG_PROC_FS
1402 struct seq_net_private p;
1403 unsigned int bucket;
1406 static struct hlist_node *get_first(struct seq_file *seq)
1408 struct iter_state *st = seq->private;
1410 struct nfnl_queue_net *q;
1415 net = seq_file_net(seq);
1416 q = nfnl_queue_pernet(net);
1417 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1418 if (!hlist_empty(&q->instance_table[st->bucket]))
1419 return q->instance_table[st->bucket].first;
1424 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1426 struct iter_state *st = seq->private;
1427 struct net *net = seq_file_net(seq);
1431 struct nfnl_queue_net *q;
1433 if (++st->bucket >= INSTANCE_BUCKETS)
1436 q = nfnl_queue_pernet(net);
1437 h = q->instance_table[st->bucket].first;
1442 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1444 struct hlist_node *head;
1445 head = get_first(seq);
1448 while (pos && (head = get_next(seq, head)))
1450 return pos ? NULL : head;
1453 static void *seq_start(struct seq_file *s, loff_t *pos)
1454 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1456 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1457 return get_idx(s, *pos);
1460 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1463 return get_next(s, v);
1466 static void seq_stop(struct seq_file *s, void *v)
1467 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1469 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1472 static int seq_show(struct seq_file *s, void *v)
1474 const struct nfqnl_instance *inst = v;
1476 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1478 inst->peer_portid, inst->queue_total,
1479 inst->copy_mode, inst->copy_range,
1480 inst->queue_dropped, inst->queue_user_dropped,
1481 inst->id_sequence, 1);
1485 static const struct seq_operations nfqnl_seq_ops = {
1491 #endif /* PROC_FS */
1493 static int __net_init nfnl_queue_net_init(struct net *net)
1496 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1498 for (i = 0; i < INSTANCE_BUCKETS; i++)
1499 INIT_HLIST_HEAD(&q->instance_table[i]);
1501 spin_lock_init(&q->instances_lock);
1503 #ifdef CONFIG_PROC_FS
1504 if (!proc_create_net("nfnetlink_queue", 0440, net->nf.proc_netfilter,
1505 &nfqnl_seq_ops, sizeof(struct iter_state)))
1508 nf_register_queue_handler(net, &nfqh);
1512 static void __net_exit nfnl_queue_net_exit(struct net *net)
1514 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1517 nf_unregister_queue_handler(net);
1518 #ifdef CONFIG_PROC_FS
1519 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1521 for (i = 0; i < INSTANCE_BUCKETS; i++)
1522 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1525 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1530 static struct pernet_operations nfnl_queue_net_ops = {
1531 .init = nfnl_queue_net_init,
1532 .exit = nfnl_queue_net_exit,
1533 .exit_batch = nfnl_queue_net_exit_batch,
1534 .id = &nfnl_queue_net_id,
1535 .size = sizeof(struct nfnl_queue_net),
1538 static int __init nfnetlink_queue_init(void)
1542 status = register_pernet_subsys(&nfnl_queue_net_ops);
1544 pr_err("failed to register pernet ops\n");
1548 netlink_register_notifier(&nfqnl_rtnl_notifier);
1549 status = nfnetlink_subsys_register(&nfqnl_subsys);
1551 pr_err("failed to create netlink socket\n");
1552 goto cleanup_netlink_notifier;
1555 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1557 pr_err("failed to register netdevice notifier\n");
1558 goto cleanup_netlink_subsys;
1563 cleanup_netlink_subsys:
1564 nfnetlink_subsys_unregister(&nfqnl_subsys);
1565 cleanup_netlink_notifier:
1566 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1567 unregister_pernet_subsys(&nfnl_queue_net_ops);
1572 static void __exit nfnetlink_queue_fini(void)
1574 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1575 nfnetlink_subsys_unregister(&nfqnl_subsys);
1576 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1577 unregister_pernet_subsys(&nfnl_queue_net_ops);
1579 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1582 MODULE_DESCRIPTION("netfilter packet queue handler");
1583 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1584 MODULE_LICENSE("GPL");
1585 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1587 module_init(nfnetlink_queue_init);
1588 module_exit(nfnetlink_queue_fini);