2 * This is a module which is used for queueing packets and communicating with
3 * userspace via nfnetlink.
5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
6 * (C) 2007 by Patrick McHardy <kaber@trash.net>
8 * Based on the old ipv4-only ip_queue.c:
9 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
10 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
12 * This program is free software; you can redistribute it and/or modify
13 * it under the terms of the GNU General Public License version 2 as
14 * published by the Free Software Foundation.
17 #include <linux/module.h>
18 #include <linux/skbuff.h>
19 #include <linux/init.h>
20 #include <linux/spinlock.h>
21 #include <linux/slab.h>
22 #include <linux/notifier.h>
23 #include <linux/netdevice.h>
24 #include <linux/netfilter.h>
25 #include <linux/proc_fs.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/netfilter_ipv6.h>
28 #include <linux/netfilter/nfnetlink.h>
29 #include <linux/netfilter/nfnetlink_queue.h>
30 #include <linux/list.h>
32 #include <net/netfilter/nf_queue.h>
33 #include <net/netns/generic.h>
34 #include <net/netfilter/nfnetlink_queue.h>
36 #include <linux/atomic.h>
38 #ifdef CONFIG_BRIDGE_NETFILTER
39 #include "../bridge/br_private.h"
42 #define NFQNL_QMAX_DEFAULT 1024
44 struct nfqnl_instance {
45 struct hlist_node hlist; /* global list of queues */
49 unsigned int queue_maxlen;
50 unsigned int copy_range;
51 unsigned int queue_dropped;
52 unsigned int queue_user_dropped;
55 u_int16_t queue_num; /* number of this queue */
57 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
59 * Following fields are dirtied for each queued packet,
60 * keep them in same cache line if possible.
63 unsigned int queue_total;
64 unsigned int id_sequence; /* 'sequence' of pkt ids */
65 struct list_head queue_list; /* packets in queue */
68 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
70 static int nfnl_queue_net_id __read_mostly;
72 #define INSTANCE_BUCKETS 16
73 struct nfnl_queue_net {
74 spinlock_t instances_lock;
75 struct hlist_head instance_table[INSTANCE_BUCKETS];
78 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
80 return net_generic(net, nfnl_queue_net_id);
83 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
85 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
88 static struct nfqnl_instance *
89 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
91 struct hlist_head *head;
92 struct nfqnl_instance *inst;
94 head = &q->instance_table[instance_hashfn(queue_num)];
95 hlist_for_each_entry_rcu(inst, head, hlist) {
96 if (inst->queue_num == queue_num)
102 static struct nfqnl_instance *
103 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num,
106 struct nfqnl_instance *inst;
110 spin_lock(&q->instances_lock);
111 if (instance_lookup(q, queue_num)) {
116 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
122 inst->queue_num = queue_num;
123 inst->peer_portid = portid;
124 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
125 inst->copy_range = 0xffff;
126 inst->copy_mode = NFQNL_COPY_NONE;
127 spin_lock_init(&inst->lock);
128 INIT_LIST_HEAD(&inst->queue_list);
130 if (!try_module_get(THIS_MODULE)) {
135 h = instance_hashfn(queue_num);
136 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
138 spin_unlock(&q->instances_lock);
145 spin_unlock(&q->instances_lock);
149 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
153 instance_destroy_rcu(struct rcu_head *head)
155 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
158 nfqnl_flush(inst, NULL, 0);
160 module_put(THIS_MODULE);
164 __instance_destroy(struct nfqnl_instance *inst)
166 hlist_del_rcu(&inst->hlist);
167 call_rcu(&inst->rcu, instance_destroy_rcu);
171 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
173 spin_lock(&q->instances_lock);
174 __instance_destroy(inst);
175 spin_unlock(&q->instances_lock);
179 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
181 list_add_tail(&entry->list, &queue->queue_list);
182 queue->queue_total++;
186 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
188 list_del(&entry->list);
189 queue->queue_total--;
192 static struct nf_queue_entry *
193 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
195 struct nf_queue_entry *entry = NULL, *i;
197 spin_lock_bh(&queue->lock);
199 list_for_each_entry(i, &queue->queue_list, list) {
207 __dequeue_entry(queue, entry);
209 spin_unlock_bh(&queue->lock);
215 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
217 struct nf_queue_entry *entry, *next;
219 spin_lock_bh(&queue->lock);
220 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
221 if (!cmpfn || cmpfn(entry, data)) {
222 list_del(&entry->list);
223 queue->queue_total--;
224 nf_reinject(entry, NF_DROP);
227 spin_unlock_bh(&queue->lock);
231 nfqnl_zcopy(struct sk_buff *to, const struct sk_buff *from, int len, int hlen)
234 int plen = 0; /* length of skb->head fragment */
238 /* dont bother with small payloads */
239 if (len <= skb_tailroom(to)) {
240 skb_copy_bits(from, 0, skb_put(to, len), len);
245 skb_copy_bits(from, 0, skb_put(to, hlen), hlen);
248 plen = min_t(int, skb_headlen(from), len);
250 page = virt_to_head_page(from->head);
251 offset = from->data - (unsigned char *)page_address(page);
252 __skb_fill_page_desc(to, 0, page, offset, plen);
259 to->truesize += len + plen;
260 to->len += len + plen;
261 to->data_len += len + plen;
263 for (i = 0; i < skb_shinfo(from)->nr_frags; i++) {
266 skb_shinfo(to)->frags[j] = skb_shinfo(from)->frags[i];
267 skb_shinfo(to)->frags[j].size = min_t(int, skb_shinfo(to)->frags[j].size, len);
268 len -= skb_shinfo(to)->frags[j].size;
272 skb_shinfo(to)->nr_frags = j;
275 static int nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet)
279 if (packet->ip_summed == CHECKSUM_PARTIAL)
280 flags = NFQA_SKB_CSUMNOTREADY;
281 if (skb_is_gso(packet))
282 flags |= NFQA_SKB_GSO;
284 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
287 static struct sk_buff *
288 nfqnl_build_packet_message(struct nfqnl_instance *queue,
289 struct nf_queue_entry *entry,
290 __be32 **packet_id_ptr)
293 size_t data_len = 0, cap_len = 0;
297 struct nfqnl_msg_packet_hdr *pmsg;
298 struct nlmsghdr *nlh;
299 struct nfgenmsg *nfmsg;
300 struct sk_buff *entskb = entry->skb;
301 struct net_device *indev;
302 struct net_device *outdev;
303 struct nf_conn *ct = NULL;
304 enum ip_conntrack_info uninitialized_var(ctinfo);
306 size = nlmsg_total_size(sizeof(struct nfgenmsg))
307 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
308 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
309 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
310 #ifdef CONFIG_BRIDGE_NETFILTER
311 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
312 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
314 + nla_total_size(sizeof(u_int32_t)) /* mark */
315 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
316 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
317 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
319 if (entskb->tstamp.tv64)
320 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
322 outdev = entry->outdev;
324 switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
325 case NFQNL_COPY_META:
326 case NFQNL_COPY_NONE:
329 case NFQNL_COPY_PACKET:
330 if (!(queue->flags & NFQA_CFG_F_GSO) &&
331 entskb->ip_summed == CHECKSUM_PARTIAL &&
332 skb_checksum_help(entskb))
335 data_len = ACCESS_ONCE(queue->copy_range);
336 if (data_len == 0 || data_len > entskb->len)
337 data_len = entskb->len;
340 if (!entskb->head_frag ||
341 skb_headlen(entskb) < L1_CACHE_BYTES ||
342 skb_shinfo(entskb)->nr_frags >= MAX_SKB_FRAGS)
343 hlen = skb_headlen(entskb);
345 if (skb_has_frag_list(entskb))
347 hlen = min_t(int, data_len, hlen);
348 size += sizeof(struct nlattr) + hlen;
349 cap_len = entskb->len;
353 if (queue->flags & NFQA_CFG_F_CONNTRACK)
354 ct = nfqnl_ct_get(entskb, &size, &ctinfo);
356 skb = nfnetlink_alloc_skb(&init_net, size, queue->peer_portid,
361 nlh = nlmsg_put(skb, 0, 0,
362 NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
363 sizeof(struct nfgenmsg), 0);
368 nfmsg = nlmsg_data(nlh);
369 nfmsg->nfgen_family = entry->pf;
370 nfmsg->version = NFNETLINK_V0;
371 nfmsg->res_id = htons(queue->queue_num);
373 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
374 pmsg = nla_data(nla);
375 pmsg->hw_protocol = entskb->protocol;
376 pmsg->hook = entry->hook;
377 *packet_id_ptr = &pmsg->packet_id;
379 indev = entry->indev;
381 #ifndef CONFIG_BRIDGE_NETFILTER
382 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
383 goto nla_put_failure;
385 if (entry->pf == PF_BRIDGE) {
386 /* Case 1: indev is physical input device, we need to
387 * look for bridge group (when called from
388 * netfilter_bridge) */
389 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
390 htonl(indev->ifindex)) ||
391 /* this is the bridge group "brX" */
392 /* rcu_read_lock()ed by __nf_queue */
393 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
394 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
395 goto nla_put_failure;
397 /* Case 2: indev is bridge group, we need to look for
398 * physical device (when called from ipv4) */
399 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
400 htonl(indev->ifindex)))
401 goto nla_put_failure;
402 if (entskb->nf_bridge && entskb->nf_bridge->physindev &&
403 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
404 htonl(entskb->nf_bridge->physindev->ifindex)))
405 goto nla_put_failure;
411 #ifndef CONFIG_BRIDGE_NETFILTER
412 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
413 goto nla_put_failure;
415 if (entry->pf == PF_BRIDGE) {
416 /* Case 1: outdev is physical output device, we need to
417 * look for bridge group (when called from
418 * netfilter_bridge) */
419 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
420 htonl(outdev->ifindex)) ||
421 /* this is the bridge group "brX" */
422 /* rcu_read_lock()ed by __nf_queue */
423 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
424 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
425 goto nla_put_failure;
427 /* Case 2: outdev is bridge group, we need to look for
428 * physical output device (when called from ipv4) */
429 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
430 htonl(outdev->ifindex)))
431 goto nla_put_failure;
432 if (entskb->nf_bridge && entskb->nf_bridge->physoutdev &&
433 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
434 htonl(entskb->nf_bridge->physoutdev->ifindex)))
435 goto nla_put_failure;
441 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
442 goto nla_put_failure;
444 if (indev && entskb->dev &&
445 entskb->mac_header != entskb->network_header) {
446 struct nfqnl_msg_packet_hw phw;
447 int len = dev_parse_header(entskb, phw.hw_addr);
449 phw.hw_addrlen = htons(len);
450 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
451 goto nla_put_failure;
455 if (entskb->tstamp.tv64) {
456 struct nfqnl_msg_packet_timestamp ts;
457 struct timeval tv = ktime_to_timeval(entskb->tstamp);
458 ts.sec = cpu_to_be64(tv.tv_sec);
459 ts.usec = cpu_to_be64(tv.tv_usec);
461 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
462 goto nla_put_failure;
465 if (ct && nfqnl_ct_put(skb, ct, ctinfo) < 0)
466 goto nla_put_failure;
468 if (cap_len > 0 && nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
469 goto nla_put_failure;
471 if (nfqnl_put_packet_info(skb, entskb))
472 goto nla_put_failure;
477 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
478 goto nla_put_failure;
480 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
481 nla->nla_type = NFQA_PAYLOAD;
482 nla->nla_len = nla_attr_size(data_len);
484 nfqnl_zcopy(skb, entskb, data_len, hlen);
487 nlh->nlmsg_len = skb->len;
492 net_err_ratelimited("nf_queue: error creating packet message\n");
497 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
498 struct nf_queue_entry *entry)
500 struct sk_buff *nskb;
502 __be32 *packet_id_ptr;
505 nskb = nfqnl_build_packet_message(queue, entry, &packet_id_ptr);
510 spin_lock_bh(&queue->lock);
512 if (!queue->peer_portid) {
514 goto err_out_free_nskb;
516 if (queue->queue_total >= queue->queue_maxlen) {
517 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
521 queue->queue_dropped++;
522 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
525 goto err_out_free_nskb;
527 entry->id = ++queue->id_sequence;
528 *packet_id_ptr = htonl(entry->id);
530 /* nfnetlink_unicast will either free the nskb or add it to a socket */
531 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
533 queue->queue_user_dropped++;
537 __enqueue_entry(queue, entry);
539 spin_unlock_bh(&queue->lock);
545 spin_unlock_bh(&queue->lock);
547 nf_reinject(entry, NF_ACCEPT);
552 static struct nf_queue_entry *
553 nf_queue_entry_dup(struct nf_queue_entry *e)
555 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
557 if (nf_queue_entry_get_refs(entry))
564 #ifdef CONFIG_BRIDGE_NETFILTER
565 /* When called from bridge netfilter, skb->data must point to MAC header
566 * before calling skb_gso_segment(). Else, original MAC header is lost
567 * and segmented skbs will be sent to wrong destination.
569 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
572 __skb_push(skb, skb->network_header - skb->mac_header);
575 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
578 __skb_pull(skb, skb->network_header - skb->mac_header);
581 #define nf_bridge_adjust_skb_data(s) do {} while (0)
582 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
585 static void free_entry(struct nf_queue_entry *entry)
587 nf_queue_entry_release_refs(entry);
592 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
593 struct sk_buff *skb, struct nf_queue_entry *entry)
596 struct nf_queue_entry *entry_seg;
598 nf_bridge_adjust_segmented_data(skb);
600 if (skb->next == NULL) { /* last packet, no need to copy entry */
601 struct sk_buff *gso_skb = entry->skb;
603 ret = __nfqnl_enqueue_packet(net, queue, entry);
605 entry->skb = gso_skb;
611 entry_seg = nf_queue_entry_dup(entry);
613 entry_seg->skb = skb;
614 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
616 free_entry(entry_seg);
622 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
625 struct nfqnl_instance *queue;
626 struct sk_buff *skb, *segs;
628 struct net *net = dev_net(entry->indev ?
629 entry->indev : entry->outdev);
630 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
632 /* rcu_read_lock()ed by nf_hook_slow() */
633 queue = instance_lookup(q, queuenum);
637 if (queue->copy_mode == NFQNL_COPY_NONE)
640 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(entry->skb))
641 return __nfqnl_enqueue_packet(net, queue, entry);
647 skb->protocol = htons(ETH_P_IP);
650 skb->protocol = htons(ETH_P_IPV6);
654 nf_bridge_adjust_skb_data(skb);
655 segs = skb_gso_segment(skb, 0);
656 /* Does not use PTR_ERR to limit the number of error codes that can be
657 * returned by nf_queue. For instance, callers rely on -ECANCELED to
658 * mean 'ignore this hook'.
665 struct sk_buff *nskb = segs->next;
667 err = __nfqnl_enqueue_packet_gso(net, queue,
677 if (err) /* some segments are already queued */
683 nf_bridge_adjust_segmented_data(skb);
688 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
690 struct sk_buff *nskb;
693 if (pskb_trim(e->skb, data_len))
695 } else if (diff > 0) {
696 if (data_len > 0xFFFF)
698 if (diff > skb_tailroom(e->skb)) {
699 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
702 printk(KERN_WARNING "nf_queue: OOM "
703 "in mangle, dropping packet\n");
709 skb_put(e->skb, diff);
711 if (!skb_make_writable(e->skb, data_len))
713 skb_copy_to_linear_data(e->skb, data, data_len);
714 e->skb->ip_summed = CHECKSUM_NONE;
719 nfqnl_set_mode(struct nfqnl_instance *queue,
720 unsigned char mode, unsigned int range)
724 spin_lock_bh(&queue->lock);
726 case NFQNL_COPY_NONE:
727 case NFQNL_COPY_META:
728 queue->copy_mode = mode;
729 queue->copy_range = 0;
732 case NFQNL_COPY_PACKET:
733 queue->copy_mode = mode;
734 /* We're using struct nlattr which has 16bit nla_len. Note that
735 * nla_len includes the header length. Thus, the maximum packet
736 * length that we support is 65531 bytes. We send truncated
737 * packets if the specified length is larger than that.
739 if (range > 0xffff - NLA_HDRLEN)
740 queue->copy_range = 0xffff - NLA_HDRLEN;
742 queue->copy_range = range;
749 spin_unlock_bh(&queue->lock);
755 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
758 if (entry->indev->ifindex == ifindex)
761 if (entry->outdev->ifindex == ifindex)
763 #ifdef CONFIG_BRIDGE_NETFILTER
764 if (entry->skb->nf_bridge) {
765 if (entry->skb->nf_bridge->physindev &&
766 entry->skb->nf_bridge->physindev->ifindex == ifindex)
768 if (entry->skb->nf_bridge->physoutdev &&
769 entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
776 /* drop all packets with either indev or outdev == ifindex from all queue
779 nfqnl_dev_drop(struct net *net, int ifindex)
782 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
786 for (i = 0; i < INSTANCE_BUCKETS; i++) {
787 struct nfqnl_instance *inst;
788 struct hlist_head *head = &q->instance_table[i];
790 hlist_for_each_entry_rcu(inst, head, hlist)
791 nfqnl_flush(inst, dev_cmp, ifindex);
797 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
800 nfqnl_rcv_dev_event(struct notifier_block *this,
801 unsigned long event, void *ptr)
803 struct net_device *dev = ptr;
805 /* Drop any packets associated with the downed device */
806 if (event == NETDEV_DOWN)
807 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
811 static struct notifier_block nfqnl_dev_notifier = {
812 .notifier_call = nfqnl_rcv_dev_event,
816 nfqnl_rcv_nl_event(struct notifier_block *this,
817 unsigned long event, void *ptr)
819 struct netlink_notify *n = ptr;
820 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
822 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
825 /* destroy all instances for this portid */
826 spin_lock(&q->instances_lock);
827 for (i = 0; i < INSTANCE_BUCKETS; i++) {
828 struct hlist_node *t2;
829 struct nfqnl_instance *inst;
830 struct hlist_head *head = &q->instance_table[i];
832 hlist_for_each_entry_safe(inst, t2, head, hlist) {
833 if (n->portid == inst->peer_portid)
834 __instance_destroy(inst);
837 spin_unlock(&q->instances_lock);
842 static struct notifier_block nfqnl_rtnl_notifier = {
843 .notifier_call = nfqnl_rcv_nl_event,
846 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
847 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
848 [NFQA_MARK] = { .type = NLA_U32 },
849 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
850 [NFQA_CT] = { .type = NLA_UNSPEC },
853 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
854 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
855 [NFQA_MARK] = { .type = NLA_U32 },
858 static struct nfqnl_instance *
859 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, int nlportid)
861 struct nfqnl_instance *queue;
863 queue = instance_lookup(q, queue_num);
865 return ERR_PTR(-ENODEV);
867 if (queue->peer_portid != nlportid)
868 return ERR_PTR(-EPERM);
873 static struct nfqnl_msg_verdict_hdr*
874 verdicthdr_get(const struct nlattr * const nfqa[])
876 struct nfqnl_msg_verdict_hdr *vhdr;
877 unsigned int verdict;
879 if (!nfqa[NFQA_VERDICT_HDR])
882 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
883 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
884 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
889 static int nfq_id_after(unsigned int id, unsigned int max)
891 return (int)(id - max) > 0;
895 nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
896 const struct nlmsghdr *nlh,
897 const struct nlattr * const nfqa[])
899 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
900 struct nf_queue_entry *entry, *tmp;
901 unsigned int verdict, maxid;
902 struct nfqnl_msg_verdict_hdr *vhdr;
903 struct nfqnl_instance *queue;
904 LIST_HEAD(batch_list);
905 u16 queue_num = ntohs(nfmsg->res_id);
907 struct net *net = sock_net(ctnl);
908 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
910 queue = verdict_instance_lookup(q, queue_num,
911 NETLINK_CB(skb).portid);
913 return PTR_ERR(queue);
915 vhdr = verdicthdr_get(nfqa);
919 verdict = ntohl(vhdr->verdict);
920 maxid = ntohl(vhdr->id);
922 spin_lock_bh(&queue->lock);
924 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
925 if (nfq_id_after(entry->id, maxid))
927 __dequeue_entry(queue, entry);
928 list_add_tail(&entry->list, &batch_list);
931 spin_unlock_bh(&queue->lock);
933 if (list_empty(&batch_list))
936 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
938 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
939 nf_reinject(entry, verdict);
945 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
946 const struct nlmsghdr *nlh,
947 const struct nlattr * const nfqa[])
949 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
950 u_int16_t queue_num = ntohs(nfmsg->res_id);
952 struct nfqnl_msg_verdict_hdr *vhdr;
953 struct nfqnl_instance *queue;
954 unsigned int verdict;
955 struct nf_queue_entry *entry;
956 enum ip_conntrack_info uninitialized_var(ctinfo);
957 struct nf_conn *ct = NULL;
959 struct net *net = sock_net(ctnl);
960 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
962 queue = instance_lookup(q, queue_num);
964 queue = verdict_instance_lookup(q, queue_num,
965 NETLINK_CB(skb).portid);
967 return PTR_ERR(queue);
969 vhdr = verdicthdr_get(nfqa);
973 verdict = ntohl(vhdr->verdict);
975 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
980 if (nfqa[NFQA_CT] && (queue->flags & NFQA_CFG_F_CONNTRACK))
981 ct = nfqnl_ct_parse(entry->skb, nfqa[NFQA_CT], &ctinfo);
983 if (nfqa[NFQA_PAYLOAD]) {
984 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
985 int diff = payload_len - entry->skb->len;
987 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
988 payload_len, entry, diff) < 0)
992 nfqnl_ct_seq_adjust(skb, ct, ctinfo, diff);
997 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
999 nf_reinject(entry, verdict);
1004 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1005 const struct nlmsghdr *nlh,
1006 const struct nlattr * const nfqa[])
1011 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1012 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1013 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1016 static const struct nf_queue_handler nfqh = {
1017 .outfn = &nfqnl_enqueue_packet,
1021 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1022 const struct nlmsghdr *nlh,
1023 const struct nlattr * const nfqa[])
1025 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1026 u_int16_t queue_num = ntohs(nfmsg->res_id);
1027 struct nfqnl_instance *queue;
1028 struct nfqnl_msg_config_cmd *cmd = NULL;
1029 struct net *net = sock_net(ctnl);
1030 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1033 if (nfqa[NFQA_CFG_CMD]) {
1034 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1036 /* Obsolete commands without queue context */
1037 switch (cmd->command) {
1038 case NFQNL_CFG_CMD_PF_BIND: return 0;
1039 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1044 queue = instance_lookup(q, queue_num);
1045 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1047 goto err_out_unlock;
1051 switch (cmd->command) {
1052 case NFQNL_CFG_CMD_BIND:
1055 goto err_out_unlock;
1057 queue = instance_create(q, queue_num,
1058 NETLINK_CB(skb).portid);
1059 if (IS_ERR(queue)) {
1060 ret = PTR_ERR(queue);
1061 goto err_out_unlock;
1064 case NFQNL_CFG_CMD_UNBIND:
1067 goto err_out_unlock;
1069 instance_destroy(q, queue);
1071 case NFQNL_CFG_CMD_PF_BIND:
1072 case NFQNL_CFG_CMD_PF_UNBIND:
1080 if (nfqa[NFQA_CFG_PARAMS]) {
1081 struct nfqnl_msg_config_params *params;
1085 goto err_out_unlock;
1087 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1088 nfqnl_set_mode(queue, params->copy_mode,
1089 ntohl(params->copy_range));
1092 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1093 __be32 *queue_maxlen;
1097 goto err_out_unlock;
1099 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1100 spin_lock_bh(&queue->lock);
1101 queue->queue_maxlen = ntohl(*queue_maxlen);
1102 spin_unlock_bh(&queue->lock);
1105 if (nfqa[NFQA_CFG_FLAGS]) {
1110 goto err_out_unlock;
1113 if (!nfqa[NFQA_CFG_MASK]) {
1114 /* A mask is needed to specify which flags are being
1118 goto err_out_unlock;
1121 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1122 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1124 if (flags >= NFQA_CFG_F_MAX) {
1126 goto err_out_unlock;
1129 spin_lock_bh(&queue->lock);
1130 queue->flags &= ~mask;
1131 queue->flags |= flags & mask;
1132 spin_unlock_bh(&queue->lock);
1140 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1141 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1142 .attr_count = NFQA_MAX, },
1143 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1144 .attr_count = NFQA_MAX,
1145 .policy = nfqa_verdict_policy },
1146 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1147 .attr_count = NFQA_CFG_MAX,
1148 .policy = nfqa_cfg_policy },
1149 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1150 .attr_count = NFQA_MAX,
1151 .policy = nfqa_verdict_batch_policy },
1154 static const struct nfnetlink_subsystem nfqnl_subsys = {
1156 .subsys_id = NFNL_SUBSYS_QUEUE,
1157 .cb_count = NFQNL_MSG_MAX,
1161 #ifdef CONFIG_PROC_FS
1163 struct seq_net_private p;
1164 unsigned int bucket;
1167 static struct hlist_node *get_first(struct seq_file *seq)
1169 struct iter_state *st = seq->private;
1171 struct nfnl_queue_net *q;
1176 net = seq_file_net(seq);
1177 q = nfnl_queue_pernet(net);
1178 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1179 if (!hlist_empty(&q->instance_table[st->bucket]))
1180 return q->instance_table[st->bucket].first;
1185 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1187 struct iter_state *st = seq->private;
1188 struct net *net = seq_file_net(seq);
1192 struct nfnl_queue_net *q;
1194 if (++st->bucket >= INSTANCE_BUCKETS)
1197 q = nfnl_queue_pernet(net);
1198 h = q->instance_table[st->bucket].first;
1203 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1205 struct hlist_node *head;
1206 head = get_first(seq);
1209 while (pos && (head = get_next(seq, head)))
1211 return pos ? NULL : head;
1214 static void *seq_start(struct seq_file *s, loff_t *pos)
1215 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1217 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1218 return get_idx(s, *pos);
1221 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1224 return get_next(s, v);
1227 static void seq_stop(struct seq_file *s, void *v)
1228 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1230 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1233 static int seq_show(struct seq_file *s, void *v)
1235 const struct nfqnl_instance *inst = v;
1237 return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
1239 inst->peer_portid, inst->queue_total,
1240 inst->copy_mode, inst->copy_range,
1241 inst->queue_dropped, inst->queue_user_dropped,
1242 inst->id_sequence, 1);
1245 static const struct seq_operations nfqnl_seq_ops = {
1252 static int nfqnl_open(struct inode *inode, struct file *file)
1254 return seq_open_net(inode, file, &nfqnl_seq_ops,
1255 sizeof(struct iter_state));
1258 static const struct file_operations nfqnl_file_ops = {
1259 .owner = THIS_MODULE,
1262 .llseek = seq_lseek,
1263 .release = seq_release_net,
1266 #endif /* PROC_FS */
1268 static int __net_init nfnl_queue_net_init(struct net *net)
1271 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1273 for (i = 0; i < INSTANCE_BUCKETS; i++)
1274 INIT_HLIST_HEAD(&q->instance_table[i]);
1276 spin_lock_init(&q->instances_lock);
1278 #ifdef CONFIG_PROC_FS
1279 if (!proc_create("nfnetlink_queue", 0440,
1280 net->nf.proc_netfilter, &nfqnl_file_ops))
1286 static void __net_exit nfnl_queue_net_exit(struct net *net)
1288 #ifdef CONFIG_PROC_FS
1289 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1293 static struct pernet_operations nfnl_queue_net_ops = {
1294 .init = nfnl_queue_net_init,
1295 .exit = nfnl_queue_net_exit,
1296 .id = &nfnl_queue_net_id,
1297 .size = sizeof(struct nfnl_queue_net),
1300 static int __init nfnetlink_queue_init(void)
1302 int status = -ENOMEM;
1304 netlink_register_notifier(&nfqnl_rtnl_notifier);
1305 status = nfnetlink_subsys_register(&nfqnl_subsys);
1307 pr_err("nf_queue: failed to create netlink socket\n");
1308 goto cleanup_netlink_notifier;
1311 status = register_pernet_subsys(&nfnl_queue_net_ops);
1313 pr_err("nf_queue: failed to register pernet ops\n");
1314 goto cleanup_subsys;
1316 register_netdevice_notifier(&nfqnl_dev_notifier);
1317 nf_register_queue_handler(&nfqh);
1321 nfnetlink_subsys_unregister(&nfqnl_subsys);
1322 cleanup_netlink_notifier:
1323 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1327 static void __exit nfnetlink_queue_fini(void)
1329 nf_unregister_queue_handler();
1330 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1331 unregister_pernet_subsys(&nfnl_queue_net_ops);
1332 nfnetlink_subsys_unregister(&nfqnl_subsys);
1333 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1335 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1338 MODULE_DESCRIPTION("netfilter packet queue handler");
1339 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1340 MODULE_LICENSE("GPL");
1341 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1343 module_init(nfnetlink_queue_init);
1344 module_exit(nfnetlink_queue_fini);