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 */
239 /* dont bother with small payloads */
240 if (len <= skb_tailroom(to))
241 return skb_copy_bits(from, 0, skb_put(to, len), len);
244 ret = skb_copy_bits(from, 0, skb_put(to, hlen), hlen);
249 plen = min_t(int, skb_headlen(from), len);
251 page = virt_to_head_page(from->head);
252 offset = from->data - (unsigned char *)page_address(page);
253 __skb_fill_page_desc(to, 0, page, offset, plen);
260 to->truesize += len + plen;
261 to->len += len + plen;
262 to->data_len += len + plen;
264 if (unlikely(skb_orphan_frags(from, GFP_ATOMIC))) {
269 for (i = 0; i < skb_shinfo(from)->nr_frags; i++) {
272 skb_shinfo(to)->frags[j] = skb_shinfo(from)->frags[i];
273 skb_shinfo(to)->frags[j].size = min_t(int, skb_shinfo(to)->frags[j].size, len);
274 len -= skb_shinfo(to)->frags[j].size;
278 skb_shinfo(to)->nr_frags = j;
283 static int nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet)
287 if (packet->ip_summed == CHECKSUM_PARTIAL)
288 flags = NFQA_SKB_CSUMNOTREADY;
289 if (skb_is_gso(packet))
290 flags |= NFQA_SKB_GSO;
292 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
295 static struct sk_buff *
296 nfqnl_build_packet_message(struct nfqnl_instance *queue,
297 struct nf_queue_entry *entry,
298 __be32 **packet_id_ptr)
301 size_t data_len = 0, cap_len = 0;
305 struct nfqnl_msg_packet_hdr *pmsg;
306 struct nlmsghdr *nlh;
307 struct nfgenmsg *nfmsg;
308 struct sk_buff *entskb = entry->skb;
309 struct net_device *indev;
310 struct net_device *outdev;
311 struct nf_conn *ct = NULL;
312 enum ip_conntrack_info uninitialized_var(ctinfo);
314 size = nlmsg_total_size(sizeof(struct nfgenmsg))
315 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
316 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
317 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
318 #ifdef CONFIG_BRIDGE_NETFILTER
319 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
320 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
322 + nla_total_size(sizeof(u_int32_t)) /* mark */
323 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
324 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
325 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
327 if (entskb->tstamp.tv64)
328 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
330 outdev = entry->outdev;
332 switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
333 case NFQNL_COPY_META:
334 case NFQNL_COPY_NONE:
337 case NFQNL_COPY_PACKET:
338 if (!(queue->flags & NFQA_CFG_F_GSO) &&
339 entskb->ip_summed == CHECKSUM_PARTIAL &&
340 skb_checksum_help(entskb))
343 data_len = ACCESS_ONCE(queue->copy_range);
344 if (data_len == 0 || data_len > entskb->len)
345 data_len = entskb->len;
348 if (!entskb->head_frag ||
349 skb_headlen(entskb) < L1_CACHE_BYTES ||
350 skb_shinfo(entskb)->nr_frags >= MAX_SKB_FRAGS)
351 hlen = skb_headlen(entskb);
353 if (skb_has_frag_list(entskb))
355 hlen = min_t(int, data_len, hlen);
356 size += sizeof(struct nlattr) + hlen;
357 cap_len = entskb->len;
361 if (queue->flags & NFQA_CFG_F_CONNTRACK)
362 ct = nfqnl_ct_get(entskb, &size, &ctinfo);
364 skb = nfnetlink_alloc_skb(&init_net, size, queue->peer_portid,
367 skb_tx_error(entskb);
371 nlh = nlmsg_put(skb, 0, 0,
372 NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
373 sizeof(struct nfgenmsg), 0);
375 skb_tx_error(entskb);
379 nfmsg = nlmsg_data(nlh);
380 nfmsg->nfgen_family = entry->pf;
381 nfmsg->version = NFNETLINK_V0;
382 nfmsg->res_id = htons(queue->queue_num);
384 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
385 pmsg = nla_data(nla);
386 pmsg->hw_protocol = entskb->protocol;
387 pmsg->hook = entry->hook;
388 *packet_id_ptr = &pmsg->packet_id;
390 indev = entry->indev;
392 #ifndef CONFIG_BRIDGE_NETFILTER
393 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
394 goto nla_put_failure;
396 if (entry->pf == PF_BRIDGE) {
397 /* Case 1: indev is physical input device, we need to
398 * look for bridge group (when called from
399 * netfilter_bridge) */
400 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
401 htonl(indev->ifindex)) ||
402 /* this is the bridge group "brX" */
403 /* rcu_read_lock()ed by __nf_queue */
404 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
405 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
406 goto nla_put_failure;
408 /* Case 2: indev is bridge group, we need to look for
409 * physical device (when called from ipv4) */
410 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
411 htonl(indev->ifindex)))
412 goto nla_put_failure;
413 if (entskb->nf_bridge && entskb->nf_bridge->physindev &&
414 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
415 htonl(entskb->nf_bridge->physindev->ifindex)))
416 goto nla_put_failure;
422 #ifndef CONFIG_BRIDGE_NETFILTER
423 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
424 goto nla_put_failure;
426 if (entry->pf == PF_BRIDGE) {
427 /* Case 1: outdev is physical output device, we need to
428 * look for bridge group (when called from
429 * netfilter_bridge) */
430 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
431 htonl(outdev->ifindex)) ||
432 /* this is the bridge group "brX" */
433 /* rcu_read_lock()ed by __nf_queue */
434 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
435 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
436 goto nla_put_failure;
438 /* Case 2: outdev is bridge group, we need to look for
439 * physical output device (when called from ipv4) */
440 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
441 htonl(outdev->ifindex)))
442 goto nla_put_failure;
443 if (entskb->nf_bridge && entskb->nf_bridge->physoutdev &&
444 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
445 htonl(entskb->nf_bridge->physoutdev->ifindex)))
446 goto nla_put_failure;
452 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
453 goto nla_put_failure;
455 if (indev && entskb->dev &&
456 entskb->mac_header != entskb->network_header) {
457 struct nfqnl_msg_packet_hw phw;
458 int len = dev_parse_header(entskb, phw.hw_addr);
460 phw.hw_addrlen = htons(len);
461 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
462 goto nla_put_failure;
466 if (entskb->tstamp.tv64) {
467 struct nfqnl_msg_packet_timestamp ts;
468 struct timeval tv = ktime_to_timeval(entskb->tstamp);
469 ts.sec = cpu_to_be64(tv.tv_sec);
470 ts.usec = cpu_to_be64(tv.tv_usec);
472 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
473 goto nla_put_failure;
476 if (ct && nfqnl_ct_put(skb, ct, ctinfo) < 0)
477 goto nla_put_failure;
479 if (cap_len > 0 && nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
480 goto nla_put_failure;
482 if (nfqnl_put_packet_info(skb, entskb))
483 goto nla_put_failure;
488 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
489 goto nla_put_failure;
491 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
492 nla->nla_type = NFQA_PAYLOAD;
493 nla->nla_len = nla_attr_size(data_len);
495 if (nfqnl_zcopy(skb, entskb, data_len, hlen))
496 goto nla_put_failure;
499 nlh->nlmsg_len = skb->len;
503 skb_tx_error(entskb);
505 net_err_ratelimited("nf_queue: error creating packet message\n");
510 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
511 struct nf_queue_entry *entry)
513 struct sk_buff *nskb;
515 __be32 *packet_id_ptr;
518 nskb = nfqnl_build_packet_message(queue, entry, &packet_id_ptr);
523 spin_lock_bh(&queue->lock);
525 if (!queue->peer_portid) {
527 goto err_out_free_nskb;
529 if (queue->queue_total >= queue->queue_maxlen) {
530 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
534 queue->queue_dropped++;
535 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
538 goto err_out_free_nskb;
540 entry->id = ++queue->id_sequence;
541 *packet_id_ptr = htonl(entry->id);
543 /* nfnetlink_unicast will either free the nskb or add it to a socket */
544 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
546 queue->queue_user_dropped++;
550 __enqueue_entry(queue, entry);
552 spin_unlock_bh(&queue->lock);
558 spin_unlock_bh(&queue->lock);
560 nf_reinject(entry, NF_ACCEPT);
565 static struct nf_queue_entry *
566 nf_queue_entry_dup(struct nf_queue_entry *e)
568 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
570 if (nf_queue_entry_get_refs(entry))
577 #ifdef CONFIG_BRIDGE_NETFILTER
578 /* When called from bridge netfilter, skb->data must point to MAC header
579 * before calling skb_gso_segment(). Else, original MAC header is lost
580 * and segmented skbs will be sent to wrong destination.
582 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
585 __skb_push(skb, skb->network_header - skb->mac_header);
588 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
591 __skb_pull(skb, skb->network_header - skb->mac_header);
594 #define nf_bridge_adjust_skb_data(s) do {} while (0)
595 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
598 static void free_entry(struct nf_queue_entry *entry)
600 nf_queue_entry_release_refs(entry);
605 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
606 struct sk_buff *skb, struct nf_queue_entry *entry)
609 struct nf_queue_entry *entry_seg;
611 nf_bridge_adjust_segmented_data(skb);
613 if (skb->next == NULL) { /* last packet, no need to copy entry */
614 struct sk_buff *gso_skb = entry->skb;
616 ret = __nfqnl_enqueue_packet(net, queue, entry);
618 entry->skb = gso_skb;
624 entry_seg = nf_queue_entry_dup(entry);
626 entry_seg->skb = skb;
627 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
629 free_entry(entry_seg);
635 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
638 struct nfqnl_instance *queue;
639 struct sk_buff *skb, *segs;
641 struct net *net = dev_net(entry->indev ?
642 entry->indev : entry->outdev);
643 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
645 /* rcu_read_lock()ed by nf_hook_slow() */
646 queue = instance_lookup(q, queuenum);
650 if (queue->copy_mode == NFQNL_COPY_NONE)
657 skb->protocol = htons(ETH_P_IP);
660 skb->protocol = htons(ETH_P_IPV6);
664 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
665 return __nfqnl_enqueue_packet(net, queue, entry);
667 nf_bridge_adjust_skb_data(skb);
668 segs = skb_gso_segment(skb, 0);
669 /* Does not use PTR_ERR to limit the number of error codes that can be
670 * returned by nf_queue. For instance, callers rely on -ECANCELED to
671 * mean 'ignore this hook'.
678 struct sk_buff *nskb = segs->next;
680 err = __nfqnl_enqueue_packet_gso(net, queue,
690 if (err) /* some segments are already queued */
696 nf_bridge_adjust_segmented_data(skb);
701 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
703 struct sk_buff *nskb;
706 if (pskb_trim(e->skb, data_len))
708 } else if (diff > 0) {
709 if (data_len > 0xFFFF)
711 if (diff > skb_tailroom(e->skb)) {
712 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
715 printk(KERN_WARNING "nf_queue: OOM "
716 "in mangle, dropping packet\n");
722 skb_put(e->skb, diff);
724 if (!skb_make_writable(e->skb, data_len))
726 skb_copy_to_linear_data(e->skb, data, data_len);
727 e->skb->ip_summed = CHECKSUM_NONE;
732 nfqnl_set_mode(struct nfqnl_instance *queue,
733 unsigned char mode, unsigned int range)
737 spin_lock_bh(&queue->lock);
739 case NFQNL_COPY_NONE:
740 case NFQNL_COPY_META:
741 queue->copy_mode = mode;
742 queue->copy_range = 0;
745 case NFQNL_COPY_PACKET:
746 queue->copy_mode = mode;
747 /* We're using struct nlattr which has 16bit nla_len. Note that
748 * nla_len includes the header length. Thus, the maximum packet
749 * length that we support is 65531 bytes. We send truncated
750 * packets if the specified length is larger than that.
752 if (range > 0xffff - NLA_HDRLEN)
753 queue->copy_range = 0xffff - NLA_HDRLEN;
755 queue->copy_range = range;
762 spin_unlock_bh(&queue->lock);
768 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
771 if (entry->indev->ifindex == ifindex)
774 if (entry->outdev->ifindex == ifindex)
776 #ifdef CONFIG_BRIDGE_NETFILTER
777 if (entry->skb->nf_bridge) {
778 if (entry->skb->nf_bridge->physindev &&
779 entry->skb->nf_bridge->physindev->ifindex == ifindex)
781 if (entry->skb->nf_bridge->physoutdev &&
782 entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
789 /* drop all packets with either indev or outdev == ifindex from all queue
792 nfqnl_dev_drop(struct net *net, int ifindex)
795 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
799 for (i = 0; i < INSTANCE_BUCKETS; i++) {
800 struct nfqnl_instance *inst;
801 struct hlist_head *head = &q->instance_table[i];
803 hlist_for_each_entry_rcu(inst, head, hlist)
804 nfqnl_flush(inst, dev_cmp, ifindex);
810 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
813 nfqnl_rcv_dev_event(struct notifier_block *this,
814 unsigned long event, void *ptr)
816 struct net_device *dev = ptr;
818 /* Drop any packets associated with the downed device */
819 if (event == NETDEV_DOWN)
820 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
824 static struct notifier_block nfqnl_dev_notifier = {
825 .notifier_call = nfqnl_rcv_dev_event,
829 nfqnl_rcv_nl_event(struct notifier_block *this,
830 unsigned long event, void *ptr)
832 struct netlink_notify *n = ptr;
833 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
835 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
838 /* destroy all instances for this portid */
839 spin_lock(&q->instances_lock);
840 for (i = 0; i < INSTANCE_BUCKETS; i++) {
841 struct hlist_node *t2;
842 struct nfqnl_instance *inst;
843 struct hlist_head *head = &q->instance_table[i];
845 hlist_for_each_entry_safe(inst, t2, head, hlist) {
846 if (n->portid == inst->peer_portid)
847 __instance_destroy(inst);
850 spin_unlock(&q->instances_lock);
855 static struct notifier_block nfqnl_rtnl_notifier = {
856 .notifier_call = nfqnl_rcv_nl_event,
859 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
860 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
861 [NFQA_MARK] = { .type = NLA_U32 },
862 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
863 [NFQA_CT] = { .type = NLA_UNSPEC },
866 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
867 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
868 [NFQA_MARK] = { .type = NLA_U32 },
871 static struct nfqnl_instance *
872 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, int nlportid)
874 struct nfqnl_instance *queue;
876 queue = instance_lookup(q, queue_num);
878 return ERR_PTR(-ENODEV);
880 if (queue->peer_portid != nlportid)
881 return ERR_PTR(-EPERM);
886 static struct nfqnl_msg_verdict_hdr*
887 verdicthdr_get(const struct nlattr * const nfqa[])
889 struct nfqnl_msg_verdict_hdr *vhdr;
890 unsigned int verdict;
892 if (!nfqa[NFQA_VERDICT_HDR])
895 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
896 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
897 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
902 static int nfq_id_after(unsigned int id, unsigned int max)
904 return (int)(id - max) > 0;
908 nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
909 const struct nlmsghdr *nlh,
910 const struct nlattr * const nfqa[])
912 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
913 struct nf_queue_entry *entry, *tmp;
914 unsigned int verdict, maxid;
915 struct nfqnl_msg_verdict_hdr *vhdr;
916 struct nfqnl_instance *queue;
917 LIST_HEAD(batch_list);
918 u16 queue_num = ntohs(nfmsg->res_id);
920 struct net *net = sock_net(ctnl);
921 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
923 queue = verdict_instance_lookup(q, queue_num,
924 NETLINK_CB(skb).portid);
926 return PTR_ERR(queue);
928 vhdr = verdicthdr_get(nfqa);
932 verdict = ntohl(vhdr->verdict);
933 maxid = ntohl(vhdr->id);
935 spin_lock_bh(&queue->lock);
937 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
938 if (nfq_id_after(entry->id, maxid))
940 __dequeue_entry(queue, entry);
941 list_add_tail(&entry->list, &batch_list);
944 spin_unlock_bh(&queue->lock);
946 if (list_empty(&batch_list))
949 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
951 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
952 nf_reinject(entry, verdict);
958 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
959 const struct nlmsghdr *nlh,
960 const struct nlattr * const nfqa[])
962 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
963 u_int16_t queue_num = ntohs(nfmsg->res_id);
965 struct nfqnl_msg_verdict_hdr *vhdr;
966 struct nfqnl_instance *queue;
967 unsigned int verdict;
968 struct nf_queue_entry *entry;
969 enum ip_conntrack_info uninitialized_var(ctinfo);
970 struct nf_conn *ct = NULL;
972 struct net *net = sock_net(ctnl);
973 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
975 queue = instance_lookup(q, queue_num);
977 queue = verdict_instance_lookup(q, queue_num,
978 NETLINK_CB(skb).portid);
980 return PTR_ERR(queue);
982 vhdr = verdicthdr_get(nfqa);
986 verdict = ntohl(vhdr->verdict);
988 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
993 if (nfqa[NFQA_CT] && (queue->flags & NFQA_CFG_F_CONNTRACK))
994 ct = nfqnl_ct_parse(entry->skb, nfqa[NFQA_CT], &ctinfo);
996 if (nfqa[NFQA_PAYLOAD]) {
997 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
998 int diff = payload_len - entry->skb->len;
1000 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1001 payload_len, entry, diff) < 0)
1005 nfqnl_ct_seq_adjust(skb, ct, ctinfo, diff);
1009 if (nfqa[NFQA_MARK])
1010 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1012 nf_reinject(entry, verdict);
1017 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1018 const struct nlmsghdr *nlh,
1019 const struct nlattr * const nfqa[])
1024 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1025 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1026 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1029 static const struct nf_queue_handler nfqh = {
1030 .outfn = &nfqnl_enqueue_packet,
1034 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1035 const struct nlmsghdr *nlh,
1036 const struct nlattr * const nfqa[])
1038 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1039 u_int16_t queue_num = ntohs(nfmsg->res_id);
1040 struct nfqnl_instance *queue;
1041 struct nfqnl_msg_config_cmd *cmd = NULL;
1042 struct net *net = sock_net(ctnl);
1043 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1046 if (nfqa[NFQA_CFG_CMD]) {
1047 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1049 /* Obsolete commands without queue context */
1050 switch (cmd->command) {
1051 case NFQNL_CFG_CMD_PF_BIND: return 0;
1052 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1057 queue = instance_lookup(q, queue_num);
1058 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1060 goto err_out_unlock;
1064 switch (cmd->command) {
1065 case NFQNL_CFG_CMD_BIND:
1068 goto err_out_unlock;
1070 queue = instance_create(q, queue_num,
1071 NETLINK_CB(skb).portid);
1072 if (IS_ERR(queue)) {
1073 ret = PTR_ERR(queue);
1074 goto err_out_unlock;
1077 case NFQNL_CFG_CMD_UNBIND:
1080 goto err_out_unlock;
1082 instance_destroy(q, queue);
1084 case NFQNL_CFG_CMD_PF_BIND:
1085 case NFQNL_CFG_CMD_PF_UNBIND:
1093 if (nfqa[NFQA_CFG_PARAMS]) {
1094 struct nfqnl_msg_config_params *params;
1098 goto err_out_unlock;
1100 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1101 nfqnl_set_mode(queue, params->copy_mode,
1102 ntohl(params->copy_range));
1105 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1106 __be32 *queue_maxlen;
1110 goto err_out_unlock;
1112 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1113 spin_lock_bh(&queue->lock);
1114 queue->queue_maxlen = ntohl(*queue_maxlen);
1115 spin_unlock_bh(&queue->lock);
1118 if (nfqa[NFQA_CFG_FLAGS]) {
1123 goto err_out_unlock;
1126 if (!nfqa[NFQA_CFG_MASK]) {
1127 /* A mask is needed to specify which flags are being
1131 goto err_out_unlock;
1134 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1135 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1137 if (flags >= NFQA_CFG_F_MAX) {
1139 goto err_out_unlock;
1142 spin_lock_bh(&queue->lock);
1143 queue->flags &= ~mask;
1144 queue->flags |= flags & mask;
1145 spin_unlock_bh(&queue->lock);
1153 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1154 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1155 .attr_count = NFQA_MAX, },
1156 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1157 .attr_count = NFQA_MAX,
1158 .policy = nfqa_verdict_policy },
1159 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1160 .attr_count = NFQA_CFG_MAX,
1161 .policy = nfqa_cfg_policy },
1162 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1163 .attr_count = NFQA_MAX,
1164 .policy = nfqa_verdict_batch_policy },
1167 static const struct nfnetlink_subsystem nfqnl_subsys = {
1169 .subsys_id = NFNL_SUBSYS_QUEUE,
1170 .cb_count = NFQNL_MSG_MAX,
1174 #ifdef CONFIG_PROC_FS
1176 struct seq_net_private p;
1177 unsigned int bucket;
1180 static struct hlist_node *get_first(struct seq_file *seq)
1182 struct iter_state *st = seq->private;
1184 struct nfnl_queue_net *q;
1189 net = seq_file_net(seq);
1190 q = nfnl_queue_pernet(net);
1191 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1192 if (!hlist_empty(&q->instance_table[st->bucket]))
1193 return q->instance_table[st->bucket].first;
1198 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1200 struct iter_state *st = seq->private;
1201 struct net *net = seq_file_net(seq);
1205 struct nfnl_queue_net *q;
1207 if (++st->bucket >= INSTANCE_BUCKETS)
1210 q = nfnl_queue_pernet(net);
1211 h = q->instance_table[st->bucket].first;
1216 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1218 struct hlist_node *head;
1219 head = get_first(seq);
1222 while (pos && (head = get_next(seq, head)))
1224 return pos ? NULL : head;
1227 static void *seq_start(struct seq_file *s, loff_t *pos)
1228 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1230 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1231 return get_idx(s, *pos);
1234 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1237 return get_next(s, v);
1240 static void seq_stop(struct seq_file *s, void *v)
1241 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1243 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1246 static int seq_show(struct seq_file *s, void *v)
1248 const struct nfqnl_instance *inst = v;
1250 return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
1252 inst->peer_portid, inst->queue_total,
1253 inst->copy_mode, inst->copy_range,
1254 inst->queue_dropped, inst->queue_user_dropped,
1255 inst->id_sequence, 1);
1258 static const struct seq_operations nfqnl_seq_ops = {
1265 static int nfqnl_open(struct inode *inode, struct file *file)
1267 return seq_open_net(inode, file, &nfqnl_seq_ops,
1268 sizeof(struct iter_state));
1271 static const struct file_operations nfqnl_file_ops = {
1272 .owner = THIS_MODULE,
1275 .llseek = seq_lseek,
1276 .release = seq_release_net,
1279 #endif /* PROC_FS */
1281 static int __net_init nfnl_queue_net_init(struct net *net)
1284 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1286 for (i = 0; i < INSTANCE_BUCKETS; i++)
1287 INIT_HLIST_HEAD(&q->instance_table[i]);
1289 spin_lock_init(&q->instances_lock);
1291 #ifdef CONFIG_PROC_FS
1292 if (!proc_create("nfnetlink_queue", 0440,
1293 net->nf.proc_netfilter, &nfqnl_file_ops))
1299 static void __net_exit nfnl_queue_net_exit(struct net *net)
1301 #ifdef CONFIG_PROC_FS
1302 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1306 static struct pernet_operations nfnl_queue_net_ops = {
1307 .init = nfnl_queue_net_init,
1308 .exit = nfnl_queue_net_exit,
1309 .id = &nfnl_queue_net_id,
1310 .size = sizeof(struct nfnl_queue_net),
1313 static int __init nfnetlink_queue_init(void)
1315 int status = -ENOMEM;
1317 netlink_register_notifier(&nfqnl_rtnl_notifier);
1318 status = nfnetlink_subsys_register(&nfqnl_subsys);
1320 pr_err("nf_queue: failed to create netlink socket\n");
1321 goto cleanup_netlink_notifier;
1324 status = register_pernet_subsys(&nfnl_queue_net_ops);
1326 pr_err("nf_queue: failed to register pernet ops\n");
1327 goto cleanup_subsys;
1329 register_netdevice_notifier(&nfqnl_dev_notifier);
1330 nf_register_queue_handler(&nfqh);
1334 nfnetlink_subsys_unregister(&nfqnl_subsys);
1335 cleanup_netlink_notifier:
1336 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1340 static void __exit nfnetlink_queue_fini(void)
1342 nf_unregister_queue_handler();
1343 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1344 unregister_pernet_subsys(&nfnl_queue_net_ops);
1345 nfnetlink_subsys_unregister(&nfqnl_subsys);
1346 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1348 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1351 MODULE_DESCRIPTION("netfilter packet queue handler");
1352 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1353 MODULE_LICENSE("GPL");
1354 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1356 module_init(nfnetlink_queue_init);
1357 module_exit(nfnetlink_queue_fini);