2 * This is a module which is used for logging packets to userspace via
5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
6 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
8 * Based on the old ipv4-only ipt_ULOG.c:
9 * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License version 2 as
13 * published by the Free Software Foundation.
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
18 #include <linux/module.h>
19 #include <linux/skbuff.h>
20 #include <linux/if_arp.h>
21 #include <linux/init.h>
23 #include <linux/ipv6.h>
24 #include <linux/netdevice.h>
25 #include <linux/netfilter.h>
26 #include <linux/netfilter_bridge.h>
27 #include <net/netlink.h>
28 #include <linux/netfilter/nfnetlink.h>
29 #include <linux/netfilter/nfnetlink_log.h>
30 #include <linux/netfilter/nf_conntrack_common.h>
31 #include <linux/spinlock.h>
32 #include <linux/sysctl.h>
33 #include <linux/proc_fs.h>
34 #include <linux/security.h>
35 #include <linux/list.h>
36 #include <linux/slab.h>
38 #include <net/netfilter/nf_log.h>
39 #include <net/netns/generic.h>
41 #include <linux/atomic.h>
42 #include <linux/refcount.h>
45 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
46 #include "../bridge/br_private.h"
49 #define NFULNL_COPY_DISABLED 0xff
50 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
51 #define NFULNL_TIMEOUT_DEFAULT 100 /* every second */
52 #define NFULNL_QTHRESH_DEFAULT 100 /* 100 packets */
53 /* max packet size is limited by 16-bit struct nfattr nfa_len field */
54 #define NFULNL_COPY_RANGE_MAX (0xFFFF - NLA_HDRLEN)
56 #define PRINTR(x, args...) do { if (net_ratelimit()) \
57 printk(x, ## args); } while (0);
59 struct nfulnl_instance {
60 struct hlist_node hlist; /* global list of instances */
62 refcount_t use; /* use count */
64 unsigned int qlen; /* number of nlmsgs in skb */
65 struct sk_buff *skb; /* pre-allocatd skb */
66 struct timer_list timer;
68 struct user_namespace *peer_user_ns; /* User namespace of the peer process */
69 u32 peer_portid; /* PORTID of the peer process */
71 /* configurable parameters */
72 unsigned int flushtimeout; /* timeout until queue flush */
73 unsigned int nlbufsiz; /* netlink buffer allocation size */
74 unsigned int qthreshold; /* threshold of the queue */
76 u_int32_t seq; /* instance-local sequential counter */
77 u_int16_t group_num; /* number of this queue */
83 #define INSTANCE_BUCKETS 16
85 static unsigned int nfnl_log_net_id __read_mostly;
88 spinlock_t instances_lock;
89 struct hlist_head instance_table[INSTANCE_BUCKETS];
93 static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
95 return net_generic(net, nfnl_log_net_id);
98 static inline u_int8_t instance_hashfn(u_int16_t group_num)
100 return ((group_num & 0xff) % INSTANCE_BUCKETS);
103 static struct nfulnl_instance *
104 __instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
106 struct hlist_head *head;
107 struct nfulnl_instance *inst;
109 head = &log->instance_table[instance_hashfn(group_num)];
110 hlist_for_each_entry_rcu(inst, head, hlist) {
111 if (inst->group_num == group_num)
118 instance_get(struct nfulnl_instance *inst)
120 refcount_inc(&inst->use);
123 static struct nfulnl_instance *
124 instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
126 struct nfulnl_instance *inst;
129 inst = __instance_lookup(log, group_num);
130 if (inst && !refcount_inc_not_zero(&inst->use))
132 rcu_read_unlock_bh();
137 static void nfulnl_instance_free_rcu(struct rcu_head *head)
139 struct nfulnl_instance *inst =
140 container_of(head, struct nfulnl_instance, rcu);
144 module_put(THIS_MODULE);
148 instance_put(struct nfulnl_instance *inst)
150 if (inst && refcount_dec_and_test(&inst->use))
151 call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
154 static void nfulnl_timer(struct timer_list *t);
156 static struct nfulnl_instance *
157 instance_create(struct net *net, u_int16_t group_num,
158 u32 portid, struct user_namespace *user_ns)
160 struct nfulnl_instance *inst;
161 struct nfnl_log_net *log = nfnl_log_pernet(net);
164 spin_lock_bh(&log->instances_lock);
165 if (__instance_lookup(log, group_num)) {
170 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
176 if (!try_module_get(THIS_MODULE)) {
182 INIT_HLIST_NODE(&inst->hlist);
183 spin_lock_init(&inst->lock);
184 /* needs to be two, since we _put() after creation */
185 refcount_set(&inst->use, 2);
187 timer_setup(&inst->timer, nfulnl_timer, 0);
189 inst->net = get_net(net);
190 inst->peer_user_ns = user_ns;
191 inst->peer_portid = portid;
192 inst->group_num = group_num;
194 inst->qthreshold = NFULNL_QTHRESH_DEFAULT;
195 inst->flushtimeout = NFULNL_TIMEOUT_DEFAULT;
196 inst->nlbufsiz = NFULNL_NLBUFSIZ_DEFAULT;
197 inst->copy_mode = NFULNL_COPY_PACKET;
198 inst->copy_range = NFULNL_COPY_RANGE_MAX;
200 hlist_add_head_rcu(&inst->hlist,
201 &log->instance_table[instance_hashfn(group_num)]);
204 spin_unlock_bh(&log->instances_lock);
209 spin_unlock_bh(&log->instances_lock);
213 static void __nfulnl_flush(struct nfulnl_instance *inst);
215 /* called with BH disabled */
217 __instance_destroy(struct nfulnl_instance *inst)
219 /* first pull it out of the global list */
220 hlist_del_rcu(&inst->hlist);
222 /* then flush all pending packets from skb */
224 spin_lock(&inst->lock);
226 /* lockless readers wont be able to use us */
227 inst->copy_mode = NFULNL_COPY_DISABLED;
230 __nfulnl_flush(inst);
231 spin_unlock(&inst->lock);
233 /* and finally put the refcount */
238 instance_destroy(struct nfnl_log_net *log,
239 struct nfulnl_instance *inst)
241 spin_lock_bh(&log->instances_lock);
242 __instance_destroy(inst);
243 spin_unlock_bh(&log->instances_lock);
247 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
252 spin_lock_bh(&inst->lock);
255 case NFULNL_COPY_NONE:
256 case NFULNL_COPY_META:
257 inst->copy_mode = mode;
258 inst->copy_range = 0;
261 case NFULNL_COPY_PACKET:
262 inst->copy_mode = mode;
264 range = NFULNL_COPY_RANGE_MAX;
265 inst->copy_range = min_t(unsigned int,
266 range, NFULNL_COPY_RANGE_MAX);
274 spin_unlock_bh(&inst->lock);
280 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
284 spin_lock_bh(&inst->lock);
285 if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
287 else if (nlbufsiz > 131072)
290 inst->nlbufsiz = nlbufsiz;
293 spin_unlock_bh(&inst->lock);
299 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
301 spin_lock_bh(&inst->lock);
302 inst->flushtimeout = timeout;
303 spin_unlock_bh(&inst->lock);
307 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
309 spin_lock_bh(&inst->lock);
310 inst->qthreshold = qthresh;
311 spin_unlock_bh(&inst->lock);
315 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
317 spin_lock_bh(&inst->lock);
319 spin_unlock_bh(&inst->lock);
324 static struct sk_buff *
325 nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
326 unsigned int pkt_size)
331 /* alloc skb which should be big enough for a whole multipart
332 * message. WARNING: has to be <= 128k due to slab restrictions */
334 n = max(inst_size, pkt_size);
335 skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
338 /* try to allocate only as much as we need for current
341 skb = alloc_skb(pkt_size, GFP_ATOMIC);
349 __nfulnl_send(struct nfulnl_instance *inst)
351 if (inst->qlen > 1) {
352 struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
354 sizeof(struct nfgenmsg),
356 if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
357 inst->skb->len, skb_tailroom(inst->skb))) {
358 kfree_skb(inst->skb);
362 nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid);
369 __nfulnl_flush(struct nfulnl_instance *inst)
371 /* timer holds a reference */
372 if (del_timer(&inst->timer))
379 nfulnl_timer(struct timer_list *t)
381 struct nfulnl_instance *inst = from_timer(inst, t, timer);
383 spin_lock_bh(&inst->lock);
386 spin_unlock_bh(&inst->lock);
390 /* This is an inline function, we don't really care about a long
391 * list of arguments */
393 __build_packet_message(struct nfnl_log_net *log,
394 struct nfulnl_instance *inst,
395 const struct sk_buff *skb,
396 unsigned int data_len,
398 unsigned int hooknum,
399 const struct net_device *indev,
400 const struct net_device *outdev,
401 const char *prefix, unsigned int plen,
402 const struct nfnl_ct_hook *nfnl_ct,
403 struct nf_conn *ct, enum ip_conntrack_info ctinfo)
405 struct nfulnl_msg_packet_hdr pmsg;
406 struct nlmsghdr *nlh;
407 sk_buff_data_t old_tail = inst->skb->tail;
409 const unsigned char *hwhdrp;
411 nlh = nfnl_msg_put(inst->skb, 0, 0,
412 nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
413 0, pf, NFNETLINK_V0, htons(inst->group_num));
417 memset(&pmsg, 0, sizeof(pmsg));
418 pmsg.hw_protocol = skb->protocol;
421 if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
422 goto nla_put_failure;
425 nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
426 goto nla_put_failure;
429 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
430 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
431 htonl(indev->ifindex)))
432 goto nla_put_failure;
434 if (pf == PF_BRIDGE) {
435 /* Case 1: outdev is physical input device, we need to
436 * look for bridge group (when called from
437 * netfilter_bridge) */
438 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
439 htonl(indev->ifindex)) ||
440 /* this is the bridge group "brX" */
441 /* rcu_read_lock()ed by nf_hook_thresh or
444 nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
445 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
446 goto nla_put_failure;
448 struct net_device *physindev;
450 /* Case 2: indev is bridge group, we need to look for
451 * physical device (when called from ipv4) */
452 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
453 htonl(indev->ifindex)))
454 goto nla_put_failure;
456 physindev = nf_bridge_get_physindev(skb);
458 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
459 htonl(physindev->ifindex)))
460 goto nla_put_failure;
466 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
467 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
468 htonl(outdev->ifindex)))
469 goto nla_put_failure;
471 if (pf == PF_BRIDGE) {
472 /* Case 1: outdev is physical output device, we need to
473 * look for bridge group (when called from
474 * netfilter_bridge) */
475 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
476 htonl(outdev->ifindex)) ||
477 /* this is the bridge group "brX" */
478 /* rcu_read_lock()ed by nf_hook_thresh or
481 nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
482 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
483 goto nla_put_failure;
485 struct net_device *physoutdev;
487 /* Case 2: indev is a bridge group, we need to look
488 * for physical device (when called from ipv4) */
489 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
490 htonl(outdev->ifindex)))
491 goto nla_put_failure;
493 physoutdev = nf_bridge_get_physoutdev(skb);
495 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
496 htonl(physoutdev->ifindex)))
497 goto nla_put_failure;
503 nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
504 goto nla_put_failure;
506 if (indev && skb->dev &&
507 skb_mac_header_was_set(skb) &&
508 skb_mac_header_len(skb) != 0) {
509 struct nfulnl_msg_packet_hw phw;
512 memset(&phw, 0, sizeof(phw));
513 len = dev_parse_header(skb, phw.hw_addr);
515 phw.hw_addrlen = htons(len);
516 if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
517 goto nla_put_failure;
521 if (indev && skb_mac_header_was_set(skb)) {
522 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
523 nla_put_be16(inst->skb, NFULA_HWLEN,
524 htons(skb->dev->hard_header_len)))
525 goto nla_put_failure;
527 hwhdrp = skb_mac_header(skb);
529 if (skb->dev->type == ARPHRD_SIT)
532 if (hwhdrp >= skb->head &&
533 nla_put(inst->skb, NFULA_HWHEADER,
534 skb->dev->hard_header_len, hwhdrp))
535 goto nla_put_failure;
539 struct nfulnl_msg_packet_timestamp ts;
540 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
541 ts.sec = cpu_to_be64(kts.tv_sec);
542 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
544 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
545 goto nla_put_failure;
550 if (sk && sk_fullsock(sk)) {
551 read_lock_bh(&sk->sk_callback_lock);
552 if (sk->sk_socket && sk->sk_socket->file) {
553 struct file *file = sk->sk_socket->file;
554 const struct cred *cred = file->f_cred;
555 struct user_namespace *user_ns = inst->peer_user_ns;
556 __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
557 __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
558 read_unlock_bh(&sk->sk_callback_lock);
559 if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
560 nla_put_be32(inst->skb, NFULA_GID, gid))
561 goto nla_put_failure;
563 read_unlock_bh(&sk->sk_callback_lock);
566 /* local sequence number */
567 if ((inst->flags & NFULNL_CFG_F_SEQ) &&
568 nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
569 goto nla_put_failure;
571 /* global sequence number */
572 if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
573 nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
574 htonl(atomic_inc_return(&log->global_seq))))
575 goto nla_put_failure;
577 if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
578 NFULA_CT, NFULA_CT_INFO) < 0)
579 goto nla_put_failure;
583 int size = nla_attr_size(data_len);
585 if (skb_tailroom(inst->skb) < nla_total_size(data_len))
586 goto nla_put_failure;
588 nla = skb_put(inst->skb, nla_total_size(data_len));
589 nla->nla_type = NFULA_PAYLOAD;
592 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
596 nlh->nlmsg_len = inst->skb->tail - old_tail;
600 PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
604 static const struct nf_loginfo default_loginfo = {
605 .type = NF_LOG_TYPE_ULOG,
615 /* log handler for internal netfilter logging api */
617 nfulnl_log_packet(struct net *net,
619 unsigned int hooknum,
620 const struct sk_buff *skb,
621 const struct net_device *in,
622 const struct net_device *out,
623 const struct nf_loginfo *li_user,
627 unsigned int data_len;
628 struct nfulnl_instance *inst;
629 const struct nf_loginfo *li;
630 unsigned int qthreshold;
631 unsigned int plen = 0;
632 struct nfnl_log_net *log = nfnl_log_pernet(net);
633 const struct nfnl_ct_hook *nfnl_ct = NULL;
634 enum ip_conntrack_info ctinfo = 0;
635 struct nf_conn *ct = NULL;
637 if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
640 li = &default_loginfo;
642 inst = instance_lookup_get(log, li->u.ulog.group);
647 plen = strlen(prefix) + 1;
649 /* FIXME: do we want to make the size calculation conditional based on
650 * what is actually present? way more branches and checks, but more
651 * memory efficient... */
652 size = nlmsg_total_size(sizeof(struct nfgenmsg))
653 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
654 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
655 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
656 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
657 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
658 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
660 + nla_total_size(sizeof(u_int32_t)) /* mark */
661 + nla_total_size(sizeof(u_int32_t)) /* uid */
662 + nla_total_size(sizeof(u_int32_t)) /* gid */
663 + nla_total_size(plen) /* prefix */
664 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
665 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
666 + nla_total_size(sizeof(struct nfgenmsg)); /* NLMSG_DONE */
668 if (in && skb_mac_header_was_set(skb)) {
669 size += nla_total_size(skb->dev->hard_header_len)
670 + nla_total_size(sizeof(u_int16_t)) /* hwtype */
671 + nla_total_size(sizeof(u_int16_t)); /* hwlen */
674 spin_lock_bh(&inst->lock);
676 if (inst->flags & NFULNL_CFG_F_SEQ)
677 size += nla_total_size(sizeof(u_int32_t));
678 if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
679 size += nla_total_size(sizeof(u_int32_t));
680 if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
681 nfnl_ct = rcu_dereference(nfnl_ct_hook);
682 if (nfnl_ct != NULL) {
683 ct = nfnl_ct->get_ct(skb, &ctinfo);
685 size += nfnl_ct->build_size(ct);
689 qthreshold = inst->qthreshold;
690 /* per-rule qthreshold overrides per-instance */
691 if (li->u.ulog.qthreshold)
692 if (qthreshold > li->u.ulog.qthreshold)
693 qthreshold = li->u.ulog.qthreshold;
696 switch (inst->copy_mode) {
697 case NFULNL_COPY_META:
698 case NFULNL_COPY_NONE:
702 case NFULNL_COPY_PACKET:
703 data_len = inst->copy_range;
704 if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
705 (li->u.ulog.copy_len < data_len))
706 data_len = li->u.ulog.copy_len;
708 if (data_len > skb->len)
711 size += nla_total_size(data_len);
714 case NFULNL_COPY_DISABLED:
716 goto unlock_and_release;
719 if (inst->skb && size > skb_tailroom(inst->skb)) {
720 /* either the queue len is too high or we don't have
721 * enough room in the skb left. flush to userspace. */
722 __nfulnl_flush(inst);
726 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
727 inst->nlbufsiz, size);
734 __build_packet_message(log, inst, skb, data_len, pf,
735 hooknum, in, out, prefix, plen,
736 nfnl_ct, ct, ctinfo);
738 if (inst->qlen >= qthreshold)
739 __nfulnl_flush(inst);
740 /* timer_pending always called within inst->lock, so there
741 * is no chance of a race here */
742 else if (!timer_pending(&inst->timer)) {
744 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
745 add_timer(&inst->timer);
749 spin_unlock_bh(&inst->lock);
754 /* FIXME: statistics */
755 goto unlock_and_release;
759 nfulnl_rcv_nl_event(struct notifier_block *this,
760 unsigned long event, void *ptr)
762 struct netlink_notify *n = ptr;
763 struct nfnl_log_net *log = nfnl_log_pernet(n->net);
765 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
768 /* destroy all instances for this portid */
769 spin_lock_bh(&log->instances_lock);
770 for (i = 0; i < INSTANCE_BUCKETS; i++) {
771 struct hlist_node *t2;
772 struct nfulnl_instance *inst;
773 struct hlist_head *head = &log->instance_table[i];
775 hlist_for_each_entry_safe(inst, t2, head, hlist) {
776 if (n->portid == inst->peer_portid)
777 __instance_destroy(inst);
780 spin_unlock_bh(&log->instances_lock);
785 static struct notifier_block nfulnl_rtnl_notifier = {
786 .notifier_call = nfulnl_rcv_nl_event,
789 static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl,
790 struct sk_buff *skb, const struct nlmsghdr *nlh,
791 const struct nlattr * const nfqa[],
792 struct netlink_ext_ack *extack)
797 static struct nf_logger nfulnl_logger __read_mostly = {
798 .name = "nfnetlink_log",
799 .type = NF_LOG_TYPE_ULOG,
800 .logfn = nfulnl_log_packet,
804 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
805 [NFULA_CFG_CMD] = { .len = sizeof(struct nfulnl_msg_config_cmd) },
806 [NFULA_CFG_MODE] = { .len = sizeof(struct nfulnl_msg_config_mode) },
807 [NFULA_CFG_TIMEOUT] = { .type = NLA_U32 },
808 [NFULA_CFG_QTHRESH] = { .type = NLA_U32 },
809 [NFULA_CFG_NLBUFSIZ] = { .type = NLA_U32 },
810 [NFULA_CFG_FLAGS] = { .type = NLA_U16 },
813 static int nfulnl_recv_config(struct net *net, struct sock *ctnl,
814 struct sk_buff *skb, const struct nlmsghdr *nlh,
815 const struct nlattr * const nfula[],
816 struct netlink_ext_ack *extack)
818 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
819 u_int16_t group_num = ntohs(nfmsg->res_id);
820 struct nfulnl_instance *inst;
821 struct nfulnl_msg_config_cmd *cmd = NULL;
822 struct nfnl_log_net *log = nfnl_log_pernet(net);
826 if (nfula[NFULA_CFG_CMD]) {
827 u_int8_t pf = nfmsg->nfgen_family;
828 cmd = nla_data(nfula[NFULA_CFG_CMD]);
830 /* Commands without queue context */
831 switch (cmd->command) {
832 case NFULNL_CFG_CMD_PF_BIND:
833 return nf_log_bind_pf(net, pf, &nfulnl_logger);
834 case NFULNL_CFG_CMD_PF_UNBIND:
835 nf_log_unbind_pf(net, pf);
840 inst = instance_lookup_get(log, group_num);
841 if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
846 /* Check if we support these flags in first place, dependencies should
847 * be there too not to break atomicity.
849 if (nfula[NFULA_CFG_FLAGS]) {
850 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
852 if ((flags & NFULNL_CFG_F_CONNTRACK) &&
853 !rcu_access_pointer(nfnl_ct_hook)) {
854 #ifdef CONFIG_MODULES
855 nfnl_unlock(NFNL_SUBSYS_ULOG);
856 request_module("ip_conntrack_netlink");
857 nfnl_lock(NFNL_SUBSYS_ULOG);
858 if (rcu_access_pointer(nfnl_ct_hook)) {
869 switch (cmd->command) {
870 case NFULNL_CFG_CMD_BIND:
876 inst = instance_create(net, group_num,
877 NETLINK_CB(skb).portid,
878 sk_user_ns(NETLINK_CB(skb).sk));
884 case NFULNL_CFG_CMD_UNBIND:
890 instance_destroy(log, inst);
901 if (nfula[NFULA_CFG_MODE]) {
902 struct nfulnl_msg_config_mode *params =
903 nla_data(nfula[NFULA_CFG_MODE]);
905 nfulnl_set_mode(inst, params->copy_mode,
906 ntohl(params->copy_range));
909 if (nfula[NFULA_CFG_TIMEOUT]) {
910 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
912 nfulnl_set_timeout(inst, ntohl(timeout));
915 if (nfula[NFULA_CFG_NLBUFSIZ]) {
916 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
918 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
921 if (nfula[NFULA_CFG_QTHRESH]) {
922 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
924 nfulnl_set_qthresh(inst, ntohl(qthresh));
927 if (nfula[NFULA_CFG_FLAGS])
928 nfulnl_set_flags(inst, flags);
936 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
937 [NFULNL_MSG_PACKET] = { .call = nfulnl_recv_unsupp,
938 .attr_count = NFULA_MAX, },
939 [NFULNL_MSG_CONFIG] = { .call = nfulnl_recv_config,
940 .attr_count = NFULA_CFG_MAX,
941 .policy = nfula_cfg_policy },
944 static const struct nfnetlink_subsystem nfulnl_subsys = {
946 .subsys_id = NFNL_SUBSYS_ULOG,
947 .cb_count = NFULNL_MSG_MAX,
951 #ifdef CONFIG_PROC_FS
953 struct seq_net_private p;
957 static struct hlist_node *get_first(struct net *net, struct iter_state *st)
959 struct nfnl_log_net *log;
963 log = nfnl_log_pernet(net);
965 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
966 struct hlist_head *head = &log->instance_table[st->bucket];
968 if (!hlist_empty(head))
969 return rcu_dereference_bh(hlist_first_rcu(head));
974 static struct hlist_node *get_next(struct net *net, struct iter_state *st,
975 struct hlist_node *h)
977 h = rcu_dereference_bh(hlist_next_rcu(h));
979 struct nfnl_log_net *log;
980 struct hlist_head *head;
982 if (++st->bucket >= INSTANCE_BUCKETS)
985 log = nfnl_log_pernet(net);
986 head = &log->instance_table[st->bucket];
987 h = rcu_dereference_bh(hlist_first_rcu(head));
992 static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
995 struct hlist_node *head;
996 head = get_first(net, st);
999 while (pos && (head = get_next(net, st, head)))
1001 return pos ? NULL : head;
1004 static void *seq_start(struct seq_file *s, loff_t *pos)
1008 return get_idx(seq_file_net(s), s->private, *pos);
1011 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1014 return get_next(seq_file_net(s), s->private, v);
1017 static void seq_stop(struct seq_file *s, void *v)
1020 rcu_read_unlock_bh();
1023 static int seq_show(struct seq_file *s, void *v)
1025 const struct nfulnl_instance *inst = v;
1027 seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1029 inst->peer_portid, inst->qlen,
1030 inst->copy_mode, inst->copy_range,
1031 inst->flushtimeout, refcount_read(&inst->use));
1036 static const struct seq_operations nful_seq_ops = {
1042 #endif /* PROC_FS */
1044 static int __net_init nfnl_log_net_init(struct net *net)
1047 struct nfnl_log_net *log = nfnl_log_pernet(net);
1048 #ifdef CONFIG_PROC_FS
1049 struct proc_dir_entry *proc;
1054 for (i = 0; i < INSTANCE_BUCKETS; i++)
1055 INIT_HLIST_HEAD(&log->instance_table[i]);
1056 spin_lock_init(&log->instances_lock);
1058 #ifdef CONFIG_PROC_FS
1059 proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1060 &nful_seq_ops, sizeof(struct iter_state));
1064 root_uid = make_kuid(net->user_ns, 0);
1065 root_gid = make_kgid(net->user_ns, 0);
1066 if (uid_valid(root_uid) && gid_valid(root_gid))
1067 proc_set_user(proc, root_uid, root_gid);
1072 static void __net_exit nfnl_log_net_exit(struct net *net)
1074 struct nfnl_log_net *log = nfnl_log_pernet(net);
1077 #ifdef CONFIG_PROC_FS
1078 remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1080 nf_log_unset(net, &nfulnl_logger);
1081 for (i = 0; i < INSTANCE_BUCKETS; i++)
1082 WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1085 static struct pernet_operations nfnl_log_net_ops = {
1086 .init = nfnl_log_net_init,
1087 .exit = nfnl_log_net_exit,
1088 .id = &nfnl_log_net_id,
1089 .size = sizeof(struct nfnl_log_net),
1092 static int __init nfnetlink_log_init(void)
1096 status = register_pernet_subsys(&nfnl_log_net_ops);
1098 pr_err("failed to register pernet ops\n");
1102 netlink_register_notifier(&nfulnl_rtnl_notifier);
1103 status = nfnetlink_subsys_register(&nfulnl_subsys);
1105 pr_err("failed to create netlink socket\n");
1106 goto cleanup_netlink_notifier;
1109 status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1111 pr_err("failed to register logger\n");
1112 goto cleanup_subsys;
1118 nfnetlink_subsys_unregister(&nfulnl_subsys);
1119 cleanup_netlink_notifier:
1120 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1121 unregister_pernet_subsys(&nfnl_log_net_ops);
1126 static void __exit nfnetlink_log_fini(void)
1128 nfnetlink_subsys_unregister(&nfulnl_subsys);
1129 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1130 unregister_pernet_subsys(&nfnl_log_net_ops);
1131 nf_log_unregister(&nfulnl_logger);
1134 MODULE_DESCRIPTION("netfilter userspace logging");
1135 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1136 MODULE_LICENSE("GPL");
1137 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1138 MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1139 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1140 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1141 MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1142 MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1144 module_init(nfnetlink_log_init);
1145 module_exit(nfnetlink_log_fini);