1 // SPDX-License-Identifier: GPL-2.0-only
3 * This is a module which is used for logging packets to userspace via
6 * (C) 2005 by Harald Welte <laforge@netfilter.org>
7 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
9 * Based on the old ipv4-only ipt_ULOG.c:
10 * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
15 #include <linux/module.h>
16 #include <linux/skbuff.h>
17 #include <linux/if_arp.h>
18 #include <linux/init.h>
20 #include <linux/ipv6.h>
21 #include <linux/netdevice.h>
22 #include <linux/netfilter.h>
23 #include <linux/netfilter_bridge.h>
24 #include <net/netlink.h>
25 #include <linux/netfilter/nfnetlink.h>
26 #include <linux/netfilter/nfnetlink_log.h>
27 #include <linux/netfilter/nf_conntrack_common.h>
28 #include <linux/spinlock.h>
29 #include <linux/sysctl.h>
30 #include <linux/proc_fs.h>
31 #include <linux/security.h>
32 #include <linux/list.h>
33 #include <linux/slab.h>
35 #include <net/netfilter/nf_log.h>
36 #include <net/netns/generic.h>
38 #include <linux/atomic.h>
39 #include <linux/refcount.h>
42 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
43 #include "../bridge/br_private.h"
46 #define NFULNL_COPY_DISABLED 0xff
47 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
48 #define NFULNL_TIMEOUT_DEFAULT 100 /* every second */
49 #define NFULNL_QTHRESH_DEFAULT 100 /* 100 packets */
50 /* max packet size is limited by 16-bit struct nfattr nfa_len field */
51 #define NFULNL_COPY_RANGE_MAX (0xFFFF - NLA_HDRLEN)
53 #define PRINTR(x, args...) do { if (net_ratelimit()) \
54 printk(x, ## args); } while (0);
56 struct nfulnl_instance {
57 struct hlist_node hlist; /* global list of instances */
59 refcount_t use; /* use count */
61 unsigned int qlen; /* number of nlmsgs in skb */
62 struct sk_buff *skb; /* pre-allocatd skb */
63 struct timer_list timer;
65 struct user_namespace *peer_user_ns; /* User namespace of the peer process */
66 u32 peer_portid; /* PORTID of the peer process */
68 /* configurable parameters */
69 unsigned int flushtimeout; /* timeout until queue flush */
70 unsigned int nlbufsiz; /* netlink buffer allocation size */
71 unsigned int qthreshold; /* threshold of the queue */
73 u_int32_t seq; /* instance-local sequential counter */
74 u_int16_t group_num; /* number of this queue */
80 #define INSTANCE_BUCKETS 16
82 static unsigned int nfnl_log_net_id __read_mostly;
85 spinlock_t instances_lock;
86 struct hlist_head instance_table[INSTANCE_BUCKETS];
90 static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
92 return net_generic(net, nfnl_log_net_id);
95 static inline u_int8_t instance_hashfn(u_int16_t group_num)
97 return ((group_num & 0xff) % INSTANCE_BUCKETS);
100 static struct nfulnl_instance *
101 __instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
103 struct hlist_head *head;
104 struct nfulnl_instance *inst;
106 head = &log->instance_table[instance_hashfn(group_num)];
107 hlist_for_each_entry_rcu(inst, head, hlist) {
108 if (inst->group_num == group_num)
115 instance_get(struct nfulnl_instance *inst)
117 refcount_inc(&inst->use);
120 static struct nfulnl_instance *
121 instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
123 struct nfulnl_instance *inst;
126 inst = __instance_lookup(log, group_num);
127 if (inst && !refcount_inc_not_zero(&inst->use))
129 rcu_read_unlock_bh();
134 static void nfulnl_instance_free_rcu(struct rcu_head *head)
136 struct nfulnl_instance *inst =
137 container_of(head, struct nfulnl_instance, rcu);
141 module_put(THIS_MODULE);
145 instance_put(struct nfulnl_instance *inst)
147 if (inst && refcount_dec_and_test(&inst->use))
148 call_rcu(&inst->rcu, nfulnl_instance_free_rcu);
151 static void nfulnl_timer(struct timer_list *t);
153 static struct nfulnl_instance *
154 instance_create(struct net *net, u_int16_t group_num,
155 u32 portid, struct user_namespace *user_ns)
157 struct nfulnl_instance *inst;
158 struct nfnl_log_net *log = nfnl_log_pernet(net);
161 spin_lock_bh(&log->instances_lock);
162 if (__instance_lookup(log, group_num)) {
167 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
173 if (!try_module_get(THIS_MODULE)) {
179 INIT_HLIST_NODE(&inst->hlist);
180 spin_lock_init(&inst->lock);
181 /* needs to be two, since we _put() after creation */
182 refcount_set(&inst->use, 2);
184 timer_setup(&inst->timer, nfulnl_timer, 0);
186 inst->net = get_net(net);
187 inst->peer_user_ns = user_ns;
188 inst->peer_portid = portid;
189 inst->group_num = group_num;
191 inst->qthreshold = NFULNL_QTHRESH_DEFAULT;
192 inst->flushtimeout = NFULNL_TIMEOUT_DEFAULT;
193 inst->nlbufsiz = NFULNL_NLBUFSIZ_DEFAULT;
194 inst->copy_mode = NFULNL_COPY_PACKET;
195 inst->copy_range = NFULNL_COPY_RANGE_MAX;
197 hlist_add_head_rcu(&inst->hlist,
198 &log->instance_table[instance_hashfn(group_num)]);
201 spin_unlock_bh(&log->instances_lock);
206 spin_unlock_bh(&log->instances_lock);
210 static void __nfulnl_flush(struct nfulnl_instance *inst);
212 /* called with BH disabled */
214 __instance_destroy(struct nfulnl_instance *inst)
216 /* first pull it out of the global list */
217 hlist_del_rcu(&inst->hlist);
219 /* then flush all pending packets from skb */
221 spin_lock(&inst->lock);
223 /* lockless readers wont be able to use us */
224 inst->copy_mode = NFULNL_COPY_DISABLED;
227 __nfulnl_flush(inst);
228 spin_unlock(&inst->lock);
230 /* and finally put the refcount */
235 instance_destroy(struct nfnl_log_net *log,
236 struct nfulnl_instance *inst)
238 spin_lock_bh(&log->instances_lock);
239 __instance_destroy(inst);
240 spin_unlock_bh(&log->instances_lock);
244 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
249 spin_lock_bh(&inst->lock);
252 case NFULNL_COPY_NONE:
253 case NFULNL_COPY_META:
254 inst->copy_mode = mode;
255 inst->copy_range = 0;
258 case NFULNL_COPY_PACKET:
259 inst->copy_mode = mode;
261 range = NFULNL_COPY_RANGE_MAX;
262 inst->copy_range = min_t(unsigned int,
263 range, NFULNL_COPY_RANGE_MAX);
271 spin_unlock_bh(&inst->lock);
277 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
281 spin_lock_bh(&inst->lock);
282 if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
284 else if (nlbufsiz > 131072)
287 inst->nlbufsiz = nlbufsiz;
290 spin_unlock_bh(&inst->lock);
296 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
298 spin_lock_bh(&inst->lock);
299 inst->flushtimeout = timeout;
300 spin_unlock_bh(&inst->lock);
304 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
306 spin_lock_bh(&inst->lock);
307 inst->qthreshold = qthresh;
308 spin_unlock_bh(&inst->lock);
312 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
314 spin_lock_bh(&inst->lock);
316 spin_unlock_bh(&inst->lock);
321 static struct sk_buff *
322 nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
323 unsigned int pkt_size)
328 /* alloc skb which should be big enough for a whole multipart
329 * message. WARNING: has to be <= 128k due to slab restrictions */
331 n = max(inst_size, pkt_size);
332 skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
335 /* try to allocate only as much as we need for current
338 skb = alloc_skb(pkt_size, GFP_ATOMIC);
346 __nfulnl_send(struct nfulnl_instance *inst)
348 if (inst->qlen > 1) {
349 struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
351 sizeof(struct nfgenmsg),
353 if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
354 inst->skb->len, skb_tailroom(inst->skb))) {
355 kfree_skb(inst->skb);
359 nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid);
366 __nfulnl_flush(struct nfulnl_instance *inst)
368 /* timer holds a reference */
369 if (del_timer(&inst->timer))
376 nfulnl_timer(struct timer_list *t)
378 struct nfulnl_instance *inst = from_timer(inst, t, timer);
380 spin_lock_bh(&inst->lock);
383 spin_unlock_bh(&inst->lock);
387 static u32 nfulnl_get_bridge_size(const struct sk_buff *skb)
391 if (!skb_mac_header_was_set(skb))
394 if (skb_vlan_tag_present(skb)) {
395 size += nla_total_size(0); /* nested */
396 size += nla_total_size(sizeof(u16)); /* id */
397 size += nla_total_size(sizeof(u16)); /* tag */
400 if (skb->network_header > skb->mac_header)
401 size += nla_total_size(skb->network_header - skb->mac_header);
406 static int nfulnl_put_bridge(struct nfulnl_instance *inst, const struct sk_buff *skb)
408 if (!skb_mac_header_was_set(skb))
411 if (skb_vlan_tag_present(skb)) {
414 nest = nla_nest_start(inst->skb, NFULA_VLAN);
416 goto nla_put_failure;
418 if (nla_put_be16(inst->skb, NFULA_VLAN_TCI, htons(skb->vlan_tci)) ||
419 nla_put_be16(inst->skb, NFULA_VLAN_PROTO, skb->vlan_proto))
420 goto nla_put_failure;
422 nla_nest_end(inst->skb, nest);
425 if (skb->mac_header < skb->network_header) {
426 int len = (int)(skb->network_header - skb->mac_header);
428 if (nla_put(inst->skb, NFULA_L2HDR, len, skb_mac_header(skb)))
429 goto nla_put_failure;
438 /* This is an inline function, we don't really care about a long
439 * list of arguments */
441 __build_packet_message(struct nfnl_log_net *log,
442 struct nfulnl_instance *inst,
443 const struct sk_buff *skb,
444 unsigned int data_len,
446 unsigned int hooknum,
447 const struct net_device *indev,
448 const struct net_device *outdev,
449 const char *prefix, unsigned int plen,
450 const struct nfnl_ct_hook *nfnl_ct,
451 struct nf_conn *ct, enum ip_conntrack_info ctinfo)
453 struct nfulnl_msg_packet_hdr pmsg;
454 struct nlmsghdr *nlh;
455 struct nfgenmsg *nfmsg;
456 sk_buff_data_t old_tail = inst->skb->tail;
458 const unsigned char *hwhdrp;
460 nlh = nlmsg_put(inst->skb, 0, 0,
461 nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
462 sizeof(struct nfgenmsg), 0);
465 nfmsg = nlmsg_data(nlh);
466 nfmsg->nfgen_family = pf;
467 nfmsg->version = NFNETLINK_V0;
468 nfmsg->res_id = htons(inst->group_num);
470 memset(&pmsg, 0, sizeof(pmsg));
471 pmsg.hw_protocol = skb->protocol;
474 if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
475 goto nla_put_failure;
478 nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
479 goto nla_put_failure;
482 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
483 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
484 htonl(indev->ifindex)))
485 goto nla_put_failure;
487 if (pf == PF_BRIDGE) {
488 /* Case 1: outdev is physical input device, we need to
489 * look for bridge group (when called from
490 * netfilter_bridge) */
491 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
492 htonl(indev->ifindex)) ||
493 /* this is the bridge group "brX" */
494 /* rcu_read_lock()ed by nf_hook_thresh or
497 nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
498 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
499 goto nla_put_failure;
501 struct net_device *physindev;
503 /* Case 2: indev is bridge group, we need to look for
504 * physical device (when called from ipv4) */
505 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
506 htonl(indev->ifindex)))
507 goto nla_put_failure;
509 physindev = nf_bridge_get_physindev(skb);
511 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
512 htonl(physindev->ifindex)))
513 goto nla_put_failure;
519 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
520 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
521 htonl(outdev->ifindex)))
522 goto nla_put_failure;
524 if (pf == PF_BRIDGE) {
525 /* Case 1: outdev is physical output device, we need to
526 * look for bridge group (when called from
527 * netfilter_bridge) */
528 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
529 htonl(outdev->ifindex)) ||
530 /* this is the bridge group "brX" */
531 /* rcu_read_lock()ed by nf_hook_thresh or
534 nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
535 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
536 goto nla_put_failure;
538 struct net_device *physoutdev;
540 /* Case 2: indev is a bridge group, we need to look
541 * for physical device (when called from ipv4) */
542 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
543 htonl(outdev->ifindex)))
544 goto nla_put_failure;
546 physoutdev = nf_bridge_get_physoutdev(skb);
548 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
549 htonl(physoutdev->ifindex)))
550 goto nla_put_failure;
556 nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
557 goto nla_put_failure;
559 if (indev && skb->dev &&
560 skb->mac_header != skb->network_header) {
561 struct nfulnl_msg_packet_hw phw;
564 memset(&phw, 0, sizeof(phw));
565 len = dev_parse_header(skb, phw.hw_addr);
567 phw.hw_addrlen = htons(len);
568 if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
569 goto nla_put_failure;
573 if (indev && skb_mac_header_was_set(skb)) {
574 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
575 nla_put_be16(inst->skb, NFULA_HWLEN,
576 htons(skb->dev->hard_header_len)))
577 goto nla_put_failure;
579 hwhdrp = skb_mac_header(skb);
581 if (skb->dev->type == ARPHRD_SIT)
584 if (hwhdrp >= skb->head &&
585 nla_put(inst->skb, NFULA_HWHEADER,
586 skb->dev->hard_header_len, hwhdrp))
587 goto nla_put_failure;
590 if (hooknum <= NF_INET_FORWARD && skb->tstamp) {
591 struct nfulnl_msg_packet_timestamp ts;
592 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
593 ts.sec = cpu_to_be64(kts.tv_sec);
594 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
596 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
597 goto nla_put_failure;
602 if (sk && sk_fullsock(sk)) {
603 read_lock_bh(&sk->sk_callback_lock);
604 if (sk->sk_socket && sk->sk_socket->file) {
605 struct file *file = sk->sk_socket->file;
606 const struct cred *cred = file->f_cred;
607 struct user_namespace *user_ns = inst->peer_user_ns;
608 __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
609 __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
610 read_unlock_bh(&sk->sk_callback_lock);
611 if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
612 nla_put_be32(inst->skb, NFULA_GID, gid))
613 goto nla_put_failure;
615 read_unlock_bh(&sk->sk_callback_lock);
618 /* local sequence number */
619 if ((inst->flags & NFULNL_CFG_F_SEQ) &&
620 nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
621 goto nla_put_failure;
623 /* global sequence number */
624 if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
625 nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
626 htonl(atomic_inc_return(&log->global_seq))))
627 goto nla_put_failure;
629 if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
630 NFULA_CT, NFULA_CT_INFO) < 0)
631 goto nla_put_failure;
633 if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) &&
634 nfulnl_put_bridge(inst, skb) < 0)
635 goto nla_put_failure;
639 int size = nla_attr_size(data_len);
641 if (skb_tailroom(inst->skb) < nla_total_size(data_len))
642 goto nla_put_failure;
644 nla = skb_put(inst->skb, nla_total_size(data_len));
645 nla->nla_type = NFULA_PAYLOAD;
648 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
652 nlh->nlmsg_len = inst->skb->tail - old_tail;
656 PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
660 static const struct nf_loginfo default_loginfo = {
661 .type = NF_LOG_TYPE_ULOG,
671 /* log handler for internal netfilter logging api */
673 nfulnl_log_packet(struct net *net,
675 unsigned int hooknum,
676 const struct sk_buff *skb,
677 const struct net_device *in,
678 const struct net_device *out,
679 const struct nf_loginfo *li_user,
683 unsigned int data_len;
684 struct nfulnl_instance *inst;
685 const struct nf_loginfo *li;
686 unsigned int qthreshold;
687 unsigned int plen = 0;
688 struct nfnl_log_net *log = nfnl_log_pernet(net);
689 const struct nfnl_ct_hook *nfnl_ct = NULL;
690 struct nf_conn *ct = NULL;
691 enum ip_conntrack_info ctinfo;
693 if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
696 li = &default_loginfo;
698 inst = instance_lookup_get(log, li->u.ulog.group);
703 plen = strlen(prefix) + 1;
705 /* FIXME: do we want to make the size calculation conditional based on
706 * what is actually present? way more branches and checks, but more
707 * memory efficient... */
708 size = nlmsg_total_size(sizeof(struct nfgenmsg))
709 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
710 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
711 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
712 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
713 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
714 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
716 + nla_total_size(sizeof(u_int32_t)) /* mark */
717 + nla_total_size(sizeof(u_int32_t)) /* uid */
718 + nla_total_size(sizeof(u_int32_t)) /* gid */
719 + nla_total_size(plen) /* prefix */
720 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
721 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
722 + nla_total_size(sizeof(struct nfgenmsg)); /* NLMSG_DONE */
724 if (in && skb_mac_header_was_set(skb)) {
725 size += nla_total_size(skb->dev->hard_header_len)
726 + nla_total_size(sizeof(u_int16_t)) /* hwtype */
727 + nla_total_size(sizeof(u_int16_t)); /* hwlen */
730 spin_lock_bh(&inst->lock);
732 if (inst->flags & NFULNL_CFG_F_SEQ)
733 size += nla_total_size(sizeof(u_int32_t));
734 if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
735 size += nla_total_size(sizeof(u_int32_t));
736 if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
737 nfnl_ct = rcu_dereference(nfnl_ct_hook);
738 if (nfnl_ct != NULL) {
739 ct = nfnl_ct->get_ct(skb, &ctinfo);
741 size += nfnl_ct->build_size(ct);
744 if (pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE)
745 size += nfulnl_get_bridge_size(skb);
747 qthreshold = inst->qthreshold;
748 /* per-rule qthreshold overrides per-instance */
749 if (li->u.ulog.qthreshold)
750 if (qthreshold > li->u.ulog.qthreshold)
751 qthreshold = li->u.ulog.qthreshold;
754 switch (inst->copy_mode) {
755 case NFULNL_COPY_META:
756 case NFULNL_COPY_NONE:
760 case NFULNL_COPY_PACKET:
761 data_len = inst->copy_range;
762 if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
763 (li->u.ulog.copy_len < data_len))
764 data_len = li->u.ulog.copy_len;
766 if (data_len > skb->len)
769 size += nla_total_size(data_len);
772 case NFULNL_COPY_DISABLED:
774 goto unlock_and_release;
777 if (inst->skb && size > skb_tailroom(inst->skb)) {
778 /* either the queue len is too high or we don't have
779 * enough room in the skb left. flush to userspace. */
780 __nfulnl_flush(inst);
784 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
785 inst->nlbufsiz, size);
792 __build_packet_message(log, inst, skb, data_len, pf,
793 hooknum, in, out, prefix, plen,
794 nfnl_ct, ct, ctinfo);
796 if (inst->qlen >= qthreshold)
797 __nfulnl_flush(inst);
798 /* timer_pending always called within inst->lock, so there
799 * is no chance of a race here */
800 else if (!timer_pending(&inst->timer)) {
802 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
803 add_timer(&inst->timer);
807 spin_unlock_bh(&inst->lock);
812 /* FIXME: statistics */
813 goto unlock_and_release;
817 nfulnl_rcv_nl_event(struct notifier_block *this,
818 unsigned long event, void *ptr)
820 struct netlink_notify *n = ptr;
821 struct nfnl_log_net *log = nfnl_log_pernet(n->net);
823 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
826 /* destroy all instances for this portid */
827 spin_lock_bh(&log->instances_lock);
828 for (i = 0; i < INSTANCE_BUCKETS; i++) {
829 struct hlist_node *t2;
830 struct nfulnl_instance *inst;
831 struct hlist_head *head = &log->instance_table[i];
833 hlist_for_each_entry_safe(inst, t2, head, hlist) {
834 if (n->portid == inst->peer_portid)
835 __instance_destroy(inst);
838 spin_unlock_bh(&log->instances_lock);
843 static struct notifier_block nfulnl_rtnl_notifier = {
844 .notifier_call = nfulnl_rcv_nl_event,
847 static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl,
848 struct sk_buff *skb, const struct nlmsghdr *nlh,
849 const struct nlattr * const nfqa[],
850 struct netlink_ext_ack *extack)
855 static struct nf_logger nfulnl_logger __read_mostly = {
856 .name = "nfnetlink_log",
857 .type = NF_LOG_TYPE_ULOG,
858 .logfn = nfulnl_log_packet,
862 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
863 [NFULA_CFG_CMD] = { .len = sizeof(struct nfulnl_msg_config_cmd) },
864 [NFULA_CFG_MODE] = { .len = sizeof(struct nfulnl_msg_config_mode) },
865 [NFULA_CFG_TIMEOUT] = { .type = NLA_U32 },
866 [NFULA_CFG_QTHRESH] = { .type = NLA_U32 },
867 [NFULA_CFG_NLBUFSIZ] = { .type = NLA_U32 },
868 [NFULA_CFG_FLAGS] = { .type = NLA_U16 },
871 static int nfulnl_recv_config(struct net *net, struct sock *ctnl,
872 struct sk_buff *skb, const struct nlmsghdr *nlh,
873 const struct nlattr * const nfula[],
874 struct netlink_ext_ack *extack)
876 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
877 u_int16_t group_num = ntohs(nfmsg->res_id);
878 struct nfulnl_instance *inst;
879 struct nfulnl_msg_config_cmd *cmd = NULL;
880 struct nfnl_log_net *log = nfnl_log_pernet(net);
884 if (nfula[NFULA_CFG_CMD]) {
885 u_int8_t pf = nfmsg->nfgen_family;
886 cmd = nla_data(nfula[NFULA_CFG_CMD]);
888 /* Commands without queue context */
889 switch (cmd->command) {
890 case NFULNL_CFG_CMD_PF_BIND:
891 return nf_log_bind_pf(net, pf, &nfulnl_logger);
892 case NFULNL_CFG_CMD_PF_UNBIND:
893 nf_log_unbind_pf(net, pf);
898 inst = instance_lookup_get(log, group_num);
899 if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
904 /* Check if we support these flags in first place, dependencies should
905 * be there too not to break atomicity.
907 if (nfula[NFULA_CFG_FLAGS]) {
908 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
910 if ((flags & NFULNL_CFG_F_CONNTRACK) &&
911 !rcu_access_pointer(nfnl_ct_hook)) {
912 #ifdef CONFIG_MODULES
913 nfnl_unlock(NFNL_SUBSYS_ULOG);
914 request_module("ip_conntrack_netlink");
915 nfnl_lock(NFNL_SUBSYS_ULOG);
916 if (rcu_access_pointer(nfnl_ct_hook)) {
927 switch (cmd->command) {
928 case NFULNL_CFG_CMD_BIND:
934 inst = instance_create(net, group_num,
935 NETLINK_CB(skb).portid,
936 sk_user_ns(NETLINK_CB(skb).sk));
942 case NFULNL_CFG_CMD_UNBIND:
948 instance_destroy(log, inst);
959 if (nfula[NFULA_CFG_MODE]) {
960 struct nfulnl_msg_config_mode *params =
961 nla_data(nfula[NFULA_CFG_MODE]);
963 nfulnl_set_mode(inst, params->copy_mode,
964 ntohl(params->copy_range));
967 if (nfula[NFULA_CFG_TIMEOUT]) {
968 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
970 nfulnl_set_timeout(inst, ntohl(timeout));
973 if (nfula[NFULA_CFG_NLBUFSIZ]) {
974 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
976 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
979 if (nfula[NFULA_CFG_QTHRESH]) {
980 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
982 nfulnl_set_qthresh(inst, ntohl(qthresh));
985 if (nfula[NFULA_CFG_FLAGS])
986 nfulnl_set_flags(inst, flags);
994 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
995 [NFULNL_MSG_PACKET] = { .call = nfulnl_recv_unsupp,
996 .attr_count = NFULA_MAX, },
997 [NFULNL_MSG_CONFIG] = { .call = nfulnl_recv_config,
998 .attr_count = NFULA_CFG_MAX,
999 .policy = nfula_cfg_policy },
1002 static const struct nfnetlink_subsystem nfulnl_subsys = {
1004 .subsys_id = NFNL_SUBSYS_ULOG,
1005 .cb_count = NFULNL_MSG_MAX,
1009 #ifdef CONFIG_PROC_FS
1011 struct seq_net_private p;
1012 unsigned int bucket;
1015 static struct hlist_node *get_first(struct net *net, struct iter_state *st)
1017 struct nfnl_log_net *log;
1021 log = nfnl_log_pernet(net);
1023 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1024 struct hlist_head *head = &log->instance_table[st->bucket];
1026 if (!hlist_empty(head))
1027 return rcu_dereference_bh(hlist_first_rcu(head));
1032 static struct hlist_node *get_next(struct net *net, struct iter_state *st,
1033 struct hlist_node *h)
1035 h = rcu_dereference_bh(hlist_next_rcu(h));
1037 struct nfnl_log_net *log;
1038 struct hlist_head *head;
1040 if (++st->bucket >= INSTANCE_BUCKETS)
1043 log = nfnl_log_pernet(net);
1044 head = &log->instance_table[st->bucket];
1045 h = rcu_dereference_bh(hlist_first_rcu(head));
1050 static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1053 struct hlist_node *head;
1054 head = get_first(net, st);
1057 while (pos && (head = get_next(net, st, head)))
1059 return pos ? NULL : head;
1062 static void *seq_start(struct seq_file *s, loff_t *pos)
1066 return get_idx(seq_file_net(s), s->private, *pos);
1069 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1072 return get_next(seq_file_net(s), s->private, v);
1075 static void seq_stop(struct seq_file *s, void *v)
1078 rcu_read_unlock_bh();
1081 static int seq_show(struct seq_file *s, void *v)
1083 const struct nfulnl_instance *inst = v;
1085 seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1087 inst->peer_portid, inst->qlen,
1088 inst->copy_mode, inst->copy_range,
1089 inst->flushtimeout, refcount_read(&inst->use));
1094 static const struct seq_operations nful_seq_ops = {
1100 #endif /* PROC_FS */
1102 static int __net_init nfnl_log_net_init(struct net *net)
1105 struct nfnl_log_net *log = nfnl_log_pernet(net);
1106 #ifdef CONFIG_PROC_FS
1107 struct proc_dir_entry *proc;
1112 for (i = 0; i < INSTANCE_BUCKETS; i++)
1113 INIT_HLIST_HEAD(&log->instance_table[i]);
1114 spin_lock_init(&log->instances_lock);
1116 #ifdef CONFIG_PROC_FS
1117 proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1118 &nful_seq_ops, sizeof(struct iter_state));
1122 root_uid = make_kuid(net->user_ns, 0);
1123 root_gid = make_kgid(net->user_ns, 0);
1124 if (uid_valid(root_uid) && gid_valid(root_gid))
1125 proc_set_user(proc, root_uid, root_gid);
1130 static void __net_exit nfnl_log_net_exit(struct net *net)
1132 struct nfnl_log_net *log = nfnl_log_pernet(net);
1135 #ifdef CONFIG_PROC_FS
1136 remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1138 nf_log_unset(net, &nfulnl_logger);
1139 for (i = 0; i < INSTANCE_BUCKETS; i++)
1140 WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1143 static struct pernet_operations nfnl_log_net_ops = {
1144 .init = nfnl_log_net_init,
1145 .exit = nfnl_log_net_exit,
1146 .id = &nfnl_log_net_id,
1147 .size = sizeof(struct nfnl_log_net),
1150 static int __init nfnetlink_log_init(void)
1154 status = register_pernet_subsys(&nfnl_log_net_ops);
1156 pr_err("failed to register pernet ops\n");
1160 netlink_register_notifier(&nfulnl_rtnl_notifier);
1161 status = nfnetlink_subsys_register(&nfulnl_subsys);
1163 pr_err("failed to create netlink socket\n");
1164 goto cleanup_netlink_notifier;
1167 status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1169 pr_err("failed to register logger\n");
1170 goto cleanup_subsys;
1176 nfnetlink_subsys_unregister(&nfulnl_subsys);
1177 cleanup_netlink_notifier:
1178 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1179 unregister_pernet_subsys(&nfnl_log_net_ops);
1184 static void __exit nfnetlink_log_fini(void)
1186 nfnetlink_subsys_unregister(&nfulnl_subsys);
1187 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1188 unregister_pernet_subsys(&nfnl_log_net_ops);
1189 nf_log_unregister(&nfulnl_logger);
1192 MODULE_DESCRIPTION("netfilter userspace logging");
1193 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1194 MODULE_LICENSE("GPL");
1195 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1196 MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1197 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1198 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1199 MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1200 MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1202 module_init(nfnetlink_log_init);
1203 module_exit(nfnetlink_log_fini);