1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * RAW - implementation of IP "raw" sockets.
10 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 * Alan Cox : verify_area() fixed up
14 * Alan Cox : ICMP error handling
15 * Alan Cox : EMSGSIZE if you send too big a packet
16 * Alan Cox : Now uses generic datagrams and shared
17 * skbuff library. No more peek crashes,
19 * Alan Cox : Checks sk->broadcast.
20 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
21 * Alan Cox : Raw passes ip options too
22 * Alan Cox : Setsocketopt added
23 * Alan Cox : Fixed error return for broadcasts
24 * Alan Cox : Removed wake_up calls
25 * Alan Cox : Use ttl/tos
26 * Alan Cox : Cleaned up old debugging
27 * Alan Cox : Use new kernel side addresses
28 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
29 * Alan Cox : BSD style RAW socket demultiplexing.
30 * Alan Cox : Beginnings of mrouted support.
31 * Alan Cox : Added IP_HDRINCL option.
32 * Alan Cox : Skip broadcast check if BSDism set.
33 * David S. Miller : New socket lookup architecture.
36 #include <linux/types.h>
37 #include <linux/atomic.h>
38 #include <asm/byteorder.h>
39 #include <asm/current.h>
40 #include <linux/uaccess.h>
41 #include <asm/ioctls.h>
42 #include <linux/stddef.h>
43 #include <linux/slab.h>
44 #include <linux/errno.h>
45 #include <linux/kernel.h>
46 #include <linux/export.h>
47 #include <linux/spinlock.h>
48 #include <linux/sockios.h>
49 #include <linux/socket.h>
51 #include <linux/mroute.h>
52 #include <linux/netdevice.h>
53 #include <linux/in_route.h>
54 #include <linux/route.h>
55 #include <linux/skbuff.h>
56 #include <linux/igmp.h>
57 #include <net/net_namespace.h>
61 #include <linux/net.h>
67 #include <net/tcp_states.h>
68 #include <net/inet_common.h>
69 #include <net/checksum.h>
71 #include <linux/rtnetlink.h>
72 #include <linux/proc_fs.h>
73 #include <linux/seq_file.h>
74 #include <linux/netfilter.h>
75 #include <linux/netfilter_ipv4.h>
76 #include <linux/compat.h>
77 #include <linux/uio.h>
88 struct raw_hashinfo raw_v4_hashinfo;
89 EXPORT_SYMBOL_GPL(raw_v4_hashinfo);
91 int raw_hash_sk(struct sock *sk)
93 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
94 struct hlist_head *hlist;
96 hlist = &h->ht[raw_hashfunc(sock_net(sk), inet_sk(sk)->inet_num)];
99 sk_add_node_rcu(sk, hlist);
100 sock_set_flag(sk, SOCK_RCU_FREE);
101 spin_unlock(&h->lock);
102 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
106 EXPORT_SYMBOL_GPL(raw_hash_sk);
108 void raw_unhash_sk(struct sock *sk)
110 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
113 if (sk_del_node_init_rcu(sk))
114 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
115 spin_unlock(&h->lock);
117 EXPORT_SYMBOL_GPL(raw_unhash_sk);
119 bool raw_v4_match(struct net *net, const struct sock *sk, unsigned short num,
120 __be32 raddr, __be32 laddr, int dif, int sdif)
122 const struct inet_sock *inet = inet_sk(sk);
124 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
125 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
126 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
127 raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif))
131 EXPORT_SYMBOL_GPL(raw_v4_match);
137 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
140 const struct icmphdr *hdr;
142 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
143 sizeof(_hdr), &_hdr);
147 if (hdr->type < 32) {
148 __u32 data = raw_sk(sk)->filter.data;
150 return ((1U << hdr->type) & data) != 0;
153 /* Do not block unknown ICMP types */
157 /* IP input processing comes here for RAW socket delivery.
158 * Caller owns SKB, so we must make clones.
160 * RFC 1122: SHOULD pass TOS value up to the transport layer.
161 * -> It does. And not only TOS, but all IP header.
163 static int raw_v4_input(struct net *net, struct sk_buff *skb,
164 const struct iphdr *iph, int hash)
166 int sdif = inet_sdif(skb);
167 struct hlist_head *hlist;
168 int dif = inet_iif(skb);
172 hlist = &raw_v4_hashinfo.ht[hash];
174 sk_for_each_rcu(sk, hlist) {
175 if (!raw_v4_match(net, sk, iph->protocol,
176 iph->saddr, iph->daddr, dif, sdif))
179 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
180 ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
181 skb->dev->ifindex, sdif)) {
182 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
184 /* Not releasing hash table! */
193 int raw_local_deliver(struct sk_buff *skb, int protocol)
195 struct net *net = dev_net(skb->dev);
197 return raw_v4_input(net, skb, ip_hdr(skb),
198 raw_hashfunc(net, protocol));
201 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
203 struct inet_sock *inet = inet_sk(sk);
204 const int type = icmp_hdr(skb)->type;
205 const int code = icmp_hdr(skb)->code;
210 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
211 ipv4_sk_update_pmtu(skb, sk, info);
212 else if (type == ICMP_REDIRECT) {
213 ipv4_sk_redirect(skb, sk);
217 /* Report error on raw socket, if:
218 1. User requested ip_recverr.
219 2. Socket is connected (otherwise the error indication
220 is useless without ip_recverr and error is hard.
222 recverr = inet_test_bit(RECVERR, sk);
223 if (!recverr && sk->sk_state != TCP_ESTABLISHED)
228 case ICMP_TIME_EXCEEDED:
231 case ICMP_SOURCE_QUENCH:
233 case ICMP_PARAMETERPROB:
237 case ICMP_DEST_UNREACH:
239 if (code > NR_ICMP_UNREACH)
241 if (code == ICMP_FRAG_NEEDED) {
242 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
245 err = icmp_err_convert[code].errno;
246 harderr = icmp_err_convert[code].fatal;
251 const struct iphdr *iph = (const struct iphdr *)skb->data;
252 u8 *payload = skb->data + (iph->ihl << 2);
254 if (inet_test_bit(HDRINCL, sk))
256 ip_icmp_error(sk, skb, err, 0, info, payload);
259 if (recverr || harderr) {
265 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
267 struct net *net = dev_net(skb->dev);
268 int dif = skb->dev->ifindex;
269 int sdif = inet_sdif(skb);
270 struct hlist_head *hlist;
271 const struct iphdr *iph;
275 hash = raw_hashfunc(net, protocol);
276 hlist = &raw_v4_hashinfo.ht[hash];
279 sk_for_each_rcu(sk, hlist) {
280 iph = (const struct iphdr *)skb->data;
281 if (!raw_v4_match(net, sk, iph->protocol,
282 iph->daddr, iph->saddr, dif, sdif))
284 raw_err(sk, skb, info);
289 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
291 enum skb_drop_reason reason;
293 /* Charge it to the socket. */
295 ipv4_pktinfo_prepare(sk, skb);
296 if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) {
297 kfree_skb_reason(skb, reason);
301 return NET_RX_SUCCESS;
304 int raw_rcv(struct sock *sk, struct sk_buff *skb)
306 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
307 atomic_inc(&sk->sk_drops);
308 kfree_skb_reason(skb, SKB_DROP_REASON_XFRM_POLICY);
313 skb_push(skb, skb->data - skb_network_header(skb));
315 raw_rcv_skb(sk, skb);
319 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
320 struct msghdr *msg, size_t length,
321 struct rtable **rtp, unsigned int flags,
322 const struct sockcm_cookie *sockc)
324 struct inet_sock *inet = inet_sk(sk);
325 struct net *net = sock_net(sk);
330 struct rtable *rt = *rtp;
333 if (length > rt->dst.dev->mtu) {
334 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
338 if (length < sizeof(struct iphdr))
344 hlen = LL_RESERVED_SPACE(rt->dst.dev);
345 tlen = rt->dst.dev->needed_tailroom;
346 skb = sock_alloc_send_skb(sk,
347 length + hlen + tlen + 15,
348 flags & MSG_DONTWAIT, &err);
351 skb_reserve(skb, hlen);
353 skb->priority = READ_ONCE(sk->sk_priority);
354 skb->mark = sockc->mark;
355 skb->tstamp = sockc->transmit_time;
356 skb_dst_set(skb, &rt->dst);
359 skb_reset_network_header(skb);
361 skb_put(skb, length);
363 skb->ip_summed = CHECKSUM_NONE;
365 skb_setup_tx_timestamp(skb, sockc->tsflags);
367 if (flags & MSG_CONFIRM)
368 skb_set_dst_pending_confirm(skb, 1);
370 skb->transport_header = skb->network_header;
372 if (memcpy_from_msg(iph, msg, length))
375 iphlen = iph->ihl * 4;
378 * We don't want to modify the ip header, but we do need to
379 * be sure that it won't cause problems later along the network
380 * stack. Specifically we want to make sure that iph->ihl is a
381 * sane value. If ihl points beyond the length of the buffer passed
382 * in, reject the frame as invalid
388 if (iphlen >= sizeof(*iph)) {
390 iph->saddr = fl4->saddr;
392 iph->tot_len = htons(length);
394 ip_select_ident(net, skb, NULL);
396 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
397 skb->transport_header += iphlen;
398 if (iph->protocol == IPPROTO_ICMP &&
399 length >= iphlen + sizeof(struct icmphdr))
400 icmp_out_count(net, ((struct icmphdr *)
401 skb_transport_header(skb))->type);
404 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
405 net, sk, skb, NULL, rt->dst.dev,
408 err = net_xmit_errno(err);
417 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
418 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk))
423 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
427 if (fl4->flowi4_proto != IPPROTO_ICMP)
430 /* We only need the first two bytes. */
433 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
437 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
438 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
443 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
446 struct raw_frag_vec *rfv = from;
448 if (offset < rfv->hlen) {
449 int copy = min(rfv->hlen - offset, len);
451 if (skb->ip_summed == CHECKSUM_PARTIAL)
452 memcpy(to, rfv->hdr.c + offset, copy);
454 skb->csum = csum_block_add(
456 csum_partial_copy_nocheck(rfv->hdr.c + offset,
471 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
474 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
476 struct inet_sock *inet = inet_sk(sk);
477 struct net *net = sock_net(sk);
478 struct ipcm_cookie ipc;
479 struct rtable *rt = NULL;
486 struct ip_options_data opt_copy;
487 struct raw_frag_vec rfv;
494 hdrincl = inet_test_bit(HDRINCL, sk);
501 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
502 goto out; /* compatibility */
505 * Get and verify the address.
508 if (msg->msg_namelen) {
509 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
511 if (msg->msg_namelen < sizeof(*usin))
513 if (usin->sin_family != AF_INET) {
514 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
515 __func__, current->comm);
517 if (usin->sin_family)
520 daddr = usin->sin_addr.s_addr;
521 /* ANK: I did not forget to get protocol from port field.
522 * I just do not know, who uses this weirdness.
523 * IP_HDRINCL is much more convenient.
527 if (sk->sk_state != TCP_ESTABLISHED)
529 daddr = inet->inet_daddr;
532 ipcm_init_sk(&ipc, inet);
533 /* Keep backward compat */
535 ipc.protocol = IPPROTO_RAW;
537 if (msg->msg_controllen) {
538 err = ip_cmsg_send(sk, msg, &ipc, false);
551 struct ip_options_rcu *inet_opt;
554 inet_opt = rcu_dereference(inet->inet_opt);
556 memcpy(&opt_copy, inet_opt,
557 sizeof(*inet_opt) + inet_opt->opt.optlen);
558 ipc.opt = &opt_copy.opt;
565 /* Linux does not mangle headers on raw sockets,
566 * so that IP options + IP_HDRINCL is non-sense.
570 if (ipc.opt->opt.srr) {
573 daddr = ipc.opt->opt.faddr;
576 tos = get_rttos(&ipc, inet);
577 scope = ip_sendmsg_scope(inet, &ipc, msg);
579 if (ipv4_is_multicast(daddr)) {
580 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
581 ipc.oif = inet->mc_index;
583 saddr = inet->mc_addr;
584 } else if (!ipc.oif) {
585 ipc.oif = inet->uc_index;
586 } else if (ipv4_is_lbcast(daddr) && inet->uc_index) {
587 /* oif is set, packet is to local broadcast
588 * and uc_index is set. oif is most likely set
589 * by sk_bound_dev_if. If uc_index != oif check if the
590 * oif is an L3 master and uc_index is an L3 slave.
591 * If so, we want to allow the send using the uc_index.
593 if (ipc.oif != inet->uc_index &&
594 ipc.oif == l3mdev_master_ifindex_by_index(sock_net(sk),
596 ipc.oif = inet->uc_index;
600 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos, scope,
601 hdrincl ? ipc.protocol : sk->sk_protocol,
602 inet_sk_flowi_flags(sk) |
603 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
604 daddr, saddr, 0, 0, sk->sk_uid);
610 err = raw_probe_proto_opt(&rfv, &fl4);
615 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
616 rt = ip_route_output_flow(net, &fl4, sk);
624 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
627 if (msg->msg_flags & MSG_CONFIRM)
632 err = raw_send_hdrinc(sk, &fl4, msg, len,
633 &rt, msg->msg_flags, &ipc.sockc);
637 ipc.addr = fl4.daddr;
639 err = ip_append_data(sk, &fl4, raw_getfrag,
641 &ipc, &rt, msg->msg_flags);
643 ip_flush_pending_frames(sk);
644 else if (!(msg->msg_flags & MSG_MORE)) {
645 err = ip_push_pending_frames(sk, &fl4);
646 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk))
662 if (msg->msg_flags & MSG_PROBE)
663 dst_confirm_neigh(&rt->dst, &fl4.daddr);
664 if (!(msg->msg_flags & MSG_PROBE) || len)
665 goto back_from_confirm;
670 static void raw_close(struct sock *sk, long timeout)
673 * Raw sockets may have direct kernel references. Kill them.
675 ip_ra_control(sk, 0, NULL);
677 sk_common_release(sk);
680 static void raw_destroy(struct sock *sk)
683 ip_flush_pending_frames(sk);
687 /* This gets rid of all the nasties in af_inet. -DaveM */
688 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
690 struct inet_sock *inet = inet_sk(sk);
691 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
692 struct net *net = sock_net(sk);
693 u32 tb_id = RT_TABLE_LOCAL;
698 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
701 if (sk->sk_bound_dev_if)
702 tb_id = l3mdev_fib_table_by_index(net,
703 sk->sk_bound_dev_if) ? : tb_id;
705 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
707 ret = -EADDRNOTAVAIL;
708 if (!inet_addr_valid_or_nonlocal(net, inet, addr->sin_addr.s_addr,
712 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
713 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
714 inet->inet_saddr = 0; /* Use device */
723 * This should be easy, if there is something there
724 * we return it, otherwise we block.
727 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
728 int flags, int *addr_len)
730 struct inet_sock *inet = inet_sk(sk);
732 int err = -EOPNOTSUPP;
733 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
739 if (flags & MSG_ERRQUEUE) {
740 err = ip_recv_error(sk, msg, len, addr_len);
744 skb = skb_recv_datagram(sk, flags, &err);
750 msg->msg_flags |= MSG_TRUNC;
754 err = skb_copy_datagram_msg(skb, 0, msg, copied);
758 sock_recv_cmsgs(msg, sk, skb);
760 /* Copy the address. */
762 sin->sin_family = AF_INET;
763 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
765 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
766 *addr_len = sizeof(*sin);
768 if (inet_cmsg_flags(inet))
769 ip_cmsg_recv(msg, skb);
770 if (flags & MSG_TRUNC)
773 skb_free_datagram(sk, skb);
780 static int raw_sk_init(struct sock *sk)
782 struct raw_sock *rp = raw_sk(sk);
784 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
785 memset(&rp->filter, 0, sizeof(rp->filter));
789 static int raw_seticmpfilter(struct sock *sk, sockptr_t optval, int optlen)
791 if (optlen > sizeof(struct icmp_filter))
792 optlen = sizeof(struct icmp_filter);
793 if (copy_from_sockptr(&raw_sk(sk)->filter, optval, optlen))
798 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
800 int len, ret = -EFAULT;
802 if (get_user(len, optlen))
807 if (len > sizeof(struct icmp_filter))
808 len = sizeof(struct icmp_filter);
810 if (put_user(len, optlen) ||
811 copy_to_user(optval, &raw_sk(sk)->filter, len))
817 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
818 sockptr_t optval, unsigned int optlen)
820 if (optname == ICMP_FILTER) {
821 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
824 return raw_seticmpfilter(sk, optval, optlen);
829 static int raw_setsockopt(struct sock *sk, int level, int optname,
830 sockptr_t optval, unsigned int optlen)
832 if (level != SOL_RAW)
833 return ip_setsockopt(sk, level, optname, optval, optlen);
834 return do_raw_setsockopt(sk, level, optname, optval, optlen);
837 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
838 char __user *optval, int __user *optlen)
840 if (optname == ICMP_FILTER) {
841 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
844 return raw_geticmpfilter(sk, optval, optlen);
849 static int raw_getsockopt(struct sock *sk, int level, int optname,
850 char __user *optval, int __user *optlen)
852 if (level != SOL_RAW)
853 return ip_getsockopt(sk, level, optname, optval, optlen);
854 return do_raw_getsockopt(sk, level, optname, optval, optlen);
857 static int raw_ioctl(struct sock *sk, int cmd, int *karg)
861 *karg = sk_wmem_alloc_get(sk);
867 spin_lock_bh(&sk->sk_receive_queue.lock);
868 skb = skb_peek(&sk->sk_receive_queue);
873 spin_unlock_bh(&sk->sk_receive_queue.lock);
878 #ifdef CONFIG_IP_MROUTE
879 return ipmr_ioctl(sk, cmd, karg);
887 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
894 #ifdef CONFIG_IP_MROUTE
895 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
903 int raw_abort(struct sock *sk, int err)
909 __udp_disconnect(sk, 0);
915 EXPORT_SYMBOL_GPL(raw_abort);
917 struct proto raw_prot = {
919 .owner = THIS_MODULE,
921 .destroy = raw_destroy,
922 .connect = ip4_datagram_connect,
923 .disconnect = __udp_disconnect,
926 .setsockopt = raw_setsockopt,
927 .getsockopt = raw_getsockopt,
928 .sendmsg = raw_sendmsg,
929 .recvmsg = raw_recvmsg,
931 .backlog_rcv = raw_rcv_skb,
932 .release_cb = ip4_datagram_release_cb,
934 .unhash = raw_unhash_sk,
935 .obj_size = sizeof(struct raw_sock),
936 .useroffset = offsetof(struct raw_sock, filter),
937 .usersize = sizeof_field(struct raw_sock, filter),
938 .h.raw_hash = &raw_v4_hashinfo,
940 .compat_ioctl = compat_raw_ioctl,
942 .diag_destroy = raw_abort,
945 #ifdef CONFIG_PROC_FS
946 static struct sock *raw_get_first(struct seq_file *seq, int bucket)
948 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
949 struct raw_iter_state *state = raw_seq_private(seq);
950 struct hlist_head *hlist;
953 for (state->bucket = bucket; state->bucket < RAW_HTABLE_SIZE;
955 hlist = &h->ht[state->bucket];
956 sk_for_each(sk, hlist) {
957 if (sock_net(sk) == seq_file_net(seq))
964 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
966 struct raw_iter_state *state = raw_seq_private(seq);
970 } while (sk && sock_net(sk) != seq_file_net(seq));
973 return raw_get_first(seq, state->bucket + 1);
977 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
979 struct sock *sk = raw_get_first(seq, 0);
982 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
984 return pos ? NULL : sk;
987 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
990 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
994 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
996 EXPORT_SYMBOL_GPL(raw_seq_start);
998 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1002 if (v == SEQ_START_TOKEN)
1003 sk = raw_get_first(seq, 0);
1005 sk = raw_get_next(seq, v);
1009 EXPORT_SYMBOL_GPL(raw_seq_next);
1011 void raw_seq_stop(struct seq_file *seq, void *v)
1012 __releases(&h->lock)
1014 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
1016 spin_unlock(&h->lock);
1018 EXPORT_SYMBOL_GPL(raw_seq_stop);
1020 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1022 struct inet_sock *inet = inet_sk(sp);
1023 __be32 dest = inet->inet_daddr,
1024 src = inet->inet_rcv_saddr;
1026 srcp = inet->inet_num;
1028 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1029 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1030 i, src, srcp, dest, destp, sp->sk_state,
1031 sk_wmem_alloc_get(sp),
1032 sk_rmem_alloc_get(sp),
1034 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1036 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1039 static int raw_seq_show(struct seq_file *seq, void *v)
1041 if (v == SEQ_START_TOKEN)
1042 seq_printf(seq, " sl local_address rem_address st tx_queue "
1043 "rx_queue tr tm->when retrnsmt uid timeout "
1044 "inode ref pointer drops\n");
1046 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1050 static const struct seq_operations raw_seq_ops = {
1051 .start = raw_seq_start,
1052 .next = raw_seq_next,
1053 .stop = raw_seq_stop,
1054 .show = raw_seq_show,
1057 static __net_init int raw_init_net(struct net *net)
1059 if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops,
1060 sizeof(struct raw_iter_state), &raw_v4_hashinfo))
1066 static __net_exit void raw_exit_net(struct net *net)
1068 remove_proc_entry("raw", net->proc_net);
1071 static __net_initdata struct pernet_operations raw_net_ops = {
1072 .init = raw_init_net,
1073 .exit = raw_exit_net,
1076 int __init raw_proc_init(void)
1079 return register_pernet_subsys(&raw_net_ops);
1082 void __init raw_proc_exit(void)
1084 unregister_pernet_subsys(&raw_net_ops);
1086 #endif /* CONFIG_PROC_FS */
1088 static void raw_sysctl_init_net(struct net *net)
1090 #ifdef CONFIG_NET_L3_MASTER_DEV
1091 net->ipv4.sysctl_raw_l3mdev_accept = 1;
1095 static int __net_init raw_sysctl_init(struct net *net)
1097 raw_sysctl_init_net(net);
1101 static struct pernet_operations __net_initdata raw_sysctl_ops = {
1102 .init = raw_sysctl_init,
1105 void __init raw_init(void)
1107 raw_sysctl_init_net(&init_net);
1108 if (register_pernet_subsys(&raw_sysctl_ops))
1109 panic("RAW: failed to init sysctl parameters.\n");