2 * inet_diag.c Module for monitoring INET transport protocols sockets.
4 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
12 #include <linux/kernel.h>
13 #include <linux/module.h>
14 #include <linux/types.h>
15 #include <linux/fcntl.h>
16 #include <linux/random.h>
17 #include <linux/slab.h>
18 #include <linux/cache.h>
19 #include <linux/init.h>
20 #include <linux/time.h>
25 #include <net/inet_common.h>
26 #include <net/inet_connection_sock.h>
27 #include <net/inet_hashtables.h>
28 #include <net/inet_timewait_sock.h>
29 #include <net/inet6_hashtables.h>
30 #include <net/netlink.h>
32 #include <linux/inet.h>
33 #include <linux/stddef.h>
35 #include <linux/inet_diag.h>
36 #include <linux/sock_diag.h>
38 static const struct inet_diag_handler **inet_diag_table;
40 struct inet_diag_entry {
51 static DEFINE_MUTEX(inet_diag_table_mutex);
53 static const struct inet_diag_handler *inet_diag_lock_handler(int proto)
55 if (!inet_diag_table[proto])
56 request_module("net-pf-%d-proto-%d-type-%d-%d", PF_NETLINK,
57 NETLINK_SOCK_DIAG, AF_INET, proto);
59 mutex_lock(&inet_diag_table_mutex);
60 if (!inet_diag_table[proto])
61 return ERR_PTR(-ENOENT);
63 return inet_diag_table[proto];
66 static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
68 mutex_unlock(&inet_diag_table_mutex);
71 void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk)
73 r->idiag_family = sk->sk_family;
75 r->id.idiag_sport = htons(sk->sk_num);
76 r->id.idiag_dport = sk->sk_dport;
77 r->id.idiag_if = sk->sk_bound_dev_if;
78 sock_diag_save_cookie(sk, r->id.idiag_cookie);
80 #if IS_ENABLED(CONFIG_IPV6)
81 if (sk->sk_family == AF_INET6) {
82 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr;
83 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr;
87 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
88 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
90 r->id.idiag_src[0] = sk->sk_rcv_saddr;
91 r->id.idiag_dst[0] = sk->sk_daddr;
94 EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill);
96 static size_t inet_sk_attr_size(struct sock *sk,
97 const struct inet_diag_req_v2 *req,
100 const struct inet_diag_handler *handler;
103 handler = inet_diag_table[req->sdiag_protocol];
104 if (handler && handler->idiag_get_aux_size)
105 aux = handler->idiag_get_aux_size(sk, net_admin);
107 return nla_total_size(sizeof(struct tcp_info))
108 + nla_total_size(sizeof(struct inet_diag_msg))
109 + inet_diag_msg_attrs_size()
110 + nla_total_size(sizeof(struct inet_diag_meminfo))
111 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
112 + nla_total_size(TCP_CA_NAME_MAX)
113 + nla_total_size(sizeof(struct tcpvegas_info))
118 int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb,
119 struct inet_diag_msg *r, int ext,
120 struct user_namespace *user_ns,
123 const struct inet_sock *inet = inet_sk(sk);
125 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
128 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
129 * hence this needs to be included regardless of socket family.
131 if (ext & (1 << (INET_DIAG_TOS - 1)))
132 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0)
135 #if IS_ENABLED(CONFIG_IPV6)
136 if (r->idiag_family == AF_INET6) {
137 if (ext & (1 << (INET_DIAG_TCLASS - 1)))
138 if (nla_put_u8(skb, INET_DIAG_TCLASS,
139 inet6_sk(sk)->tclass) < 0)
142 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
143 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
148 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, sk->sk_mark))
151 if (ext & (1 << (INET_DIAG_CLASS_ID - 1)) ||
152 ext & (1 << (INET_DIAG_TCLASS - 1))) {
155 #ifdef CONFIG_SOCK_CGROUP_DATA
156 classid = sock_cgroup_classid(&sk->sk_cgrp_data);
158 /* Fallback to socket priority if class id isn't set.
159 * Classful qdiscs use it as direct reference to class.
160 * For cgroup2 classid is always zero.
163 classid = sk->sk_priority;
165 if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid))
169 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
170 r->idiag_inode = sock_i_ino(sk);
176 EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill);
178 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
179 struct sk_buff *skb, const struct inet_diag_req_v2 *req,
180 struct user_namespace *user_ns,
181 u32 portid, u32 seq, u16 nlmsg_flags,
182 const struct nlmsghdr *unlh,
185 const struct tcp_congestion_ops *ca_ops;
186 const struct inet_diag_handler *handler;
187 int ext = req->idiag_ext;
188 struct inet_diag_msg *r;
189 struct nlmsghdr *nlh;
193 handler = inet_diag_table[req->sdiag_protocol];
196 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
202 BUG_ON(!sk_fullsock(sk));
204 inet_diag_msg_common_fill(r, sk);
205 r->idiag_state = sk->sk_state;
207 r->idiag_retrans = 0;
209 if (inet_diag_msg_attrs_fill(sk, skb, r, ext, user_ns, net_admin))
212 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
213 struct inet_diag_meminfo minfo = {
214 .idiag_rmem = sk_rmem_alloc_get(sk),
215 .idiag_wmem = sk->sk_wmem_queued,
216 .idiag_fmem = sk->sk_forward_alloc,
217 .idiag_tmem = sk_wmem_alloc_get(sk),
220 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
224 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
225 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
229 * RAW sockets might have user-defined protocols assigned,
230 * so report the one supplied on socket creation.
232 if (sk->sk_type == SOCK_RAW) {
233 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
238 handler->idiag_get_info(sk, r, NULL);
242 if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
243 icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
244 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
246 r->idiag_retrans = icsk->icsk_retransmits;
248 jiffies_to_msecs(icsk->icsk_timeout - jiffies);
249 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
251 r->idiag_retrans = icsk->icsk_probes_out;
253 jiffies_to_msecs(icsk->icsk_timeout - jiffies);
254 } else if (timer_pending(&sk->sk_timer)) {
256 r->idiag_retrans = icsk->icsk_probes_out;
258 jiffies_to_msecs(sk->sk_timer.expires - jiffies);
261 r->idiag_expires = 0;
264 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
265 attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
266 handler->idiag_info_size,
271 info = nla_data(attr);
274 if (ext & (1 << (INET_DIAG_CONG - 1))) {
278 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
280 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
286 handler->idiag_get_info(sk, r, info);
288 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
289 if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
292 if (sk->sk_state < TCP_TIME_WAIT) {
293 union tcp_cc_info info;
298 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
299 if (ca_ops && ca_ops->get_info)
300 sz = ca_ops->get_info(sk, ext, &attr, &info);
302 if (sz && nla_put(skb, attr, sz, &info) < 0)
311 nlmsg_cancel(skb, nlh);
314 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
316 static int inet_csk_diag_fill(struct sock *sk,
318 const struct inet_diag_req_v2 *req,
319 struct user_namespace *user_ns,
320 u32 portid, u32 seq, u16 nlmsg_flags,
321 const struct nlmsghdr *unlh,
324 return inet_sk_diag_fill(sk, inet_csk(sk), skb, req, user_ns,
325 portid, seq, nlmsg_flags, unlh, net_admin);
328 static int inet_twsk_diag_fill(struct sock *sk,
330 u32 portid, u32 seq, u16 nlmsg_flags,
331 const struct nlmsghdr *unlh)
333 struct inet_timewait_sock *tw = inet_twsk(sk);
334 struct inet_diag_msg *r;
335 struct nlmsghdr *nlh;
338 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
344 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
346 tmo = tw->tw_timer.expires - jiffies;
350 inet_diag_msg_common_fill(r, sk);
351 r->idiag_retrans = 0;
353 r->idiag_state = tw->tw_substate;
355 r->idiag_expires = jiffies_to_msecs(tmo);
365 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
366 u32 portid, u32 seq, u16 nlmsg_flags,
367 const struct nlmsghdr *unlh, bool net_admin)
369 struct request_sock *reqsk = inet_reqsk(sk);
370 struct inet_diag_msg *r;
371 struct nlmsghdr *nlh;
374 nlh = nlmsg_put(skb, portid, seq, unlh->nlmsg_type, sizeof(*r),
380 inet_diag_msg_common_fill(r, sk);
381 r->idiag_state = TCP_SYN_RECV;
383 r->idiag_retrans = reqsk->num_retrans;
385 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
386 offsetof(struct sock, sk_cookie));
388 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
389 r->idiag_expires = (tmo >= 0) ? jiffies_to_msecs(tmo) : 0;
395 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
396 inet_rsk(reqsk)->ir_mark)) {
397 nlmsg_cancel(skb, nlh);
405 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
406 const struct inet_diag_req_v2 *r,
407 struct user_namespace *user_ns,
408 u32 portid, u32 seq, u16 nlmsg_flags,
409 const struct nlmsghdr *unlh, bool net_admin)
411 if (sk->sk_state == TCP_TIME_WAIT)
412 return inet_twsk_diag_fill(sk, skb, portid, seq,
415 if (sk->sk_state == TCP_NEW_SYN_RECV)
416 return inet_req_diag_fill(sk, skb, portid, seq,
417 nlmsg_flags, unlh, net_admin);
419 return inet_csk_diag_fill(sk, skb, r, user_ns, portid, seq,
420 nlmsg_flags, unlh, net_admin);
423 struct sock *inet_diag_find_one_icsk(struct net *net,
424 struct inet_hashinfo *hashinfo,
425 const struct inet_diag_req_v2 *req)
430 if (req->sdiag_family == AF_INET)
431 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
432 req->id.idiag_dport, req->id.idiag_src[0],
433 req->id.idiag_sport, req->id.idiag_if);
434 #if IS_ENABLED(CONFIG_IPV6)
435 else if (req->sdiag_family == AF_INET6) {
436 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
437 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
438 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
439 req->id.idiag_dport, req->id.idiag_src[3],
440 req->id.idiag_sport, req->id.idiag_if);
442 sk = inet6_lookup(net, hashinfo, NULL, 0,
443 (struct in6_addr *)req->id.idiag_dst,
445 (struct in6_addr *)req->id.idiag_src,
452 return ERR_PTR(-EINVAL);
456 return ERR_PTR(-ENOENT);
458 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
460 return ERR_PTR(-ENOENT);
465 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
467 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
468 struct sk_buff *in_skb,
469 const struct nlmsghdr *nlh,
470 const struct inet_diag_req_v2 *req)
472 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
473 struct net *net = sock_net(in_skb->sk);
478 sk = inet_diag_find_one_icsk(net, hashinfo, req);
482 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
488 err = sk_diag_fill(sk, rep, req,
489 sk_user_ns(NETLINK_CB(in_skb).sk),
490 NETLINK_CB(in_skb).portid,
491 nlh->nlmsg_seq, 0, nlh, net_admin);
493 WARN_ON(err == -EMSGSIZE);
497 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
508 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
510 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
511 const struct nlmsghdr *nlh,
512 const struct inet_diag_req_v2 *req)
514 const struct inet_diag_handler *handler;
517 handler = inet_diag_lock_handler(req->sdiag_protocol);
519 err = PTR_ERR(handler);
520 else if (cmd == SOCK_DIAG_BY_FAMILY)
521 err = handler->dump_one(in_skb, nlh, req);
522 else if (cmd == SOCK_DESTROY && handler->destroy)
523 err = handler->destroy(in_skb, req);
526 inet_diag_unlock_handler(handler);
531 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
533 int words = bits >> 5;
538 if (memcmp(a1, a2, words << 2))
548 mask = htonl((0xffffffff) << (32 - bits));
550 if ((w1 ^ w2) & mask)
557 static int inet_diag_bc_run(const struct nlattr *_bc,
558 const struct inet_diag_entry *entry)
560 const void *bc = nla_data(_bc);
561 int len = nla_len(_bc);
565 const struct inet_diag_bc_op *op = bc;
568 case INET_DIAG_BC_NOP:
570 case INET_DIAG_BC_JMP:
573 case INET_DIAG_BC_S_GE:
574 yes = entry->sport >= op[1].no;
576 case INET_DIAG_BC_S_LE:
577 yes = entry->sport <= op[1].no;
579 case INET_DIAG_BC_D_GE:
580 yes = entry->dport >= op[1].no;
582 case INET_DIAG_BC_D_LE:
583 yes = entry->dport <= op[1].no;
585 case INET_DIAG_BC_AUTO:
586 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
588 case INET_DIAG_BC_S_COND:
589 case INET_DIAG_BC_D_COND: {
590 const struct inet_diag_hostcond *cond;
593 cond = (const struct inet_diag_hostcond *)(op + 1);
594 if (cond->port != -1 &&
595 cond->port != (op->code == INET_DIAG_BC_S_COND ?
596 entry->sport : entry->dport)) {
601 if (op->code == INET_DIAG_BC_S_COND)
606 if (cond->family != AF_UNSPEC &&
607 cond->family != entry->family) {
608 if (entry->family == AF_INET6 &&
609 cond->family == AF_INET) {
610 if (addr[0] == 0 && addr[1] == 0 &&
611 addr[2] == htonl(0xffff) &&
612 bitstring_match(addr + 3,
621 if (cond->prefix_len == 0)
623 if (bitstring_match(addr, cond->addr,
629 case INET_DIAG_BC_DEV_COND: {
632 ifindex = *((const u32 *)(op + 1));
633 if (ifindex != entry->ifindex)
637 case INET_DIAG_BC_MARK_COND: {
638 struct inet_diag_markcond *cond;
640 cond = (struct inet_diag_markcond *)(op + 1);
641 if ((entry->mark & cond->mask) != cond->mark)
658 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
660 static void entry_fill_addrs(struct inet_diag_entry *entry,
661 const struct sock *sk)
663 #if IS_ENABLED(CONFIG_IPV6)
664 if (sk->sk_family == AF_INET6) {
665 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
666 entry->daddr = sk->sk_v6_daddr.s6_addr32;
670 entry->saddr = &sk->sk_rcv_saddr;
671 entry->daddr = &sk->sk_daddr;
675 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
677 struct inet_sock *inet = inet_sk(sk);
678 struct inet_diag_entry entry;
683 entry.family = sk->sk_family;
684 entry_fill_addrs(&entry, sk);
685 entry.sport = inet->inet_num;
686 entry.dport = ntohs(inet->inet_dport);
687 entry.ifindex = sk->sk_bound_dev_if;
688 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
690 entry.mark = sk->sk_mark;
691 else if (sk->sk_state == TCP_NEW_SYN_RECV)
692 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
696 return inet_diag_bc_run(bc, &entry);
698 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
700 static int valid_cc(const void *bc, int len, int cc)
703 const struct inet_diag_bc_op *op = bc;
709 if (op->yes < 4 || op->yes & 3)
717 /* data is u32 ifindex */
718 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
721 /* Check ifindex space. */
722 *min_len += sizeof(u32);
728 /* Validate an inet_diag_hostcond. */
729 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
732 struct inet_diag_hostcond *cond;
735 /* Check hostcond space. */
736 *min_len += sizeof(struct inet_diag_hostcond);
739 cond = (struct inet_diag_hostcond *)(op + 1);
741 /* Check address family and address length. */
742 switch (cond->family) {
747 addr_len = sizeof(struct in_addr);
750 addr_len = sizeof(struct in6_addr);
755 *min_len += addr_len;
759 /* Check prefix length (in bits) vs address length (in bytes). */
760 if (cond->prefix_len > 8 * addr_len)
766 /* Validate a port comparison operator. */
767 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
768 int len, int *min_len)
770 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
771 *min_len += sizeof(struct inet_diag_bc_op);
777 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
780 *min_len += sizeof(struct inet_diag_markcond);
781 return len >= *min_len;
784 static int inet_diag_bc_audit(const struct nlattr *attr,
785 const struct sk_buff *skb)
787 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
788 const void *bytecode, *bc;
789 int bytecode_len, len;
791 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
794 bytecode = bc = nla_data(attr);
795 len = bytecode_len = nla_len(attr);
798 int min_len = sizeof(struct inet_diag_bc_op);
799 const struct inet_diag_bc_op *op = bc;
802 case INET_DIAG_BC_S_COND:
803 case INET_DIAG_BC_D_COND:
804 if (!valid_hostcond(bc, len, &min_len))
807 case INET_DIAG_BC_DEV_COND:
808 if (!valid_devcond(bc, len, &min_len))
811 case INET_DIAG_BC_S_GE:
812 case INET_DIAG_BC_S_LE:
813 case INET_DIAG_BC_D_GE:
814 case INET_DIAG_BC_D_LE:
815 if (!valid_port_comparison(bc, len, &min_len))
818 case INET_DIAG_BC_MARK_COND:
821 if (!valid_markcond(bc, len, &min_len))
824 case INET_DIAG_BC_AUTO:
825 case INET_DIAG_BC_JMP:
826 case INET_DIAG_BC_NOP:
832 if (op->code != INET_DIAG_BC_NOP) {
833 if (op->no < min_len || op->no > len + 4 || op->no & 3)
836 !valid_cc(bytecode, bytecode_len, len - op->no))
840 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
845 return len == 0 ? 0 : -EINVAL;
848 static int inet_csk_diag_dump(struct sock *sk,
850 struct netlink_callback *cb,
851 const struct inet_diag_req_v2 *r,
852 const struct nlattr *bc,
855 if (!inet_diag_bc_sk(bc, sk))
858 return inet_csk_diag_fill(sk, skb, r,
859 sk_user_ns(NETLINK_CB(cb->skb).sk),
860 NETLINK_CB(cb->skb).portid,
861 cb->nlh->nlmsg_seq, NLM_F_MULTI, cb->nlh,
865 static void twsk_build_assert(void)
867 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
868 offsetof(struct sock, sk_family));
870 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
871 offsetof(struct inet_sock, inet_num));
873 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
874 offsetof(struct inet_sock, inet_dport));
876 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
877 offsetof(struct inet_sock, inet_rcv_saddr));
879 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
880 offsetof(struct inet_sock, inet_daddr));
882 #if IS_ENABLED(CONFIG_IPV6)
883 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
884 offsetof(struct sock, sk_v6_rcv_saddr));
886 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
887 offsetof(struct sock, sk_v6_daddr));
891 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
892 struct netlink_callback *cb,
893 const struct inet_diag_req_v2 *r, struct nlattr *bc)
895 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
896 struct net *net = sock_net(skb->sk);
897 u32 idiag_states = r->idiag_states;
898 int i, num, s_i, s_num;
901 if (idiag_states & TCPF_SYN_RECV)
902 idiag_states |= TCPF_NEW_SYN_RECV;
904 s_num = num = cb->args[2];
906 if (cb->args[0] == 0) {
907 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
910 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
911 struct inet_listen_hashbucket *ilb;
912 struct hlist_nulls_node *node;
915 ilb = &hashinfo->listening_hash[i];
916 spin_lock(&ilb->lock);
917 sk_nulls_for_each(sk, node, &ilb->nulls_head) {
918 struct inet_sock *inet = inet_sk(sk);
920 if (!net_eq(sock_net(sk), net))
928 if (r->sdiag_family != AF_UNSPEC &&
929 sk->sk_family != r->sdiag_family)
932 if (r->id.idiag_sport != inet->inet_sport &&
936 if (inet_csk_diag_dump(sk, skb, cb, r,
937 bc, net_admin) < 0) {
938 spin_unlock(&ilb->lock);
945 spin_unlock(&ilb->lock);
951 s_i = num = s_num = 0;
954 if (!(idiag_states & ~TCPF_LISTEN))
958 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
959 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
960 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
961 struct hlist_nulls_node *node;
962 struct sock *sk_arr[SKARR_SZ];
963 int num_arr[SKARR_SZ];
966 if (hlist_nulls_empty(&head->chain))
976 sk_nulls_for_each(sk, node, &head->chain) {
979 if (!net_eq(sock_net(sk), net))
983 state = (sk->sk_state == TCP_TIME_WAIT) ?
984 inet_twsk(sk)->tw_substate : sk->sk_state;
985 if (!(idiag_states & (1 << state)))
987 if (r->sdiag_family != AF_UNSPEC &&
988 sk->sk_family != r->sdiag_family)
990 if (r->id.idiag_sport != htons(sk->sk_num) &&
993 if (r->id.idiag_dport != sk->sk_dport &&
998 if (!inet_diag_bc_sk(bc, sk))
1001 if (!refcount_inc_not_zero(&sk->sk_refcnt))
1004 num_arr[accum] = num;
1006 if (++accum == SKARR_SZ)
1011 spin_unlock_bh(lock);
1013 for (idx = 0; idx < accum; idx++) {
1015 res = sk_diag_fill(sk_arr[idx], skb, r,
1016 sk_user_ns(NETLINK_CB(cb->skb).sk),
1017 NETLINK_CB(cb->skb).portid,
1018 cb->nlh->nlmsg_seq, NLM_F_MULTI,
1019 cb->nlh, net_admin);
1023 sock_gen_put(sk_arr[idx]);
1028 if (accum == SKARR_SZ) {
1040 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1042 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1043 const struct inet_diag_req_v2 *r,
1046 const struct inet_diag_handler *handler;
1049 handler = inet_diag_lock_handler(r->sdiag_protocol);
1050 if (!IS_ERR(handler))
1051 handler->dump(skb, cb, r, bc);
1053 err = PTR_ERR(handler);
1054 inet_diag_unlock_handler(handler);
1056 return err ? : skb->len;
1059 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1061 int hdrlen = sizeof(struct inet_diag_req_v2);
1062 struct nlattr *bc = NULL;
1064 if (nlmsg_attrlen(cb->nlh, hdrlen))
1065 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1067 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh), bc);
1070 static int inet_diag_type2proto(int type)
1073 case TCPDIAG_GETSOCK:
1075 case DCCPDIAG_GETSOCK:
1076 return IPPROTO_DCCP;
1082 static int inet_diag_dump_compat(struct sk_buff *skb,
1083 struct netlink_callback *cb)
1085 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1086 int hdrlen = sizeof(struct inet_diag_req);
1087 struct inet_diag_req_v2 req;
1088 struct nlattr *bc = NULL;
1090 req.sdiag_family = AF_UNSPEC; /* compatibility */
1091 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1092 req.idiag_ext = rc->idiag_ext;
1093 req.idiag_states = rc->idiag_states;
1096 if (nlmsg_attrlen(cb->nlh, hdrlen))
1097 bc = nlmsg_find_attr(cb->nlh, hdrlen, INET_DIAG_REQ_BYTECODE);
1099 return __inet_diag_dump(skb, cb, &req, bc);
1102 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1103 const struct nlmsghdr *nlh)
1105 struct inet_diag_req *rc = nlmsg_data(nlh);
1106 struct inet_diag_req_v2 req;
1108 req.sdiag_family = rc->idiag_family;
1109 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1110 req.idiag_ext = rc->idiag_ext;
1111 req.idiag_states = rc->idiag_states;
1114 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh, &req);
1117 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1119 int hdrlen = sizeof(struct inet_diag_req);
1120 struct net *net = sock_net(skb->sk);
1122 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1123 nlmsg_len(nlh) < hdrlen)
1126 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1127 if (nlmsg_attrlen(nlh, hdrlen)) {
1128 struct nlattr *attr;
1131 attr = nlmsg_find_attr(nlh, hdrlen,
1132 INET_DIAG_REQ_BYTECODE);
1133 err = inet_diag_bc_audit(attr, skb);
1138 struct netlink_dump_control c = {
1139 .dump = inet_diag_dump_compat,
1141 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1145 return inet_diag_get_exact_compat(skb, nlh);
1148 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1150 int hdrlen = sizeof(struct inet_diag_req_v2);
1151 struct net *net = sock_net(skb->sk);
1153 if (nlmsg_len(h) < hdrlen)
1156 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1157 h->nlmsg_flags & NLM_F_DUMP) {
1158 if (nlmsg_attrlen(h, hdrlen)) {
1159 struct nlattr *attr;
1162 attr = nlmsg_find_attr(h, hdrlen,
1163 INET_DIAG_REQ_BYTECODE);
1164 err = inet_diag_bc_audit(attr, skb);
1169 struct netlink_dump_control c = {
1170 .dump = inet_diag_dump,
1172 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1176 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, nlmsg_data(h));
1180 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1182 const struct inet_diag_handler *handler;
1183 struct nlmsghdr *nlh;
1184 struct nlattr *attr;
1185 struct inet_diag_msg *r;
1189 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1193 r = nlmsg_data(nlh);
1194 memset(r, 0, sizeof(*r));
1195 inet_diag_msg_common_fill(r, sk);
1196 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1197 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1198 r->idiag_state = sk->sk_state;
1200 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1201 nlmsg_cancel(skb, nlh);
1205 handler = inet_diag_lock_handler(sk->sk_protocol);
1206 if (IS_ERR(handler)) {
1207 inet_diag_unlock_handler(handler);
1208 nlmsg_cancel(skb, nlh);
1209 return PTR_ERR(handler);
1212 attr = handler->idiag_info_size
1213 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1214 handler->idiag_info_size,
1218 info = nla_data(attr);
1220 handler->idiag_get_info(sk, r, info);
1221 inet_diag_unlock_handler(handler);
1223 nlmsg_end(skb, nlh);
1227 static const struct sock_diag_handler inet_diag_handler = {
1229 .dump = inet_diag_handler_cmd,
1230 .get_info = inet_diag_handler_get_info,
1231 .destroy = inet_diag_handler_cmd,
1234 static const struct sock_diag_handler inet6_diag_handler = {
1236 .dump = inet_diag_handler_cmd,
1237 .get_info = inet_diag_handler_get_info,
1238 .destroy = inet_diag_handler_cmd,
1241 int inet_diag_register(const struct inet_diag_handler *h)
1243 const __u16 type = h->idiag_type;
1246 if (type >= IPPROTO_MAX)
1249 mutex_lock(&inet_diag_table_mutex);
1251 if (!inet_diag_table[type]) {
1252 inet_diag_table[type] = h;
1255 mutex_unlock(&inet_diag_table_mutex);
1259 EXPORT_SYMBOL_GPL(inet_diag_register);
1261 void inet_diag_unregister(const struct inet_diag_handler *h)
1263 const __u16 type = h->idiag_type;
1265 if (type >= IPPROTO_MAX)
1268 mutex_lock(&inet_diag_table_mutex);
1269 inet_diag_table[type] = NULL;
1270 mutex_unlock(&inet_diag_table_mutex);
1272 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1274 static int __init inet_diag_init(void)
1276 const int inet_diag_table_size = (IPPROTO_MAX *
1277 sizeof(struct inet_diag_handler *));
1280 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1281 if (!inet_diag_table)
1284 err = sock_diag_register(&inet_diag_handler);
1288 err = sock_diag_register(&inet6_diag_handler);
1292 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1297 sock_diag_unregister(&inet_diag_handler);
1299 kfree(inet_diag_table);
1303 static void __exit inet_diag_exit(void)
1305 sock_diag_unregister(&inet6_diag_handler);
1306 sock_diag_unregister(&inet_diag_handler);
1307 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1308 kfree(inet_diag_table);
1311 module_init(inet_diag_init);
1312 module_exit(inet_diag_exit);
1313 MODULE_LICENSE("GPL");
1314 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1315 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);