1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * inet_diag.c Module for monitoring INET transport protocols sockets.
5 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
8 #include <linux/kernel.h>
9 #include <linux/module.h>
10 #include <linux/types.h>
11 #include <linux/fcntl.h>
12 #include <linux/random.h>
13 #include <linux/slab.h>
14 #include <linux/cache.h>
15 #include <linux/init.h>
16 #include <linux/time.h>
21 #include <net/inet_common.h>
22 #include <net/inet_connection_sock.h>
23 #include <net/inet_hashtables.h>
24 #include <net/inet_timewait_sock.h>
25 #include <net/inet6_hashtables.h>
26 #include <net/bpf_sk_storage.h>
27 #include <net/netlink.h>
29 #include <linux/inet.h>
30 #include <linux/stddef.h>
32 #include <linux/inet_diag.h>
33 #include <linux/sock_diag.h>
35 static const struct inet_diag_handler **inet_diag_table;
37 struct inet_diag_entry {
46 #ifdef CONFIG_SOCK_CGROUP_DATA
51 static DEFINE_MUTEX(inet_diag_table_mutex);
53 static const struct inet_diag_handler *inet_diag_lock_handler(int proto)
55 if (proto < 0 || proto >= IPPROTO_MAX) {
56 mutex_lock(&inet_diag_table_mutex);
57 return ERR_PTR(-ENOENT);
60 if (!inet_diag_table[proto])
61 sock_load_diag_module(AF_INET, proto);
63 mutex_lock(&inet_diag_table_mutex);
64 if (!inet_diag_table[proto])
65 return ERR_PTR(-ENOENT);
67 return inet_diag_table[proto];
70 static void inet_diag_unlock_handler(const struct inet_diag_handler *handler)
72 mutex_unlock(&inet_diag_table_mutex);
75 void inet_diag_msg_common_fill(struct inet_diag_msg *r, struct sock *sk)
77 r->idiag_family = sk->sk_family;
79 r->id.idiag_sport = htons(sk->sk_num);
80 r->id.idiag_dport = sk->sk_dport;
81 r->id.idiag_if = sk->sk_bound_dev_if;
82 sock_diag_save_cookie(sk, r->id.idiag_cookie);
84 #if IS_ENABLED(CONFIG_IPV6)
85 if (sk->sk_family == AF_INET6) {
86 *(struct in6_addr *)r->id.idiag_src = sk->sk_v6_rcv_saddr;
87 *(struct in6_addr *)r->id.idiag_dst = sk->sk_v6_daddr;
91 memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src));
92 memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst));
94 r->id.idiag_src[0] = sk->sk_rcv_saddr;
95 r->id.idiag_dst[0] = sk->sk_daddr;
98 EXPORT_SYMBOL_GPL(inet_diag_msg_common_fill);
100 static size_t inet_sk_attr_size(struct sock *sk,
101 const struct inet_diag_req_v2 *req,
104 const struct inet_diag_handler *handler;
107 handler = inet_diag_table[req->sdiag_protocol];
108 if (handler && handler->idiag_get_aux_size)
109 aux = handler->idiag_get_aux_size(sk, net_admin);
111 return nla_total_size(sizeof(struct tcp_info))
112 + nla_total_size(sizeof(struct inet_diag_msg))
113 + inet_diag_msg_attrs_size()
114 + nla_total_size(sizeof(struct inet_diag_meminfo))
115 + nla_total_size(SK_MEMINFO_VARS * sizeof(u32))
116 + nla_total_size(TCP_CA_NAME_MAX)
117 + nla_total_size(sizeof(struct tcpvegas_info))
122 int inet_diag_msg_attrs_fill(struct sock *sk, struct sk_buff *skb,
123 struct inet_diag_msg *r, int ext,
124 struct user_namespace *user_ns,
127 const struct inet_sock *inet = inet_sk(sk);
128 struct inet_diag_sockopt inet_sockopt;
130 if (nla_put_u8(skb, INET_DIAG_SHUTDOWN, sk->sk_shutdown))
133 /* IPv6 dual-stack sockets use inet->tos for IPv4 connections,
134 * hence this needs to be included regardless of socket family.
136 if (ext & (1 << (INET_DIAG_TOS - 1)))
137 if (nla_put_u8(skb, INET_DIAG_TOS, inet->tos) < 0)
140 #if IS_ENABLED(CONFIG_IPV6)
141 if (r->idiag_family == AF_INET6) {
142 if (ext & (1 << (INET_DIAG_TCLASS - 1)))
143 if (nla_put_u8(skb, INET_DIAG_TCLASS,
144 inet6_sk(sk)->tclass) < 0)
147 if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
148 nla_put_u8(skb, INET_DIAG_SKV6ONLY, ipv6_only_sock(sk)))
153 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK, sk->sk_mark))
156 if (ext & (1 << (INET_DIAG_CLASS_ID - 1)) ||
157 ext & (1 << (INET_DIAG_TCLASS - 1))) {
160 #ifdef CONFIG_SOCK_CGROUP_DATA
161 classid = sock_cgroup_classid(&sk->sk_cgrp_data);
163 /* Fallback to socket priority if class id isn't set.
164 * Classful qdiscs use it as direct reference to class.
165 * For cgroup2 classid is always zero.
168 classid = sk->sk_priority;
170 if (nla_put_u32(skb, INET_DIAG_CLASS_ID, classid))
174 #ifdef CONFIG_SOCK_CGROUP_DATA
175 if (nla_put_u64_64bit(skb, INET_DIAG_CGROUP_ID,
176 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)),
181 r->idiag_uid = from_kuid_munged(user_ns, sock_i_uid(sk));
182 r->idiag_inode = sock_i_ino(sk);
184 memset(&inet_sockopt, 0, sizeof(inet_sockopt));
185 inet_sockopt.recverr = inet->recverr;
186 inet_sockopt.is_icsk = inet->is_icsk;
187 inet_sockopt.freebind = inet->freebind;
188 inet_sockopt.hdrincl = inet->hdrincl;
189 inet_sockopt.mc_loop = inet->mc_loop;
190 inet_sockopt.transparent = inet->transparent;
191 inet_sockopt.mc_all = inet->mc_all;
192 inet_sockopt.nodefrag = inet->nodefrag;
193 inet_sockopt.bind_address_no_port = inet->bind_address_no_port;
194 inet_sockopt.recverr_rfc4884 = inet->recverr_rfc4884;
195 inet_sockopt.defer_connect = inet->defer_connect;
196 if (nla_put(skb, INET_DIAG_SOCKOPT, sizeof(inet_sockopt),
204 EXPORT_SYMBOL_GPL(inet_diag_msg_attrs_fill);
206 static int inet_diag_parse_attrs(const struct nlmsghdr *nlh, int hdrlen,
207 struct nlattr **req_nlas)
212 nlmsg_for_each_attr(nla, nlh, hdrlen, remaining) {
213 int type = nla_type(nla);
215 if (type == INET_DIAG_REQ_PROTOCOL && nla_len(nla) != sizeof(u32))
218 if (type < __INET_DIAG_REQ_MAX)
219 req_nlas[type] = nla;
224 static int inet_diag_get_protocol(const struct inet_diag_req_v2 *req,
225 const struct inet_diag_dump_data *data)
227 if (data->req_nlas[INET_DIAG_REQ_PROTOCOL])
228 return nla_get_u32(data->req_nlas[INET_DIAG_REQ_PROTOCOL]);
229 return req->sdiag_protocol;
232 #define MAX_DUMP_ALLOC_SIZE (KMALLOC_MAX_SIZE - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
234 int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
235 struct sk_buff *skb, struct netlink_callback *cb,
236 const struct inet_diag_req_v2 *req,
237 u16 nlmsg_flags, bool net_admin)
239 const struct tcp_congestion_ops *ca_ops;
240 const struct inet_diag_handler *handler;
241 struct inet_diag_dump_data *cb_data;
242 int ext = req->idiag_ext;
243 struct inet_diag_msg *r;
244 struct nlmsghdr *nlh;
249 handler = inet_diag_table[inet_diag_get_protocol(req, cb_data)];
252 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
253 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
258 BUG_ON(!sk_fullsock(sk));
260 inet_diag_msg_common_fill(r, sk);
261 r->idiag_state = sk->sk_state;
263 r->idiag_retrans = 0;
265 if (inet_diag_msg_attrs_fill(sk, skb, r, ext,
266 sk_user_ns(NETLINK_CB(cb->skb).sk),
270 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
271 struct inet_diag_meminfo minfo = {
272 .idiag_rmem = sk_rmem_alloc_get(sk),
273 .idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
274 .idiag_fmem = sk->sk_forward_alloc,
275 .idiag_tmem = sk_wmem_alloc_get(sk),
278 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
282 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
283 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
287 * RAW sockets might have user-defined protocols assigned,
288 * so report the one supplied on socket creation.
290 if (sk->sk_type == SOCK_RAW) {
291 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
296 handler->idiag_get_info(sk, r, NULL);
300 if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
301 icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
302 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
304 r->idiag_retrans = icsk->icsk_retransmits;
306 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
307 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
309 r->idiag_retrans = icsk->icsk_probes_out;
311 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
312 } else if (timer_pending(&sk->sk_timer)) {
314 r->idiag_retrans = icsk->icsk_probes_out;
316 jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
319 r->idiag_expires = 0;
322 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
323 attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
324 handler->idiag_info_size,
329 info = nla_data(attr);
332 if (ext & (1 << (INET_DIAG_CONG - 1))) {
336 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
338 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
344 handler->idiag_get_info(sk, r, info);
346 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
347 if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
350 if (sk->sk_state < TCP_TIME_WAIT) {
351 union tcp_cc_info info;
356 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
357 if (ca_ops && ca_ops->get_info)
358 sz = ca_ops->get_info(sk, ext, &attr, &info);
360 if (sz && nla_put(skb, attr, sz, &info) < 0)
364 /* Keep it at the end for potential retry with a larger skb,
365 * or else do best-effort fitting, which is only done for the
368 if (cb_data->bpf_stg_diag) {
369 bool first_nlmsg = ((unsigned char *)nlh == skb->data);
370 unsigned int prev_min_dump_alloc;
371 unsigned int total_nla_size = 0;
372 unsigned int msg_len;
375 msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh;
376 err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb,
377 INET_DIAG_SK_BPF_STORAGES,
383 total_nla_size += msg_len;
384 prev_min_dump_alloc = cb->min_dump_alloc;
385 if (total_nla_size > prev_min_dump_alloc)
386 cb->min_dump_alloc = min_t(u32, total_nla_size,
387 MAX_DUMP_ALLOC_SIZE);
392 if (cb->min_dump_alloc > prev_min_dump_alloc)
393 /* Retry with pskb_expand_head() with
394 * __GFP_DIRECT_RECLAIM
398 WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc);
400 /* Send what we have for this sk
401 * and move on to the next sk in the following
411 nlmsg_cancel(skb, nlh);
414 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
416 static int inet_twsk_diag_fill(struct sock *sk,
418 struct netlink_callback *cb,
419 u16 nlmsg_flags, bool net_admin)
421 struct inet_timewait_sock *tw = inet_twsk(sk);
422 struct inet_diag_msg *r;
423 struct nlmsghdr *nlh;
426 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
427 cb->nlh->nlmsg_seq, cb->nlh->nlmsg_type,
428 sizeof(*r), nlmsg_flags);
433 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
435 inet_diag_msg_common_fill(r, sk);
436 r->idiag_retrans = 0;
438 r->idiag_state = tw->tw_substate;
440 tmo = tw->tw_timer.expires - jiffies;
441 r->idiag_expires = jiffies_delta_to_msecs(tmo);
447 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
449 nlmsg_cancel(skb, nlh);
457 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
458 struct netlink_callback *cb,
459 u16 nlmsg_flags, bool net_admin)
461 struct request_sock *reqsk = inet_reqsk(sk);
462 struct inet_diag_msg *r;
463 struct nlmsghdr *nlh;
466 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
467 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
472 inet_diag_msg_common_fill(r, sk);
473 r->idiag_state = TCP_SYN_RECV;
475 r->idiag_retrans = reqsk->num_retrans;
477 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
478 offsetof(struct sock, sk_cookie));
480 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
481 r->idiag_expires = jiffies_delta_to_msecs(tmo);
487 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
488 inet_rsk(reqsk)->ir_mark)) {
489 nlmsg_cancel(skb, nlh);
497 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
498 struct netlink_callback *cb,
499 const struct inet_diag_req_v2 *r,
500 u16 nlmsg_flags, bool net_admin)
502 if (sk->sk_state == TCP_TIME_WAIT)
503 return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
505 if (sk->sk_state == TCP_NEW_SYN_RECV)
506 return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
508 return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
512 struct sock *inet_diag_find_one_icsk(struct net *net,
513 struct inet_hashinfo *hashinfo,
514 const struct inet_diag_req_v2 *req)
519 if (req->sdiag_family == AF_INET)
520 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
521 req->id.idiag_dport, req->id.idiag_src[0],
522 req->id.idiag_sport, req->id.idiag_if);
523 #if IS_ENABLED(CONFIG_IPV6)
524 else if (req->sdiag_family == AF_INET6) {
525 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
526 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
527 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
528 req->id.idiag_dport, req->id.idiag_src[3],
529 req->id.idiag_sport, req->id.idiag_if);
531 sk = inet6_lookup(net, hashinfo, NULL, 0,
532 (struct in6_addr *)req->id.idiag_dst,
534 (struct in6_addr *)req->id.idiag_src,
541 return ERR_PTR(-EINVAL);
545 return ERR_PTR(-ENOENT);
547 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
549 return ERR_PTR(-ENOENT);
554 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
556 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
557 struct netlink_callback *cb,
558 const struct inet_diag_req_v2 *req)
560 struct sk_buff *in_skb = cb->skb;
561 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
562 struct net *net = sock_net(in_skb->sk);
567 sk = inet_diag_find_one_icsk(net, hashinfo, req);
571 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
577 err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
579 WARN_ON(err == -EMSGSIZE);
583 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
594 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
596 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
597 const struct nlmsghdr *nlh,
599 const struct inet_diag_req_v2 *req)
601 const struct inet_diag_handler *handler;
602 struct inet_diag_dump_data dump_data;
605 memset(&dump_data, 0, sizeof(dump_data));
606 err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
610 protocol = inet_diag_get_protocol(req, &dump_data);
612 handler = inet_diag_lock_handler(protocol);
613 if (IS_ERR(handler)) {
614 err = PTR_ERR(handler);
615 } else if (cmd == SOCK_DIAG_BY_FAMILY) {
616 struct netlink_callback cb = {
621 err = handler->dump_one(&cb, req);
622 } else if (cmd == SOCK_DESTROY && handler->destroy) {
623 err = handler->destroy(in_skb, req);
627 inet_diag_unlock_handler(handler);
632 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
634 int words = bits >> 5;
639 if (memcmp(a1, a2, words << 2))
649 mask = htonl((0xffffffff) << (32 - bits));
651 if ((w1 ^ w2) & mask)
658 static int inet_diag_bc_run(const struct nlattr *_bc,
659 const struct inet_diag_entry *entry)
661 const void *bc = nla_data(_bc);
662 int len = nla_len(_bc);
666 const struct inet_diag_bc_op *op = bc;
669 case INET_DIAG_BC_NOP:
671 case INET_DIAG_BC_JMP:
674 case INET_DIAG_BC_S_EQ:
675 yes = entry->sport == op[1].no;
677 case INET_DIAG_BC_S_GE:
678 yes = entry->sport >= op[1].no;
680 case INET_DIAG_BC_S_LE:
681 yes = entry->sport <= op[1].no;
683 case INET_DIAG_BC_D_EQ:
684 yes = entry->dport == op[1].no;
686 case INET_DIAG_BC_D_GE:
687 yes = entry->dport >= op[1].no;
689 case INET_DIAG_BC_D_LE:
690 yes = entry->dport <= op[1].no;
692 case INET_DIAG_BC_AUTO:
693 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
695 case INET_DIAG_BC_S_COND:
696 case INET_DIAG_BC_D_COND: {
697 const struct inet_diag_hostcond *cond;
700 cond = (const struct inet_diag_hostcond *)(op + 1);
701 if (cond->port != -1 &&
702 cond->port != (op->code == INET_DIAG_BC_S_COND ?
703 entry->sport : entry->dport)) {
708 if (op->code == INET_DIAG_BC_S_COND)
713 if (cond->family != AF_UNSPEC &&
714 cond->family != entry->family) {
715 if (entry->family == AF_INET6 &&
716 cond->family == AF_INET) {
717 if (addr[0] == 0 && addr[1] == 0 &&
718 addr[2] == htonl(0xffff) &&
719 bitstring_match(addr + 3,
728 if (cond->prefix_len == 0)
730 if (bitstring_match(addr, cond->addr,
736 case INET_DIAG_BC_DEV_COND: {
739 ifindex = *((const u32 *)(op + 1));
740 if (ifindex != entry->ifindex)
744 case INET_DIAG_BC_MARK_COND: {
745 struct inet_diag_markcond *cond;
747 cond = (struct inet_diag_markcond *)(op + 1);
748 if ((entry->mark & cond->mask) != cond->mark)
752 #ifdef CONFIG_SOCK_CGROUP_DATA
753 case INET_DIAG_BC_CGROUP_COND: {
756 cgroup_id = get_unaligned((const u64 *)(op + 1));
757 if (cgroup_id != entry->cgroup_id)
775 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
777 static void entry_fill_addrs(struct inet_diag_entry *entry,
778 const struct sock *sk)
780 #if IS_ENABLED(CONFIG_IPV6)
781 if (sk->sk_family == AF_INET6) {
782 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
783 entry->daddr = sk->sk_v6_daddr.s6_addr32;
787 entry->saddr = &sk->sk_rcv_saddr;
788 entry->daddr = &sk->sk_daddr;
792 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
794 struct inet_sock *inet = inet_sk(sk);
795 struct inet_diag_entry entry;
800 entry.family = sk->sk_family;
801 entry_fill_addrs(&entry, sk);
802 entry.sport = inet->inet_num;
803 entry.dport = ntohs(inet->inet_dport);
804 entry.ifindex = sk->sk_bound_dev_if;
805 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
807 entry.mark = sk->sk_mark;
808 else if (sk->sk_state == TCP_NEW_SYN_RECV)
809 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
810 else if (sk->sk_state == TCP_TIME_WAIT)
811 entry.mark = inet_twsk(sk)->tw_mark;
814 #ifdef CONFIG_SOCK_CGROUP_DATA
815 entry.cgroup_id = sk_fullsock(sk) ?
816 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
819 return inet_diag_bc_run(bc, &entry);
821 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
823 static int valid_cc(const void *bc, int len, int cc)
826 const struct inet_diag_bc_op *op = bc;
832 if (op->yes < 4 || op->yes & 3)
840 /* data is u32 ifindex */
841 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
844 /* Check ifindex space. */
845 *min_len += sizeof(u32);
851 /* Validate an inet_diag_hostcond. */
852 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
855 struct inet_diag_hostcond *cond;
858 /* Check hostcond space. */
859 *min_len += sizeof(struct inet_diag_hostcond);
862 cond = (struct inet_diag_hostcond *)(op + 1);
864 /* Check address family and address length. */
865 switch (cond->family) {
870 addr_len = sizeof(struct in_addr);
873 addr_len = sizeof(struct in6_addr);
878 *min_len += addr_len;
882 /* Check prefix length (in bits) vs address length (in bytes). */
883 if (cond->prefix_len > 8 * addr_len)
889 /* Validate a port comparison operator. */
890 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
891 int len, int *min_len)
893 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
894 *min_len += sizeof(struct inet_diag_bc_op);
900 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
903 *min_len += sizeof(struct inet_diag_markcond);
904 return len >= *min_len;
907 #ifdef CONFIG_SOCK_CGROUP_DATA
908 static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
911 *min_len += sizeof(u64);
912 return len >= *min_len;
916 static int inet_diag_bc_audit(const struct nlattr *attr,
917 const struct sk_buff *skb)
919 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
920 const void *bytecode, *bc;
921 int bytecode_len, len;
923 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
926 bytecode = bc = nla_data(attr);
927 len = bytecode_len = nla_len(attr);
930 int min_len = sizeof(struct inet_diag_bc_op);
931 const struct inet_diag_bc_op *op = bc;
934 case INET_DIAG_BC_S_COND:
935 case INET_DIAG_BC_D_COND:
936 if (!valid_hostcond(bc, len, &min_len))
939 case INET_DIAG_BC_DEV_COND:
940 if (!valid_devcond(bc, len, &min_len))
943 case INET_DIAG_BC_S_EQ:
944 case INET_DIAG_BC_S_GE:
945 case INET_DIAG_BC_S_LE:
946 case INET_DIAG_BC_D_EQ:
947 case INET_DIAG_BC_D_GE:
948 case INET_DIAG_BC_D_LE:
949 if (!valid_port_comparison(bc, len, &min_len))
952 case INET_DIAG_BC_MARK_COND:
955 if (!valid_markcond(bc, len, &min_len))
958 #ifdef CONFIG_SOCK_CGROUP_DATA
959 case INET_DIAG_BC_CGROUP_COND:
960 if (!valid_cgroupcond(bc, len, &min_len))
964 case INET_DIAG_BC_AUTO:
965 case INET_DIAG_BC_JMP:
966 case INET_DIAG_BC_NOP:
972 if (op->code != INET_DIAG_BC_NOP) {
973 if (op->no < min_len || op->no > len + 4 || op->no & 3)
976 !valid_cc(bytecode, bytecode_len, len - op->no))
980 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
985 return len == 0 ? 0 : -EINVAL;
988 static void twsk_build_assert(void)
990 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
991 offsetof(struct sock, sk_family));
993 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
994 offsetof(struct inet_sock, inet_num));
996 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
997 offsetof(struct inet_sock, inet_dport));
999 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
1000 offsetof(struct inet_sock, inet_rcv_saddr));
1002 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
1003 offsetof(struct inet_sock, inet_daddr));
1005 #if IS_ENABLED(CONFIG_IPV6)
1006 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
1007 offsetof(struct sock, sk_v6_rcv_saddr));
1009 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1010 offsetof(struct sock, sk_v6_daddr));
1014 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1015 struct netlink_callback *cb,
1016 const struct inet_diag_req_v2 *r)
1018 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1019 struct inet_diag_dump_data *cb_data = cb->data;
1020 struct net *net = sock_net(skb->sk);
1021 u32 idiag_states = r->idiag_states;
1022 int i, num, s_i, s_num;
1026 bc = cb_data->inet_diag_nla_bc;
1027 if (idiag_states & TCPF_SYN_RECV)
1028 idiag_states |= TCPF_NEW_SYN_RECV;
1030 s_num = num = cb->args[2];
1032 if (cb->args[0] == 0) {
1033 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1034 goto skip_listen_ht;
1036 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
1037 struct inet_listen_hashbucket *ilb;
1038 struct hlist_nulls_node *node;
1041 ilb = &hashinfo->listening_hash[i];
1042 spin_lock(&ilb->lock);
1043 sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1044 struct inet_sock *inet = inet_sk(sk);
1046 if (!net_eq(sock_net(sk), net))
1054 if (r->sdiag_family != AF_UNSPEC &&
1055 sk->sk_family != r->sdiag_family)
1058 if (r->id.idiag_sport != inet->inet_sport &&
1062 if (!inet_diag_bc_sk(bc, sk))
1065 if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1068 spin_unlock(&ilb->lock);
1075 spin_unlock(&ilb->lock);
1081 s_i = num = s_num = 0;
1084 if (!(idiag_states & ~TCPF_LISTEN))
1088 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1089 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1090 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1091 struct hlist_nulls_node *node;
1092 struct sock *sk_arr[SKARR_SZ];
1093 int num_arr[SKARR_SZ];
1094 int idx, accum, res;
1096 if (hlist_nulls_empty(&head->chain))
1106 sk_nulls_for_each(sk, node, &head->chain) {
1109 if (!net_eq(sock_net(sk), net))
1113 state = (sk->sk_state == TCP_TIME_WAIT) ?
1114 inet_twsk(sk)->tw_substate : sk->sk_state;
1115 if (!(idiag_states & (1 << state)))
1117 if (r->sdiag_family != AF_UNSPEC &&
1118 sk->sk_family != r->sdiag_family)
1120 if (r->id.idiag_sport != htons(sk->sk_num) &&
1123 if (r->id.idiag_dport != sk->sk_dport &&
1126 twsk_build_assert();
1128 if (!inet_diag_bc_sk(bc, sk))
1131 if (!refcount_inc_not_zero(&sk->sk_refcnt))
1134 num_arr[accum] = num;
1136 if (++accum == SKARR_SZ)
1141 spin_unlock_bh(lock);
1143 for (idx = 0; idx < accum; idx++) {
1145 res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1146 NLM_F_MULTI, net_admin);
1150 sock_gen_put(sk_arr[idx]);
1155 if (accum == SKARR_SZ) {
1167 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1169 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1170 const struct inet_diag_req_v2 *r)
1172 struct inet_diag_dump_data *cb_data = cb->data;
1173 const struct inet_diag_handler *handler;
1174 u32 prev_min_dump_alloc;
1175 int protocol, err = 0;
1177 protocol = inet_diag_get_protocol(r, cb_data);
1180 prev_min_dump_alloc = cb->min_dump_alloc;
1181 handler = inet_diag_lock_handler(protocol);
1182 if (!IS_ERR(handler))
1183 handler->dump(skb, cb, r);
1185 err = PTR_ERR(handler);
1186 inet_diag_unlock_handler(handler);
1188 /* The skb is not large enough to fit one sk info and
1189 * inet_sk_diag_fill() has requested for a larger skb.
1191 if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1192 err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1197 return err ? : skb->len;
1200 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1202 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1205 static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1207 const struct nlmsghdr *nlh = cb->nlh;
1208 struct inet_diag_dump_data *cb_data;
1209 struct sk_buff *skb = cb->skb;
1213 cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1217 err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1222 nla = cb_data->inet_diag_nla_bc;
1224 err = inet_diag_bc_audit(nla, skb);
1231 nla = cb_data->inet_diag_nla_bpf_stgs;
1233 struct bpf_sk_storage_diag *bpf_stg_diag;
1235 bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1236 if (IS_ERR(bpf_stg_diag)) {
1238 return PTR_ERR(bpf_stg_diag);
1240 cb_data->bpf_stg_diag = bpf_stg_diag;
1247 static int inet_diag_dump_start(struct netlink_callback *cb)
1249 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1252 static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1254 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1257 static int inet_diag_dump_done(struct netlink_callback *cb)
1259 struct inet_diag_dump_data *cb_data = cb->data;
1261 bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1267 static int inet_diag_type2proto(int type)
1270 case TCPDIAG_GETSOCK:
1272 case DCCPDIAG_GETSOCK:
1273 return IPPROTO_DCCP;
1279 static int inet_diag_dump_compat(struct sk_buff *skb,
1280 struct netlink_callback *cb)
1282 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1283 struct inet_diag_req_v2 req;
1285 req.sdiag_family = AF_UNSPEC; /* compatibility */
1286 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1287 req.idiag_ext = rc->idiag_ext;
1288 req.idiag_states = rc->idiag_states;
1291 return __inet_diag_dump(skb, cb, &req);
1294 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1295 const struct nlmsghdr *nlh)
1297 struct inet_diag_req *rc = nlmsg_data(nlh);
1298 struct inet_diag_req_v2 req;
1300 req.sdiag_family = rc->idiag_family;
1301 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1302 req.idiag_ext = rc->idiag_ext;
1303 req.idiag_states = rc->idiag_states;
1306 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1307 sizeof(struct inet_diag_req), &req);
1310 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1312 int hdrlen = sizeof(struct inet_diag_req);
1313 struct net *net = sock_net(skb->sk);
1315 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1316 nlmsg_len(nlh) < hdrlen)
1319 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1320 struct netlink_dump_control c = {
1321 .start = inet_diag_dump_start_compat,
1322 .done = inet_diag_dump_done,
1323 .dump = inet_diag_dump_compat,
1325 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1328 return inet_diag_get_exact_compat(skb, nlh);
1331 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1333 int hdrlen = sizeof(struct inet_diag_req_v2);
1334 struct net *net = sock_net(skb->sk);
1336 if (nlmsg_len(h) < hdrlen)
1339 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1340 h->nlmsg_flags & NLM_F_DUMP) {
1341 struct netlink_dump_control c = {
1342 .start = inet_diag_dump_start,
1343 .done = inet_diag_dump_done,
1344 .dump = inet_diag_dump,
1346 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1349 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1354 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1356 const struct inet_diag_handler *handler;
1357 struct nlmsghdr *nlh;
1358 struct nlattr *attr;
1359 struct inet_diag_msg *r;
1363 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1367 r = nlmsg_data(nlh);
1368 memset(r, 0, sizeof(*r));
1369 inet_diag_msg_common_fill(r, sk);
1370 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1371 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1372 r->idiag_state = sk->sk_state;
1374 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1375 nlmsg_cancel(skb, nlh);
1379 handler = inet_diag_lock_handler(sk->sk_protocol);
1380 if (IS_ERR(handler)) {
1381 inet_diag_unlock_handler(handler);
1382 nlmsg_cancel(skb, nlh);
1383 return PTR_ERR(handler);
1386 attr = handler->idiag_info_size
1387 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1388 handler->idiag_info_size,
1392 info = nla_data(attr);
1394 handler->idiag_get_info(sk, r, info);
1395 inet_diag_unlock_handler(handler);
1397 nlmsg_end(skb, nlh);
1401 static const struct sock_diag_handler inet_diag_handler = {
1403 .dump = inet_diag_handler_cmd,
1404 .get_info = inet_diag_handler_get_info,
1405 .destroy = inet_diag_handler_cmd,
1408 static const struct sock_diag_handler inet6_diag_handler = {
1410 .dump = inet_diag_handler_cmd,
1411 .get_info = inet_diag_handler_get_info,
1412 .destroy = inet_diag_handler_cmd,
1415 int inet_diag_register(const struct inet_diag_handler *h)
1417 const __u16 type = h->idiag_type;
1420 if (type >= IPPROTO_MAX)
1423 mutex_lock(&inet_diag_table_mutex);
1425 if (!inet_diag_table[type]) {
1426 inet_diag_table[type] = h;
1429 mutex_unlock(&inet_diag_table_mutex);
1433 EXPORT_SYMBOL_GPL(inet_diag_register);
1435 void inet_diag_unregister(const struct inet_diag_handler *h)
1437 const __u16 type = h->idiag_type;
1439 if (type >= IPPROTO_MAX)
1442 mutex_lock(&inet_diag_table_mutex);
1443 inet_diag_table[type] = NULL;
1444 mutex_unlock(&inet_diag_table_mutex);
1446 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1448 static int __init inet_diag_init(void)
1450 const int inet_diag_table_size = (IPPROTO_MAX *
1451 sizeof(struct inet_diag_handler *));
1454 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1455 if (!inet_diag_table)
1458 err = sock_diag_register(&inet_diag_handler);
1462 err = sock_diag_register(&inet6_diag_handler);
1466 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1471 sock_diag_unregister(&inet_diag_handler);
1473 kfree(inet_diag_table);
1477 static void __exit inet_diag_exit(void)
1479 sock_diag_unregister(&inet6_diag_handler);
1480 sock_diag_unregister(&inet_diag_handler);
1481 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1482 kfree(inet_diag_table);
1485 module_init(inet_diag_init);
1486 module_exit(inet_diag_exit);
1487 MODULE_LICENSE("GPL");
1488 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1489 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);