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;
264 r->idiag_expires = 0;
266 if (inet_diag_msg_attrs_fill(sk, skb, r, ext,
267 sk_user_ns(NETLINK_CB(cb->skb).sk),
271 if (ext & (1 << (INET_DIAG_MEMINFO - 1))) {
272 struct inet_diag_meminfo minfo = {
273 .idiag_rmem = sk_rmem_alloc_get(sk),
274 .idiag_wmem = READ_ONCE(sk->sk_wmem_queued),
275 .idiag_fmem = sk->sk_forward_alloc,
276 .idiag_tmem = sk_wmem_alloc_get(sk),
279 if (nla_put(skb, INET_DIAG_MEMINFO, sizeof(minfo), &minfo) < 0)
283 if (ext & (1 << (INET_DIAG_SKMEMINFO - 1)))
284 if (sock_diag_put_meminfo(sk, skb, INET_DIAG_SKMEMINFO))
288 * RAW sockets might have user-defined protocols assigned,
289 * so report the one supplied on socket creation.
291 if (sk->sk_type == SOCK_RAW) {
292 if (nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))
297 handler->idiag_get_info(sk, r, NULL);
301 if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
302 icsk->icsk_pending == ICSK_TIME_REO_TIMEOUT ||
303 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
305 r->idiag_retrans = icsk->icsk_retransmits;
307 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
308 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
310 r->idiag_retrans = icsk->icsk_probes_out;
312 jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
313 } else if (timer_pending(&sk->sk_timer)) {
315 r->idiag_retrans = icsk->icsk_probes_out;
317 jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
320 if ((ext & (1 << (INET_DIAG_INFO - 1))) && handler->idiag_info_size) {
321 attr = nla_reserve_64bit(skb, INET_DIAG_INFO,
322 handler->idiag_info_size,
327 info = nla_data(attr);
330 if (ext & (1 << (INET_DIAG_CONG - 1))) {
334 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
336 err = nla_put_string(skb, INET_DIAG_CONG, ca_ops->name);
342 handler->idiag_get_info(sk, r, info);
344 if (ext & (1 << (INET_DIAG_INFO - 1)) && handler->idiag_get_aux)
345 if (handler->idiag_get_aux(sk, net_admin, skb) < 0)
348 if (sk->sk_state < TCP_TIME_WAIT) {
349 union tcp_cc_info info;
354 ca_ops = READ_ONCE(icsk->icsk_ca_ops);
355 if (ca_ops && ca_ops->get_info)
356 sz = ca_ops->get_info(sk, ext, &attr, &info);
358 if (sz && nla_put(skb, attr, sz, &info) < 0)
362 /* Keep it at the end for potential retry with a larger skb,
363 * or else do best-effort fitting, which is only done for the
366 if (cb_data->bpf_stg_diag) {
367 bool first_nlmsg = ((unsigned char *)nlh == skb->data);
368 unsigned int prev_min_dump_alloc;
369 unsigned int total_nla_size = 0;
370 unsigned int msg_len;
373 msg_len = skb_tail_pointer(skb) - (unsigned char *)nlh;
374 err = bpf_sk_storage_diag_put(cb_data->bpf_stg_diag, sk, skb,
375 INET_DIAG_SK_BPF_STORAGES,
381 total_nla_size += msg_len;
382 prev_min_dump_alloc = cb->min_dump_alloc;
383 if (total_nla_size > prev_min_dump_alloc)
384 cb->min_dump_alloc = min_t(u32, total_nla_size,
385 MAX_DUMP_ALLOC_SIZE);
390 if (cb->min_dump_alloc > prev_min_dump_alloc)
391 /* Retry with pskb_expand_head() with
392 * __GFP_DIRECT_RECLAIM
396 WARN_ON_ONCE(total_nla_size <= prev_min_dump_alloc);
398 /* Send what we have for this sk
399 * and move on to the next sk in the following
409 nlmsg_cancel(skb, nlh);
412 EXPORT_SYMBOL_GPL(inet_sk_diag_fill);
414 static int inet_twsk_diag_fill(struct sock *sk,
416 struct netlink_callback *cb,
419 struct inet_timewait_sock *tw = inet_twsk(sk);
420 struct inet_diag_msg *r;
421 struct nlmsghdr *nlh;
424 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid,
425 cb->nlh->nlmsg_seq, cb->nlh->nlmsg_type,
426 sizeof(*r), nlmsg_flags);
431 BUG_ON(tw->tw_state != TCP_TIME_WAIT);
433 inet_diag_msg_common_fill(r, sk);
434 r->idiag_retrans = 0;
436 r->idiag_state = tw->tw_substate;
438 tmo = tw->tw_timer.expires - jiffies;
439 r->idiag_expires = jiffies_delta_to_msecs(tmo);
449 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
450 struct netlink_callback *cb,
451 u16 nlmsg_flags, bool net_admin)
453 struct request_sock *reqsk = inet_reqsk(sk);
454 struct inet_diag_msg *r;
455 struct nlmsghdr *nlh;
458 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
459 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
464 inet_diag_msg_common_fill(r, sk);
465 r->idiag_state = TCP_SYN_RECV;
467 r->idiag_retrans = reqsk->num_retrans;
469 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
470 offsetof(struct sock, sk_cookie));
472 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
473 r->idiag_expires = jiffies_delta_to_msecs(tmo);
479 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
480 inet_rsk(reqsk)->ir_mark)) {
481 nlmsg_cancel(skb, nlh);
489 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
490 struct netlink_callback *cb,
491 const struct inet_diag_req_v2 *r,
492 u16 nlmsg_flags, bool net_admin)
494 if (sk->sk_state == TCP_TIME_WAIT)
495 return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags);
497 if (sk->sk_state == TCP_NEW_SYN_RECV)
498 return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
500 return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
504 struct sock *inet_diag_find_one_icsk(struct net *net,
505 struct inet_hashinfo *hashinfo,
506 const struct inet_diag_req_v2 *req)
511 if (req->sdiag_family == AF_INET)
512 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
513 req->id.idiag_dport, req->id.idiag_src[0],
514 req->id.idiag_sport, req->id.idiag_if);
515 #if IS_ENABLED(CONFIG_IPV6)
516 else if (req->sdiag_family == AF_INET6) {
517 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
518 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
519 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
520 req->id.idiag_dport, req->id.idiag_src[3],
521 req->id.idiag_sport, req->id.idiag_if);
523 sk = inet6_lookup(net, hashinfo, NULL, 0,
524 (struct in6_addr *)req->id.idiag_dst,
526 (struct in6_addr *)req->id.idiag_src,
533 return ERR_PTR(-EINVAL);
537 return ERR_PTR(-ENOENT);
539 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
541 return ERR_PTR(-ENOENT);
546 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
548 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
549 struct netlink_callback *cb,
550 const struct inet_diag_req_v2 *req)
552 struct sk_buff *in_skb = cb->skb;
553 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
554 struct net *net = sock_net(in_skb->sk);
559 sk = inet_diag_find_one_icsk(net, hashinfo, req);
563 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
569 err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
571 WARN_ON(err == -EMSGSIZE);
575 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
586 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
588 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
589 const struct nlmsghdr *nlh,
591 const struct inet_diag_req_v2 *req)
593 const struct inet_diag_handler *handler;
594 struct inet_diag_dump_data dump_data;
597 memset(&dump_data, 0, sizeof(dump_data));
598 err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
602 protocol = inet_diag_get_protocol(req, &dump_data);
604 handler = inet_diag_lock_handler(protocol);
605 if (IS_ERR(handler)) {
606 err = PTR_ERR(handler);
607 } else if (cmd == SOCK_DIAG_BY_FAMILY) {
608 struct netlink_callback cb = {
613 err = handler->dump_one(&cb, req);
614 } else if (cmd == SOCK_DESTROY && handler->destroy) {
615 err = handler->destroy(in_skb, req);
619 inet_diag_unlock_handler(handler);
624 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
626 int words = bits >> 5;
631 if (memcmp(a1, a2, words << 2))
641 mask = htonl((0xffffffff) << (32 - bits));
643 if ((w1 ^ w2) & mask)
650 static int inet_diag_bc_run(const struct nlattr *_bc,
651 const struct inet_diag_entry *entry)
653 const void *bc = nla_data(_bc);
654 int len = nla_len(_bc);
658 const struct inet_diag_bc_op *op = bc;
661 case INET_DIAG_BC_NOP:
663 case INET_DIAG_BC_JMP:
666 case INET_DIAG_BC_S_EQ:
667 yes = entry->sport == op[1].no;
669 case INET_DIAG_BC_S_GE:
670 yes = entry->sport >= op[1].no;
672 case INET_DIAG_BC_S_LE:
673 yes = entry->sport <= op[1].no;
675 case INET_DIAG_BC_D_EQ:
676 yes = entry->dport == op[1].no;
678 case INET_DIAG_BC_D_GE:
679 yes = entry->dport >= op[1].no;
681 case INET_DIAG_BC_D_LE:
682 yes = entry->dport <= op[1].no;
684 case INET_DIAG_BC_AUTO:
685 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
687 case INET_DIAG_BC_S_COND:
688 case INET_DIAG_BC_D_COND: {
689 const struct inet_diag_hostcond *cond;
692 cond = (const struct inet_diag_hostcond *)(op + 1);
693 if (cond->port != -1 &&
694 cond->port != (op->code == INET_DIAG_BC_S_COND ?
695 entry->sport : entry->dport)) {
700 if (op->code == INET_DIAG_BC_S_COND)
705 if (cond->family != AF_UNSPEC &&
706 cond->family != entry->family) {
707 if (entry->family == AF_INET6 &&
708 cond->family == AF_INET) {
709 if (addr[0] == 0 && addr[1] == 0 &&
710 addr[2] == htonl(0xffff) &&
711 bitstring_match(addr + 3,
720 if (cond->prefix_len == 0)
722 if (bitstring_match(addr, cond->addr,
728 case INET_DIAG_BC_DEV_COND: {
731 ifindex = *((const u32 *)(op + 1));
732 if (ifindex != entry->ifindex)
736 case INET_DIAG_BC_MARK_COND: {
737 struct inet_diag_markcond *cond;
739 cond = (struct inet_diag_markcond *)(op + 1);
740 if ((entry->mark & cond->mask) != cond->mark)
744 #ifdef CONFIG_SOCK_CGROUP_DATA
745 case INET_DIAG_BC_CGROUP_COND: {
748 cgroup_id = get_unaligned((const u64 *)(op + 1));
749 if (cgroup_id != entry->cgroup_id)
767 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
769 static void entry_fill_addrs(struct inet_diag_entry *entry,
770 const struct sock *sk)
772 #if IS_ENABLED(CONFIG_IPV6)
773 if (sk->sk_family == AF_INET6) {
774 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
775 entry->daddr = sk->sk_v6_daddr.s6_addr32;
779 entry->saddr = &sk->sk_rcv_saddr;
780 entry->daddr = &sk->sk_daddr;
784 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
786 struct inet_sock *inet = inet_sk(sk);
787 struct inet_diag_entry entry;
792 entry.family = sk->sk_family;
793 entry_fill_addrs(&entry, sk);
794 entry.sport = inet->inet_num;
795 entry.dport = ntohs(inet->inet_dport);
796 entry.ifindex = sk->sk_bound_dev_if;
797 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
799 entry.mark = sk->sk_mark;
800 else if (sk->sk_state == TCP_NEW_SYN_RECV)
801 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
804 #ifdef CONFIG_SOCK_CGROUP_DATA
805 entry.cgroup_id = sk_fullsock(sk) ?
806 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
809 return inet_diag_bc_run(bc, &entry);
811 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
813 static int valid_cc(const void *bc, int len, int cc)
816 const struct inet_diag_bc_op *op = bc;
822 if (op->yes < 4 || op->yes & 3)
830 /* data is u32 ifindex */
831 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
834 /* Check ifindex space. */
835 *min_len += sizeof(u32);
841 /* Validate an inet_diag_hostcond. */
842 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
845 struct inet_diag_hostcond *cond;
848 /* Check hostcond space. */
849 *min_len += sizeof(struct inet_diag_hostcond);
852 cond = (struct inet_diag_hostcond *)(op + 1);
854 /* Check address family and address length. */
855 switch (cond->family) {
860 addr_len = sizeof(struct in_addr);
863 addr_len = sizeof(struct in6_addr);
868 *min_len += addr_len;
872 /* Check prefix length (in bits) vs address length (in bytes). */
873 if (cond->prefix_len > 8 * addr_len)
879 /* Validate a port comparison operator. */
880 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
881 int len, int *min_len)
883 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
884 *min_len += sizeof(struct inet_diag_bc_op);
890 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
893 *min_len += sizeof(struct inet_diag_markcond);
894 return len >= *min_len;
897 #ifdef CONFIG_SOCK_CGROUP_DATA
898 static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
901 *min_len += sizeof(u64);
902 return len >= *min_len;
906 static int inet_diag_bc_audit(const struct nlattr *attr,
907 const struct sk_buff *skb)
909 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
910 const void *bytecode, *bc;
911 int bytecode_len, len;
913 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
916 bytecode = bc = nla_data(attr);
917 len = bytecode_len = nla_len(attr);
920 int min_len = sizeof(struct inet_diag_bc_op);
921 const struct inet_diag_bc_op *op = bc;
924 case INET_DIAG_BC_S_COND:
925 case INET_DIAG_BC_D_COND:
926 if (!valid_hostcond(bc, len, &min_len))
929 case INET_DIAG_BC_DEV_COND:
930 if (!valid_devcond(bc, len, &min_len))
933 case INET_DIAG_BC_S_EQ:
934 case INET_DIAG_BC_S_GE:
935 case INET_DIAG_BC_S_LE:
936 case INET_DIAG_BC_D_EQ:
937 case INET_DIAG_BC_D_GE:
938 case INET_DIAG_BC_D_LE:
939 if (!valid_port_comparison(bc, len, &min_len))
942 case INET_DIAG_BC_MARK_COND:
945 if (!valid_markcond(bc, len, &min_len))
948 #ifdef CONFIG_SOCK_CGROUP_DATA
949 case INET_DIAG_BC_CGROUP_COND:
950 if (!valid_cgroupcond(bc, len, &min_len))
954 case INET_DIAG_BC_AUTO:
955 case INET_DIAG_BC_JMP:
956 case INET_DIAG_BC_NOP:
962 if (op->code != INET_DIAG_BC_NOP) {
963 if (op->no < min_len || op->no > len + 4 || op->no & 3)
966 !valid_cc(bytecode, bytecode_len, len - op->no))
970 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
975 return len == 0 ? 0 : -EINVAL;
978 static void twsk_build_assert(void)
980 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
981 offsetof(struct sock, sk_family));
983 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
984 offsetof(struct inet_sock, inet_num));
986 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
987 offsetof(struct inet_sock, inet_dport));
989 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
990 offsetof(struct inet_sock, inet_rcv_saddr));
992 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
993 offsetof(struct inet_sock, inet_daddr));
995 #if IS_ENABLED(CONFIG_IPV6)
996 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
997 offsetof(struct sock, sk_v6_rcv_saddr));
999 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1000 offsetof(struct sock, sk_v6_daddr));
1004 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1005 struct netlink_callback *cb,
1006 const struct inet_diag_req_v2 *r)
1008 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1009 struct inet_diag_dump_data *cb_data = cb->data;
1010 struct net *net = sock_net(skb->sk);
1011 u32 idiag_states = r->idiag_states;
1012 int i, num, s_i, s_num;
1016 bc = cb_data->inet_diag_nla_bc;
1017 if (idiag_states & TCPF_SYN_RECV)
1018 idiag_states |= TCPF_NEW_SYN_RECV;
1020 s_num = num = cb->args[2];
1022 if (cb->args[0] == 0) {
1023 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1024 goto skip_listen_ht;
1026 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
1027 struct inet_listen_hashbucket *ilb;
1028 struct hlist_nulls_node *node;
1031 ilb = &hashinfo->listening_hash[i];
1032 spin_lock(&ilb->lock);
1033 sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1034 struct inet_sock *inet = inet_sk(sk);
1036 if (!net_eq(sock_net(sk), net))
1044 if (r->sdiag_family != AF_UNSPEC &&
1045 sk->sk_family != r->sdiag_family)
1048 if (r->id.idiag_sport != inet->inet_sport &&
1052 if (!inet_diag_bc_sk(bc, sk))
1055 if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1058 spin_unlock(&ilb->lock);
1065 spin_unlock(&ilb->lock);
1071 s_i = num = s_num = 0;
1074 if (!(idiag_states & ~TCPF_LISTEN))
1078 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1079 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1080 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1081 struct hlist_nulls_node *node;
1082 struct sock *sk_arr[SKARR_SZ];
1083 int num_arr[SKARR_SZ];
1084 int idx, accum, res;
1086 if (hlist_nulls_empty(&head->chain))
1096 sk_nulls_for_each(sk, node, &head->chain) {
1099 if (!net_eq(sock_net(sk), net))
1103 state = (sk->sk_state == TCP_TIME_WAIT) ?
1104 inet_twsk(sk)->tw_substate : sk->sk_state;
1105 if (!(idiag_states & (1 << state)))
1107 if (r->sdiag_family != AF_UNSPEC &&
1108 sk->sk_family != r->sdiag_family)
1110 if (r->id.idiag_sport != htons(sk->sk_num) &&
1113 if (r->id.idiag_dport != sk->sk_dport &&
1116 twsk_build_assert();
1118 if (!inet_diag_bc_sk(bc, sk))
1121 if (!refcount_inc_not_zero(&sk->sk_refcnt))
1124 num_arr[accum] = num;
1126 if (++accum == SKARR_SZ)
1131 spin_unlock_bh(lock);
1133 for (idx = 0; idx < accum; idx++) {
1135 res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1136 NLM_F_MULTI, net_admin);
1140 sock_gen_put(sk_arr[idx]);
1145 if (accum == SKARR_SZ) {
1157 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1159 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1160 const struct inet_diag_req_v2 *r)
1162 struct inet_diag_dump_data *cb_data = cb->data;
1163 const struct inet_diag_handler *handler;
1164 u32 prev_min_dump_alloc;
1165 int protocol, err = 0;
1167 protocol = inet_diag_get_protocol(r, cb_data);
1170 prev_min_dump_alloc = cb->min_dump_alloc;
1171 handler = inet_diag_lock_handler(protocol);
1172 if (!IS_ERR(handler))
1173 handler->dump(skb, cb, r);
1175 err = PTR_ERR(handler);
1176 inet_diag_unlock_handler(handler);
1178 /* The skb is not large enough to fit one sk info and
1179 * inet_sk_diag_fill() has requested for a larger skb.
1181 if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1182 err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1187 return err ? : skb->len;
1190 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1192 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1195 static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1197 const struct nlmsghdr *nlh = cb->nlh;
1198 struct inet_diag_dump_data *cb_data;
1199 struct sk_buff *skb = cb->skb;
1203 cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1207 err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1212 nla = cb_data->inet_diag_nla_bc;
1214 err = inet_diag_bc_audit(nla, skb);
1221 nla = cb_data->inet_diag_nla_bpf_stgs;
1223 struct bpf_sk_storage_diag *bpf_stg_diag;
1225 bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1226 if (IS_ERR(bpf_stg_diag)) {
1228 return PTR_ERR(bpf_stg_diag);
1230 cb_data->bpf_stg_diag = bpf_stg_diag;
1237 static int inet_diag_dump_start(struct netlink_callback *cb)
1239 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1242 static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1244 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1247 static int inet_diag_dump_done(struct netlink_callback *cb)
1249 struct inet_diag_dump_data *cb_data = cb->data;
1251 bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1257 static int inet_diag_type2proto(int type)
1260 case TCPDIAG_GETSOCK:
1262 case DCCPDIAG_GETSOCK:
1263 return IPPROTO_DCCP;
1269 static int inet_diag_dump_compat(struct sk_buff *skb,
1270 struct netlink_callback *cb)
1272 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1273 struct inet_diag_req_v2 req;
1275 req.sdiag_family = AF_UNSPEC; /* compatibility */
1276 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1277 req.idiag_ext = rc->idiag_ext;
1278 req.idiag_states = rc->idiag_states;
1281 return __inet_diag_dump(skb, cb, &req);
1284 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1285 const struct nlmsghdr *nlh)
1287 struct inet_diag_req *rc = nlmsg_data(nlh);
1288 struct inet_diag_req_v2 req;
1290 req.sdiag_family = rc->idiag_family;
1291 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1292 req.idiag_ext = rc->idiag_ext;
1293 req.idiag_states = rc->idiag_states;
1296 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1297 sizeof(struct inet_diag_req), &req);
1300 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1302 int hdrlen = sizeof(struct inet_diag_req);
1303 struct net *net = sock_net(skb->sk);
1305 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1306 nlmsg_len(nlh) < hdrlen)
1309 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1310 struct netlink_dump_control c = {
1311 .start = inet_diag_dump_start_compat,
1312 .done = inet_diag_dump_done,
1313 .dump = inet_diag_dump_compat,
1315 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1318 return inet_diag_get_exact_compat(skb, nlh);
1321 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1323 int hdrlen = sizeof(struct inet_diag_req_v2);
1324 struct net *net = sock_net(skb->sk);
1326 if (nlmsg_len(h) < hdrlen)
1329 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1330 h->nlmsg_flags & NLM_F_DUMP) {
1331 struct netlink_dump_control c = {
1332 .start = inet_diag_dump_start,
1333 .done = inet_diag_dump_done,
1334 .dump = inet_diag_dump,
1336 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1339 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1344 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1346 const struct inet_diag_handler *handler;
1347 struct nlmsghdr *nlh;
1348 struct nlattr *attr;
1349 struct inet_diag_msg *r;
1353 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1357 r = nlmsg_data(nlh);
1358 memset(r, 0, sizeof(*r));
1359 inet_diag_msg_common_fill(r, sk);
1360 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1361 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1362 r->idiag_state = sk->sk_state;
1364 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1365 nlmsg_cancel(skb, nlh);
1369 handler = inet_diag_lock_handler(sk->sk_protocol);
1370 if (IS_ERR(handler)) {
1371 inet_diag_unlock_handler(handler);
1372 nlmsg_cancel(skb, nlh);
1373 return PTR_ERR(handler);
1376 attr = handler->idiag_info_size
1377 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1378 handler->idiag_info_size,
1382 info = nla_data(attr);
1384 handler->idiag_get_info(sk, r, info);
1385 inet_diag_unlock_handler(handler);
1387 nlmsg_end(skb, nlh);
1391 static const struct sock_diag_handler inet_diag_handler = {
1393 .dump = inet_diag_handler_cmd,
1394 .get_info = inet_diag_handler_get_info,
1395 .destroy = inet_diag_handler_cmd,
1398 static const struct sock_diag_handler inet6_diag_handler = {
1400 .dump = inet_diag_handler_cmd,
1401 .get_info = inet_diag_handler_get_info,
1402 .destroy = inet_diag_handler_cmd,
1405 int inet_diag_register(const struct inet_diag_handler *h)
1407 const __u16 type = h->idiag_type;
1410 if (type >= IPPROTO_MAX)
1413 mutex_lock(&inet_diag_table_mutex);
1415 if (!inet_diag_table[type]) {
1416 inet_diag_table[type] = h;
1419 mutex_unlock(&inet_diag_table_mutex);
1423 EXPORT_SYMBOL_GPL(inet_diag_register);
1425 void inet_diag_unregister(const struct inet_diag_handler *h)
1427 const __u16 type = h->idiag_type;
1429 if (type >= IPPROTO_MAX)
1432 mutex_lock(&inet_diag_table_mutex);
1433 inet_diag_table[type] = NULL;
1434 mutex_unlock(&inet_diag_table_mutex);
1436 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1438 static int __init inet_diag_init(void)
1440 const int inet_diag_table_size = (IPPROTO_MAX *
1441 sizeof(struct inet_diag_handler *));
1444 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1445 if (!inet_diag_table)
1448 err = sock_diag_register(&inet_diag_handler);
1452 err = sock_diag_register(&inet6_diag_handler);
1456 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1461 sock_diag_unregister(&inet_diag_handler);
1463 kfree(inet_diag_table);
1467 static void __exit inet_diag_exit(void)
1469 sock_diag_unregister(&inet6_diag_handler);
1470 sock_diag_unregister(&inet_diag_handler);
1471 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1472 kfree(inet_diag_table);
1475 module_init(inet_diag_init);
1476 module_exit(inet_diag_exit);
1477 MODULE_LICENSE("GPL");
1478 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1479 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);