1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * inet_diag.c Module for monitoring INET transport protocols sockets.
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,
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);
451 static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
452 struct netlink_callback *cb,
453 u16 nlmsg_flags, bool net_admin)
455 struct request_sock *reqsk = inet_reqsk(sk);
456 struct inet_diag_msg *r;
457 struct nlmsghdr *nlh;
460 nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
461 cb->nlh->nlmsg_type, sizeof(*r), nlmsg_flags);
466 inet_diag_msg_common_fill(r, sk);
467 r->idiag_state = TCP_SYN_RECV;
469 r->idiag_retrans = reqsk->num_retrans;
471 BUILD_BUG_ON(offsetof(struct inet_request_sock, ir_cookie) !=
472 offsetof(struct sock, sk_cookie));
474 tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
475 r->idiag_expires = jiffies_delta_to_msecs(tmo);
481 if (net_admin && nla_put_u32(skb, INET_DIAG_MARK,
482 inet_rsk(reqsk)->ir_mark)) {
483 nlmsg_cancel(skb, nlh);
491 static int sk_diag_fill(struct sock *sk, struct sk_buff *skb,
492 struct netlink_callback *cb,
493 const struct inet_diag_req_v2 *r,
494 u16 nlmsg_flags, bool net_admin)
496 if (sk->sk_state == TCP_TIME_WAIT)
497 return inet_twsk_diag_fill(sk, skb, cb, nlmsg_flags);
499 if (sk->sk_state == TCP_NEW_SYN_RECV)
500 return inet_req_diag_fill(sk, skb, cb, nlmsg_flags, net_admin);
502 return inet_sk_diag_fill(sk, inet_csk(sk), skb, cb, r, nlmsg_flags,
506 struct sock *inet_diag_find_one_icsk(struct net *net,
507 struct inet_hashinfo *hashinfo,
508 const struct inet_diag_req_v2 *req)
513 if (req->sdiag_family == AF_INET)
514 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[0],
515 req->id.idiag_dport, req->id.idiag_src[0],
516 req->id.idiag_sport, req->id.idiag_if);
517 #if IS_ENABLED(CONFIG_IPV6)
518 else if (req->sdiag_family == AF_INET6) {
519 if (ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_dst) &&
520 ipv6_addr_v4mapped((struct in6_addr *)req->id.idiag_src))
521 sk = inet_lookup(net, hashinfo, NULL, 0, req->id.idiag_dst[3],
522 req->id.idiag_dport, req->id.idiag_src[3],
523 req->id.idiag_sport, req->id.idiag_if);
525 sk = inet6_lookup(net, hashinfo, NULL, 0,
526 (struct in6_addr *)req->id.idiag_dst,
528 (struct in6_addr *)req->id.idiag_src,
535 return ERR_PTR(-EINVAL);
539 return ERR_PTR(-ENOENT);
541 if (sock_diag_check_cookie(sk, req->id.idiag_cookie)) {
543 return ERR_PTR(-ENOENT);
548 EXPORT_SYMBOL_GPL(inet_diag_find_one_icsk);
550 int inet_diag_dump_one_icsk(struct inet_hashinfo *hashinfo,
551 struct netlink_callback *cb,
552 const struct inet_diag_req_v2 *req)
554 struct sk_buff *in_skb = cb->skb;
555 bool net_admin = netlink_net_capable(in_skb, CAP_NET_ADMIN);
556 struct net *net = sock_net(in_skb->sk);
561 sk = inet_diag_find_one_icsk(net, hashinfo, req);
565 rep = nlmsg_new(inet_sk_attr_size(sk, req, net_admin), GFP_KERNEL);
571 err = sk_diag_fill(sk, rep, cb, req, 0, net_admin);
573 WARN_ON(err == -EMSGSIZE);
577 err = netlink_unicast(net->diag_nlsk, rep, NETLINK_CB(in_skb).portid,
588 EXPORT_SYMBOL_GPL(inet_diag_dump_one_icsk);
590 static int inet_diag_cmd_exact(int cmd, struct sk_buff *in_skb,
591 const struct nlmsghdr *nlh,
593 const struct inet_diag_req_v2 *req)
595 const struct inet_diag_handler *handler;
596 struct inet_diag_dump_data dump_data;
599 memset(&dump_data, 0, sizeof(dump_data));
600 err = inet_diag_parse_attrs(nlh, hdrlen, dump_data.req_nlas);
604 protocol = inet_diag_get_protocol(req, &dump_data);
606 handler = inet_diag_lock_handler(protocol);
607 if (IS_ERR(handler)) {
608 err = PTR_ERR(handler);
609 } else if (cmd == SOCK_DIAG_BY_FAMILY) {
610 struct netlink_callback cb = {
615 err = handler->dump_one(&cb, req);
616 } else if (cmd == SOCK_DESTROY && handler->destroy) {
617 err = handler->destroy(in_skb, req);
621 inet_diag_unlock_handler(handler);
626 static int bitstring_match(const __be32 *a1, const __be32 *a2, int bits)
628 int words = bits >> 5;
633 if (memcmp(a1, a2, words << 2))
643 mask = htonl((0xffffffff) << (32 - bits));
645 if ((w1 ^ w2) & mask)
652 static int inet_diag_bc_run(const struct nlattr *_bc,
653 const struct inet_diag_entry *entry)
655 const void *bc = nla_data(_bc);
656 int len = nla_len(_bc);
660 const struct inet_diag_bc_op *op = bc;
663 case INET_DIAG_BC_NOP:
665 case INET_DIAG_BC_JMP:
668 case INET_DIAG_BC_S_EQ:
669 yes = entry->sport == op[1].no;
671 case INET_DIAG_BC_S_GE:
672 yes = entry->sport >= op[1].no;
674 case INET_DIAG_BC_S_LE:
675 yes = entry->sport <= op[1].no;
677 case INET_DIAG_BC_D_EQ:
678 yes = entry->dport == op[1].no;
680 case INET_DIAG_BC_D_GE:
681 yes = entry->dport >= op[1].no;
683 case INET_DIAG_BC_D_LE:
684 yes = entry->dport <= op[1].no;
686 case INET_DIAG_BC_AUTO:
687 yes = !(entry->userlocks & SOCK_BINDPORT_LOCK);
689 case INET_DIAG_BC_S_COND:
690 case INET_DIAG_BC_D_COND: {
691 const struct inet_diag_hostcond *cond;
694 cond = (const struct inet_diag_hostcond *)(op + 1);
695 if (cond->port != -1 &&
696 cond->port != (op->code == INET_DIAG_BC_S_COND ?
697 entry->sport : entry->dport)) {
702 if (op->code == INET_DIAG_BC_S_COND)
707 if (cond->family != AF_UNSPEC &&
708 cond->family != entry->family) {
709 if (entry->family == AF_INET6 &&
710 cond->family == AF_INET) {
711 if (addr[0] == 0 && addr[1] == 0 &&
712 addr[2] == htonl(0xffff) &&
713 bitstring_match(addr + 3,
722 if (cond->prefix_len == 0)
724 if (bitstring_match(addr, cond->addr,
730 case INET_DIAG_BC_DEV_COND: {
733 ifindex = *((const u32 *)(op + 1));
734 if (ifindex != entry->ifindex)
738 case INET_DIAG_BC_MARK_COND: {
739 struct inet_diag_markcond *cond;
741 cond = (struct inet_diag_markcond *)(op + 1);
742 if ((entry->mark & cond->mask) != cond->mark)
746 #ifdef CONFIG_SOCK_CGROUP_DATA
747 case INET_DIAG_BC_CGROUP_COND: {
750 cgroup_id = get_unaligned((const u64 *)(op + 1));
751 if (cgroup_id != entry->cgroup_id)
769 /* This helper is available for all sockets (ESTABLISH, TIMEWAIT, SYN_RECV)
771 static void entry_fill_addrs(struct inet_diag_entry *entry,
772 const struct sock *sk)
774 #if IS_ENABLED(CONFIG_IPV6)
775 if (sk->sk_family == AF_INET6) {
776 entry->saddr = sk->sk_v6_rcv_saddr.s6_addr32;
777 entry->daddr = sk->sk_v6_daddr.s6_addr32;
781 entry->saddr = &sk->sk_rcv_saddr;
782 entry->daddr = &sk->sk_daddr;
786 int inet_diag_bc_sk(const struct nlattr *bc, struct sock *sk)
788 struct inet_sock *inet = inet_sk(sk);
789 struct inet_diag_entry entry;
794 entry.family = sk->sk_family;
795 entry_fill_addrs(&entry, sk);
796 entry.sport = inet->inet_num;
797 entry.dport = ntohs(inet->inet_dport);
798 entry.ifindex = sk->sk_bound_dev_if;
799 entry.userlocks = sk_fullsock(sk) ? sk->sk_userlocks : 0;
801 entry.mark = sk->sk_mark;
802 else if (sk->sk_state == TCP_NEW_SYN_RECV)
803 entry.mark = inet_rsk(inet_reqsk(sk))->ir_mark;
806 #ifdef CONFIG_SOCK_CGROUP_DATA
807 entry.cgroup_id = sk_fullsock(sk) ?
808 cgroup_id(sock_cgroup_ptr(&sk->sk_cgrp_data)) : 0;
811 return inet_diag_bc_run(bc, &entry);
813 EXPORT_SYMBOL_GPL(inet_diag_bc_sk);
815 static int valid_cc(const void *bc, int len, int cc)
818 const struct inet_diag_bc_op *op = bc;
824 if (op->yes < 4 || op->yes & 3)
832 /* data is u32 ifindex */
833 static bool valid_devcond(const struct inet_diag_bc_op *op, int len,
836 /* Check ifindex space. */
837 *min_len += sizeof(u32);
843 /* Validate an inet_diag_hostcond. */
844 static bool valid_hostcond(const struct inet_diag_bc_op *op, int len,
847 struct inet_diag_hostcond *cond;
850 /* Check hostcond space. */
851 *min_len += sizeof(struct inet_diag_hostcond);
854 cond = (struct inet_diag_hostcond *)(op + 1);
856 /* Check address family and address length. */
857 switch (cond->family) {
862 addr_len = sizeof(struct in_addr);
865 addr_len = sizeof(struct in6_addr);
870 *min_len += addr_len;
874 /* Check prefix length (in bits) vs address length (in bytes). */
875 if (cond->prefix_len > 8 * addr_len)
881 /* Validate a port comparison operator. */
882 static bool valid_port_comparison(const struct inet_diag_bc_op *op,
883 int len, int *min_len)
885 /* Port comparisons put the port in a follow-on inet_diag_bc_op. */
886 *min_len += sizeof(struct inet_diag_bc_op);
892 static bool valid_markcond(const struct inet_diag_bc_op *op, int len,
895 *min_len += sizeof(struct inet_diag_markcond);
896 return len >= *min_len;
899 #ifdef CONFIG_SOCK_CGROUP_DATA
900 static bool valid_cgroupcond(const struct inet_diag_bc_op *op, int len,
903 *min_len += sizeof(u64);
904 return len >= *min_len;
908 static int inet_diag_bc_audit(const struct nlattr *attr,
909 const struct sk_buff *skb)
911 bool net_admin = netlink_net_capable(skb, CAP_NET_ADMIN);
912 const void *bytecode, *bc;
913 int bytecode_len, len;
915 if (!attr || nla_len(attr) < sizeof(struct inet_diag_bc_op))
918 bytecode = bc = nla_data(attr);
919 len = bytecode_len = nla_len(attr);
922 int min_len = sizeof(struct inet_diag_bc_op);
923 const struct inet_diag_bc_op *op = bc;
926 case INET_DIAG_BC_S_COND:
927 case INET_DIAG_BC_D_COND:
928 if (!valid_hostcond(bc, len, &min_len))
931 case INET_DIAG_BC_DEV_COND:
932 if (!valid_devcond(bc, len, &min_len))
935 case INET_DIAG_BC_S_EQ:
936 case INET_DIAG_BC_S_GE:
937 case INET_DIAG_BC_S_LE:
938 case INET_DIAG_BC_D_EQ:
939 case INET_DIAG_BC_D_GE:
940 case INET_DIAG_BC_D_LE:
941 if (!valid_port_comparison(bc, len, &min_len))
944 case INET_DIAG_BC_MARK_COND:
947 if (!valid_markcond(bc, len, &min_len))
950 #ifdef CONFIG_SOCK_CGROUP_DATA
951 case INET_DIAG_BC_CGROUP_COND:
952 if (!valid_cgroupcond(bc, len, &min_len))
956 case INET_DIAG_BC_AUTO:
957 case INET_DIAG_BC_JMP:
958 case INET_DIAG_BC_NOP:
964 if (op->code != INET_DIAG_BC_NOP) {
965 if (op->no < min_len || op->no > len + 4 || op->no & 3)
968 !valid_cc(bytecode, bytecode_len, len - op->no))
972 if (op->yes < min_len || op->yes > len + 4 || op->yes & 3)
977 return len == 0 ? 0 : -EINVAL;
980 static void twsk_build_assert(void)
982 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_family) !=
983 offsetof(struct sock, sk_family));
985 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_num) !=
986 offsetof(struct inet_sock, inet_num));
988 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_dport) !=
989 offsetof(struct inet_sock, inet_dport));
991 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_rcv_saddr) !=
992 offsetof(struct inet_sock, inet_rcv_saddr));
994 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_daddr) !=
995 offsetof(struct inet_sock, inet_daddr));
997 #if IS_ENABLED(CONFIG_IPV6)
998 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_rcv_saddr) !=
999 offsetof(struct sock, sk_v6_rcv_saddr));
1001 BUILD_BUG_ON(offsetof(struct inet_timewait_sock, tw_v6_daddr) !=
1002 offsetof(struct sock, sk_v6_daddr));
1006 void inet_diag_dump_icsk(struct inet_hashinfo *hashinfo, struct sk_buff *skb,
1007 struct netlink_callback *cb,
1008 const struct inet_diag_req_v2 *r)
1010 bool net_admin = netlink_net_capable(cb->skb, CAP_NET_ADMIN);
1011 struct inet_diag_dump_data *cb_data = cb->data;
1012 struct net *net = sock_net(skb->sk);
1013 u32 idiag_states = r->idiag_states;
1014 int i, num, s_i, s_num;
1018 bc = cb_data->inet_diag_nla_bc;
1019 if (idiag_states & TCPF_SYN_RECV)
1020 idiag_states |= TCPF_NEW_SYN_RECV;
1022 s_num = num = cb->args[2];
1024 if (cb->args[0] == 0) {
1025 if (!(idiag_states & TCPF_LISTEN) || r->id.idiag_dport)
1026 goto skip_listen_ht;
1028 for (i = s_i; i < INET_LHTABLE_SIZE; i++) {
1029 struct inet_listen_hashbucket *ilb;
1030 struct hlist_nulls_node *node;
1033 ilb = &hashinfo->listening_hash[i];
1034 spin_lock(&ilb->lock);
1035 sk_nulls_for_each(sk, node, &ilb->nulls_head) {
1036 struct inet_sock *inet = inet_sk(sk);
1038 if (!net_eq(sock_net(sk), net))
1046 if (r->sdiag_family != AF_UNSPEC &&
1047 sk->sk_family != r->sdiag_family)
1050 if (r->id.idiag_sport != inet->inet_sport &&
1054 if (!inet_diag_bc_sk(bc, sk))
1057 if (inet_sk_diag_fill(sk, inet_csk(sk), skb,
1060 spin_unlock(&ilb->lock);
1067 spin_unlock(&ilb->lock);
1073 s_i = num = s_num = 0;
1076 if (!(idiag_states & ~TCPF_LISTEN))
1080 for (i = s_i; i <= hashinfo->ehash_mask; i++) {
1081 struct inet_ehash_bucket *head = &hashinfo->ehash[i];
1082 spinlock_t *lock = inet_ehash_lockp(hashinfo, i);
1083 struct hlist_nulls_node *node;
1084 struct sock *sk_arr[SKARR_SZ];
1085 int num_arr[SKARR_SZ];
1086 int idx, accum, res;
1088 if (hlist_nulls_empty(&head->chain))
1098 sk_nulls_for_each(sk, node, &head->chain) {
1101 if (!net_eq(sock_net(sk), net))
1105 state = (sk->sk_state == TCP_TIME_WAIT) ?
1106 inet_twsk(sk)->tw_substate : sk->sk_state;
1107 if (!(idiag_states & (1 << state)))
1109 if (r->sdiag_family != AF_UNSPEC &&
1110 sk->sk_family != r->sdiag_family)
1112 if (r->id.idiag_sport != htons(sk->sk_num) &&
1115 if (r->id.idiag_dport != sk->sk_dport &&
1118 twsk_build_assert();
1120 if (!inet_diag_bc_sk(bc, sk))
1123 if (!refcount_inc_not_zero(&sk->sk_refcnt))
1126 num_arr[accum] = num;
1128 if (++accum == SKARR_SZ)
1133 spin_unlock_bh(lock);
1135 for (idx = 0; idx < accum; idx++) {
1137 res = sk_diag_fill(sk_arr[idx], skb, cb, r,
1138 NLM_F_MULTI, net_admin);
1142 sock_gen_put(sk_arr[idx]);
1147 if (accum == SKARR_SZ) {
1159 EXPORT_SYMBOL_GPL(inet_diag_dump_icsk);
1161 static int __inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
1162 const struct inet_diag_req_v2 *r)
1164 struct inet_diag_dump_data *cb_data = cb->data;
1165 const struct inet_diag_handler *handler;
1166 u32 prev_min_dump_alloc;
1167 int protocol, err = 0;
1169 protocol = inet_diag_get_protocol(r, cb_data);
1172 prev_min_dump_alloc = cb->min_dump_alloc;
1173 handler = inet_diag_lock_handler(protocol);
1174 if (!IS_ERR(handler))
1175 handler->dump(skb, cb, r);
1177 err = PTR_ERR(handler);
1178 inet_diag_unlock_handler(handler);
1180 /* The skb is not large enough to fit one sk info and
1181 * inet_sk_diag_fill() has requested for a larger skb.
1183 if (!skb->len && cb->min_dump_alloc > prev_min_dump_alloc) {
1184 err = pskb_expand_head(skb, 0, cb->min_dump_alloc, GFP_KERNEL);
1189 return err ? : skb->len;
1192 static int inet_diag_dump(struct sk_buff *skb, struct netlink_callback *cb)
1194 return __inet_diag_dump(skb, cb, nlmsg_data(cb->nlh));
1197 static int __inet_diag_dump_start(struct netlink_callback *cb, int hdrlen)
1199 const struct nlmsghdr *nlh = cb->nlh;
1200 struct inet_diag_dump_data *cb_data;
1201 struct sk_buff *skb = cb->skb;
1205 cb_data = kzalloc(sizeof(*cb_data), GFP_KERNEL);
1209 err = inet_diag_parse_attrs(nlh, hdrlen, cb_data->req_nlas);
1214 nla = cb_data->inet_diag_nla_bc;
1216 err = inet_diag_bc_audit(nla, skb);
1223 nla = cb_data->inet_diag_nla_bpf_stgs;
1225 struct bpf_sk_storage_diag *bpf_stg_diag;
1227 bpf_stg_diag = bpf_sk_storage_diag_alloc(nla);
1228 if (IS_ERR(bpf_stg_diag)) {
1230 return PTR_ERR(bpf_stg_diag);
1232 cb_data->bpf_stg_diag = bpf_stg_diag;
1239 static int inet_diag_dump_start(struct netlink_callback *cb)
1241 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req_v2));
1244 static int inet_diag_dump_start_compat(struct netlink_callback *cb)
1246 return __inet_diag_dump_start(cb, sizeof(struct inet_diag_req));
1249 static int inet_diag_dump_done(struct netlink_callback *cb)
1251 struct inet_diag_dump_data *cb_data = cb->data;
1253 bpf_sk_storage_diag_free(cb_data->bpf_stg_diag);
1259 static int inet_diag_type2proto(int type)
1262 case TCPDIAG_GETSOCK:
1264 case DCCPDIAG_GETSOCK:
1265 return IPPROTO_DCCP;
1271 static int inet_diag_dump_compat(struct sk_buff *skb,
1272 struct netlink_callback *cb)
1274 struct inet_diag_req *rc = nlmsg_data(cb->nlh);
1275 struct inet_diag_req_v2 req;
1277 req.sdiag_family = AF_UNSPEC; /* compatibility */
1278 req.sdiag_protocol = inet_diag_type2proto(cb->nlh->nlmsg_type);
1279 req.idiag_ext = rc->idiag_ext;
1280 req.idiag_states = rc->idiag_states;
1283 return __inet_diag_dump(skb, cb, &req);
1286 static int inet_diag_get_exact_compat(struct sk_buff *in_skb,
1287 const struct nlmsghdr *nlh)
1289 struct inet_diag_req *rc = nlmsg_data(nlh);
1290 struct inet_diag_req_v2 req;
1292 req.sdiag_family = rc->idiag_family;
1293 req.sdiag_protocol = inet_diag_type2proto(nlh->nlmsg_type);
1294 req.idiag_ext = rc->idiag_ext;
1295 req.idiag_states = rc->idiag_states;
1298 return inet_diag_cmd_exact(SOCK_DIAG_BY_FAMILY, in_skb, nlh,
1299 sizeof(struct inet_diag_req), &req);
1302 static int inet_diag_rcv_msg_compat(struct sk_buff *skb, struct nlmsghdr *nlh)
1304 int hdrlen = sizeof(struct inet_diag_req);
1305 struct net *net = sock_net(skb->sk);
1307 if (nlh->nlmsg_type >= INET_DIAG_GETSOCK_MAX ||
1308 nlmsg_len(nlh) < hdrlen)
1311 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1312 struct netlink_dump_control c = {
1313 .start = inet_diag_dump_start_compat,
1314 .done = inet_diag_dump_done,
1315 .dump = inet_diag_dump_compat,
1317 return netlink_dump_start(net->diag_nlsk, skb, nlh, &c);
1320 return inet_diag_get_exact_compat(skb, nlh);
1323 static int inet_diag_handler_cmd(struct sk_buff *skb, struct nlmsghdr *h)
1325 int hdrlen = sizeof(struct inet_diag_req_v2);
1326 struct net *net = sock_net(skb->sk);
1328 if (nlmsg_len(h) < hdrlen)
1331 if (h->nlmsg_type == SOCK_DIAG_BY_FAMILY &&
1332 h->nlmsg_flags & NLM_F_DUMP) {
1333 struct netlink_dump_control c = {
1334 .start = inet_diag_dump_start,
1335 .done = inet_diag_dump_done,
1336 .dump = inet_diag_dump,
1338 return netlink_dump_start(net->diag_nlsk, skb, h, &c);
1341 return inet_diag_cmd_exact(h->nlmsg_type, skb, h, hdrlen,
1346 int inet_diag_handler_get_info(struct sk_buff *skb, struct sock *sk)
1348 const struct inet_diag_handler *handler;
1349 struct nlmsghdr *nlh;
1350 struct nlattr *attr;
1351 struct inet_diag_msg *r;
1355 nlh = nlmsg_put(skb, 0, 0, SOCK_DIAG_BY_FAMILY, sizeof(*r), 0);
1359 r = nlmsg_data(nlh);
1360 memset(r, 0, sizeof(*r));
1361 inet_diag_msg_common_fill(r, sk);
1362 if (sk->sk_type == SOCK_DGRAM || sk->sk_type == SOCK_STREAM)
1363 r->id.idiag_sport = inet_sk(sk)->inet_sport;
1364 r->idiag_state = sk->sk_state;
1366 if ((err = nla_put_u8(skb, INET_DIAG_PROTOCOL, sk->sk_protocol))) {
1367 nlmsg_cancel(skb, nlh);
1371 handler = inet_diag_lock_handler(sk->sk_protocol);
1372 if (IS_ERR(handler)) {
1373 inet_diag_unlock_handler(handler);
1374 nlmsg_cancel(skb, nlh);
1375 return PTR_ERR(handler);
1378 attr = handler->idiag_info_size
1379 ? nla_reserve_64bit(skb, INET_DIAG_INFO,
1380 handler->idiag_info_size,
1384 info = nla_data(attr);
1386 handler->idiag_get_info(sk, r, info);
1387 inet_diag_unlock_handler(handler);
1389 nlmsg_end(skb, nlh);
1393 static const struct sock_diag_handler inet_diag_handler = {
1395 .dump = inet_diag_handler_cmd,
1396 .get_info = inet_diag_handler_get_info,
1397 .destroy = inet_diag_handler_cmd,
1400 static const struct sock_diag_handler inet6_diag_handler = {
1402 .dump = inet_diag_handler_cmd,
1403 .get_info = inet_diag_handler_get_info,
1404 .destroy = inet_diag_handler_cmd,
1407 int inet_diag_register(const struct inet_diag_handler *h)
1409 const __u16 type = h->idiag_type;
1412 if (type >= IPPROTO_MAX)
1415 mutex_lock(&inet_diag_table_mutex);
1417 if (!inet_diag_table[type]) {
1418 inet_diag_table[type] = h;
1421 mutex_unlock(&inet_diag_table_mutex);
1425 EXPORT_SYMBOL_GPL(inet_diag_register);
1427 void inet_diag_unregister(const struct inet_diag_handler *h)
1429 const __u16 type = h->idiag_type;
1431 if (type >= IPPROTO_MAX)
1434 mutex_lock(&inet_diag_table_mutex);
1435 inet_diag_table[type] = NULL;
1436 mutex_unlock(&inet_diag_table_mutex);
1438 EXPORT_SYMBOL_GPL(inet_diag_unregister);
1440 static int __init inet_diag_init(void)
1442 const int inet_diag_table_size = (IPPROTO_MAX *
1443 sizeof(struct inet_diag_handler *));
1446 inet_diag_table = kzalloc(inet_diag_table_size, GFP_KERNEL);
1447 if (!inet_diag_table)
1450 err = sock_diag_register(&inet_diag_handler);
1454 err = sock_diag_register(&inet6_diag_handler);
1458 sock_diag_register_inet_compat(inet_diag_rcv_msg_compat);
1463 sock_diag_unregister(&inet_diag_handler);
1465 kfree(inet_diag_table);
1469 static void __exit inet_diag_exit(void)
1471 sock_diag_unregister(&inet6_diag_handler);
1472 sock_diag_unregister(&inet_diag_handler);
1473 sock_diag_unregister_inet_compat(inet_diag_rcv_msg_compat);
1474 kfree(inet_diag_table);
1477 module_init(inet_diag_init);
1478 module_exit(inet_diag_exit);
1479 MODULE_LICENSE("GPL");
1480 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 2 /* AF_INET */);
1481 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_NETLINK, NETLINK_SOCK_DIAG, 10 /* AF_INET6 */);