]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/ipv6/inet6_connection_sock.c
soreuseport: fast reuseport UDP socket selection
[mirror_ubuntu-jammy-kernel.git] / net / ipv6 / inet6_connection_sock.c
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Support for INET6 connection oriented protocols.
7 *
8 * Authors: See the TCPv6 sources
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or(at your option) any later version.
14 */
15
16 #include <linux/module.h>
17 #include <linux/in6.h>
18 #include <linux/ipv6.h>
19 #include <linux/jhash.h>
20 #include <linux/slab.h>
21
22 #include <net/addrconf.h>
23 #include <net/inet_connection_sock.h>
24 #include <net/inet_ecn.h>
25 #include <net/inet_hashtables.h>
26 #include <net/ip6_route.h>
27 #include <net/sock.h>
28 #include <net/inet6_connection_sock.h>
29
30 int inet6_csk_bind_conflict(const struct sock *sk,
31 const struct inet_bind_bucket *tb, bool relax)
32 {
33 const struct sock *sk2;
34 int reuse = sk->sk_reuse;
35 int reuseport = sk->sk_reuseport;
36 kuid_t uid = sock_i_uid((struct sock *)sk);
37
38 /* We must walk the whole port owner list in this case. -DaveM */
39 /*
40 * See comment in inet_csk_bind_conflict about sock lookup
41 * vs net namespaces issues.
42 */
43 sk_for_each_bound(sk2, &tb->owners) {
44 if (sk != sk2 &&
45 (!sk->sk_bound_dev_if ||
46 !sk2->sk_bound_dev_if ||
47 sk->sk_bound_dev_if == sk2->sk_bound_dev_if)) {
48 if ((!reuse || !sk2->sk_reuse ||
49 sk2->sk_state == TCP_LISTEN) &&
50 (!reuseport || !sk2->sk_reuseport ||
51 (sk2->sk_state != TCP_TIME_WAIT &&
52 !uid_eq(uid,
53 sock_i_uid((struct sock *)sk2))))) {
54 if (ipv6_rcv_saddr_equal(sk, sk2, true))
55 break;
56 }
57 if (!relax && reuse && sk2->sk_reuse &&
58 sk2->sk_state != TCP_LISTEN &&
59 ipv6_rcv_saddr_equal(sk, sk2, true))
60 break;
61 }
62 }
63
64 return sk2 != NULL;
65 }
66 EXPORT_SYMBOL_GPL(inet6_csk_bind_conflict);
67
68 struct dst_entry *inet6_csk_route_req(const struct sock *sk,
69 struct flowi6 *fl6,
70 const struct request_sock *req,
71 u8 proto)
72 {
73 struct inet_request_sock *ireq = inet_rsk(req);
74 const struct ipv6_pinfo *np = inet6_sk(sk);
75 struct in6_addr *final_p, final;
76 struct dst_entry *dst;
77
78 memset(fl6, 0, sizeof(*fl6));
79 fl6->flowi6_proto = proto;
80 fl6->daddr = ireq->ir_v6_rmt_addr;
81 rcu_read_lock();
82 final_p = fl6_update_dst(fl6, rcu_dereference(np->opt), &final);
83 rcu_read_unlock();
84 fl6->saddr = ireq->ir_v6_loc_addr;
85 fl6->flowi6_oif = ireq->ir_iif;
86 fl6->flowi6_mark = ireq->ir_mark;
87 fl6->fl6_dport = ireq->ir_rmt_port;
88 fl6->fl6_sport = htons(ireq->ir_num);
89 security_req_classify_flow(req, flowi6_to_flowi(fl6));
90
91 dst = ip6_dst_lookup_flow(sk, fl6, final_p);
92 if (IS_ERR(dst))
93 return NULL;
94
95 return dst;
96 }
97 EXPORT_SYMBOL(inet6_csk_route_req);
98
99 void inet6_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr)
100 {
101 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) uaddr;
102
103 sin6->sin6_family = AF_INET6;
104 sin6->sin6_addr = sk->sk_v6_daddr;
105 sin6->sin6_port = inet_sk(sk)->inet_dport;
106 /* We do not store received flowlabel for TCP */
107 sin6->sin6_flowinfo = 0;
108 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
109 sk->sk_bound_dev_if);
110 }
111 EXPORT_SYMBOL_GPL(inet6_csk_addr2sockaddr);
112
113 static inline
114 struct dst_entry *__inet6_csk_dst_check(struct sock *sk, u32 cookie)
115 {
116 return __sk_dst_check(sk, cookie);
117 }
118
119 static struct dst_entry *inet6_csk_route_socket(struct sock *sk,
120 struct flowi6 *fl6)
121 {
122 struct inet_sock *inet = inet_sk(sk);
123 struct ipv6_pinfo *np = inet6_sk(sk);
124 struct in6_addr *final_p, final;
125 struct dst_entry *dst;
126
127 memset(fl6, 0, sizeof(*fl6));
128 fl6->flowi6_proto = sk->sk_protocol;
129 fl6->daddr = sk->sk_v6_daddr;
130 fl6->saddr = np->saddr;
131 fl6->flowlabel = np->flow_label;
132 IP6_ECN_flow_xmit(sk, fl6->flowlabel);
133 fl6->flowi6_oif = sk->sk_bound_dev_if;
134 fl6->flowi6_mark = sk->sk_mark;
135 fl6->fl6_sport = inet->inet_sport;
136 fl6->fl6_dport = inet->inet_dport;
137 security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
138
139 rcu_read_lock();
140 final_p = fl6_update_dst(fl6, rcu_dereference(np->opt), &final);
141 rcu_read_unlock();
142
143 dst = __inet6_csk_dst_check(sk, np->dst_cookie);
144 if (!dst) {
145 dst = ip6_dst_lookup_flow(sk, fl6, final_p);
146
147 if (!IS_ERR(dst))
148 ip6_dst_store(sk, dst, NULL, NULL);
149 }
150 return dst;
151 }
152
153 int inet6_csk_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl_unused)
154 {
155 struct ipv6_pinfo *np = inet6_sk(sk);
156 struct flowi6 fl6;
157 struct dst_entry *dst;
158 int res;
159
160 dst = inet6_csk_route_socket(sk, &fl6);
161 if (IS_ERR(dst)) {
162 sk->sk_err_soft = -PTR_ERR(dst);
163 sk->sk_route_caps = 0;
164 kfree_skb(skb);
165 return PTR_ERR(dst);
166 }
167
168 rcu_read_lock();
169 skb_dst_set_noref(skb, dst);
170
171 /* Restore final destination back after routing done */
172 fl6.daddr = sk->sk_v6_daddr;
173
174 res = ip6_xmit(sk, skb, &fl6, rcu_dereference(np->opt),
175 np->tclass);
176 rcu_read_unlock();
177 return res;
178 }
179 EXPORT_SYMBOL_GPL(inet6_csk_xmit);
180
181 struct dst_entry *inet6_csk_update_pmtu(struct sock *sk, u32 mtu)
182 {
183 struct flowi6 fl6;
184 struct dst_entry *dst = inet6_csk_route_socket(sk, &fl6);
185
186 if (IS_ERR(dst))
187 return NULL;
188 dst->ops->update_pmtu(dst, sk, NULL, mtu);
189
190 dst = inet6_csk_route_socket(sk, &fl6);
191 return IS_ERR(dst) ? NULL : dst;
192 }
193 EXPORT_SYMBOL_GPL(inet6_csk_update_pmtu);