This commit was manufactured by cvs2svn to create branch 'vserver'.
[linux-2.6.git] / net / ipv4 / inet_connection_sock.c
diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c
new file mode 100644 (file)
index 0000000..9a01bb8
--- /dev/null
@@ -0,0 +1,699 @@
+/*
+ * INET                An implementation of the TCP/IP protocol suite for the LINUX
+ *             operating system.  INET is implemented using the  BSD Socket
+ *             interface as the means of communication with the user level.
+ *
+ *             Support for INET connection oriented protocols.
+ *
+ * Authors:    See the TCP sources
+ *
+ *             This program is free software; you can redistribute it and/or
+ *             modify it under the terms of the GNU General Public License
+ *             as published by the Free Software Foundation; either version
+ *             2 of the License, or(at your option) any later version.
+ */
+
+#include <linux/config.h>
+#include <linux/module.h>
+#include <linux/jhash.h>
+
+#include <net/inet_connection_sock.h>
+#include <net/inet_hashtables.h>
+#include <net/inet_timewait_sock.h>
+#include <net/ip.h>
+#include <net/route.h>
+#include <net/tcp_states.h>
+#include <net/xfrm.h>
+
+#ifdef INET_CSK_DEBUG
+const char inet_csk_timer_bug_msg[] = "inet_csk BUG: unknown timer value\n";
+EXPORT_SYMBOL(inet_csk_timer_bug_msg);
+#endif
+
+/*
+ * This array holds the first and last local port number.
+ * For high-usage systems, use sysctl to change this to
+ * 32768-61000
+ */
+int sysctl_local_port_range[2] = { 1024, 4999 };
+
+int inet_csk_bind_conflict(const struct sock *sk,
+                          const struct inet_bind_bucket *tb)
+{
+       const u32 sk_rcv_saddr = inet_rcv_saddr(sk);
+       struct sock *sk2;
+       struct hlist_node *node;
+       int reuse = sk->sk_reuse;
+
+       sk_for_each_bound(sk2, node, &tb->owners) {
+               if (sk != sk2 &&
+                   !inet_v6_ipv6only(sk2) &&
+                   (!sk->sk_bound_dev_if ||
+                    !sk2->sk_bound_dev_if ||
+                    sk->sk_bound_dev_if == sk2->sk_bound_dev_if)) {
+                       if (!reuse || !sk2->sk_reuse ||
+                           sk2->sk_state == TCP_LISTEN) {
+                               const u32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
+                               if (!sk2_rcv_saddr || !sk_rcv_saddr ||
+                                   sk2_rcv_saddr == sk_rcv_saddr)
+                                       break;
+                       }
+               }
+       }
+       return node != NULL;
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_bind_conflict);
+
+/* Obtain a reference to a local port for the given sock,
+ * if snum is zero it means select any available local port.
+ */
+int inet_csk_get_port(struct inet_hashinfo *hashinfo,
+                     struct sock *sk, unsigned short snum,
+                     int (*bind_conflict)(const struct sock *sk,
+                                          const struct inet_bind_bucket *tb))
+{
+       struct inet_bind_hashbucket *head;
+       struct hlist_node *node;
+       struct inet_bind_bucket *tb;
+       int ret;
+
+       local_bh_disable();
+       if (!snum) {
+               int low = sysctl_local_port_range[0];
+               int high = sysctl_local_port_range[1];
+               int remaining = (high - low) + 1;
+               int rover = net_random() % (high - low) + low;
+
+               do {
+                       head = &hashinfo->bhash[inet_bhashfn(rover, hashinfo->bhash_size)];
+                       spin_lock(&head->lock);
+                       inet_bind_bucket_for_each(tb, node, &head->chain)
+                               if (tb->port == rover)
+                                       goto next;
+                       break;
+               next:
+                       spin_unlock(&head->lock);
+                       if (++rover > high)
+                               rover = low;
+               } while (--remaining > 0);
+
+               /* Exhausted local port range during search?  It is not
+                * possible for us to be holding one of the bind hash
+                * locks if this test triggers, because if 'remaining'
+                * drops to zero, we broke out of the do/while loop at
+                * the top level, not from the 'break;' statement.
+                */
+               ret = 1;
+               if (remaining <= 0)
+                       goto fail;
+
+               /* OK, here is the one we will use.  HEAD is
+                * non-NULL and we hold it's mutex.
+                */
+               snum = rover;
+       } else {
+               head = &hashinfo->bhash[inet_bhashfn(snum, hashinfo->bhash_size)];
+               spin_lock(&head->lock);
+               inet_bind_bucket_for_each(tb, node, &head->chain)
+                       if (tb->port == snum)
+                               goto tb_found;
+       }
+       tb = NULL;
+       goto tb_not_found;
+tb_found:
+       if (!hlist_empty(&tb->owners)) {
+               if (sk->sk_reuse > 1)
+                       goto success;
+               if (tb->fastreuse > 0 &&
+                   sk->sk_reuse && sk->sk_state != TCP_LISTEN) {
+                       goto success;
+               } else {
+                       ret = 1;
+                       if (bind_conflict(sk, tb))
+                               goto fail_unlock;
+               }
+       }
+tb_not_found:
+       ret = 1;
+       if (!tb && (tb = inet_bind_bucket_create(hashinfo->bind_bucket_cachep, head, snum)) == NULL)
+               goto fail_unlock;
+       if (hlist_empty(&tb->owners)) {
+               if (sk->sk_reuse && sk->sk_state != TCP_LISTEN)
+                       tb->fastreuse = 1;
+               else
+                       tb->fastreuse = 0;
+       } else if (tb->fastreuse &&
+                  (!sk->sk_reuse || sk->sk_state == TCP_LISTEN))
+               tb->fastreuse = 0;
+success:
+       if (!inet_csk(sk)->icsk_bind_hash)
+               inet_bind_hash(sk, tb, snum);
+       BUG_TRAP(inet_csk(sk)->icsk_bind_hash == tb);
+       ret = 0;
+
+fail_unlock:
+       spin_unlock(&head->lock);
+fail:
+       local_bh_enable();
+       return ret;
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_get_port);
+
+/*
+ * Wait for an incoming connection, avoid race conditions. This must be called
+ * with the socket locked.
+ */
+static int inet_csk_wait_for_connect(struct sock *sk, long timeo)
+{
+       struct inet_connection_sock *icsk = inet_csk(sk);
+       DEFINE_WAIT(wait);
+       int err;
+
+       /*
+        * True wake-one mechanism for incoming connections: only
+        * one process gets woken up, not the 'whole herd'.
+        * Since we do not 'race & poll' for established sockets
+        * anymore, the common case will execute the loop only once.
+        *
+        * Subtle issue: "add_wait_queue_exclusive()" will be added
+        * after any current non-exclusive waiters, and we know that
+        * it will always _stay_ after any new non-exclusive waiters
+        * because all non-exclusive waiters are added at the
+        * beginning of the wait-queue. As such, it's ok to "drop"
+        * our exclusiveness temporarily when we get woken up without
+        * having to remove and re-insert us on the wait queue.
+        */
+       for (;;) {
+               prepare_to_wait_exclusive(sk->sk_sleep, &wait,
+                                         TASK_INTERRUPTIBLE);
+               release_sock(sk);
+               if (reqsk_queue_empty(&icsk->icsk_accept_queue))
+                       timeo = schedule_timeout(timeo);
+               lock_sock(sk);
+               err = 0;
+               if (!reqsk_queue_empty(&icsk->icsk_accept_queue))
+                       break;
+               err = -EINVAL;
+               if (sk->sk_state != TCP_LISTEN)
+                       break;
+               err = sock_intr_errno(timeo);
+               if (signal_pending(current))
+                       break;
+               err = -EAGAIN;
+               if (!timeo)
+                       break;
+       }
+       finish_wait(sk->sk_sleep, &wait);
+       return err;
+}
+
+/*
+ * This will accept the next outstanding connection.
+ */
+struct sock *inet_csk_accept(struct sock *sk, int flags, int *err)
+{
+       struct inet_connection_sock *icsk = inet_csk(sk);
+       struct sock *newsk;
+       int error;
+
+       lock_sock(sk);
+
+       /* We need to make sure that this socket is listening,
+        * and that it has something pending.
+        */
+       error = -EINVAL;
+       if (sk->sk_state != TCP_LISTEN)
+               goto out_err;
+
+       /* Find already established connection */
+       if (reqsk_queue_empty(&icsk->icsk_accept_queue)) {
+               long timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
+
+               /* If this is a non blocking socket don't sleep */
+               error = -EAGAIN;
+               if (!timeo)
+                       goto out_err;
+
+               error = inet_csk_wait_for_connect(sk, timeo);
+               if (error)
+                       goto out_err;
+       }
+
+       newsk = reqsk_queue_get_child(&icsk->icsk_accept_queue, sk);
+       BUG_TRAP(newsk->sk_state != TCP_SYN_RECV);
+out:
+       release_sock(sk);
+       return newsk;
+out_err:
+       newsk = NULL;
+       *err = error;
+       goto out;
+}
+
+EXPORT_SYMBOL(inet_csk_accept);
+
+/*
+ * Using different timers for retransmit, delayed acks and probes
+ * We may wish use just one timer maintaining a list of expire jiffies 
+ * to optimize.
+ */
+void inet_csk_init_xmit_timers(struct sock *sk,
+                              void (*retransmit_handler)(unsigned long),
+                              void (*delack_handler)(unsigned long),
+                              void (*keepalive_handler)(unsigned long))
+{
+       struct inet_connection_sock *icsk = inet_csk(sk);
+
+       init_timer(&icsk->icsk_retransmit_timer);
+       init_timer(&icsk->icsk_delack_timer);
+       init_timer(&sk->sk_timer);
+
+       icsk->icsk_retransmit_timer.function = retransmit_handler;
+       icsk->icsk_delack_timer.function     = delack_handler;
+       sk->sk_timer.function                = keepalive_handler;
+
+       icsk->icsk_retransmit_timer.data = 
+               icsk->icsk_delack_timer.data =
+                       sk->sk_timer.data  = (unsigned long)sk;
+
+       icsk->icsk_pending = icsk->icsk_ack.pending = 0;
+}
+
+EXPORT_SYMBOL(inet_csk_init_xmit_timers);
+
+void inet_csk_clear_xmit_timers(struct sock *sk)
+{
+       struct inet_connection_sock *icsk = inet_csk(sk);
+
+       icsk->icsk_pending = icsk->icsk_ack.pending = icsk->icsk_ack.blocked = 0;
+
+       sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
+       sk_stop_timer(sk, &icsk->icsk_delack_timer);
+       sk_stop_timer(sk, &sk->sk_timer);
+}
+
+EXPORT_SYMBOL(inet_csk_clear_xmit_timers);
+
+void inet_csk_delete_keepalive_timer(struct sock *sk)
+{
+       sk_stop_timer(sk, &sk->sk_timer);
+}
+
+EXPORT_SYMBOL(inet_csk_delete_keepalive_timer);
+
+void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long len)
+{
+       sk_reset_timer(sk, &sk->sk_timer, jiffies + len);
+}
+
+EXPORT_SYMBOL(inet_csk_reset_keepalive_timer);
+
+struct dst_entry* inet_csk_route_req(struct sock *sk,
+                                    const struct request_sock *req)
+{
+       struct rtable *rt;
+       const struct inet_request_sock *ireq = inet_rsk(req);
+       struct ip_options *opt = inet_rsk(req)->opt;
+       struct flowi fl = { .oif = sk->sk_bound_dev_if,
+                           .nl_u = { .ip4_u =
+                                     { .daddr = ((opt && opt->srr) ?
+                                                 opt->faddr :
+                                                 ireq->rmt_addr),
+                                       .saddr = ireq->loc_addr,
+                                       .tos = RT_CONN_FLAGS(sk) } },
+                           .proto = sk->sk_protocol,
+                           .uli_u = { .ports =
+                                      { .sport = inet_sk(sk)->sport,
+                                        .dport = ireq->rmt_port } } };
+
+       if (ip_route_output_flow(&rt, &fl, sk, 0)) {
+               IP_INC_STATS_BH(IPSTATS_MIB_OUTNOROUTES);
+               return NULL;
+       }
+       if (opt && opt->is_strictroute && rt->rt_dst != rt->rt_gateway) {
+               ip_rt_put(rt);
+               IP_INC_STATS_BH(IPSTATS_MIB_OUTNOROUTES);
+               return NULL;
+       }
+       return &rt->u.dst;
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_route_req);
+
+static inline u32 inet_synq_hash(const u32 raddr, const u16 rport,
+                                const u32 rnd, const u16 synq_hsize)
+{
+       return jhash_2words(raddr, (u32)rport, rnd) & (synq_hsize - 1);
+}
+
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+#define AF_INET_FAMILY(fam) ((fam) == AF_INET)
+#else
+#define AF_INET_FAMILY(fam) 1
+#endif
+
+struct request_sock *inet_csk_search_req(const struct sock *sk,
+                                        struct request_sock ***prevp,
+                                        const __u16 rport, const __u32 raddr,
+                                        const __u32 laddr)
+{
+       const struct inet_connection_sock *icsk = inet_csk(sk);
+       struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
+       struct request_sock *req, **prev;
+
+       for (prev = &lopt->syn_table[inet_synq_hash(raddr, rport, lopt->hash_rnd,
+                                                   lopt->nr_table_entries)];
+            (req = *prev) != NULL;
+            prev = &req->dl_next) {
+               const struct inet_request_sock *ireq = inet_rsk(req);
+
+               if (ireq->rmt_port == rport &&
+                   ireq->rmt_addr == raddr &&
+                   ireq->loc_addr == laddr &&
+                   AF_INET_FAMILY(req->rsk_ops->family)) {
+                       BUG_TRAP(!req->sk);
+                       *prevp = prev;
+                       break;
+               }
+       }
+
+       return req;
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_search_req);
+
+void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
+                                  unsigned long timeout)
+{
+       struct inet_connection_sock *icsk = inet_csk(sk);
+       struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
+       const u32 h = inet_synq_hash(inet_rsk(req)->rmt_addr, inet_rsk(req)->rmt_port,
+                                    lopt->hash_rnd, lopt->nr_table_entries);
+
+       reqsk_queue_hash_req(&icsk->icsk_accept_queue, h, req, timeout);
+       inet_csk_reqsk_queue_added(sk, timeout);
+}
+
+/* Only thing we need from tcp.h */
+extern int sysctl_tcp_synack_retries;
+
+EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_hash_add);
+
+void inet_csk_reqsk_queue_prune(struct sock *parent,
+                               const unsigned long interval,
+                               const unsigned long timeout,
+                               const unsigned long max_rto)
+{
+       struct inet_connection_sock *icsk = inet_csk(parent);
+       struct request_sock_queue *queue = &icsk->icsk_accept_queue;
+       struct listen_sock *lopt = queue->listen_opt;
+       int max_retries = icsk->icsk_syn_retries ? : sysctl_tcp_synack_retries;
+       int thresh = max_retries;
+       unsigned long now = jiffies;
+       struct request_sock **reqp, *req;
+       int i, budget;
+
+       if (lopt == NULL || lopt->qlen == 0)
+               return;
+
+       /* Normally all the openreqs are young and become mature
+        * (i.e. converted to established socket) for first timeout.
+        * If synack was not acknowledged for 3 seconds, it means
+        * one of the following things: synack was lost, ack was lost,
+        * rtt is high or nobody planned to ack (i.e. synflood).
+        * When server is a bit loaded, queue is populated with old
+        * open requests, reducing effective size of queue.
+        * When server is well loaded, queue size reduces to zero
+        * after several minutes of work. It is not synflood,
+        * it is normal operation. The solution is pruning
+        * too old entries overriding normal timeout, when
+        * situation becomes dangerous.
+        *
+        * Essentially, we reserve half of room for young
+        * embrions; and abort old ones without pity, if old
+        * ones are about to clog our table.
+        */
+       if (lopt->qlen>>(lopt->max_qlen_log-1)) {
+               int young = (lopt->qlen_young<<1);
+
+               while (thresh > 2) {
+                       if (lopt->qlen < young)
+                               break;
+                       thresh--;
+                       young <<= 1;
+               }
+       }
+
+       if (queue->rskq_defer_accept)
+               max_retries = queue->rskq_defer_accept;
+
+       budget = 2 * (lopt->nr_table_entries / (timeout / interval));
+       i = lopt->clock_hand;
+
+       do {
+               reqp=&lopt->syn_table[i];
+               while ((req = *reqp) != NULL) {
+                       if (time_after_eq(now, req->expires)) {
+                               if ((req->retrans < thresh ||
+                                    (inet_rsk(req)->acked && req->retrans < max_retries))
+                                   && !req->rsk_ops->rtx_syn_ack(parent, req, NULL)) {
+                                       unsigned long timeo;
+
+                                       if (req->retrans++ == 0)
+                                               lopt->qlen_young--;
+                                       timeo = min((timeout << req->retrans), max_rto);
+                                       req->expires = now + timeo;
+                                       reqp = &req->dl_next;
+                                       continue;
+                               }
+
+                               /* Drop this request */
+                               inet_csk_reqsk_queue_unlink(parent, req, reqp);
+                               reqsk_queue_removed(queue, req);
+                               reqsk_free(req);
+                               continue;
+                       }
+                       reqp = &req->dl_next;
+               }
+
+               i = (i + 1) & (lopt->nr_table_entries - 1);
+
+       } while (--budget > 0);
+
+       lopt->clock_hand = i;
+
+       if (lopt->qlen)
+               inet_csk_reset_keepalive_timer(parent, interval);
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_reqsk_queue_prune);
+
+struct sock *inet_csk_clone(struct sock *sk, const struct request_sock *req,
+                           const gfp_t priority)
+{
+       struct sock *newsk = sk_clone(sk, priority);
+
+       if (newsk != NULL) {
+               struct inet_connection_sock *newicsk = inet_csk(newsk);
+
+               newsk->sk_state = TCP_SYN_RECV;
+               newicsk->icsk_bind_hash = NULL;
+
+               inet_sk(newsk)->dport = inet_rsk(req)->rmt_port;
+               newsk->sk_write_space = sk_stream_write_space;
+
+               newicsk->icsk_retransmits = 0;
+               newicsk->icsk_backoff     = 0;
+               newicsk->icsk_probes_out  = 0;
+
+               /* Deinitialize accept_queue to trap illegal accesses. */
+               memset(&newicsk->icsk_accept_queue, 0, sizeof(newicsk->icsk_accept_queue));
+       }
+       return newsk;
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_clone);
+
+/*
+ * At this point, there should be no process reference to this
+ * socket, and thus no user references at all.  Therefore we
+ * can assume the socket waitqueue is inactive and nobody will
+ * try to jump onto it.
+ */
+void inet_csk_destroy_sock(struct sock *sk)
+{
+       BUG_TRAP(sk->sk_state == TCP_CLOSE);
+       BUG_TRAP(sock_flag(sk, SOCK_DEAD));
+
+       /* It cannot be in hash table! */
+       BUG_TRAP(sk_unhashed(sk));
+
+       /* If it has not 0 inet_sk(sk)->num, it must be bound */
+       BUG_TRAP(!inet_sk(sk)->num || inet_csk(sk)->icsk_bind_hash);
+
+       sk->sk_prot->destroy(sk);
+
+       sk_stream_kill_queues(sk);
+
+       xfrm_sk_free_policy(sk);
+
+       sk_refcnt_debug_release(sk);
+
+       atomic_dec(sk->sk_prot->orphan_count);
+       sock_put(sk);
+}
+
+EXPORT_SYMBOL(inet_csk_destroy_sock);
+
+int inet_csk_listen_start(struct sock *sk, const int nr_table_entries)
+{
+       struct inet_sock *inet = inet_sk(sk);
+       struct inet_connection_sock *icsk = inet_csk(sk);
+       int rc = reqsk_queue_alloc(&icsk->icsk_accept_queue, nr_table_entries);
+
+       if (rc != 0)
+               return rc;
+
+       sk->sk_max_ack_backlog = 0;
+       sk->sk_ack_backlog = 0;
+       inet_csk_delack_init(sk);
+
+       /* There is race window here: we announce ourselves listening,
+        * but this transition is still not validated by get_port().
+        * It is OK, because this socket enters to hash table only
+        * after validation is complete.
+        */
+       sk->sk_state = TCP_LISTEN;
+       if (!sk->sk_prot->get_port(sk, inet->num)) {
+               inet->sport = htons(inet->num);
+
+               sk_dst_reset(sk);
+               sk->sk_prot->hash(sk);
+
+               return 0;
+       }
+
+       sk->sk_state = TCP_CLOSE;
+       __reqsk_queue_destroy(&icsk->icsk_accept_queue);
+       return -EADDRINUSE;
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_listen_start);
+
+/*
+ *     This routine closes sockets which have been at least partially
+ *     opened, but not yet accepted.
+ */
+void inet_csk_listen_stop(struct sock *sk)
+{
+       struct inet_connection_sock *icsk = inet_csk(sk);
+       struct request_sock *acc_req;
+       struct request_sock *req;
+
+       inet_csk_delete_keepalive_timer(sk);
+
+       /* make all the listen_opt local to us */
+       acc_req = reqsk_queue_yank_acceptq(&icsk->icsk_accept_queue);
+
+       /* Following specs, it would be better either to send FIN
+        * (and enter FIN-WAIT-1, it is normal close)
+        * or to send active reset (abort).
+        * Certainly, it is pretty dangerous while synflood, but it is
+        * bad justification for our negligence 8)
+        * To be honest, we are not able to make either
+        * of the variants now.                 --ANK
+        */
+       reqsk_queue_destroy(&icsk->icsk_accept_queue);
+
+       while ((req = acc_req) != NULL) {
+               struct sock *child = req->sk;
+
+               acc_req = req->dl_next;
+
+               local_bh_disable();
+               bh_lock_sock(child);
+               BUG_TRAP(!sock_owned_by_user(child));
+               sock_hold(child);
+
+               sk->sk_prot->disconnect(child, O_NONBLOCK);
+
+               sock_orphan(child);
+
+               atomic_inc(sk->sk_prot->orphan_count);
+
+               inet_csk_destroy_sock(child);
+
+               bh_unlock_sock(child);
+               local_bh_enable();
+               sock_put(child);
+
+               sk_acceptq_removed(sk);
+               __reqsk_free(req);
+       }
+       BUG_TRAP(!sk->sk_ack_backlog);
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_listen_stop);
+
+void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr)
+{
+       struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
+       const struct inet_sock *inet = inet_sk(sk);
+
+       sin->sin_family         = AF_INET;
+       sin->sin_addr.s_addr    = inet->daddr;
+       sin->sin_port           = inet->dport;
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_addr2sockaddr);
+
+int inet_csk_ctl_sock_create(struct socket **sock, unsigned short family,
+                            unsigned short type, unsigned char protocol)
+{
+       int rc = sock_create_kern(family, type, protocol, sock);
+
+       if (rc == 0) {
+               (*sock)->sk->sk_allocation = GFP_ATOMIC;
+               inet_sk((*sock)->sk)->uc_ttl = -1;
+               /*
+                * Unhash it so that IP input processing does not even see it,
+                * we do not wish this socket to see incoming packets.
+                */
+               (*sock)->sk->sk_prot->unhash((*sock)->sk);
+       }
+       return rc;
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_ctl_sock_create);
+
+#ifdef CONFIG_COMPAT
+int inet_csk_compat_getsockopt(struct sock *sk, int level, int optname,
+                              char __user *optval, int __user *optlen)
+{
+       const struct inet_connection_sock *icsk = inet_csk(sk);
+
+       if (icsk->icsk_af_ops->compat_getsockopt != NULL)
+               return icsk->icsk_af_ops->compat_getsockopt(sk, level, optname,
+                                                           optval, optlen);
+       return icsk->icsk_af_ops->getsockopt(sk, level, optname,
+                                            optval, optlen);
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_compat_getsockopt);
+
+int inet_csk_compat_setsockopt(struct sock *sk, int level, int optname,
+                              char __user *optval, int optlen)
+{
+       const struct inet_connection_sock *icsk = inet_csk(sk);
+
+       if (icsk->icsk_af_ops->compat_setsockopt != NULL)
+               return icsk->icsk_af_ops->compat_setsockopt(sk, level, optname,
+                                                           optval, optlen);
+       return icsk->icsk_af_ops->setsockopt(sk, level, optname,
+                                            optval, optlen);
+}
+
+EXPORT_SYMBOL_GPL(inet_csk_compat_setsockopt);
+#endif