1// SPDX-License-Identifier: GPL-2.0
  2/* Copyright (c) 2021 Cong Wang <cong.wang@bytedance.com> */
  3
  4#include <linux/skmsg.h>
  5#include <linux/bpf.h>
  6#include <net/sock.h>
  7#include <net/af_unix.h>
  8
  9#define unix_sk_has_data(__sk, __psock)					\
 10		({	!skb_queue_empty(&__sk->sk_receive_queue) ||	\
 11			!skb_queue_empty(&__psock->ingress_skb) ||	\
 12			!list_empty(&__psock->ingress_msg);		\
 13		})
 14
 15static int unix_msg_wait_data(struct sock *sk, struct sk_psock *psock,
 16			      long timeo)
 17{
 18	DEFINE_WAIT_FUNC(wait, woken_wake_function);
 19	struct unix_sock *u = unix_sk(sk);
 20	int ret = 0;
 21
 22	if (sk->sk_shutdown & RCV_SHUTDOWN)
 23		return 1;
 24
 25	if (!timeo)
 26		return ret;
 27
 28	add_wait_queue(sk_sleep(sk), &wait);
 29	sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
 30	if (!unix_sk_has_data(sk, psock)) {
 31		mutex_unlock(&u->iolock);
 32		wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
 33		mutex_lock(&u->iolock);
 34		ret = unix_sk_has_data(sk, psock);
 35	}
 36	sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
 37	remove_wait_queue(sk_sleep(sk), &wait);
 38	return ret;
 39}
 40
 41static int __unix_recvmsg(struct sock *sk, struct msghdr *msg,
 42			  size_t len, int flags)
 43{
 44	if (sk->sk_type == SOCK_DGRAM)
 45		return __unix_dgram_recvmsg(sk, msg, len, flags);
 46	else
 47		return __unix_stream_recvmsg(sk, msg, len, flags);
 48}
 49
 50static int unix_bpf_recvmsg(struct sock *sk, struct msghdr *msg,
 51			    size_t len, int flags, int *addr_len)
 52{
 53	struct unix_sock *u = unix_sk(sk);
 54	struct sk_psock *psock;
 55	int copied;
 56
 57	if (flags & MSG_OOB)
 58		return -EOPNOTSUPP;
 59
 60	if (!len)
 61		return 0;
 62
 63	psock = sk_psock_get(sk);
 64	if (unlikely(!psock))
 65		return __unix_recvmsg(sk, msg, len, flags);
 66
 67	mutex_lock(&u->iolock);
 68	if (!skb_queue_empty(&sk->sk_receive_queue) &&
 69	    sk_psock_queue_empty(psock)) {
 70		mutex_unlock(&u->iolock);
 71		sk_psock_put(sk, psock);
 72		return __unix_recvmsg(sk, msg, len, flags);
 73	}
 74
 75msg_bytes_ready:
 76	copied = sk_msg_recvmsg(sk, psock, msg, len, flags);
 77	if (!copied) {
 78		long timeo;
 79		int data;
 80
 81		timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
 82		data = unix_msg_wait_data(sk, psock, timeo);
 83		if (data) {
 84			if (!sk_psock_queue_empty(psock))
 85				goto msg_bytes_ready;
 86			mutex_unlock(&u->iolock);
 87			sk_psock_put(sk, psock);
 88			return __unix_recvmsg(sk, msg, len, flags);
 89		}
 90		copied = -EAGAIN;
 91	}
 92	mutex_unlock(&u->iolock);
 93	sk_psock_put(sk, psock);
 94	return copied;
 95}
 96
 97static struct proto *unix_dgram_prot_saved __read_mostly;
 98static DEFINE_SPINLOCK(unix_dgram_prot_lock);
 99static struct proto unix_dgram_bpf_prot;
100
101static struct proto *unix_stream_prot_saved __read_mostly;
102static DEFINE_SPINLOCK(unix_stream_prot_lock);
103static struct proto unix_stream_bpf_prot;
104
105static void unix_dgram_bpf_rebuild_protos(struct proto *prot, const struct proto *base)
106{
107	*prot        = *base;
108	prot->close  = sock_map_close;
109	prot->recvmsg = unix_bpf_recvmsg;
110	prot->sock_is_readable = sk_msg_is_readable;
111}
112
113static void unix_stream_bpf_rebuild_protos(struct proto *prot,
114					   const struct proto *base)
115{
116	*prot        = *base;
117	prot->close  = sock_map_close;
118	prot->recvmsg = unix_bpf_recvmsg;
119	prot->sock_is_readable = sk_msg_is_readable;
120	prot->unhash  = sock_map_unhash;
121}
122
123static void unix_dgram_bpf_check_needs_rebuild(struct proto *ops)
124{
125	if (unlikely(ops != smp_load_acquire(&unix_dgram_prot_saved))) {
126		spin_lock_bh(&unix_dgram_prot_lock);
127		if (likely(ops != unix_dgram_prot_saved)) {
128			unix_dgram_bpf_rebuild_protos(&unix_dgram_bpf_prot, ops);
129			smp_store_release(&unix_dgram_prot_saved, ops);
130		}
131		spin_unlock_bh(&unix_dgram_prot_lock);
132	}
133}
134
135static void unix_stream_bpf_check_needs_rebuild(struct proto *ops)
136{
137	if (unlikely(ops != smp_load_acquire(&unix_stream_prot_saved))) {
138		spin_lock_bh(&unix_stream_prot_lock);
139		if (likely(ops != unix_stream_prot_saved)) {
140			unix_stream_bpf_rebuild_protos(&unix_stream_bpf_prot, ops);
141			smp_store_release(&unix_stream_prot_saved, ops);
142		}
143		spin_unlock_bh(&unix_stream_prot_lock);
144	}
145}
146
147int unix_dgram_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore)
148{
149	if (sk->sk_type != SOCK_DGRAM)
150		return -EOPNOTSUPP;
151
152	if (restore) {
153		sk->sk_write_space = psock->saved_write_space;
154		sock_replace_proto(sk, psock->sk_proto);
155		return 0;
156	}
157
158	unix_dgram_bpf_check_needs_rebuild(psock->sk_proto);
159	sock_replace_proto(sk, &unix_dgram_bpf_prot);
160	return 0;
161}
162
163int unix_stream_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore)
164{
165	struct sock *sk_pair;
166
167	/* Restore does not decrement the sk_pair reference yet because we must
168	 * keep the a reference to the socket until after an RCU grace period
169	 * and any pending sends have completed.
170	 */
171	if (restore) {
172		sk->sk_write_space = psock->saved_write_space;
173		sock_replace_proto(sk, psock->sk_proto);
174		return 0;
175	}
176
177	/* psock_update_sk_prot can be called multiple times if psock is
178	 * added to multiple maps and/or slots in the same map. There is
179	 * also an edge case where replacing a psock with itself can trigger
180	 * an extra psock_update_sk_prot during the insert process. So it
181	 * must be safe to do multiple calls. Here we need to ensure we don't
182	 * increment the refcnt through sock_hold many times. There will only
183	 * be a single matching destroy operation.
184	 */
185	if (!psock->sk_pair) {
186		sk_pair = unix_peer(sk);
187		sock_hold(sk_pair);
188		psock->sk_pair = sk_pair;
189	}
190
191	unix_stream_bpf_check_needs_rebuild(psock->sk_proto);
192	sock_replace_proto(sk, &unix_stream_bpf_prot);
193	return 0;
194}
195
196void __init unix_bpf_build_proto(void)
197{
198	unix_dgram_bpf_rebuild_protos(&unix_dgram_bpf_prot, &unix_dgram_proto);
199	unix_stream_bpf_rebuild_protos(&unix_stream_bpf_prot, &unix_stream_proto);
200
201}