unix_bpf.c

  1// SPDX-License-Identifier: GPL-2.0
  2/* Copyright (c) 2021 Cong Wang <cong.wang@bytedance.com> */
  3
  4#include <linux/bpf.h>
  5#include <linux/skmsg.h>
  6#include <net/af_unix.h>
  7
  8#include "af_unix.h"
  9
 10#define unix_sk_has_data(__sk, __psock)					\
 11		({	!skb_queue_empty(&__sk->sk_receive_queue) ||	\
 12			!skb_queue_empty(&__psock->ingress_skb) ||	\
 13			!list_empty(&__psock->ingress_msg);		\
 14		})
 15
 16static int unix_msg_wait_data(struct sock *sk, struct sk_psock *psock,
 17			      long timeo)
 18{
 19	DEFINE_WAIT_FUNC(wait, woken_wake_function);
 20	struct unix_sock *u = unix_sk(sk);
 21	int ret = 0;
 22
 23	if (sk->sk_shutdown & RCV_SHUTDOWN)
 24		return 1;
 25
 26	if (!timeo)
 27		return ret;
 28
 29	add_wait_queue(sk_sleep(sk), &wait);
 30	sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
 31	if (!unix_sk_has_data(sk, psock)) {
 32		mutex_unlock(&u->iolock);
 33		wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
 34		mutex_lock(&u->iolock);
 35		ret = unix_sk_has_data(sk, psock);
 36	}
 37	sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
 38	remove_wait_queue(sk_sleep(sk), &wait);
 39	return ret;
 40}
 41
 42static int __unix_recvmsg(struct sock *sk, struct msghdr *msg,
 43			  size_t len, int flags)
 44{
 45	if (sk->sk_type == SOCK_DGRAM)
 46		return __unix_dgram_recvmsg(sk, msg, len, flags);
 47	else
 48		return __unix_stream_recvmsg(sk, msg, len, flags);
 49}
 50
 51static int unix_bpf_recvmsg(struct sock *sk, struct msghdr *msg,
 52			    size_t len, int flags, int *addr_len)
 53{
 54	struct unix_sock *u = unix_sk(sk);
 55	struct sk_psock *psock;
 56	int copied;
 57
 58	if (flags & MSG_OOB)
 59		return -EOPNOTSUPP;
 60
 61	if (!len)
 62		return 0;
 63
 64	psock = sk_psock_get(sk);
 65	if (unlikely(!psock))
 66		return __unix_recvmsg(sk, msg, len, flags);
 67
 68	mutex_lock(&u->iolock);
 69	if (!skb_queue_empty(&sk->sk_receive_queue) &&
 70	    sk_psock_queue_empty(psock)) {
 71		mutex_unlock(&u->iolock);
 72		sk_psock_put(sk, psock);
 73		return __unix_recvmsg(sk, msg, len, flags);
 74	}
 75
 76msg_bytes_ready:
 77	copied = sk_msg_recvmsg(sk, psock, msg, len, flags);
 78	if (!copied) {
 79		long timeo;
 80		int data;
 81
 82		timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
 83		data = unix_msg_wait_data(sk, psock, timeo);
 84		if (data) {
 85			if (!sk_psock_queue_empty(psock))
 86				goto msg_bytes_ready;
 87			mutex_unlock(&u->iolock);
 88			sk_psock_put(sk, psock);
 89			return __unix_recvmsg(sk, msg, len, flags);
 90		}
 91		copied = -EAGAIN;
 92	}
 93	mutex_unlock(&u->iolock);
 94	sk_psock_put(sk, psock);
 95	return copied;
 96}
 97
 98static struct proto *unix_dgram_prot_saved __read_mostly;
 99static DEFINE_SPINLOCK(unix_dgram_prot_lock);
100static struct proto unix_dgram_bpf_prot;
101
102static struct proto *unix_stream_prot_saved __read_mostly;
103static DEFINE_SPINLOCK(unix_stream_prot_lock);
104static struct proto unix_stream_bpf_prot;
105
106static void unix_dgram_bpf_rebuild_protos(struct proto *prot, const struct proto *base)
107{
108	*prot        = *base;
109	prot->close  = sock_map_close;
110	prot->recvmsg = unix_bpf_recvmsg;
111	prot->sock_is_readable = sk_msg_is_readable;
112}
113
114static void unix_stream_bpf_rebuild_protos(struct proto *prot,
115					   const struct proto *base)
116{
117	*prot        = *base;
118	prot->close  = sock_map_close;
119	prot->recvmsg = unix_bpf_recvmsg;
120	prot->sock_is_readable = sk_msg_is_readable;
121	prot->unhash  = sock_map_unhash;
122}
123
124static void unix_dgram_bpf_check_needs_rebuild(struct proto *ops)
125{
126	if (unlikely(ops != smp_load_acquire(&unix_dgram_prot_saved))) {
127		spin_lock_bh(&unix_dgram_prot_lock);
128		if (likely(ops != unix_dgram_prot_saved)) {
129			unix_dgram_bpf_rebuild_protos(&unix_dgram_bpf_prot, ops);
130			smp_store_release(&unix_dgram_prot_saved, ops);
131		}
132		spin_unlock_bh(&unix_dgram_prot_lock);
133	}
134}
135
136static void unix_stream_bpf_check_needs_rebuild(struct proto *ops)
137{
138	if (unlikely(ops != smp_load_acquire(&unix_stream_prot_saved))) {
139		spin_lock_bh(&unix_stream_prot_lock);
140		if (likely(ops != unix_stream_prot_saved)) {
141			unix_stream_bpf_rebuild_protos(&unix_stream_bpf_prot, ops);
142			smp_store_release(&unix_stream_prot_saved, ops);
143		}
144		spin_unlock_bh(&unix_stream_prot_lock);
145	}
146}
147
148int unix_dgram_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore)
149{
150	if (sk->sk_type != SOCK_DGRAM)
151		return -EOPNOTSUPP;
152
153	if (restore) {
154		sk->sk_write_space = psock->saved_write_space;
155		sock_replace_proto(sk, psock->sk_proto);
156		return 0;
157	}
158
159	unix_dgram_bpf_check_needs_rebuild(psock->sk_proto);
160	sock_replace_proto(sk, &unix_dgram_bpf_prot);
161	return 0;
162}
163
164int unix_stream_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore)
165{
166	struct sock *sk_pair;
167
168	/* Restore does not decrement the sk_pair reference yet because we must
169	 * keep the a reference to the socket until after an RCU grace period
170	 * and any pending sends have completed.
171	 */
172	if (restore) {
173		sk->sk_write_space = psock->saved_write_space;
174		sock_replace_proto(sk, psock->sk_proto);
175		return 0;
176	}
177
178	/* psock_update_sk_prot can be called multiple times if psock is
179	 * added to multiple maps and/or slots in the same map. There is
180	 * also an edge case where replacing a psock with itself can trigger
181	 * an extra psock_update_sk_prot during the insert process. So it
182	 * must be safe to do multiple calls. Here we need to ensure we don't
183	 * increment the refcnt through sock_hold many times. There will only
184	 * be a single matching destroy operation.
185	 */
186	if (!psock->sk_pair) {
187		sk_pair = unix_peer(sk);
188		sock_hold(sk_pair);
189		psock->sk_pair = sk_pair;
190	}
191
192	unix_stream_bpf_check_needs_rebuild(psock->sk_proto);
193	sock_replace_proto(sk, &unix_stream_bpf_prot);
194	return 0;
195}
196
197void __init unix_bpf_build_proto(void)
198{
199	unix_dgram_bpf_rebuild_protos(&unix_dgram_bpf_prot, &unix_dgram_proto);
200	unix_stream_bpf_rebuild_protos(&unix_stream_bpf_prot, &unix_stream_proto);
201
202}