]>
Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 LT |
2 | /* |
3 | * NET4: Implementation of BSD Unix domain sockets. | |
4 | * | |
113aa838 | 5 | * Authors: Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 6 | * |
1da177e4 LT |
7 | * Fixes: |
8 | * Linus Torvalds : Assorted bug cures. | |
9 | * Niibe Yutaka : async I/O support. | |
10 | * Carsten Paeth : PF_UNIX check, address fixes. | |
11 | * Alan Cox : Limit size of allocated blocks. | |
12 | * Alan Cox : Fixed the stupid socketpair bug. | |
13 | * Alan Cox : BSD compatibility fine tuning. | |
14 | * Alan Cox : Fixed a bug in connect when interrupted. | |
15 | * Alan Cox : Sorted out a proper draft version of | |
16 | * file descriptor passing hacked up from | |
17 | * Mike Shaver's work. | |
18 | * Marty Leisner : Fixes to fd passing | |
19 | * Nick Nevin : recvmsg bugfix. | |
20 | * Alan Cox : Started proper garbage collector | |
21 | * Heiko EiBfeldt : Missing verify_area check | |
22 | * Alan Cox : Started POSIXisms | |
23 | * Andreas Schwab : Replace inode by dentry for proper | |
24 | * reference counting | |
25 | * Kirk Petersen : Made this a module | |
26 | * Christoph Rohland : Elegant non-blocking accept/connect algorithm. | |
27 | * Lots of bug fixes. | |
28 | * Alexey Kuznetosv : Repaired (I hope) bugs introduces | |
29 | * by above two patches. | |
30 | * Andrea Arcangeli : If possible we block in connect(2) | |
31 | * if the max backlog of the listen socket | |
32 | * is been reached. This won't break | |
33 | * old apps and it will avoid huge amount | |
34 | * of socks hashed (this for unix_gc() | |
35 | * performances reasons). | |
36 | * Security fix that limits the max | |
37 | * number of socks to 2*max_files and | |
38 | * the number of skb queueable in the | |
39 | * dgram receiver. | |
40 | * Artur Skawina : Hash function optimizations | |
41 | * Alexey Kuznetsov : Full scale SMP. Lot of bugs are introduced 8) | |
42 | * Malcolm Beattie : Set peercred for socketpair | |
43 | * Michal Ostrowski : Module initialization cleanup. | |
44 | * Arnaldo C. Melo : Remove MOD_{INC,DEC}_USE_COUNT, | |
45 | * the core infrastructure is doing that | |
46 | * for all net proto families now (2.5.69+) | |
47 | * | |
1da177e4 LT |
48 | * Known differences from reference BSD that was tested: |
49 | * | |
50 | * [TO FIX] | |
51 | * ECONNREFUSED is not returned from one end of a connected() socket to the | |
52 | * other the moment one end closes. | |
53 | * fstat() doesn't return st_dev=0, and give the blksize as high water mark | |
54 | * and a fake inode identifier (nor the BSD first socket fstat twice bug). | |
55 | * [NOT TO FIX] | |
56 | * accept() returns a path name even if the connecting socket has closed | |
57 | * in the meantime (BSD loses the path and gives up). | |
58 | * accept() returns 0 length path for an unbound connector. BSD returns 16 | |
59 | * and a null first byte in the path (but not for gethost/peername - BSD bug ??) | |
60 | * socketpair(...SOCK_RAW..) doesn't panic the kernel. | |
61 | * BSD af_unix apparently has connect forgetting to block properly. | |
62 | * (need to check this with the POSIX spec in detail) | |
63 | * | |
64 | * Differences from 2.0.0-11-... (ANK) | |
65 | * Bug fixes and improvements. | |
66 | * - client shutdown killed server socket. | |
67 | * - removed all useless cli/sti pairs. | |
68 | * | |
69 | * Semantic changes/extensions. | |
70 | * - generic control message passing. | |
71 | * - SCM_CREDENTIALS control message. | |
72 | * - "Abstract" (not FS based) socket bindings. | |
73 | * Abstract names are sequences of bytes (not zero terminated) | |
74 | * started by 0, so that this name space does not intersect | |
75 | * with BSD names. | |
76 | */ | |
77 | ||
5cc208be | 78 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
79 | ||
1da177e4 | 80 | #include <linux/module.h> |
1da177e4 | 81 | #include <linux/kernel.h> |
1da177e4 | 82 | #include <linux/signal.h> |
3f07c014 | 83 | #include <linux/sched/signal.h> |
1da177e4 LT |
84 | #include <linux/errno.h> |
85 | #include <linux/string.h> | |
86 | #include <linux/stat.h> | |
87 | #include <linux/dcache.h> | |
88 | #include <linux/namei.h> | |
89 | #include <linux/socket.h> | |
90 | #include <linux/un.h> | |
91 | #include <linux/fcntl.h> | |
92 | #include <linux/termios.h> | |
93 | #include <linux/sockios.h> | |
94 | #include <linux/net.h> | |
95 | #include <linux/in.h> | |
96 | #include <linux/fs.h> | |
97 | #include <linux/slab.h> | |
7c0f6ba6 | 98 | #include <linux/uaccess.h> |
1da177e4 LT |
99 | #include <linux/skbuff.h> |
100 | #include <linux/netdevice.h> | |
457c4cbc | 101 | #include <net/net_namespace.h> |
1da177e4 | 102 | #include <net/sock.h> |
c752f073 | 103 | #include <net/tcp_states.h> |
1da177e4 LT |
104 | #include <net/af_unix.h> |
105 | #include <linux/proc_fs.h> | |
106 | #include <linux/seq_file.h> | |
107 | #include <net/scm.h> | |
108 | #include <linux/init.h> | |
109 | #include <linux/poll.h> | |
1da177e4 LT |
110 | #include <linux/rtnetlink.h> |
111 | #include <linux/mount.h> | |
112 | #include <net/checksum.h> | |
113 | #include <linux/security.h> | |
2b15af6f | 114 | #include <linux/freezer.h> |
ba94f308 | 115 | #include <linux/file.h> |
1da177e4 | 116 | |
f4e65870 JA |
117 | #include "scm.h" |
118 | ||
7123aaa3 | 119 | struct hlist_head unix_socket_table[2 * UNIX_HASH_SIZE]; |
fa7ff56f PE |
120 | EXPORT_SYMBOL_GPL(unix_socket_table); |
121 | DEFINE_SPINLOCK(unix_table_lock); | |
122 | EXPORT_SYMBOL_GPL(unix_table_lock); | |
518de9b3 | 123 | static atomic_long_t unix_nr_socks; |
1da177e4 | 124 | |
1da177e4 | 125 | |
7123aaa3 ED |
126 | static struct hlist_head *unix_sockets_unbound(void *addr) |
127 | { | |
128 | unsigned long hash = (unsigned long)addr; | |
129 | ||
130 | hash ^= hash >> 16; | |
131 | hash ^= hash >> 8; | |
132 | hash %= UNIX_HASH_SIZE; | |
133 | return &unix_socket_table[UNIX_HASH_SIZE + hash]; | |
134 | } | |
135 | ||
136 | #define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash < UNIX_HASH_SIZE) | |
1da177e4 | 137 | |
877ce7c1 | 138 | #ifdef CONFIG_SECURITY_NETWORK |
dc49c1f9 | 139 | static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 | 140 | { |
37a9a8df | 141 | UNIXCB(skb).secid = scm->secid; |
877ce7c1 CZ |
142 | } |
143 | ||
144 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
145 | { | |
37a9a8df SS |
146 | scm->secid = UNIXCB(skb).secid; |
147 | } | |
148 | ||
149 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
150 | { | |
151 | return (scm->secid == UNIXCB(skb).secid); | |
877ce7c1 CZ |
152 | } |
153 | #else | |
dc49c1f9 | 154 | static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 CZ |
155 | { } |
156 | ||
157 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
158 | { } | |
37a9a8df SS |
159 | |
160 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
161 | { | |
162 | return true; | |
163 | } | |
877ce7c1 CZ |
164 | #endif /* CONFIG_SECURITY_NETWORK */ |
165 | ||
1da177e4 LT |
166 | /* |
167 | * SMP locking strategy: | |
fbe9cc4a | 168 | * hash table is protected with spinlock unix_table_lock |
663717f6 | 169 | * each socket state is protected by separate spin lock. |
1da177e4 LT |
170 | */ |
171 | ||
95c96174 | 172 | static inline unsigned int unix_hash_fold(__wsum n) |
1da177e4 | 173 | { |
0a13404d | 174 | unsigned int hash = (__force unsigned int)csum_fold(n); |
95c96174 | 175 | |
1da177e4 LT |
176 | hash ^= hash>>8; |
177 | return hash&(UNIX_HASH_SIZE-1); | |
178 | } | |
179 | ||
180 | #define unix_peer(sk) (unix_sk(sk)->peer) | |
181 | ||
182 | static inline int unix_our_peer(struct sock *sk, struct sock *osk) | |
183 | { | |
184 | return unix_peer(osk) == sk; | |
185 | } | |
186 | ||
187 | static inline int unix_may_send(struct sock *sk, struct sock *osk) | |
188 | { | |
6eba6a37 | 189 | return unix_peer(osk) == NULL || unix_our_peer(sk, osk); |
1da177e4 LT |
190 | } |
191 | ||
3c73419c RW |
192 | static inline int unix_recvq_full(struct sock const *sk) |
193 | { | |
194 | return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog; | |
195 | } | |
196 | ||
fa7ff56f | 197 | struct sock *unix_peer_get(struct sock *s) |
1da177e4 LT |
198 | { |
199 | struct sock *peer; | |
200 | ||
1c92b4e5 | 201 | unix_state_lock(s); |
1da177e4 LT |
202 | peer = unix_peer(s); |
203 | if (peer) | |
204 | sock_hold(peer); | |
1c92b4e5 | 205 | unix_state_unlock(s); |
1da177e4 LT |
206 | return peer; |
207 | } | |
fa7ff56f | 208 | EXPORT_SYMBOL_GPL(unix_peer_get); |
1da177e4 LT |
209 | |
210 | static inline void unix_release_addr(struct unix_address *addr) | |
211 | { | |
8c9814b9 | 212 | if (refcount_dec_and_test(&addr->refcnt)) |
1da177e4 LT |
213 | kfree(addr); |
214 | } | |
215 | ||
216 | /* | |
217 | * Check unix socket name: | |
218 | * - should be not zero length. | |
219 | * - if started by not zero, should be NULL terminated (FS object) | |
220 | * - if started by zero, it is abstract name. | |
221 | */ | |
ac7bfa62 | 222 | |
95c96174 | 223 | static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp) |
1da177e4 | 224 | { |
33c4368e KK |
225 | *hashp = 0; |
226 | ||
1da177e4 LT |
227 | if (len <= sizeof(short) || len > sizeof(*sunaddr)) |
228 | return -EINVAL; | |
229 | if (!sunaddr || sunaddr->sun_family != AF_UNIX) | |
230 | return -EINVAL; | |
231 | if (sunaddr->sun_path[0]) { | |
232 | /* | |
233 | * This may look like an off by one error but it is a bit more | |
234 | * subtle. 108 is the longest valid AF_UNIX path for a binding. | |
25985edc | 235 | * sun_path[108] doesn't as such exist. However in kernel space |
1da177e4 LT |
236 | * we are guaranteed that it is a valid memory location in our |
237 | * kernel address buffer. | |
238 | */ | |
e27dfcea | 239 | ((char *)sunaddr)[len] = 0; |
1da177e4 LT |
240 | len = strlen(sunaddr->sun_path)+1+sizeof(short); |
241 | return len; | |
242 | } | |
243 | ||
07f0757a | 244 | *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0)); |
1da177e4 LT |
245 | return len; |
246 | } | |
247 | ||
248 | static void __unix_remove_socket(struct sock *sk) | |
249 | { | |
250 | sk_del_node_init(sk); | |
251 | } | |
252 | ||
253 | static void __unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
254 | { | |
547b792c | 255 | WARN_ON(!sk_unhashed(sk)); |
1da177e4 LT |
256 | sk_add_node(sk, list); |
257 | } | |
258 | ||
259 | static inline void unix_remove_socket(struct sock *sk) | |
260 | { | |
fbe9cc4a | 261 | spin_lock(&unix_table_lock); |
1da177e4 | 262 | __unix_remove_socket(sk); |
fbe9cc4a | 263 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
264 | } |
265 | ||
266 | static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
267 | { | |
fbe9cc4a | 268 | spin_lock(&unix_table_lock); |
1da177e4 | 269 | __unix_insert_socket(list, sk); |
fbe9cc4a | 270 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
271 | } |
272 | ||
097e66c5 DL |
273 | static struct sock *__unix_find_socket_byname(struct net *net, |
274 | struct sockaddr_un *sunname, | |
95c96174 | 275 | int len, int type, unsigned int hash) |
1da177e4 LT |
276 | { |
277 | struct sock *s; | |
1da177e4 | 278 | |
b67bfe0d | 279 | sk_for_each(s, &unix_socket_table[hash ^ type]) { |
1da177e4 LT |
280 | struct unix_sock *u = unix_sk(s); |
281 | ||
878628fb | 282 | if (!net_eq(sock_net(s), net)) |
097e66c5 DL |
283 | continue; |
284 | ||
1da177e4 LT |
285 | if (u->addr->len == len && |
286 | !memcmp(u->addr->name, sunname, len)) | |
262ce0af | 287 | return s; |
1da177e4 | 288 | } |
262ce0af | 289 | return NULL; |
1da177e4 LT |
290 | } |
291 | ||
097e66c5 DL |
292 | static inline struct sock *unix_find_socket_byname(struct net *net, |
293 | struct sockaddr_un *sunname, | |
1da177e4 | 294 | int len, int type, |
95c96174 | 295 | unsigned int hash) |
1da177e4 LT |
296 | { |
297 | struct sock *s; | |
298 | ||
fbe9cc4a | 299 | spin_lock(&unix_table_lock); |
097e66c5 | 300 | s = __unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
301 | if (s) |
302 | sock_hold(s); | |
fbe9cc4a | 303 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
304 | return s; |
305 | } | |
306 | ||
6616f788 | 307 | static struct sock *unix_find_socket_byinode(struct inode *i) |
1da177e4 LT |
308 | { |
309 | struct sock *s; | |
1da177e4 | 310 | |
fbe9cc4a | 311 | spin_lock(&unix_table_lock); |
b67bfe0d | 312 | sk_for_each(s, |
1da177e4 | 313 | &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) { |
40ffe67d | 314 | struct dentry *dentry = unix_sk(s)->path.dentry; |
1da177e4 | 315 | |
beef5121 | 316 | if (dentry && d_backing_inode(dentry) == i) { |
1da177e4 LT |
317 | sock_hold(s); |
318 | goto found; | |
319 | } | |
320 | } | |
321 | s = NULL; | |
322 | found: | |
fbe9cc4a | 323 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
324 | return s; |
325 | } | |
326 | ||
7d267278 RW |
327 | /* Support code for asymmetrically connected dgram sockets |
328 | * | |
329 | * If a datagram socket is connected to a socket not itself connected | |
330 | * to the first socket (eg, /dev/log), clients may only enqueue more | |
331 | * messages if the present receive queue of the server socket is not | |
332 | * "too large". This means there's a second writeability condition | |
333 | * poll and sendmsg need to test. The dgram recv code will do a wake | |
334 | * up on the peer_wait wait queue of a socket upon reception of a | |
335 | * datagram which needs to be propagated to sleeping would-be writers | |
336 | * since these might not have sent anything so far. This can't be | |
337 | * accomplished via poll_wait because the lifetime of the server | |
338 | * socket might be less than that of its clients if these break their | |
339 | * association with it or if the server socket is closed while clients | |
340 | * are still connected to it and there's no way to inform "a polling | |
341 | * implementation" that it should let go of a certain wait queue | |
342 | * | |
ac6424b9 | 343 | * In order to propagate a wake up, a wait_queue_entry_t of the client |
7d267278 RW |
344 | * socket is enqueued on the peer_wait queue of the server socket |
345 | * whose wake function does a wake_up on the ordinary client socket | |
346 | * wait queue. This connection is established whenever a write (or | |
347 | * poll for write) hit the flow control condition and broken when the | |
348 | * association to the server socket is dissolved or after a wake up | |
349 | * was relayed. | |
350 | */ | |
351 | ||
ac6424b9 | 352 | static int unix_dgram_peer_wake_relay(wait_queue_entry_t *q, unsigned mode, int flags, |
7d267278 RW |
353 | void *key) |
354 | { | |
355 | struct unix_sock *u; | |
356 | wait_queue_head_t *u_sleep; | |
357 | ||
358 | u = container_of(q, struct unix_sock, peer_wake); | |
359 | ||
360 | __remove_wait_queue(&unix_sk(u->peer_wake.private)->peer_wait, | |
361 | q); | |
362 | u->peer_wake.private = NULL; | |
363 | ||
364 | /* relaying can only happen while the wq still exists */ | |
365 | u_sleep = sk_sleep(&u->sk); | |
366 | if (u_sleep) | |
3ad6f93e | 367 | wake_up_interruptible_poll(u_sleep, key_to_poll(key)); |
7d267278 RW |
368 | |
369 | return 0; | |
370 | } | |
371 | ||
372 | static int unix_dgram_peer_wake_connect(struct sock *sk, struct sock *other) | |
373 | { | |
374 | struct unix_sock *u, *u_other; | |
375 | int rc; | |
376 | ||
377 | u = unix_sk(sk); | |
378 | u_other = unix_sk(other); | |
379 | rc = 0; | |
380 | spin_lock(&u_other->peer_wait.lock); | |
381 | ||
382 | if (!u->peer_wake.private) { | |
383 | u->peer_wake.private = other; | |
384 | __add_wait_queue(&u_other->peer_wait, &u->peer_wake); | |
385 | ||
386 | rc = 1; | |
387 | } | |
388 | ||
389 | spin_unlock(&u_other->peer_wait.lock); | |
390 | return rc; | |
391 | } | |
392 | ||
393 | static void unix_dgram_peer_wake_disconnect(struct sock *sk, | |
394 | struct sock *other) | |
395 | { | |
396 | struct unix_sock *u, *u_other; | |
397 | ||
398 | u = unix_sk(sk); | |
399 | u_other = unix_sk(other); | |
400 | spin_lock(&u_other->peer_wait.lock); | |
401 | ||
402 | if (u->peer_wake.private == other) { | |
403 | __remove_wait_queue(&u_other->peer_wait, &u->peer_wake); | |
404 | u->peer_wake.private = NULL; | |
405 | } | |
406 | ||
407 | spin_unlock(&u_other->peer_wait.lock); | |
408 | } | |
409 | ||
410 | static void unix_dgram_peer_wake_disconnect_wakeup(struct sock *sk, | |
411 | struct sock *other) | |
412 | { | |
413 | unix_dgram_peer_wake_disconnect(sk, other); | |
414 | wake_up_interruptible_poll(sk_sleep(sk), | |
a9a08845 LT |
415 | EPOLLOUT | |
416 | EPOLLWRNORM | | |
417 | EPOLLWRBAND); | |
7d267278 RW |
418 | } |
419 | ||
420 | /* preconditions: | |
421 | * - unix_peer(sk) == other | |
422 | * - association is stable | |
423 | */ | |
424 | static int unix_dgram_peer_wake_me(struct sock *sk, struct sock *other) | |
425 | { | |
426 | int connected; | |
427 | ||
428 | connected = unix_dgram_peer_wake_connect(sk, other); | |
429 | ||
51f7e951 JB |
430 | /* If other is SOCK_DEAD, we want to make sure we signal |
431 | * POLLOUT, such that a subsequent write() can get a | |
432 | * -ECONNREFUSED. Otherwise, if we haven't queued any skbs | |
433 | * to other and its full, we will hang waiting for POLLOUT. | |
434 | */ | |
435 | if (unix_recvq_full(other) && !sock_flag(other, SOCK_DEAD)) | |
7d267278 RW |
436 | return 1; |
437 | ||
438 | if (connected) | |
439 | unix_dgram_peer_wake_disconnect(sk, other); | |
440 | ||
441 | return 0; | |
442 | } | |
443 | ||
1586a587 | 444 | static int unix_writable(const struct sock *sk) |
1da177e4 | 445 | { |
1586a587 | 446 | return sk->sk_state != TCP_LISTEN && |
14afee4b | 447 | (refcount_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf; |
1da177e4 LT |
448 | } |
449 | ||
450 | static void unix_write_space(struct sock *sk) | |
451 | { | |
43815482 ED |
452 | struct socket_wq *wq; |
453 | ||
454 | rcu_read_lock(); | |
1da177e4 | 455 | if (unix_writable(sk)) { |
43815482 | 456 | wq = rcu_dereference(sk->sk_wq); |
1ce0bf50 | 457 | if (skwq_has_sleeper(wq)) |
67426b75 | 458 | wake_up_interruptible_sync_poll(&wq->wait, |
a9a08845 | 459 | EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND); |
8d8ad9d7 | 460 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
1da177e4 | 461 | } |
43815482 | 462 | rcu_read_unlock(); |
1da177e4 LT |
463 | } |
464 | ||
465 | /* When dgram socket disconnects (or changes its peer), we clear its receive | |
466 | * queue of packets arrived from previous peer. First, it allows to do | |
467 | * flow control based only on wmem_alloc; second, sk connected to peer | |
468 | * may receive messages only from that peer. */ | |
469 | static void unix_dgram_disconnected(struct sock *sk, struct sock *other) | |
470 | { | |
b03efcfb | 471 | if (!skb_queue_empty(&sk->sk_receive_queue)) { |
1da177e4 LT |
472 | skb_queue_purge(&sk->sk_receive_queue); |
473 | wake_up_interruptible_all(&unix_sk(sk)->peer_wait); | |
474 | ||
475 | /* If one link of bidirectional dgram pipe is disconnected, | |
476 | * we signal error. Messages are lost. Do not make this, | |
477 | * when peer was not connected to us. | |
478 | */ | |
479 | if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) { | |
480 | other->sk_err = ECONNRESET; | |
481 | other->sk_error_report(other); | |
482 | } | |
483 | } | |
484 | } | |
485 | ||
486 | static void unix_sock_destructor(struct sock *sk) | |
487 | { | |
488 | struct unix_sock *u = unix_sk(sk); | |
489 | ||
490 | skb_queue_purge(&sk->sk_receive_queue); | |
491 | ||
14afee4b | 492 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
547b792c IJ |
493 | WARN_ON(!sk_unhashed(sk)); |
494 | WARN_ON(sk->sk_socket); | |
1da177e4 | 495 | if (!sock_flag(sk, SOCK_DEAD)) { |
5cc208be | 496 | pr_info("Attempt to release alive unix socket: %p\n", sk); |
1da177e4 LT |
497 | return; |
498 | } | |
499 | ||
500 | if (u->addr) | |
501 | unix_release_addr(u->addr); | |
502 | ||
518de9b3 | 503 | atomic_long_dec(&unix_nr_socks); |
6f756a8c | 504 | local_bh_disable(); |
a8076d8d | 505 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
6f756a8c | 506 | local_bh_enable(); |
1da177e4 | 507 | #ifdef UNIX_REFCNT_DEBUG |
5cc208be | 508 | pr_debug("UNIX %p is destroyed, %ld are still alive.\n", sk, |
518de9b3 | 509 | atomic_long_read(&unix_nr_socks)); |
1da177e4 LT |
510 | #endif |
511 | } | |
512 | ||
ded34e0f | 513 | static void unix_release_sock(struct sock *sk, int embrion) |
1da177e4 LT |
514 | { |
515 | struct unix_sock *u = unix_sk(sk); | |
40ffe67d | 516 | struct path path; |
1da177e4 LT |
517 | struct sock *skpair; |
518 | struct sk_buff *skb; | |
519 | int state; | |
520 | ||
521 | unix_remove_socket(sk); | |
522 | ||
523 | /* Clear state */ | |
1c92b4e5 | 524 | unix_state_lock(sk); |
1da177e4 LT |
525 | sock_orphan(sk); |
526 | sk->sk_shutdown = SHUTDOWN_MASK; | |
40ffe67d AV |
527 | path = u->path; |
528 | u->path.dentry = NULL; | |
529 | u->path.mnt = NULL; | |
1da177e4 LT |
530 | state = sk->sk_state; |
531 | sk->sk_state = TCP_CLOSE; | |
1c92b4e5 | 532 | unix_state_unlock(sk); |
1da177e4 LT |
533 | |
534 | wake_up_interruptible_all(&u->peer_wait); | |
535 | ||
e27dfcea | 536 | skpair = unix_peer(sk); |
1da177e4 | 537 | |
e27dfcea | 538 | if (skpair != NULL) { |
1da177e4 | 539 | if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) { |
1c92b4e5 | 540 | unix_state_lock(skpair); |
1da177e4 LT |
541 | /* No more writes */ |
542 | skpair->sk_shutdown = SHUTDOWN_MASK; | |
543 | if (!skb_queue_empty(&sk->sk_receive_queue) || embrion) | |
544 | skpair->sk_err = ECONNRESET; | |
1c92b4e5 | 545 | unix_state_unlock(skpair); |
1da177e4 | 546 | skpair->sk_state_change(skpair); |
8d8ad9d7 | 547 | sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP); |
1da177e4 | 548 | } |
7d267278 RW |
549 | |
550 | unix_dgram_peer_wake_disconnect(sk, skpair); | |
1da177e4 LT |
551 | sock_put(skpair); /* It may now die */ |
552 | unix_peer(sk) = NULL; | |
553 | } | |
554 | ||
555 | /* Try to flush out this socket. Throw out buffers at least */ | |
556 | ||
557 | while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) { | |
e27dfcea | 558 | if (state == TCP_LISTEN) |
1da177e4 LT |
559 | unix_release_sock(skb->sk, 1); |
560 | /* passed fds are erased in the kfree_skb hook */ | |
73ed5d25 | 561 | UNIXCB(skb).consumed = skb->len; |
1da177e4 LT |
562 | kfree_skb(skb); |
563 | } | |
564 | ||
40ffe67d AV |
565 | if (path.dentry) |
566 | path_put(&path); | |
1da177e4 LT |
567 | |
568 | sock_put(sk); | |
569 | ||
570 | /* ---- Socket is dead now and most probably destroyed ---- */ | |
571 | ||
572 | /* | |
e04dae84 | 573 | * Fixme: BSD difference: In BSD all sockets connected to us get |
1da177e4 LT |
574 | * ECONNRESET and we die on the spot. In Linux we behave |
575 | * like files and pipes do and wait for the last | |
576 | * dereference. | |
577 | * | |
578 | * Can't we simply set sock->err? | |
579 | * | |
580 | * What the above comment does talk about? --ANK(980817) | |
581 | */ | |
582 | ||
9305cfa4 | 583 | if (unix_tot_inflight) |
ac7bfa62 | 584 | unix_gc(); /* Garbage collect fds */ |
1da177e4 LT |
585 | } |
586 | ||
109f6e39 EB |
587 | static void init_peercred(struct sock *sk) |
588 | { | |
589 | put_pid(sk->sk_peer_pid); | |
590 | if (sk->sk_peer_cred) | |
591 | put_cred(sk->sk_peer_cred); | |
592 | sk->sk_peer_pid = get_pid(task_tgid(current)); | |
593 | sk->sk_peer_cred = get_current_cred(); | |
594 | } | |
595 | ||
596 | static void copy_peercred(struct sock *sk, struct sock *peersk) | |
597 | { | |
598 | put_pid(sk->sk_peer_pid); | |
599 | if (sk->sk_peer_cred) | |
600 | put_cred(sk->sk_peer_cred); | |
601 | sk->sk_peer_pid = get_pid(peersk->sk_peer_pid); | |
602 | sk->sk_peer_cred = get_cred(peersk->sk_peer_cred); | |
603 | } | |
604 | ||
1da177e4 LT |
605 | static int unix_listen(struct socket *sock, int backlog) |
606 | { | |
607 | int err; | |
608 | struct sock *sk = sock->sk; | |
609 | struct unix_sock *u = unix_sk(sk); | |
109f6e39 | 610 | struct pid *old_pid = NULL; |
1da177e4 LT |
611 | |
612 | err = -EOPNOTSUPP; | |
6eba6a37 ED |
613 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
614 | goto out; /* Only stream/seqpacket sockets accept */ | |
1da177e4 LT |
615 | err = -EINVAL; |
616 | if (!u->addr) | |
6eba6a37 | 617 | goto out; /* No listens on an unbound socket */ |
1c92b4e5 | 618 | unix_state_lock(sk); |
1da177e4 LT |
619 | if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN) |
620 | goto out_unlock; | |
621 | if (backlog > sk->sk_max_ack_backlog) | |
622 | wake_up_interruptible_all(&u->peer_wait); | |
623 | sk->sk_max_ack_backlog = backlog; | |
624 | sk->sk_state = TCP_LISTEN; | |
625 | /* set credentials so connect can copy them */ | |
109f6e39 | 626 | init_peercred(sk); |
1da177e4 LT |
627 | err = 0; |
628 | ||
629 | out_unlock: | |
1c92b4e5 | 630 | unix_state_unlock(sk); |
109f6e39 | 631 | put_pid(old_pid); |
1da177e4 LT |
632 | out: |
633 | return err; | |
634 | } | |
635 | ||
636 | static int unix_release(struct socket *); | |
637 | static int unix_bind(struct socket *, struct sockaddr *, int); | |
638 | static int unix_stream_connect(struct socket *, struct sockaddr *, | |
639 | int addr_len, int flags); | |
640 | static int unix_socketpair(struct socket *, struct socket *); | |
cdfbabfb | 641 | static int unix_accept(struct socket *, struct socket *, int, bool); |
9b2c45d4 | 642 | static int unix_getname(struct socket *, struct sockaddr *, int); |
a11e1d43 LT |
643 | static __poll_t unix_poll(struct file *, struct socket *, poll_table *); |
644 | static __poll_t unix_dgram_poll(struct file *, struct socket *, | |
645 | poll_table *); | |
1da177e4 LT |
646 | static int unix_ioctl(struct socket *, unsigned int, unsigned long); |
647 | static int unix_shutdown(struct socket *, int); | |
1b784140 YX |
648 | static int unix_stream_sendmsg(struct socket *, struct msghdr *, size_t); |
649 | static int unix_stream_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
869e7c62 HFS |
650 | static ssize_t unix_stream_sendpage(struct socket *, struct page *, int offset, |
651 | size_t size, int flags); | |
2b514574 HFS |
652 | static ssize_t unix_stream_splice_read(struct socket *, loff_t *ppos, |
653 | struct pipe_inode_info *, size_t size, | |
654 | unsigned int flags); | |
1b784140 YX |
655 | static int unix_dgram_sendmsg(struct socket *, struct msghdr *, size_t); |
656 | static int unix_dgram_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
1da177e4 LT |
657 | static int unix_dgram_connect(struct socket *, struct sockaddr *, |
658 | int, int); | |
1b784140 YX |
659 | static int unix_seqpacket_sendmsg(struct socket *, struct msghdr *, size_t); |
660 | static int unix_seqpacket_recvmsg(struct socket *, struct msghdr *, size_t, | |
661 | int); | |
1da177e4 | 662 | |
12663bfc | 663 | static int unix_set_peek_off(struct sock *sk, int val) |
f55bb7f9 PE |
664 | { |
665 | struct unix_sock *u = unix_sk(sk); | |
666 | ||
6e1ce3c3 | 667 | if (mutex_lock_interruptible(&u->iolock)) |
12663bfc SL |
668 | return -EINTR; |
669 | ||
f55bb7f9 | 670 | sk->sk_peek_off = val; |
6e1ce3c3 | 671 | mutex_unlock(&u->iolock); |
12663bfc SL |
672 | |
673 | return 0; | |
f55bb7f9 PE |
674 | } |
675 | ||
676 | ||
90ddc4f0 | 677 | static const struct proto_ops unix_stream_ops = { |
1da177e4 LT |
678 | .family = PF_UNIX, |
679 | .owner = THIS_MODULE, | |
680 | .release = unix_release, | |
681 | .bind = unix_bind, | |
682 | .connect = unix_stream_connect, | |
683 | .socketpair = unix_socketpair, | |
684 | .accept = unix_accept, | |
685 | .getname = unix_getname, | |
a11e1d43 | 686 | .poll = unix_poll, |
1da177e4 LT |
687 | .ioctl = unix_ioctl, |
688 | .listen = unix_listen, | |
689 | .shutdown = unix_shutdown, | |
690 | .setsockopt = sock_no_setsockopt, | |
691 | .getsockopt = sock_no_getsockopt, | |
692 | .sendmsg = unix_stream_sendmsg, | |
693 | .recvmsg = unix_stream_recvmsg, | |
694 | .mmap = sock_no_mmap, | |
869e7c62 | 695 | .sendpage = unix_stream_sendpage, |
2b514574 | 696 | .splice_read = unix_stream_splice_read, |
fc0d7536 | 697 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
698 | }; |
699 | ||
90ddc4f0 | 700 | static const struct proto_ops unix_dgram_ops = { |
1da177e4 LT |
701 | .family = PF_UNIX, |
702 | .owner = THIS_MODULE, | |
703 | .release = unix_release, | |
704 | .bind = unix_bind, | |
705 | .connect = unix_dgram_connect, | |
706 | .socketpair = unix_socketpair, | |
707 | .accept = sock_no_accept, | |
708 | .getname = unix_getname, | |
a11e1d43 | 709 | .poll = unix_dgram_poll, |
1da177e4 LT |
710 | .ioctl = unix_ioctl, |
711 | .listen = sock_no_listen, | |
712 | .shutdown = unix_shutdown, | |
713 | .setsockopt = sock_no_setsockopt, | |
714 | .getsockopt = sock_no_getsockopt, | |
715 | .sendmsg = unix_dgram_sendmsg, | |
716 | .recvmsg = unix_dgram_recvmsg, | |
717 | .mmap = sock_no_mmap, | |
718 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 719 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
720 | }; |
721 | ||
90ddc4f0 | 722 | static const struct proto_ops unix_seqpacket_ops = { |
1da177e4 LT |
723 | .family = PF_UNIX, |
724 | .owner = THIS_MODULE, | |
725 | .release = unix_release, | |
726 | .bind = unix_bind, | |
727 | .connect = unix_stream_connect, | |
728 | .socketpair = unix_socketpair, | |
729 | .accept = unix_accept, | |
730 | .getname = unix_getname, | |
a11e1d43 | 731 | .poll = unix_dgram_poll, |
1da177e4 LT |
732 | .ioctl = unix_ioctl, |
733 | .listen = unix_listen, | |
734 | .shutdown = unix_shutdown, | |
735 | .setsockopt = sock_no_setsockopt, | |
736 | .getsockopt = sock_no_getsockopt, | |
737 | .sendmsg = unix_seqpacket_sendmsg, | |
a05d2ad1 | 738 | .recvmsg = unix_seqpacket_recvmsg, |
1da177e4 LT |
739 | .mmap = sock_no_mmap, |
740 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 741 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
742 | }; |
743 | ||
744 | static struct proto unix_proto = { | |
248969ae ED |
745 | .name = "UNIX", |
746 | .owner = THIS_MODULE, | |
248969ae | 747 | .obj_size = sizeof(struct unix_sock), |
1da177e4 LT |
748 | }; |
749 | ||
11aa9c28 | 750 | static struct sock *unix_create1(struct net *net, struct socket *sock, int kern) |
1da177e4 LT |
751 | { |
752 | struct sock *sk = NULL; | |
753 | struct unix_sock *u; | |
754 | ||
518de9b3 ED |
755 | atomic_long_inc(&unix_nr_socks); |
756 | if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files()) | |
1da177e4 LT |
757 | goto out; |
758 | ||
11aa9c28 | 759 | sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto, kern); |
1da177e4 LT |
760 | if (!sk) |
761 | goto out; | |
762 | ||
6eba6a37 | 763 | sock_init_data(sock, sk); |
1da177e4 | 764 | |
3aa9799e | 765 | sk->sk_allocation = GFP_KERNEL_ACCOUNT; |
1da177e4 | 766 | sk->sk_write_space = unix_write_space; |
a0a53c8b | 767 | sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen; |
1da177e4 LT |
768 | sk->sk_destruct = unix_sock_destructor; |
769 | u = unix_sk(sk); | |
40ffe67d AV |
770 | u->path.dentry = NULL; |
771 | u->path.mnt = NULL; | |
fd19f329 | 772 | spin_lock_init(&u->lock); |
516e0cc5 | 773 | atomic_long_set(&u->inflight, 0); |
1fd05ba5 | 774 | INIT_LIST_HEAD(&u->link); |
6e1ce3c3 LT |
775 | mutex_init(&u->iolock); /* single task reading lock */ |
776 | mutex_init(&u->bindlock); /* single task binding lock */ | |
1da177e4 | 777 | init_waitqueue_head(&u->peer_wait); |
7d267278 | 778 | init_waitqueue_func_entry(&u->peer_wake, unix_dgram_peer_wake_relay); |
7123aaa3 | 779 | unix_insert_socket(unix_sockets_unbound(sk), sk); |
1da177e4 | 780 | out: |
284b327b | 781 | if (sk == NULL) |
518de9b3 | 782 | atomic_long_dec(&unix_nr_socks); |
920de804 ED |
783 | else { |
784 | local_bh_disable(); | |
a8076d8d | 785 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
920de804 ED |
786 | local_bh_enable(); |
787 | } | |
1da177e4 LT |
788 | return sk; |
789 | } | |
790 | ||
3f378b68 EP |
791 | static int unix_create(struct net *net, struct socket *sock, int protocol, |
792 | int kern) | |
1da177e4 LT |
793 | { |
794 | if (protocol && protocol != PF_UNIX) | |
795 | return -EPROTONOSUPPORT; | |
796 | ||
797 | sock->state = SS_UNCONNECTED; | |
798 | ||
799 | switch (sock->type) { | |
800 | case SOCK_STREAM: | |
801 | sock->ops = &unix_stream_ops; | |
802 | break; | |
803 | /* | |
804 | * Believe it or not BSD has AF_UNIX, SOCK_RAW though | |
805 | * nothing uses it. | |
806 | */ | |
807 | case SOCK_RAW: | |
e27dfcea | 808 | sock->type = SOCK_DGRAM; |
110af3ac | 809 | /* fall through */ |
1da177e4 LT |
810 | case SOCK_DGRAM: |
811 | sock->ops = &unix_dgram_ops; | |
812 | break; | |
813 | case SOCK_SEQPACKET: | |
814 | sock->ops = &unix_seqpacket_ops; | |
815 | break; | |
816 | default: | |
817 | return -ESOCKTNOSUPPORT; | |
818 | } | |
819 | ||
11aa9c28 | 820 | return unix_create1(net, sock, kern) ? 0 : -ENOMEM; |
1da177e4 LT |
821 | } |
822 | ||
823 | static int unix_release(struct socket *sock) | |
824 | { | |
825 | struct sock *sk = sock->sk; | |
826 | ||
827 | if (!sk) | |
828 | return 0; | |
829 | ||
ded34e0f | 830 | unix_release_sock(sk, 0); |
1da177e4 LT |
831 | sock->sk = NULL; |
832 | ||
ded34e0f | 833 | return 0; |
1da177e4 LT |
834 | } |
835 | ||
836 | static int unix_autobind(struct socket *sock) | |
837 | { | |
838 | struct sock *sk = sock->sk; | |
3b1e0a65 | 839 | struct net *net = sock_net(sk); |
1da177e4 LT |
840 | struct unix_sock *u = unix_sk(sk); |
841 | static u32 ordernum = 1; | |
6eba6a37 | 842 | struct unix_address *addr; |
1da177e4 | 843 | int err; |
8df73ff9 | 844 | unsigned int retries = 0; |
1da177e4 | 845 | |
6e1ce3c3 | 846 | err = mutex_lock_interruptible(&u->bindlock); |
37ab4fa7 SL |
847 | if (err) |
848 | return err; | |
1da177e4 LT |
849 | |
850 | err = 0; | |
851 | if (u->addr) | |
852 | goto out; | |
853 | ||
854 | err = -ENOMEM; | |
0da974f4 | 855 | addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL); |
1da177e4 LT |
856 | if (!addr) |
857 | goto out; | |
858 | ||
1da177e4 | 859 | addr->name->sun_family = AF_UNIX; |
8c9814b9 | 860 | refcount_set(&addr->refcnt, 1); |
1da177e4 LT |
861 | |
862 | retry: | |
863 | addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short); | |
07f0757a | 864 | addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0)); |
1da177e4 | 865 | |
fbe9cc4a | 866 | spin_lock(&unix_table_lock); |
1da177e4 LT |
867 | ordernum = (ordernum+1)&0xFFFFF; |
868 | ||
097e66c5 | 869 | if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type, |
1da177e4 | 870 | addr->hash)) { |
fbe9cc4a | 871 | spin_unlock(&unix_table_lock); |
8df73ff9 TH |
872 | /* |
873 | * __unix_find_socket_byname() may take long time if many names | |
874 | * are already in use. | |
875 | */ | |
876 | cond_resched(); | |
877 | /* Give up if all names seems to be in use. */ | |
878 | if (retries++ == 0xFFFFF) { | |
879 | err = -ENOSPC; | |
880 | kfree(addr); | |
881 | goto out; | |
882 | } | |
1da177e4 LT |
883 | goto retry; |
884 | } | |
885 | addr->hash ^= sk->sk_type; | |
886 | ||
887 | __unix_remove_socket(sk); | |
ae3b5641 | 888 | smp_store_release(&u->addr, addr); |
1da177e4 | 889 | __unix_insert_socket(&unix_socket_table[addr->hash], sk); |
fbe9cc4a | 890 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
891 | err = 0; |
892 | ||
6e1ce3c3 | 893 | out: mutex_unlock(&u->bindlock); |
1da177e4 LT |
894 | return err; |
895 | } | |
896 | ||
097e66c5 DL |
897 | static struct sock *unix_find_other(struct net *net, |
898 | struct sockaddr_un *sunname, int len, | |
95c96174 | 899 | int type, unsigned int hash, int *error) |
1da177e4 LT |
900 | { |
901 | struct sock *u; | |
421748ec | 902 | struct path path; |
1da177e4 | 903 | int err = 0; |
ac7bfa62 | 904 | |
1da177e4 | 905 | if (sunname->sun_path[0]) { |
421748ec AV |
906 | struct inode *inode; |
907 | err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path); | |
1da177e4 LT |
908 | if (err) |
909 | goto fail; | |
beef5121 | 910 | inode = d_backing_inode(path.dentry); |
421748ec | 911 | err = inode_permission(inode, MAY_WRITE); |
1da177e4 LT |
912 | if (err) |
913 | goto put_fail; | |
914 | ||
915 | err = -ECONNREFUSED; | |
421748ec | 916 | if (!S_ISSOCK(inode->i_mode)) |
1da177e4 | 917 | goto put_fail; |
6616f788 | 918 | u = unix_find_socket_byinode(inode); |
1da177e4 LT |
919 | if (!u) |
920 | goto put_fail; | |
921 | ||
922 | if (u->sk_type == type) | |
68ac1234 | 923 | touch_atime(&path); |
1da177e4 | 924 | |
421748ec | 925 | path_put(&path); |
1da177e4 | 926 | |
e27dfcea | 927 | err = -EPROTOTYPE; |
1da177e4 LT |
928 | if (u->sk_type != type) { |
929 | sock_put(u); | |
930 | goto fail; | |
931 | } | |
932 | } else { | |
933 | err = -ECONNREFUSED; | |
e27dfcea | 934 | u = unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
935 | if (u) { |
936 | struct dentry *dentry; | |
40ffe67d | 937 | dentry = unix_sk(u)->path.dentry; |
1da177e4 | 938 | if (dentry) |
68ac1234 | 939 | touch_atime(&unix_sk(u)->path); |
1da177e4 LT |
940 | } else |
941 | goto fail; | |
942 | } | |
943 | return u; | |
944 | ||
945 | put_fail: | |
421748ec | 946 | path_put(&path); |
1da177e4 | 947 | fail: |
e27dfcea | 948 | *error = err; |
1da177e4 LT |
949 | return NULL; |
950 | } | |
951 | ||
38f7bd94 | 952 | static int unix_mknod(const char *sun_path, umode_t mode, struct path *res) |
faf02010 | 953 | { |
38f7bd94 LT |
954 | struct dentry *dentry; |
955 | struct path path; | |
956 | int err = 0; | |
957 | /* | |
958 | * Get the parent directory, calculate the hash for last | |
959 | * component. | |
960 | */ | |
961 | dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0); | |
962 | err = PTR_ERR(dentry); | |
963 | if (IS_ERR(dentry)) | |
964 | return err; | |
faf02010 | 965 | |
38f7bd94 LT |
966 | /* |
967 | * All right, let's create it. | |
968 | */ | |
969 | err = security_path_mknod(&path, dentry, mode, 0); | |
faf02010 | 970 | if (!err) { |
38f7bd94 | 971 | err = vfs_mknod(d_inode(path.dentry), dentry, mode, 0); |
faf02010 | 972 | if (!err) { |
38f7bd94 | 973 | res->mnt = mntget(path.mnt); |
faf02010 AV |
974 | res->dentry = dget(dentry); |
975 | } | |
976 | } | |
38f7bd94 | 977 | done_path_create(&path, dentry); |
faf02010 AV |
978 | return err; |
979 | } | |
1da177e4 LT |
980 | |
981 | static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
982 | { | |
983 | struct sock *sk = sock->sk; | |
3b1e0a65 | 984 | struct net *net = sock_net(sk); |
1da177e4 | 985 | struct unix_sock *u = unix_sk(sk); |
e27dfcea | 986 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
dae6ad8f | 987 | char *sun_path = sunaddr->sun_path; |
38f7bd94 | 988 | int err; |
95c96174 | 989 | unsigned int hash; |
1da177e4 LT |
990 | struct unix_address *addr; |
991 | struct hlist_head *list; | |
82fe0d2b | 992 | struct path path = { }; |
1da177e4 LT |
993 | |
994 | err = -EINVAL; | |
defbcf2d MJ |
995 | if (addr_len < offsetofend(struct sockaddr_un, sun_family) || |
996 | sunaddr->sun_family != AF_UNIX) | |
1da177e4 LT |
997 | goto out; |
998 | ||
e27dfcea | 999 | if (addr_len == sizeof(short)) { |
1da177e4 LT |
1000 | err = unix_autobind(sock); |
1001 | goto out; | |
1002 | } | |
1003 | ||
1004 | err = unix_mkname(sunaddr, addr_len, &hash); | |
1005 | if (err < 0) | |
1006 | goto out; | |
1007 | addr_len = err; | |
1008 | ||
0fb44559 WC |
1009 | if (sun_path[0]) { |
1010 | umode_t mode = S_IFSOCK | | |
1011 | (SOCK_INODE(sock)->i_mode & ~current_umask()); | |
1012 | err = unix_mknod(sun_path, mode, &path); | |
1013 | if (err) { | |
1014 | if (err == -EEXIST) | |
1015 | err = -EADDRINUSE; | |
1016 | goto out; | |
1017 | } | |
1018 | } | |
1019 | ||
6e1ce3c3 | 1020 | err = mutex_lock_interruptible(&u->bindlock); |
37ab4fa7 | 1021 | if (err) |
0fb44559 | 1022 | goto out_put; |
1da177e4 LT |
1023 | |
1024 | err = -EINVAL; | |
1025 | if (u->addr) | |
1026 | goto out_up; | |
1027 | ||
1028 | err = -ENOMEM; | |
1029 | addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL); | |
1030 | if (!addr) | |
1031 | goto out_up; | |
1032 | ||
1033 | memcpy(addr->name, sunaddr, addr_len); | |
1034 | addr->len = addr_len; | |
1035 | addr->hash = hash ^ sk->sk_type; | |
8c9814b9 | 1036 | refcount_set(&addr->refcnt, 1); |
1da177e4 | 1037 | |
38f7bd94 | 1038 | if (sun_path[0]) { |
1da177e4 | 1039 | addr->hash = UNIX_HASH_SIZE; |
beef5121 | 1040 | hash = d_backing_inode(path.dentry)->i_ino & (UNIX_HASH_SIZE - 1); |
faf02010 | 1041 | spin_lock(&unix_table_lock); |
38f7bd94 | 1042 | u->path = path; |
faf02010 AV |
1043 | list = &unix_socket_table[hash]; |
1044 | } else { | |
1045 | spin_lock(&unix_table_lock); | |
1da177e4 | 1046 | err = -EADDRINUSE; |
097e66c5 | 1047 | if (__unix_find_socket_byname(net, sunaddr, addr_len, |
1da177e4 LT |
1048 | sk->sk_type, hash)) { |
1049 | unix_release_addr(addr); | |
1050 | goto out_unlock; | |
1051 | } | |
1052 | ||
1053 | list = &unix_socket_table[addr->hash]; | |
1da177e4 LT |
1054 | } |
1055 | ||
1056 | err = 0; | |
1057 | __unix_remove_socket(sk); | |
ae3b5641 | 1058 | smp_store_release(&u->addr, addr); |
1da177e4 LT |
1059 | __unix_insert_socket(list, sk); |
1060 | ||
1061 | out_unlock: | |
fbe9cc4a | 1062 | spin_unlock(&unix_table_lock); |
1da177e4 | 1063 | out_up: |
6e1ce3c3 | 1064 | mutex_unlock(&u->bindlock); |
0fb44559 WC |
1065 | out_put: |
1066 | if (err) | |
1067 | path_put(&path); | |
1da177e4 LT |
1068 | out: |
1069 | return err; | |
1da177e4 LT |
1070 | } |
1071 | ||
278a3de5 DM |
1072 | static void unix_state_double_lock(struct sock *sk1, struct sock *sk2) |
1073 | { | |
1074 | if (unlikely(sk1 == sk2) || !sk2) { | |
1075 | unix_state_lock(sk1); | |
1076 | return; | |
1077 | } | |
1078 | if (sk1 < sk2) { | |
1079 | unix_state_lock(sk1); | |
1080 | unix_state_lock_nested(sk2); | |
1081 | } else { | |
1082 | unix_state_lock(sk2); | |
1083 | unix_state_lock_nested(sk1); | |
1084 | } | |
1085 | } | |
1086 | ||
1087 | static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2) | |
1088 | { | |
1089 | if (unlikely(sk1 == sk2) || !sk2) { | |
1090 | unix_state_unlock(sk1); | |
1091 | return; | |
1092 | } | |
1093 | unix_state_unlock(sk1); | |
1094 | unix_state_unlock(sk2); | |
1095 | } | |
1096 | ||
1da177e4 LT |
1097 | static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr, |
1098 | int alen, int flags) | |
1099 | { | |
1100 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1101 | struct net *net = sock_net(sk); |
e27dfcea | 1102 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr; |
1da177e4 | 1103 | struct sock *other; |
95c96174 | 1104 | unsigned int hash; |
1da177e4 LT |
1105 | int err; |
1106 | ||
defbcf2d MJ |
1107 | err = -EINVAL; |
1108 | if (alen < offsetofend(struct sockaddr, sa_family)) | |
1109 | goto out; | |
1110 | ||
1da177e4 LT |
1111 | if (addr->sa_family != AF_UNSPEC) { |
1112 | err = unix_mkname(sunaddr, alen, &hash); | |
1113 | if (err < 0) | |
1114 | goto out; | |
1115 | alen = err; | |
1116 | ||
1117 | if (test_bit(SOCK_PASSCRED, &sock->flags) && | |
1118 | !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0) | |
1119 | goto out; | |
1120 | ||
278a3de5 | 1121 | restart: |
e27dfcea | 1122 | other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err); |
1da177e4 LT |
1123 | if (!other) |
1124 | goto out; | |
1125 | ||
278a3de5 DM |
1126 | unix_state_double_lock(sk, other); |
1127 | ||
1128 | /* Apparently VFS overslept socket death. Retry. */ | |
1129 | if (sock_flag(other, SOCK_DEAD)) { | |
1130 | unix_state_double_unlock(sk, other); | |
1131 | sock_put(other); | |
1132 | goto restart; | |
1133 | } | |
1da177e4 LT |
1134 | |
1135 | err = -EPERM; | |
1136 | if (!unix_may_send(sk, other)) | |
1137 | goto out_unlock; | |
1138 | ||
1139 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1140 | if (err) | |
1141 | goto out_unlock; | |
1142 | ||
1143 | } else { | |
1144 | /* | |
1145 | * 1003.1g breaking connected state with AF_UNSPEC | |
1146 | */ | |
1147 | other = NULL; | |
278a3de5 | 1148 | unix_state_double_lock(sk, other); |
1da177e4 LT |
1149 | } |
1150 | ||
1151 | /* | |
1152 | * If it was connected, reconnect. | |
1153 | */ | |
1154 | if (unix_peer(sk)) { | |
1155 | struct sock *old_peer = unix_peer(sk); | |
e27dfcea | 1156 | unix_peer(sk) = other; |
7d267278 RW |
1157 | unix_dgram_peer_wake_disconnect_wakeup(sk, old_peer); |
1158 | ||
278a3de5 | 1159 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1160 | |
1161 | if (other != old_peer) | |
1162 | unix_dgram_disconnected(sk, old_peer); | |
1163 | sock_put(old_peer); | |
1164 | } else { | |
e27dfcea | 1165 | unix_peer(sk) = other; |
278a3de5 | 1166 | unix_state_double_unlock(sk, other); |
1da177e4 | 1167 | } |
ac7bfa62 | 1168 | return 0; |
1da177e4 LT |
1169 | |
1170 | out_unlock: | |
278a3de5 | 1171 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1172 | sock_put(other); |
1173 | out: | |
1174 | return err; | |
1175 | } | |
1176 | ||
1177 | static long unix_wait_for_peer(struct sock *other, long timeo) | |
1178 | { | |
1179 | struct unix_sock *u = unix_sk(other); | |
1180 | int sched; | |
1181 | DEFINE_WAIT(wait); | |
1182 | ||
1183 | prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE); | |
1184 | ||
1185 | sched = !sock_flag(other, SOCK_DEAD) && | |
1186 | !(other->sk_shutdown & RCV_SHUTDOWN) && | |
3c73419c | 1187 | unix_recvq_full(other); |
1da177e4 | 1188 | |
1c92b4e5 | 1189 | unix_state_unlock(other); |
1da177e4 LT |
1190 | |
1191 | if (sched) | |
1192 | timeo = schedule_timeout(timeo); | |
1193 | ||
1194 | finish_wait(&u->peer_wait, &wait); | |
1195 | return timeo; | |
1196 | } | |
1197 | ||
1198 | static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
1199 | int addr_len, int flags) | |
1200 | { | |
e27dfcea | 1201 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
1da177e4 | 1202 | struct sock *sk = sock->sk; |
3b1e0a65 | 1203 | struct net *net = sock_net(sk); |
1da177e4 LT |
1204 | struct unix_sock *u = unix_sk(sk), *newu, *otheru; |
1205 | struct sock *newsk = NULL; | |
1206 | struct sock *other = NULL; | |
1207 | struct sk_buff *skb = NULL; | |
95c96174 | 1208 | unsigned int hash; |
1da177e4 LT |
1209 | int st; |
1210 | int err; | |
1211 | long timeo; | |
1212 | ||
1213 | err = unix_mkname(sunaddr, addr_len, &hash); | |
1214 | if (err < 0) | |
1215 | goto out; | |
1216 | addr_len = err; | |
1217 | ||
f64f9e71 JP |
1218 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr && |
1219 | (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1220 | goto out; |
1221 | ||
1222 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
1223 | ||
1224 | /* First of all allocate resources. | |
1225 | If we will make it after state is locked, | |
1226 | we will have to recheck all again in any case. | |
1227 | */ | |
1228 | ||
1229 | err = -ENOMEM; | |
1230 | ||
1231 | /* create new sock for complete connection */ | |
11aa9c28 | 1232 | newsk = unix_create1(sock_net(sk), NULL, 0); |
1da177e4 LT |
1233 | if (newsk == NULL) |
1234 | goto out; | |
1235 | ||
1236 | /* Allocate skb for sending to listening sock */ | |
1237 | skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL); | |
1238 | if (skb == NULL) | |
1239 | goto out; | |
1240 | ||
1241 | restart: | |
1242 | /* Find listening sock. */ | |
097e66c5 | 1243 | other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err); |
1da177e4 LT |
1244 | if (!other) |
1245 | goto out; | |
1246 | ||
1247 | /* Latch state of peer */ | |
1c92b4e5 | 1248 | unix_state_lock(other); |
1da177e4 LT |
1249 | |
1250 | /* Apparently VFS overslept socket death. Retry. */ | |
1251 | if (sock_flag(other, SOCK_DEAD)) { | |
1c92b4e5 | 1252 | unix_state_unlock(other); |
1da177e4 LT |
1253 | sock_put(other); |
1254 | goto restart; | |
1255 | } | |
1256 | ||
1257 | err = -ECONNREFUSED; | |
1258 | if (other->sk_state != TCP_LISTEN) | |
1259 | goto out_unlock; | |
77238f2b TS |
1260 | if (other->sk_shutdown & RCV_SHUTDOWN) |
1261 | goto out_unlock; | |
1da177e4 | 1262 | |
3c73419c | 1263 | if (unix_recvq_full(other)) { |
1da177e4 LT |
1264 | err = -EAGAIN; |
1265 | if (!timeo) | |
1266 | goto out_unlock; | |
1267 | ||
1268 | timeo = unix_wait_for_peer(other, timeo); | |
1269 | ||
1270 | err = sock_intr_errno(timeo); | |
1271 | if (signal_pending(current)) | |
1272 | goto out; | |
1273 | sock_put(other); | |
1274 | goto restart; | |
ac7bfa62 | 1275 | } |
1da177e4 LT |
1276 | |
1277 | /* Latch our state. | |
1278 | ||
e5537bfc | 1279 | It is tricky place. We need to grab our state lock and cannot |
1da177e4 LT |
1280 | drop lock on peer. It is dangerous because deadlock is |
1281 | possible. Connect to self case and simultaneous | |
1282 | attempt to connect are eliminated by checking socket | |
1283 | state. other is TCP_LISTEN, if sk is TCP_LISTEN we | |
1284 | check this before attempt to grab lock. | |
1285 | ||
1286 | Well, and we have to recheck the state after socket locked. | |
1287 | */ | |
1288 | st = sk->sk_state; | |
1289 | ||
1290 | switch (st) { | |
1291 | case TCP_CLOSE: | |
1292 | /* This is ok... continue with connect */ | |
1293 | break; | |
1294 | case TCP_ESTABLISHED: | |
1295 | /* Socket is already connected */ | |
1296 | err = -EISCONN; | |
1297 | goto out_unlock; | |
1298 | default: | |
1299 | err = -EINVAL; | |
1300 | goto out_unlock; | |
1301 | } | |
1302 | ||
1c92b4e5 | 1303 | unix_state_lock_nested(sk); |
1da177e4 LT |
1304 | |
1305 | if (sk->sk_state != st) { | |
1c92b4e5 DM |
1306 | unix_state_unlock(sk); |
1307 | unix_state_unlock(other); | |
1da177e4 LT |
1308 | sock_put(other); |
1309 | goto restart; | |
1310 | } | |
1311 | ||
3610cda5 | 1312 | err = security_unix_stream_connect(sk, other, newsk); |
1da177e4 | 1313 | if (err) { |
1c92b4e5 | 1314 | unix_state_unlock(sk); |
1da177e4 LT |
1315 | goto out_unlock; |
1316 | } | |
1317 | ||
1318 | /* The way is open! Fastly set all the necessary fields... */ | |
1319 | ||
1320 | sock_hold(sk); | |
1321 | unix_peer(newsk) = sk; | |
1322 | newsk->sk_state = TCP_ESTABLISHED; | |
1323 | newsk->sk_type = sk->sk_type; | |
109f6e39 | 1324 | init_peercred(newsk); |
1da177e4 | 1325 | newu = unix_sk(newsk); |
eaefd110 | 1326 | RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq); |
1da177e4 LT |
1327 | otheru = unix_sk(other); |
1328 | ||
ae3b5641 AV |
1329 | /* copy address information from listening to new sock |
1330 | * | |
1331 | * The contents of *(otheru->addr) and otheru->path | |
1332 | * are seen fully set up here, since we have found | |
1333 | * otheru in hash under unix_table_lock. Insertion | |
1334 | * into the hash chain we'd found it in had been done | |
1335 | * in an earlier critical area protected by unix_table_lock, | |
1336 | * the same one where we'd set *(otheru->addr) contents, | |
1337 | * as well as otheru->path and otheru->addr itself. | |
1338 | * | |
1339 | * Using smp_store_release() here to set newu->addr | |
1340 | * is enough to make those stores, as well as stores | |
1341 | * to newu->path visible to anyone who gets newu->addr | |
1342 | * by smp_load_acquire(). IOW, the same warranties | |
1343 | * as for unix_sock instances bound in unix_bind() or | |
1344 | * in unix_autobind(). | |
1345 | */ | |
40ffe67d AV |
1346 | if (otheru->path.dentry) { |
1347 | path_get(&otheru->path); | |
1348 | newu->path = otheru->path; | |
1da177e4 | 1349 | } |
ae3b5641 AV |
1350 | refcount_inc(&otheru->addr->refcnt); |
1351 | smp_store_release(&newu->addr, otheru->addr); | |
1da177e4 LT |
1352 | |
1353 | /* Set credentials */ | |
109f6e39 | 1354 | copy_peercred(sk, other); |
1da177e4 | 1355 | |
1da177e4 LT |
1356 | sock->state = SS_CONNECTED; |
1357 | sk->sk_state = TCP_ESTABLISHED; | |
830a1e5c BL |
1358 | sock_hold(newsk); |
1359 | ||
4e857c58 | 1360 | smp_mb__after_atomic(); /* sock_hold() does an atomic_inc() */ |
830a1e5c | 1361 | unix_peer(sk) = newsk; |
1da177e4 | 1362 | |
1c92b4e5 | 1363 | unix_state_unlock(sk); |
1da177e4 LT |
1364 | |
1365 | /* take ten and and send info to listening sock */ | |
1366 | spin_lock(&other->sk_receive_queue.lock); | |
1367 | __skb_queue_tail(&other->sk_receive_queue, skb); | |
1da177e4 | 1368 | spin_unlock(&other->sk_receive_queue.lock); |
1c92b4e5 | 1369 | unix_state_unlock(other); |
676d2369 | 1370 | other->sk_data_ready(other); |
1da177e4 LT |
1371 | sock_put(other); |
1372 | return 0; | |
1373 | ||
1374 | out_unlock: | |
1375 | if (other) | |
1c92b4e5 | 1376 | unix_state_unlock(other); |
1da177e4 LT |
1377 | |
1378 | out: | |
40d44446 | 1379 | kfree_skb(skb); |
1da177e4 LT |
1380 | if (newsk) |
1381 | unix_release_sock(newsk, 0); | |
1382 | if (other) | |
1383 | sock_put(other); | |
1384 | return err; | |
1385 | } | |
1386 | ||
1387 | static int unix_socketpair(struct socket *socka, struct socket *sockb) | |
1388 | { | |
e27dfcea | 1389 | struct sock *ska = socka->sk, *skb = sockb->sk; |
1da177e4 LT |
1390 | |
1391 | /* Join our sockets back to back */ | |
1392 | sock_hold(ska); | |
1393 | sock_hold(skb); | |
e27dfcea JK |
1394 | unix_peer(ska) = skb; |
1395 | unix_peer(skb) = ska; | |
109f6e39 EB |
1396 | init_peercred(ska); |
1397 | init_peercred(skb); | |
1da177e4 LT |
1398 | |
1399 | if (ska->sk_type != SOCK_DGRAM) { | |
1400 | ska->sk_state = TCP_ESTABLISHED; | |
1401 | skb->sk_state = TCP_ESTABLISHED; | |
1402 | socka->state = SS_CONNECTED; | |
1403 | sockb->state = SS_CONNECTED; | |
1404 | } | |
1405 | return 0; | |
1406 | } | |
1407 | ||
90c6bd34 DB |
1408 | static void unix_sock_inherit_flags(const struct socket *old, |
1409 | struct socket *new) | |
1410 | { | |
1411 | if (test_bit(SOCK_PASSCRED, &old->flags)) | |
1412 | set_bit(SOCK_PASSCRED, &new->flags); | |
1413 | if (test_bit(SOCK_PASSSEC, &old->flags)) | |
1414 | set_bit(SOCK_PASSSEC, &new->flags); | |
1415 | } | |
1416 | ||
cdfbabfb DH |
1417 | static int unix_accept(struct socket *sock, struct socket *newsock, int flags, |
1418 | bool kern) | |
1da177e4 LT |
1419 | { |
1420 | struct sock *sk = sock->sk; | |
1421 | struct sock *tsk; | |
1422 | struct sk_buff *skb; | |
1423 | int err; | |
1424 | ||
1425 | err = -EOPNOTSUPP; | |
6eba6a37 | 1426 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
1da177e4 LT |
1427 | goto out; |
1428 | ||
1429 | err = -EINVAL; | |
1430 | if (sk->sk_state != TCP_LISTEN) | |
1431 | goto out; | |
1432 | ||
1433 | /* If socket state is TCP_LISTEN it cannot change (for now...), | |
1434 | * so that no locks are necessary. | |
1435 | */ | |
1436 | ||
1437 | skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err); | |
1438 | if (!skb) { | |
1439 | /* This means receive shutdown. */ | |
1440 | if (err == 0) | |
1441 | err = -EINVAL; | |
1442 | goto out; | |
1443 | } | |
1444 | ||
1445 | tsk = skb->sk; | |
1446 | skb_free_datagram(sk, skb); | |
1447 | wake_up_interruptible(&unix_sk(sk)->peer_wait); | |
1448 | ||
1449 | /* attach accepted sock to socket */ | |
1c92b4e5 | 1450 | unix_state_lock(tsk); |
1da177e4 | 1451 | newsock->state = SS_CONNECTED; |
90c6bd34 | 1452 | unix_sock_inherit_flags(sock, newsock); |
1da177e4 | 1453 | sock_graft(tsk, newsock); |
1c92b4e5 | 1454 | unix_state_unlock(tsk); |
1da177e4 LT |
1455 | return 0; |
1456 | ||
1457 | out: | |
1458 | return err; | |
1459 | } | |
1460 | ||
1461 | ||
9b2c45d4 | 1462 | static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int peer) |
1da177e4 LT |
1463 | { |
1464 | struct sock *sk = sock->sk; | |
ae3b5641 | 1465 | struct unix_address *addr; |
13cfa97b | 1466 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr); |
1da177e4 LT |
1467 | int err = 0; |
1468 | ||
1469 | if (peer) { | |
1470 | sk = unix_peer_get(sk); | |
1471 | ||
1472 | err = -ENOTCONN; | |
1473 | if (!sk) | |
1474 | goto out; | |
1475 | err = 0; | |
1476 | } else { | |
1477 | sock_hold(sk); | |
1478 | } | |
1479 | ||
ae3b5641 AV |
1480 | addr = smp_load_acquire(&unix_sk(sk)->addr); |
1481 | if (!addr) { | |
1da177e4 LT |
1482 | sunaddr->sun_family = AF_UNIX; |
1483 | sunaddr->sun_path[0] = 0; | |
9b2c45d4 | 1484 | err = sizeof(short); |
1da177e4 | 1485 | } else { |
9b2c45d4 DV |
1486 | err = addr->len; |
1487 | memcpy(sunaddr, addr->name, addr->len); | |
1da177e4 | 1488 | } |
1da177e4 LT |
1489 | sock_put(sk); |
1490 | out: | |
1491 | return err; | |
1492 | } | |
1493 | ||
f78a5fda | 1494 | static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds) |
7361c36c EB |
1495 | { |
1496 | int err = 0; | |
16e57262 | 1497 | |
f78a5fda | 1498 | UNIXCB(skb).pid = get_pid(scm->pid); |
6b0ee8c0 EB |
1499 | UNIXCB(skb).uid = scm->creds.uid; |
1500 | UNIXCB(skb).gid = scm->creds.gid; | |
7361c36c | 1501 | UNIXCB(skb).fp = NULL; |
37a9a8df | 1502 | unix_get_secdata(scm, skb); |
7361c36c EB |
1503 | if (scm->fp && send_fds) |
1504 | err = unix_attach_fds(scm, skb); | |
1505 | ||
1506 | skb->destructor = unix_destruct_scm; | |
1507 | return err; | |
1508 | } | |
1509 | ||
9490f886 HFS |
1510 | static bool unix_passcred_enabled(const struct socket *sock, |
1511 | const struct sock *other) | |
1512 | { | |
1513 | return test_bit(SOCK_PASSCRED, &sock->flags) || | |
1514 | !other->sk_socket || | |
1515 | test_bit(SOCK_PASSCRED, &other->sk_socket->flags); | |
1516 | } | |
1517 | ||
16e57262 ED |
1518 | /* |
1519 | * Some apps rely on write() giving SCM_CREDENTIALS | |
1520 | * We include credentials if source or destination socket | |
1521 | * asserted SOCK_PASSCRED. | |
1522 | */ | |
1523 | static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock, | |
1524 | const struct sock *other) | |
1525 | { | |
6b0ee8c0 | 1526 | if (UNIXCB(skb).pid) |
16e57262 | 1527 | return; |
9490f886 | 1528 | if (unix_passcred_enabled(sock, other)) { |
16e57262 | 1529 | UNIXCB(skb).pid = get_pid(task_tgid(current)); |
6e0895c2 | 1530 | current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid); |
16e57262 ED |
1531 | } |
1532 | } | |
1533 | ||
9490f886 HFS |
1534 | static int maybe_init_creds(struct scm_cookie *scm, |
1535 | struct socket *socket, | |
1536 | const struct sock *other) | |
1537 | { | |
1538 | int err; | |
1539 | struct msghdr msg = { .msg_controllen = 0 }; | |
1540 | ||
1541 | err = scm_send(socket, &msg, scm, false); | |
1542 | if (err) | |
1543 | return err; | |
1544 | ||
1545 | if (unix_passcred_enabled(socket, other)) { | |
1546 | scm->pid = get_pid(task_tgid(current)); | |
1547 | current_uid_gid(&scm->creds.uid, &scm->creds.gid); | |
1548 | } | |
1549 | return err; | |
1550 | } | |
1551 | ||
1552 | static bool unix_skb_scm_eq(struct sk_buff *skb, | |
1553 | struct scm_cookie *scm) | |
1554 | { | |
1555 | const struct unix_skb_parms *u = &UNIXCB(skb); | |
1556 | ||
1557 | return u->pid == scm->pid && | |
1558 | uid_eq(u->uid, scm->creds.uid) && | |
1559 | gid_eq(u->gid, scm->creds.gid) && | |
1560 | unix_secdata_eq(scm, skb); | |
1561 | } | |
1562 | ||
1da177e4 LT |
1563 | /* |
1564 | * Send AF_UNIX data. | |
1565 | */ | |
1566 | ||
1b784140 YX |
1567 | static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg, |
1568 | size_t len) | |
1da177e4 | 1569 | { |
1da177e4 | 1570 | struct sock *sk = sock->sk; |
3b1e0a65 | 1571 | struct net *net = sock_net(sk); |
1da177e4 | 1572 | struct unix_sock *u = unix_sk(sk); |
342dfc30 | 1573 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name); |
1da177e4 LT |
1574 | struct sock *other = NULL; |
1575 | int namelen = 0; /* fake GCC */ | |
1576 | int err; | |
95c96174 | 1577 | unsigned int hash; |
f78a5fda | 1578 | struct sk_buff *skb; |
1da177e4 | 1579 | long timeo; |
7cc05662 | 1580 | struct scm_cookie scm; |
eb6a2481 | 1581 | int data_len = 0; |
7d267278 | 1582 | int sk_locked; |
1da177e4 | 1583 | |
5f23b734 | 1584 | wait_for_unix_gc(); |
7cc05662 | 1585 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1586 | if (err < 0) |
1587 | return err; | |
1588 | ||
1589 | err = -EOPNOTSUPP; | |
1590 | if (msg->msg_flags&MSG_OOB) | |
1591 | goto out; | |
1592 | ||
1593 | if (msg->msg_namelen) { | |
1594 | err = unix_mkname(sunaddr, msg->msg_namelen, &hash); | |
1595 | if (err < 0) | |
1596 | goto out; | |
1597 | namelen = err; | |
1598 | } else { | |
1599 | sunaddr = NULL; | |
1600 | err = -ENOTCONN; | |
1601 | other = unix_peer_get(sk); | |
1602 | if (!other) | |
1603 | goto out; | |
1604 | } | |
1605 | ||
f64f9e71 JP |
1606 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr |
1607 | && (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1608 | goto out; |
1609 | ||
1610 | err = -EMSGSIZE; | |
1611 | if (len > sk->sk_sndbuf - 32) | |
1612 | goto out; | |
1613 | ||
31ff6aa5 | 1614 | if (len > SKB_MAX_ALLOC) { |
eb6a2481 ED |
1615 | data_len = min_t(size_t, |
1616 | len - SKB_MAX_ALLOC, | |
1617 | MAX_SKB_FRAGS * PAGE_SIZE); | |
31ff6aa5 KT |
1618 | data_len = PAGE_ALIGN(data_len); |
1619 | ||
1620 | BUILD_BUG_ON(SKB_MAX_ALLOC < PAGE_SIZE); | |
1621 | } | |
eb6a2481 ED |
1622 | |
1623 | skb = sock_alloc_send_pskb(sk, len - data_len, data_len, | |
28d64271 ED |
1624 | msg->msg_flags & MSG_DONTWAIT, &err, |
1625 | PAGE_ALLOC_COSTLY_ORDER); | |
e27dfcea | 1626 | if (skb == NULL) |
1da177e4 LT |
1627 | goto out; |
1628 | ||
7cc05662 | 1629 | err = unix_scm_to_skb(&scm, skb, true); |
25888e30 | 1630 | if (err < 0) |
7361c36c | 1631 | goto out_free; |
877ce7c1 | 1632 | |
eb6a2481 ED |
1633 | skb_put(skb, len - data_len); |
1634 | skb->data_len = data_len; | |
1635 | skb->len = len; | |
c0371da6 | 1636 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len); |
1da177e4 LT |
1637 | if (err) |
1638 | goto out_free; | |
1639 | ||
1640 | timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); | |
1641 | ||
1642 | restart: | |
1643 | if (!other) { | |
1644 | err = -ECONNRESET; | |
1645 | if (sunaddr == NULL) | |
1646 | goto out_free; | |
1647 | ||
097e66c5 | 1648 | other = unix_find_other(net, sunaddr, namelen, sk->sk_type, |
1da177e4 | 1649 | hash, &err); |
e27dfcea | 1650 | if (other == NULL) |
1da177e4 LT |
1651 | goto out_free; |
1652 | } | |
1653 | ||
d6ae3bae AC |
1654 | if (sk_filter(other, skb) < 0) { |
1655 | /* Toss the packet but do not return any error to the sender */ | |
1656 | err = len; | |
1657 | goto out_free; | |
1658 | } | |
1659 | ||
7d267278 | 1660 | sk_locked = 0; |
1c92b4e5 | 1661 | unix_state_lock(other); |
7d267278 | 1662 | restart_locked: |
1da177e4 LT |
1663 | err = -EPERM; |
1664 | if (!unix_may_send(sk, other)) | |
1665 | goto out_unlock; | |
1666 | ||
7d267278 | 1667 | if (unlikely(sock_flag(other, SOCK_DEAD))) { |
1da177e4 LT |
1668 | /* |
1669 | * Check with 1003.1g - what should | |
1670 | * datagram error | |
1671 | */ | |
1c92b4e5 | 1672 | unix_state_unlock(other); |
1da177e4 LT |
1673 | sock_put(other); |
1674 | ||
7d267278 RW |
1675 | if (!sk_locked) |
1676 | unix_state_lock(sk); | |
1677 | ||
1da177e4 | 1678 | err = 0; |
1da177e4 | 1679 | if (unix_peer(sk) == other) { |
e27dfcea | 1680 | unix_peer(sk) = NULL; |
7d267278 RW |
1681 | unix_dgram_peer_wake_disconnect_wakeup(sk, other); |
1682 | ||
1c92b4e5 | 1683 | unix_state_unlock(sk); |
1da177e4 LT |
1684 | |
1685 | unix_dgram_disconnected(sk, other); | |
1686 | sock_put(other); | |
1687 | err = -ECONNREFUSED; | |
1688 | } else { | |
1c92b4e5 | 1689 | unix_state_unlock(sk); |
1da177e4 LT |
1690 | } |
1691 | ||
1692 | other = NULL; | |
1693 | if (err) | |
1694 | goto out_free; | |
1695 | goto restart; | |
1696 | } | |
1697 | ||
1698 | err = -EPIPE; | |
1699 | if (other->sk_shutdown & RCV_SHUTDOWN) | |
1700 | goto out_unlock; | |
1701 | ||
1702 | if (sk->sk_type != SOCK_SEQPACKET) { | |
1703 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1704 | if (err) | |
1705 | goto out_unlock; | |
1706 | } | |
1707 | ||
a5527dda RW |
1708 | /* other == sk && unix_peer(other) != sk if |
1709 | * - unix_peer(sk) == NULL, destination address bound to sk | |
1710 | * - unix_peer(sk) == sk by time of get but disconnected before lock | |
1711 | */ | |
1712 | if (other != sk && | |
1713 | unlikely(unix_peer(other) != sk && unix_recvq_full(other))) { | |
7d267278 RW |
1714 | if (timeo) { |
1715 | timeo = unix_wait_for_peer(other, timeo); | |
1716 | ||
1717 | err = sock_intr_errno(timeo); | |
1718 | if (signal_pending(current)) | |
1719 | goto out_free; | |
1720 | ||
1721 | goto restart; | |
1da177e4 LT |
1722 | } |
1723 | ||
7d267278 RW |
1724 | if (!sk_locked) { |
1725 | unix_state_unlock(other); | |
1726 | unix_state_double_lock(sk, other); | |
1727 | } | |
1da177e4 | 1728 | |
7d267278 RW |
1729 | if (unix_peer(sk) != other || |
1730 | unix_dgram_peer_wake_me(sk, other)) { | |
1731 | err = -EAGAIN; | |
1732 | sk_locked = 1; | |
1733 | goto out_unlock; | |
1734 | } | |
1da177e4 | 1735 | |
7d267278 RW |
1736 | if (!sk_locked) { |
1737 | sk_locked = 1; | |
1738 | goto restart_locked; | |
1739 | } | |
1da177e4 LT |
1740 | } |
1741 | ||
7d267278 RW |
1742 | if (unlikely(sk_locked)) |
1743 | unix_state_unlock(sk); | |
1744 | ||
3f66116e AC |
1745 | if (sock_flag(other, SOCK_RCVTSTAMP)) |
1746 | __net_timestamp(skb); | |
16e57262 | 1747 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1748 | skb_queue_tail(&other->sk_receive_queue, skb); |
1c92b4e5 | 1749 | unix_state_unlock(other); |
676d2369 | 1750 | other->sk_data_ready(other); |
1da177e4 | 1751 | sock_put(other); |
7cc05662 | 1752 | scm_destroy(&scm); |
1da177e4 LT |
1753 | return len; |
1754 | ||
1755 | out_unlock: | |
7d267278 RW |
1756 | if (sk_locked) |
1757 | unix_state_unlock(sk); | |
1c92b4e5 | 1758 | unix_state_unlock(other); |
1da177e4 LT |
1759 | out_free: |
1760 | kfree_skb(skb); | |
1761 | out: | |
1762 | if (other) | |
1763 | sock_put(other); | |
7cc05662 | 1764 | scm_destroy(&scm); |
1da177e4 LT |
1765 | return err; |
1766 | } | |
1767 | ||
e370a723 | 1768 | /* We use paged skbs for stream sockets, and limit occupancy to 32768 |
d4e9a408 | 1769 | * bytes, and a minimum of a full page. |
e370a723 ED |
1770 | */ |
1771 | #define UNIX_SKB_FRAGS_SZ (PAGE_SIZE << get_order(32768)) | |
ac7bfa62 | 1772 | |
1b784140 YX |
1773 | static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg, |
1774 | size_t len) | |
1da177e4 | 1775 | { |
1da177e4 LT |
1776 | struct sock *sk = sock->sk; |
1777 | struct sock *other = NULL; | |
6eba6a37 | 1778 | int err, size; |
f78a5fda | 1779 | struct sk_buff *skb; |
e27dfcea | 1780 | int sent = 0; |
7cc05662 | 1781 | struct scm_cookie scm; |
8ba69ba6 | 1782 | bool fds_sent = false; |
e370a723 | 1783 | int data_len; |
1da177e4 | 1784 | |
5f23b734 | 1785 | wait_for_unix_gc(); |
7cc05662 | 1786 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1787 | if (err < 0) |
1788 | return err; | |
1789 | ||
1790 | err = -EOPNOTSUPP; | |
1791 | if (msg->msg_flags&MSG_OOB) | |
1792 | goto out_err; | |
1793 | ||
1794 | if (msg->msg_namelen) { | |
1795 | err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; | |
1796 | goto out_err; | |
1797 | } else { | |
1da177e4 | 1798 | err = -ENOTCONN; |
830a1e5c | 1799 | other = unix_peer(sk); |
1da177e4 LT |
1800 | if (!other) |
1801 | goto out_err; | |
1802 | } | |
1803 | ||
1804 | if (sk->sk_shutdown & SEND_SHUTDOWN) | |
1805 | goto pipe_err; | |
1806 | ||
6eba6a37 | 1807 | while (sent < len) { |
e370a723 | 1808 | size = len - sent; |
1da177e4 LT |
1809 | |
1810 | /* Keep two messages in the pipe so it schedules better */ | |
e370a723 | 1811 | size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64); |
1da177e4 | 1812 | |
e370a723 ED |
1813 | /* allow fallback to order-0 allocations */ |
1814 | size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ); | |
ac7bfa62 | 1815 | |
e370a723 | 1816 | data_len = max_t(int, 0, size - SKB_MAX_HEAD(0)); |
1da177e4 | 1817 | |
31ff6aa5 KT |
1818 | data_len = min_t(size_t, size, PAGE_ALIGN(data_len)); |
1819 | ||
e370a723 | 1820 | skb = sock_alloc_send_pskb(sk, size - data_len, data_len, |
28d64271 ED |
1821 | msg->msg_flags & MSG_DONTWAIT, &err, |
1822 | get_order(UNIX_SKB_FRAGS_SZ)); | |
e370a723 | 1823 | if (!skb) |
1da177e4 LT |
1824 | goto out_err; |
1825 | ||
f78a5fda | 1826 | /* Only send the fds in the first buffer */ |
7cc05662 | 1827 | err = unix_scm_to_skb(&scm, skb, !fds_sent); |
25888e30 | 1828 | if (err < 0) { |
7361c36c | 1829 | kfree_skb(skb); |
f78a5fda | 1830 | goto out_err; |
6209344f | 1831 | } |
7361c36c | 1832 | fds_sent = true; |
1da177e4 | 1833 | |
e370a723 ED |
1834 | skb_put(skb, size - data_len); |
1835 | skb->data_len = data_len; | |
1836 | skb->len = size; | |
c0371da6 | 1837 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size); |
6eba6a37 | 1838 | if (err) { |
1da177e4 | 1839 | kfree_skb(skb); |
f78a5fda | 1840 | goto out_err; |
1da177e4 LT |
1841 | } |
1842 | ||
1c92b4e5 | 1843 | unix_state_lock(other); |
1da177e4 LT |
1844 | |
1845 | if (sock_flag(other, SOCK_DEAD) || | |
1846 | (other->sk_shutdown & RCV_SHUTDOWN)) | |
1847 | goto pipe_err_free; | |
1848 | ||
16e57262 | 1849 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1850 | skb_queue_tail(&other->sk_receive_queue, skb); |
1c92b4e5 | 1851 | unix_state_unlock(other); |
676d2369 | 1852 | other->sk_data_ready(other); |
e27dfcea | 1853 | sent += size; |
1da177e4 | 1854 | } |
1da177e4 | 1855 | |
7cc05662 | 1856 | scm_destroy(&scm); |
1da177e4 LT |
1857 | |
1858 | return sent; | |
1859 | ||
1860 | pipe_err_free: | |
1c92b4e5 | 1861 | unix_state_unlock(other); |
1da177e4 LT |
1862 | kfree_skb(skb); |
1863 | pipe_err: | |
6eba6a37 ED |
1864 | if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL)) |
1865 | send_sig(SIGPIPE, current, 0); | |
1da177e4 LT |
1866 | err = -EPIPE; |
1867 | out_err: | |
7cc05662 | 1868 | scm_destroy(&scm); |
1da177e4 LT |
1869 | return sent ? : err; |
1870 | } | |
1871 | ||
869e7c62 HFS |
1872 | static ssize_t unix_stream_sendpage(struct socket *socket, struct page *page, |
1873 | int offset, size_t size, int flags) | |
1874 | { | |
9490f886 HFS |
1875 | int err; |
1876 | bool send_sigpipe = false; | |
1877 | bool init_scm = true; | |
1878 | struct scm_cookie scm; | |
869e7c62 HFS |
1879 | struct sock *other, *sk = socket->sk; |
1880 | struct sk_buff *skb, *newskb = NULL, *tail = NULL; | |
1881 | ||
1882 | if (flags & MSG_OOB) | |
1883 | return -EOPNOTSUPP; | |
1884 | ||
1885 | other = unix_peer(sk); | |
1886 | if (!other || sk->sk_state != TCP_ESTABLISHED) | |
1887 | return -ENOTCONN; | |
1888 | ||
1889 | if (false) { | |
1890 | alloc_skb: | |
1891 | unix_state_unlock(other); | |
6e1ce3c3 | 1892 | mutex_unlock(&unix_sk(other)->iolock); |
869e7c62 HFS |
1893 | newskb = sock_alloc_send_pskb(sk, 0, 0, flags & MSG_DONTWAIT, |
1894 | &err, 0); | |
1895 | if (!newskb) | |
9490f886 | 1896 | goto err; |
869e7c62 HFS |
1897 | } |
1898 | ||
6e1ce3c3 | 1899 | /* we must acquire iolock as we modify already present |
869e7c62 HFS |
1900 | * skbs in the sk_receive_queue and mess with skb->len |
1901 | */ | |
6e1ce3c3 | 1902 | err = mutex_lock_interruptible(&unix_sk(other)->iolock); |
869e7c62 HFS |
1903 | if (err) { |
1904 | err = flags & MSG_DONTWAIT ? -EAGAIN : -ERESTARTSYS; | |
869e7c62 HFS |
1905 | goto err; |
1906 | } | |
1907 | ||
1908 | if (sk->sk_shutdown & SEND_SHUTDOWN) { | |
1909 | err = -EPIPE; | |
9490f886 | 1910 | send_sigpipe = true; |
869e7c62 HFS |
1911 | goto err_unlock; |
1912 | } | |
1913 | ||
1914 | unix_state_lock(other); | |
1915 | ||
1916 | if (sock_flag(other, SOCK_DEAD) || | |
1917 | other->sk_shutdown & RCV_SHUTDOWN) { | |
1918 | err = -EPIPE; | |
9490f886 | 1919 | send_sigpipe = true; |
869e7c62 HFS |
1920 | goto err_state_unlock; |
1921 | } | |
1922 | ||
9490f886 HFS |
1923 | if (init_scm) { |
1924 | err = maybe_init_creds(&scm, socket, other); | |
1925 | if (err) | |
1926 | goto err_state_unlock; | |
1927 | init_scm = false; | |
1928 | } | |
1929 | ||
869e7c62 HFS |
1930 | skb = skb_peek_tail(&other->sk_receive_queue); |
1931 | if (tail && tail == skb) { | |
1932 | skb = newskb; | |
9490f886 HFS |
1933 | } else if (!skb || !unix_skb_scm_eq(skb, &scm)) { |
1934 | if (newskb) { | |
869e7c62 | 1935 | skb = newskb; |
9490f886 HFS |
1936 | } else { |
1937 | tail = skb; | |
869e7c62 | 1938 | goto alloc_skb; |
9490f886 | 1939 | } |
869e7c62 HFS |
1940 | } else if (newskb) { |
1941 | /* this is fast path, we don't necessarily need to | |
1942 | * call to kfree_skb even though with newskb == NULL | |
1943 | * this - does no harm | |
1944 | */ | |
1945 | consume_skb(newskb); | |
8844f972 | 1946 | newskb = NULL; |
869e7c62 HFS |
1947 | } |
1948 | ||
1949 | if (skb_append_pagefrags(skb, page, offset, size)) { | |
1950 | tail = skb; | |
1951 | goto alloc_skb; | |
1952 | } | |
1953 | ||
1954 | skb->len += size; | |
1955 | skb->data_len += size; | |
1956 | skb->truesize += size; | |
14afee4b | 1957 | refcount_add(size, &sk->sk_wmem_alloc); |
869e7c62 | 1958 | |
a3a116e0 | 1959 | if (newskb) { |
9490f886 HFS |
1960 | err = unix_scm_to_skb(&scm, skb, false); |
1961 | if (err) | |
1962 | goto err_state_unlock; | |
a3a116e0 | 1963 | spin_lock(&other->sk_receive_queue.lock); |
869e7c62 | 1964 | __skb_queue_tail(&other->sk_receive_queue, newskb); |
a3a116e0 HFS |
1965 | spin_unlock(&other->sk_receive_queue.lock); |
1966 | } | |
869e7c62 HFS |
1967 | |
1968 | unix_state_unlock(other); | |
6e1ce3c3 | 1969 | mutex_unlock(&unix_sk(other)->iolock); |
869e7c62 HFS |
1970 | |
1971 | other->sk_data_ready(other); | |
9490f886 | 1972 | scm_destroy(&scm); |
869e7c62 HFS |
1973 | return size; |
1974 | ||
1975 | err_state_unlock: | |
1976 | unix_state_unlock(other); | |
1977 | err_unlock: | |
6e1ce3c3 | 1978 | mutex_unlock(&unix_sk(other)->iolock); |
869e7c62 HFS |
1979 | err: |
1980 | kfree_skb(newskb); | |
1981 | if (send_sigpipe && !(flags & MSG_NOSIGNAL)) | |
1982 | send_sig(SIGPIPE, current, 0); | |
9490f886 HFS |
1983 | if (!init_scm) |
1984 | scm_destroy(&scm); | |
869e7c62 HFS |
1985 | return err; |
1986 | } | |
1987 | ||
1b784140 YX |
1988 | static int unix_seqpacket_sendmsg(struct socket *sock, struct msghdr *msg, |
1989 | size_t len) | |
1da177e4 LT |
1990 | { |
1991 | int err; | |
1992 | struct sock *sk = sock->sk; | |
ac7bfa62 | 1993 | |
1da177e4 LT |
1994 | err = sock_error(sk); |
1995 | if (err) | |
1996 | return err; | |
1997 | ||
1998 | if (sk->sk_state != TCP_ESTABLISHED) | |
1999 | return -ENOTCONN; | |
2000 | ||
2001 | if (msg->msg_namelen) | |
2002 | msg->msg_namelen = 0; | |
2003 | ||
1b784140 | 2004 | return unix_dgram_sendmsg(sock, msg, len); |
1da177e4 | 2005 | } |
ac7bfa62 | 2006 | |
1b784140 YX |
2007 | static int unix_seqpacket_recvmsg(struct socket *sock, struct msghdr *msg, |
2008 | size_t size, int flags) | |
a05d2ad1 EB |
2009 | { |
2010 | struct sock *sk = sock->sk; | |
2011 | ||
2012 | if (sk->sk_state != TCP_ESTABLISHED) | |
2013 | return -ENOTCONN; | |
2014 | ||
1b784140 | 2015 | return unix_dgram_recvmsg(sock, msg, size, flags); |
a05d2ad1 EB |
2016 | } |
2017 | ||
1da177e4 LT |
2018 | static void unix_copy_addr(struct msghdr *msg, struct sock *sk) |
2019 | { | |
ae3b5641 | 2020 | struct unix_address *addr = smp_load_acquire(&unix_sk(sk)->addr); |
1da177e4 | 2021 | |
ae3b5641 AV |
2022 | if (addr) { |
2023 | msg->msg_namelen = addr->len; | |
2024 | memcpy(msg->msg_name, addr->name, addr->len); | |
1da177e4 LT |
2025 | } |
2026 | } | |
2027 | ||
1b784140 YX |
2028 | static int unix_dgram_recvmsg(struct socket *sock, struct msghdr *msg, |
2029 | size_t size, int flags) | |
1da177e4 | 2030 | { |
7cc05662 | 2031 | struct scm_cookie scm; |
1da177e4 LT |
2032 | struct sock *sk = sock->sk; |
2033 | struct unix_sock *u = unix_sk(sk); | |
64874280 RW |
2034 | struct sk_buff *skb, *last; |
2035 | long timeo; | |
fd69c399 | 2036 | int skip; |
1da177e4 LT |
2037 | int err; |
2038 | ||
2039 | err = -EOPNOTSUPP; | |
2040 | if (flags&MSG_OOB) | |
2041 | goto out; | |
2042 | ||
64874280 | 2043 | timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); |
1da177e4 | 2044 | |
64874280 | 2045 | do { |
6e1ce3c3 | 2046 | mutex_lock(&u->iolock); |
f55bb7f9 | 2047 | |
64874280 | 2048 | skip = sk_peek_offset(sk, flags); |
fd69c399 PA |
2049 | skb = __skb_try_recv_datagram(sk, flags, NULL, &skip, &err, |
2050 | &last); | |
64874280 RW |
2051 | if (skb) |
2052 | break; | |
2053 | ||
6e1ce3c3 | 2054 | mutex_unlock(&u->iolock); |
64874280 RW |
2055 | |
2056 | if (err != -EAGAIN) | |
2057 | break; | |
2058 | } while (timeo && | |
2059 | !__skb_wait_for_more_packets(sk, &err, &timeo, last)); | |
2060 | ||
6e1ce3c3 | 2061 | if (!skb) { /* implies iolock unlocked */ |
0a112258 FZ |
2062 | unix_state_lock(sk); |
2063 | /* Signal EOF on disconnected non-blocking SEQPACKET socket. */ | |
2064 | if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN && | |
2065 | (sk->sk_shutdown & RCV_SHUTDOWN)) | |
2066 | err = 0; | |
2067 | unix_state_unlock(sk); | |
64874280 | 2068 | goto out; |
0a112258 | 2069 | } |
1da177e4 | 2070 | |
77b75f4d RW |
2071 | if (wq_has_sleeper(&u->peer_wait)) |
2072 | wake_up_interruptible_sync_poll(&u->peer_wait, | |
a9a08845 LT |
2073 | EPOLLOUT | EPOLLWRNORM | |
2074 | EPOLLWRBAND); | |
1da177e4 LT |
2075 | |
2076 | if (msg->msg_name) | |
2077 | unix_copy_addr(msg, skb->sk); | |
2078 | ||
f55bb7f9 PE |
2079 | if (size > skb->len - skip) |
2080 | size = skb->len - skip; | |
2081 | else if (size < skb->len - skip) | |
1da177e4 LT |
2082 | msg->msg_flags |= MSG_TRUNC; |
2083 | ||
51f3d02b | 2084 | err = skb_copy_datagram_msg(skb, skip, msg, size); |
1da177e4 LT |
2085 | if (err) |
2086 | goto out_free; | |
2087 | ||
3f66116e AC |
2088 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
2089 | __sock_recv_timestamp(msg, sk, skb); | |
2090 | ||
7cc05662 CH |
2091 | memset(&scm, 0, sizeof(scm)); |
2092 | ||
2093 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); | |
2094 | unix_set_secdata(&scm, skb); | |
1da177e4 | 2095 | |
6eba6a37 | 2096 | if (!(flags & MSG_PEEK)) { |
1da177e4 | 2097 | if (UNIXCB(skb).fp) |
7cc05662 | 2098 | unix_detach_fds(&scm, skb); |
f55bb7f9 PE |
2099 | |
2100 | sk_peek_offset_bwd(sk, skb->len); | |
6eba6a37 | 2101 | } else { |
1da177e4 LT |
2102 | /* It is questionable: on PEEK we could: |
2103 | - do not return fds - good, but too simple 8) | |
2104 | - return fds, and do not return them on read (old strategy, | |
2105 | apparently wrong) | |
2106 | - clone fds (I chose it for now, it is the most universal | |
2107 | solution) | |
ac7bfa62 YH |
2108 | |
2109 | POSIX 1003.1g does not actually define this clearly | |
2110 | at all. POSIX 1003.1g doesn't define a lot of things | |
2111 | clearly however! | |
2112 | ||
1da177e4 | 2113 | */ |
f55bb7f9 PE |
2114 | |
2115 | sk_peek_offset_fwd(sk, size); | |
2116 | ||
1da177e4 | 2117 | if (UNIXCB(skb).fp) |
7cc05662 | 2118 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 2119 | } |
9f6f9af7 | 2120 | err = (flags & MSG_TRUNC) ? skb->len - skip : size; |
1da177e4 | 2121 | |
7cc05662 | 2122 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2123 | |
2124 | out_free: | |
6eba6a37 | 2125 | skb_free_datagram(sk, skb); |
6e1ce3c3 | 2126 | mutex_unlock(&u->iolock); |
1da177e4 LT |
2127 | out: |
2128 | return err; | |
2129 | } | |
2130 | ||
2131 | /* | |
79f632c7 | 2132 | * Sleep until more data has arrived. But check for races.. |
1da177e4 | 2133 | */ |
79f632c7 | 2134 | static long unix_stream_data_wait(struct sock *sk, long timeo, |
06a77b07 WC |
2135 | struct sk_buff *last, unsigned int last_len, |
2136 | bool freezable) | |
1da177e4 | 2137 | { |
2b514574 | 2138 | struct sk_buff *tail; |
1da177e4 LT |
2139 | DEFINE_WAIT(wait); |
2140 | ||
1c92b4e5 | 2141 | unix_state_lock(sk); |
1da177e4 LT |
2142 | |
2143 | for (;;) { | |
aa395145 | 2144 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 2145 | |
2b514574 HFS |
2146 | tail = skb_peek_tail(&sk->sk_receive_queue); |
2147 | if (tail != last || | |
2148 | (tail && tail->len != last_len) || | |
1da177e4 LT |
2149 | sk->sk_err || |
2150 | (sk->sk_shutdown & RCV_SHUTDOWN) || | |
2151 | signal_pending(current) || | |
2152 | !timeo) | |
2153 | break; | |
2154 | ||
9cd3e072 | 2155 | sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk); |
1c92b4e5 | 2156 | unix_state_unlock(sk); |
06a77b07 WC |
2157 | if (freezable) |
2158 | timeo = freezable_schedule_timeout(timeo); | |
2159 | else | |
2160 | timeo = schedule_timeout(timeo); | |
1c92b4e5 | 2161 | unix_state_lock(sk); |
b48732e4 MS |
2162 | |
2163 | if (sock_flag(sk, SOCK_DEAD)) | |
2164 | break; | |
2165 | ||
9cd3e072 | 2166 | sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk); |
1da177e4 LT |
2167 | } |
2168 | ||
aa395145 | 2169 | finish_wait(sk_sleep(sk), &wait); |
1c92b4e5 | 2170 | unix_state_unlock(sk); |
1da177e4 LT |
2171 | return timeo; |
2172 | } | |
2173 | ||
e370a723 ED |
2174 | static unsigned int unix_skb_len(const struct sk_buff *skb) |
2175 | { | |
2176 | return skb->len - UNIXCB(skb).consumed; | |
2177 | } | |
2178 | ||
2b514574 HFS |
2179 | struct unix_stream_read_state { |
2180 | int (*recv_actor)(struct sk_buff *, int, int, | |
2181 | struct unix_stream_read_state *); | |
2182 | struct socket *socket; | |
2183 | struct msghdr *msg; | |
2184 | struct pipe_inode_info *pipe; | |
2185 | size_t size; | |
2186 | int flags; | |
2187 | unsigned int splice_flags; | |
2188 | }; | |
2189 | ||
06a77b07 WC |
2190 | static int unix_stream_read_generic(struct unix_stream_read_state *state, |
2191 | bool freezable) | |
1da177e4 | 2192 | { |
7cc05662 | 2193 | struct scm_cookie scm; |
2b514574 | 2194 | struct socket *sock = state->socket; |
1da177e4 LT |
2195 | struct sock *sk = sock->sk; |
2196 | struct unix_sock *u = unix_sk(sk); | |
1da177e4 | 2197 | int copied = 0; |
2b514574 | 2198 | int flags = state->flags; |
de144391 | 2199 | int noblock = flags & MSG_DONTWAIT; |
2b514574 | 2200 | bool check_creds = false; |
1da177e4 LT |
2201 | int target; |
2202 | int err = 0; | |
2203 | long timeo; | |
fc0d7536 | 2204 | int skip; |
2b514574 HFS |
2205 | size_t size = state->size; |
2206 | unsigned int last_len; | |
1da177e4 | 2207 | |
1b92ee3d RW |
2208 | if (unlikely(sk->sk_state != TCP_ESTABLISHED)) { |
2209 | err = -EINVAL; | |
1da177e4 | 2210 | goto out; |
1b92ee3d | 2211 | } |
1da177e4 | 2212 | |
1b92ee3d RW |
2213 | if (unlikely(flags & MSG_OOB)) { |
2214 | err = -EOPNOTSUPP; | |
1da177e4 | 2215 | goto out; |
1b92ee3d | 2216 | } |
1da177e4 | 2217 | |
2b514574 | 2218 | target = sock_rcvlowat(sk, flags & MSG_WAITALL, size); |
de144391 | 2219 | timeo = sock_rcvtimeo(sk, noblock); |
1da177e4 | 2220 | |
2b514574 HFS |
2221 | memset(&scm, 0, sizeof(scm)); |
2222 | ||
1da177e4 LT |
2223 | /* Lock the socket to prevent queue disordering |
2224 | * while sleeps in memcpy_tomsg | |
2225 | */ | |
6e1ce3c3 | 2226 | mutex_lock(&u->iolock); |
1da177e4 | 2227 | |
a0917e0b | 2228 | skip = max(sk_peek_offset(sk, flags), 0); |
e9193d60 | 2229 | |
6eba6a37 | 2230 | do { |
1da177e4 | 2231 | int chunk; |
73ed5d25 | 2232 | bool drop_skb; |
79f632c7 | 2233 | struct sk_buff *skb, *last; |
1da177e4 | 2234 | |
18eceb81 | 2235 | redo: |
3c0d2f37 | 2236 | unix_state_lock(sk); |
b48732e4 MS |
2237 | if (sock_flag(sk, SOCK_DEAD)) { |
2238 | err = -ECONNRESET; | |
2239 | goto unlock; | |
2240 | } | |
79f632c7 | 2241 | last = skb = skb_peek(&sk->sk_receive_queue); |
2b514574 | 2242 | last_len = last ? last->len : 0; |
fc0d7536 | 2243 | again: |
6eba6a37 | 2244 | if (skb == NULL) { |
1da177e4 | 2245 | if (copied >= target) |
3c0d2f37 | 2246 | goto unlock; |
1da177e4 LT |
2247 | |
2248 | /* | |
2249 | * POSIX 1003.1g mandates this order. | |
2250 | */ | |
ac7bfa62 | 2251 | |
6eba6a37 ED |
2252 | err = sock_error(sk); |
2253 | if (err) | |
3c0d2f37 | 2254 | goto unlock; |
1da177e4 | 2255 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
3c0d2f37 MS |
2256 | goto unlock; |
2257 | ||
2258 | unix_state_unlock(sk); | |
1b92ee3d RW |
2259 | if (!timeo) { |
2260 | err = -EAGAIN; | |
1da177e4 | 2261 | break; |
1b92ee3d RW |
2262 | } |
2263 | ||
6e1ce3c3 | 2264 | mutex_unlock(&u->iolock); |
1da177e4 | 2265 | |
2b514574 | 2266 | timeo = unix_stream_data_wait(sk, timeo, last, |
06a77b07 | 2267 | last_len, freezable); |
1da177e4 | 2268 | |
3822b5c2 | 2269 | if (signal_pending(current)) { |
1da177e4 | 2270 | err = sock_intr_errno(timeo); |
fa0dc04d | 2271 | scm_destroy(&scm); |
1da177e4 LT |
2272 | goto out; |
2273 | } | |
b3ca9b02 | 2274 | |
6e1ce3c3 | 2275 | mutex_lock(&u->iolock); |
18eceb81 | 2276 | goto redo; |
2b514574 | 2277 | unlock: |
3c0d2f37 MS |
2278 | unix_state_unlock(sk); |
2279 | break; | |
1da177e4 | 2280 | } |
fc0d7536 | 2281 | |
e370a723 ED |
2282 | while (skip >= unix_skb_len(skb)) { |
2283 | skip -= unix_skb_len(skb); | |
79f632c7 | 2284 | last = skb; |
2b514574 | 2285 | last_len = skb->len; |
fc0d7536 | 2286 | skb = skb_peek_next(skb, &sk->sk_receive_queue); |
79f632c7 BP |
2287 | if (!skb) |
2288 | goto again; | |
fc0d7536 PE |
2289 | } |
2290 | ||
3c0d2f37 | 2291 | unix_state_unlock(sk); |
1da177e4 LT |
2292 | |
2293 | if (check_creds) { | |
2294 | /* Never glue messages from different writers */ | |
9490f886 | 2295 | if (!unix_skb_scm_eq(skb, &scm)) |
1da177e4 | 2296 | break; |
0e82e7f6 | 2297 | } else if (test_bit(SOCK_PASSCRED, &sock->flags)) { |
1da177e4 | 2298 | /* Copy credentials */ |
7cc05662 | 2299 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); |
37a9a8df | 2300 | unix_set_secdata(&scm, skb); |
2b514574 | 2301 | check_creds = true; |
1da177e4 LT |
2302 | } |
2303 | ||
2304 | /* Copy address just once */ | |
2b514574 HFS |
2305 | if (state->msg && state->msg->msg_name) { |
2306 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, | |
2307 | state->msg->msg_name); | |
2308 | unix_copy_addr(state->msg, skb->sk); | |
1da177e4 LT |
2309 | sunaddr = NULL; |
2310 | } | |
2311 | ||
e370a723 | 2312 | chunk = min_t(unsigned int, unix_skb_len(skb) - skip, size); |
73ed5d25 | 2313 | skb_get(skb); |
2b514574 | 2314 | chunk = state->recv_actor(skb, skip, chunk, state); |
73ed5d25 HFS |
2315 | drop_skb = !unix_skb_len(skb); |
2316 | /* skb is only safe to use if !drop_skb */ | |
2317 | consume_skb(skb); | |
2b514574 | 2318 | if (chunk < 0) { |
1da177e4 LT |
2319 | if (copied == 0) |
2320 | copied = -EFAULT; | |
2321 | break; | |
2322 | } | |
2323 | copied += chunk; | |
2324 | size -= chunk; | |
2325 | ||
73ed5d25 HFS |
2326 | if (drop_skb) { |
2327 | /* the skb was touched by a concurrent reader; | |
2328 | * we should not expect anything from this skb | |
2329 | * anymore and assume it invalid - we can be | |
2330 | * sure it was dropped from the socket queue | |
2331 | * | |
2332 | * let's report a short read | |
2333 | */ | |
2334 | err = 0; | |
2335 | break; | |
2336 | } | |
2337 | ||
1da177e4 | 2338 | /* Mark read part of skb as used */ |
6eba6a37 | 2339 | if (!(flags & MSG_PEEK)) { |
e370a723 | 2340 | UNIXCB(skb).consumed += chunk; |
1da177e4 | 2341 | |
fc0d7536 PE |
2342 | sk_peek_offset_bwd(sk, chunk); |
2343 | ||
1da177e4 | 2344 | if (UNIXCB(skb).fp) |
7cc05662 | 2345 | unix_detach_fds(&scm, skb); |
1da177e4 | 2346 | |
e370a723 | 2347 | if (unix_skb_len(skb)) |
1da177e4 | 2348 | break; |
1da177e4 | 2349 | |
6f01fd6e | 2350 | skb_unlink(skb, &sk->sk_receive_queue); |
70d4bf6d | 2351 | consume_skb(skb); |
1da177e4 | 2352 | |
7cc05662 | 2353 | if (scm.fp) |
1da177e4 | 2354 | break; |
6eba6a37 | 2355 | } else { |
1da177e4 LT |
2356 | /* It is questionable, see note in unix_dgram_recvmsg. |
2357 | */ | |
2358 | if (UNIXCB(skb).fp) | |
7cc05662 | 2359 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 2360 | |
e9193d60 | 2361 | sk_peek_offset_fwd(sk, chunk); |
fc0d7536 | 2362 | |
9f389e35 AC |
2363 | if (UNIXCB(skb).fp) |
2364 | break; | |
2365 | ||
e9193d60 | 2366 | skip = 0; |
9f389e35 AC |
2367 | last = skb; |
2368 | last_len = skb->len; | |
2369 | unix_state_lock(sk); | |
2370 | skb = skb_peek_next(skb, &sk->sk_receive_queue); | |
2371 | if (skb) | |
2372 | goto again; | |
2373 | unix_state_unlock(sk); | |
1da177e4 LT |
2374 | break; |
2375 | } | |
2376 | } while (size); | |
2377 | ||
6e1ce3c3 | 2378 | mutex_unlock(&u->iolock); |
2b514574 HFS |
2379 | if (state->msg) |
2380 | scm_recv(sock, state->msg, &scm, flags); | |
2381 | else | |
2382 | scm_destroy(&scm); | |
1da177e4 LT |
2383 | out: |
2384 | return copied ? : err; | |
2385 | } | |
2386 | ||
2b514574 HFS |
2387 | static int unix_stream_read_actor(struct sk_buff *skb, |
2388 | int skip, int chunk, | |
2389 | struct unix_stream_read_state *state) | |
2390 | { | |
2391 | int ret; | |
2392 | ||
2393 | ret = skb_copy_datagram_msg(skb, UNIXCB(skb).consumed + skip, | |
2394 | state->msg, chunk); | |
2395 | return ret ?: chunk; | |
2396 | } | |
2397 | ||
2398 | static int unix_stream_recvmsg(struct socket *sock, struct msghdr *msg, | |
2399 | size_t size, int flags) | |
2400 | { | |
2401 | struct unix_stream_read_state state = { | |
2402 | .recv_actor = unix_stream_read_actor, | |
2403 | .socket = sock, | |
2404 | .msg = msg, | |
2405 | .size = size, | |
2406 | .flags = flags | |
2407 | }; | |
2408 | ||
06a77b07 | 2409 | return unix_stream_read_generic(&state, true); |
2b514574 HFS |
2410 | } |
2411 | ||
2b514574 HFS |
2412 | static int unix_stream_splice_actor(struct sk_buff *skb, |
2413 | int skip, int chunk, | |
2414 | struct unix_stream_read_state *state) | |
2415 | { | |
2416 | return skb_splice_bits(skb, state->socket->sk, | |
2417 | UNIXCB(skb).consumed + skip, | |
25869262 | 2418 | state->pipe, chunk, state->splice_flags); |
2b514574 HFS |
2419 | } |
2420 | ||
2421 | static ssize_t unix_stream_splice_read(struct socket *sock, loff_t *ppos, | |
2422 | struct pipe_inode_info *pipe, | |
2423 | size_t size, unsigned int flags) | |
2424 | { | |
2425 | struct unix_stream_read_state state = { | |
2426 | .recv_actor = unix_stream_splice_actor, | |
2427 | .socket = sock, | |
2428 | .pipe = pipe, | |
2429 | .size = size, | |
2430 | .splice_flags = flags, | |
2431 | }; | |
2432 | ||
2433 | if (unlikely(*ppos)) | |
2434 | return -ESPIPE; | |
2435 | ||
2436 | if (sock->file->f_flags & O_NONBLOCK || | |
2437 | flags & SPLICE_F_NONBLOCK) | |
2438 | state.flags = MSG_DONTWAIT; | |
2439 | ||
06a77b07 | 2440 | return unix_stream_read_generic(&state, false); |
2b514574 HFS |
2441 | } |
2442 | ||
1da177e4 LT |
2443 | static int unix_shutdown(struct socket *sock, int mode) |
2444 | { | |
2445 | struct sock *sk = sock->sk; | |
2446 | struct sock *other; | |
2447 | ||
fc61b928 XW |
2448 | if (mode < SHUT_RD || mode > SHUT_RDWR) |
2449 | return -EINVAL; | |
2450 | /* This maps: | |
2451 | * SHUT_RD (0) -> RCV_SHUTDOWN (1) | |
2452 | * SHUT_WR (1) -> SEND_SHUTDOWN (2) | |
2453 | * SHUT_RDWR (2) -> SHUTDOWN_MASK (3) | |
2454 | */ | |
2455 | ++mode; | |
7180a031 AC |
2456 | |
2457 | unix_state_lock(sk); | |
2458 | sk->sk_shutdown |= mode; | |
2459 | other = unix_peer(sk); | |
2460 | if (other) | |
2461 | sock_hold(other); | |
2462 | unix_state_unlock(sk); | |
2463 | sk->sk_state_change(sk); | |
2464 | ||
2465 | if (other && | |
2466 | (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) { | |
2467 | ||
2468 | int peer_mode = 0; | |
2469 | ||
2470 | if (mode&RCV_SHUTDOWN) | |
2471 | peer_mode |= SEND_SHUTDOWN; | |
2472 | if (mode&SEND_SHUTDOWN) | |
2473 | peer_mode |= RCV_SHUTDOWN; | |
2474 | unix_state_lock(other); | |
2475 | other->sk_shutdown |= peer_mode; | |
2476 | unix_state_unlock(other); | |
2477 | other->sk_state_change(other); | |
2478 | if (peer_mode == SHUTDOWN_MASK) | |
2479 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP); | |
2480 | else if (peer_mode & RCV_SHUTDOWN) | |
2481 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN); | |
1da177e4 | 2482 | } |
7180a031 AC |
2483 | if (other) |
2484 | sock_put(other); | |
2485 | ||
1da177e4 LT |
2486 | return 0; |
2487 | } | |
2488 | ||
885ee74d PE |
2489 | long unix_inq_len(struct sock *sk) |
2490 | { | |
2491 | struct sk_buff *skb; | |
2492 | long amount = 0; | |
2493 | ||
2494 | if (sk->sk_state == TCP_LISTEN) | |
2495 | return -EINVAL; | |
2496 | ||
2497 | spin_lock(&sk->sk_receive_queue.lock); | |
2498 | if (sk->sk_type == SOCK_STREAM || | |
2499 | sk->sk_type == SOCK_SEQPACKET) { | |
2500 | skb_queue_walk(&sk->sk_receive_queue, skb) | |
e370a723 | 2501 | amount += unix_skb_len(skb); |
885ee74d PE |
2502 | } else { |
2503 | skb = skb_peek(&sk->sk_receive_queue); | |
2504 | if (skb) | |
2505 | amount = skb->len; | |
2506 | } | |
2507 | spin_unlock(&sk->sk_receive_queue.lock); | |
2508 | ||
2509 | return amount; | |
2510 | } | |
2511 | EXPORT_SYMBOL_GPL(unix_inq_len); | |
2512 | ||
2513 | long unix_outq_len(struct sock *sk) | |
2514 | { | |
2515 | return sk_wmem_alloc_get(sk); | |
2516 | } | |
2517 | EXPORT_SYMBOL_GPL(unix_outq_len); | |
2518 | ||
ba94f308 AV |
2519 | static int unix_open_file(struct sock *sk) |
2520 | { | |
2521 | struct path path; | |
2522 | struct file *f; | |
2523 | int fd; | |
2524 | ||
2525 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) | |
2526 | return -EPERM; | |
2527 | ||
ae3b5641 AV |
2528 | if (!smp_load_acquire(&unix_sk(sk)->addr)) |
2529 | return -ENOENT; | |
2530 | ||
ba94f308 | 2531 | path = unix_sk(sk)->path; |
ae3b5641 | 2532 | if (!path.dentry) |
ba94f308 | 2533 | return -ENOENT; |
ba94f308 AV |
2534 | |
2535 | path_get(&path); | |
ba94f308 AV |
2536 | |
2537 | fd = get_unused_fd_flags(O_CLOEXEC); | |
2538 | if (fd < 0) | |
2539 | goto out; | |
2540 | ||
2541 | f = dentry_open(&path, O_PATH, current_cred()); | |
2542 | if (IS_ERR(f)) { | |
2543 | put_unused_fd(fd); | |
2544 | fd = PTR_ERR(f); | |
2545 | goto out; | |
2546 | } | |
2547 | ||
2548 | fd_install(fd, f); | |
2549 | out: | |
2550 | path_put(&path); | |
2551 | ||
2552 | return fd; | |
2553 | } | |
2554 | ||
1da177e4 LT |
2555 | static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
2556 | { | |
2557 | struct sock *sk = sock->sk; | |
e27dfcea | 2558 | long amount = 0; |
1da177e4 LT |
2559 | int err; |
2560 | ||
6eba6a37 ED |
2561 | switch (cmd) { |
2562 | case SIOCOUTQ: | |
885ee74d | 2563 | amount = unix_outq_len(sk); |
6eba6a37 ED |
2564 | err = put_user(amount, (int __user *)arg); |
2565 | break; | |
2566 | case SIOCINQ: | |
885ee74d PE |
2567 | amount = unix_inq_len(sk); |
2568 | if (amount < 0) | |
2569 | err = amount; | |
2570 | else | |
1da177e4 | 2571 | err = put_user(amount, (int __user *)arg); |
885ee74d | 2572 | break; |
ba94f308 AV |
2573 | case SIOCUNIXFILE: |
2574 | err = unix_open_file(sk); | |
2575 | break; | |
6eba6a37 ED |
2576 | default: |
2577 | err = -ENOIOCTLCMD; | |
2578 | break; | |
1da177e4 LT |
2579 | } |
2580 | return err; | |
2581 | } | |
2582 | ||
a11e1d43 | 2583 | static __poll_t unix_poll(struct file *file, struct socket *sock, poll_table *wait) |
1da177e4 LT |
2584 | { |
2585 | struct sock *sk = sock->sk; | |
a11e1d43 LT |
2586 | __poll_t mask; |
2587 | ||
89ab066d | 2588 | sock_poll_wait(file, sock, wait); |
a11e1d43 | 2589 | mask = 0; |
1da177e4 LT |
2590 | |
2591 | /* exceptional events? */ | |
2592 | if (sk->sk_err) | |
a9a08845 | 2593 | mask |= EPOLLERR; |
1da177e4 | 2594 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
a9a08845 | 2595 | mask |= EPOLLHUP; |
f348d70a | 2596 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
a9a08845 | 2597 | mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; |
1da177e4 LT |
2598 | |
2599 | /* readable? */ | |
3ef7cf57 | 2600 | if (!skb_queue_empty_lockless(&sk->sk_receive_queue)) |
a9a08845 | 2601 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 LT |
2602 | |
2603 | /* Connection-based need to check for termination and startup */ | |
6eba6a37 ED |
2604 | if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) && |
2605 | sk->sk_state == TCP_CLOSE) | |
a9a08845 | 2606 | mask |= EPOLLHUP; |
1da177e4 LT |
2607 | |
2608 | /* | |
2609 | * we set writable also when the other side has shut down the | |
2610 | * connection. This prevents stuck sockets. | |
2611 | */ | |
2612 | if (unix_writable(sk)) | |
a9a08845 | 2613 | mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; |
1da177e4 LT |
2614 | |
2615 | return mask; | |
2616 | } | |
2617 | ||
a11e1d43 LT |
2618 | static __poll_t unix_dgram_poll(struct file *file, struct socket *sock, |
2619 | poll_table *wait) | |
3c73419c | 2620 | { |
ec0d215f | 2621 | struct sock *sk = sock->sk, *other; |
a11e1d43 LT |
2622 | unsigned int writable; |
2623 | __poll_t mask; | |
2624 | ||
89ab066d | 2625 | sock_poll_wait(file, sock, wait); |
a11e1d43 | 2626 | mask = 0; |
3c73419c RW |
2627 | |
2628 | /* exceptional events? */ | |
3ef7cf57 | 2629 | if (sk->sk_err || !skb_queue_empty_lockless(&sk->sk_error_queue)) |
a9a08845 LT |
2630 | mask |= EPOLLERR | |
2631 | (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0); | |
7d4c04fc | 2632 | |
3c73419c | 2633 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
a9a08845 | 2634 | mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; |
3c73419c | 2635 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
a9a08845 | 2636 | mask |= EPOLLHUP; |
3c73419c RW |
2637 | |
2638 | /* readable? */ | |
3ef7cf57 | 2639 | if (!skb_queue_empty_lockless(&sk->sk_receive_queue)) |
a9a08845 | 2640 | mask |= EPOLLIN | EPOLLRDNORM; |
3c73419c RW |
2641 | |
2642 | /* Connection-based need to check for termination and startup */ | |
2643 | if (sk->sk_type == SOCK_SEQPACKET) { | |
2644 | if (sk->sk_state == TCP_CLOSE) | |
a9a08845 | 2645 | mask |= EPOLLHUP; |
3c73419c RW |
2646 | /* connection hasn't started yet? */ |
2647 | if (sk->sk_state == TCP_SYN_SENT) | |
2648 | return mask; | |
2649 | } | |
2650 | ||
973a34aa | 2651 | /* No write status requested, avoid expensive OUT tests. */ |
a11e1d43 | 2652 | if (!(poll_requested_events(wait) & (EPOLLWRBAND|EPOLLWRNORM|EPOLLOUT))) |
973a34aa ED |
2653 | return mask; |
2654 | ||
ec0d215f | 2655 | writable = unix_writable(sk); |
7d267278 RW |
2656 | if (writable) { |
2657 | unix_state_lock(sk); | |
2658 | ||
2659 | other = unix_peer(sk); | |
2660 | if (other && unix_peer(other) != sk && | |
2661 | unix_recvq_full(other) && | |
2662 | unix_dgram_peer_wake_me(sk, other)) | |
2663 | writable = 0; | |
2664 | ||
2665 | unix_state_unlock(sk); | |
ec0d215f RW |
2666 | } |
2667 | ||
2668 | if (writable) | |
a9a08845 | 2669 | mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; |
3c73419c | 2670 | else |
9cd3e072 | 2671 | sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk); |
3c73419c | 2672 | |
3c73419c RW |
2673 | return mask; |
2674 | } | |
1da177e4 LT |
2675 | |
2676 | #ifdef CONFIG_PROC_FS | |
a53eb3fe | 2677 | |
7123aaa3 ED |
2678 | #define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1) |
2679 | ||
2680 | #define get_bucket(x) ((x) >> BUCKET_SPACE) | |
2681 | #define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1)) | |
2682 | #define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o)) | |
a53eb3fe | 2683 | |
7123aaa3 | 2684 | static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos) |
1da177e4 | 2685 | { |
7123aaa3 ED |
2686 | unsigned long offset = get_offset(*pos); |
2687 | unsigned long bucket = get_bucket(*pos); | |
2688 | struct sock *sk; | |
2689 | unsigned long count = 0; | |
1da177e4 | 2690 | |
7123aaa3 ED |
2691 | for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) { |
2692 | if (sock_net(sk) != seq_file_net(seq)) | |
097e66c5 | 2693 | continue; |
7123aaa3 ED |
2694 | if (++count == offset) |
2695 | break; | |
2696 | } | |
2697 | ||
2698 | return sk; | |
2699 | } | |
2700 | ||
2701 | static struct sock *unix_next_socket(struct seq_file *seq, | |
2702 | struct sock *sk, | |
2703 | loff_t *pos) | |
2704 | { | |
2705 | unsigned long bucket; | |
2706 | ||
2707 | while (sk > (struct sock *)SEQ_START_TOKEN) { | |
2708 | sk = sk_next(sk); | |
2709 | if (!sk) | |
2710 | goto next_bucket; | |
2711 | if (sock_net(sk) == seq_file_net(seq)) | |
2712 | return sk; | |
1da177e4 | 2713 | } |
7123aaa3 ED |
2714 | |
2715 | do { | |
2716 | sk = unix_from_bucket(seq, pos); | |
2717 | if (sk) | |
2718 | return sk; | |
2719 | ||
2720 | next_bucket: | |
2721 | bucket = get_bucket(*pos) + 1; | |
2722 | *pos = set_bucket_offset(bucket, 1); | |
2723 | } while (bucket < ARRAY_SIZE(unix_socket_table)); | |
2724 | ||
1da177e4 LT |
2725 | return NULL; |
2726 | } | |
2727 | ||
1da177e4 | 2728 | static void *unix_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 2729 | __acquires(unix_table_lock) |
1da177e4 | 2730 | { |
fbe9cc4a | 2731 | spin_lock(&unix_table_lock); |
7123aaa3 ED |
2732 | |
2733 | if (!*pos) | |
2734 | return SEQ_START_TOKEN; | |
2735 | ||
2736 | if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table)) | |
2737 | return NULL; | |
2738 | ||
2739 | return unix_next_socket(seq, NULL, pos); | |
1da177e4 LT |
2740 | } |
2741 | ||
2742 | static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2743 | { | |
2744 | ++*pos; | |
7123aaa3 | 2745 | return unix_next_socket(seq, v, pos); |
1da177e4 LT |
2746 | } |
2747 | ||
2748 | static void unix_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2749 | __releases(unix_table_lock) |
1da177e4 | 2750 | { |
fbe9cc4a | 2751 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
2752 | } |
2753 | ||
2754 | static int unix_seq_show(struct seq_file *seq, void *v) | |
2755 | { | |
ac7bfa62 | 2756 | |
b9f3124f | 2757 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2758 | seq_puts(seq, "Num RefCount Protocol Flags Type St " |
2759 | "Inode Path\n"); | |
2760 | else { | |
2761 | struct sock *s = v; | |
2762 | struct unix_sock *u = unix_sk(s); | |
1c92b4e5 | 2763 | unix_state_lock(s); |
1da177e4 | 2764 | |
71338aa7 | 2765 | seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu", |
1da177e4 | 2766 | s, |
41c6d650 | 2767 | refcount_read(&s->sk_refcnt), |
1da177e4 LT |
2768 | 0, |
2769 | s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0, | |
2770 | s->sk_type, | |
2771 | s->sk_socket ? | |
2772 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) : | |
2773 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING), | |
2774 | sock_i_ino(s)); | |
2775 | ||
ae3b5641 | 2776 | if (u->addr) { // under unix_table_lock here |
1da177e4 LT |
2777 | int i, len; |
2778 | seq_putc(seq, ' '); | |
2779 | ||
2780 | i = 0; | |
2781 | len = u->addr->len - sizeof(short); | |
2782 | if (!UNIX_ABSTRACT(s)) | |
2783 | len--; | |
2784 | else { | |
2785 | seq_putc(seq, '@'); | |
2786 | i++; | |
2787 | } | |
2788 | for ( ; i < len; i++) | |
e7947ea7 IB |
2789 | seq_putc(seq, u->addr->name->sun_path[i] ?: |
2790 | '@'); | |
1da177e4 | 2791 | } |
1c92b4e5 | 2792 | unix_state_unlock(s); |
1da177e4 LT |
2793 | seq_putc(seq, '\n'); |
2794 | } | |
2795 | ||
2796 | return 0; | |
2797 | } | |
2798 | ||
56b3d975 | 2799 | static const struct seq_operations unix_seq_ops = { |
1da177e4 LT |
2800 | .start = unix_seq_start, |
2801 | .next = unix_seq_next, | |
2802 | .stop = unix_seq_stop, | |
2803 | .show = unix_seq_show, | |
2804 | }; | |
1da177e4 LT |
2805 | #endif |
2806 | ||
ec1b4cf7 | 2807 | static const struct net_proto_family unix_family_ops = { |
1da177e4 LT |
2808 | .family = PF_UNIX, |
2809 | .create = unix_create, | |
2810 | .owner = THIS_MODULE, | |
2811 | }; | |
2812 | ||
097e66c5 | 2813 | |
2c8c1e72 | 2814 | static int __net_init unix_net_init(struct net *net) |
097e66c5 DL |
2815 | { |
2816 | int error = -ENOMEM; | |
2817 | ||
a0a53c8b | 2818 | net->unx.sysctl_max_dgram_qlen = 10; |
1597fbc0 PE |
2819 | if (unix_sysctl_register(net)) |
2820 | goto out; | |
d392e497 | 2821 | |
097e66c5 | 2822 | #ifdef CONFIG_PROC_FS |
c3506372 CH |
2823 | if (!proc_create_net("unix", 0, net->proc_net, &unix_seq_ops, |
2824 | sizeof(struct seq_net_private))) { | |
1597fbc0 | 2825 | unix_sysctl_unregister(net); |
097e66c5 | 2826 | goto out; |
1597fbc0 | 2827 | } |
097e66c5 DL |
2828 | #endif |
2829 | error = 0; | |
2830 | out: | |
48dcc33e | 2831 | return error; |
097e66c5 DL |
2832 | } |
2833 | ||
2c8c1e72 | 2834 | static void __net_exit unix_net_exit(struct net *net) |
097e66c5 | 2835 | { |
1597fbc0 | 2836 | unix_sysctl_unregister(net); |
ece31ffd | 2837 | remove_proc_entry("unix", net->proc_net); |
097e66c5 DL |
2838 | } |
2839 | ||
2840 | static struct pernet_operations unix_net_ops = { | |
2841 | .init = unix_net_init, | |
2842 | .exit = unix_net_exit, | |
2843 | }; | |
2844 | ||
1da177e4 LT |
2845 | static int __init af_unix_init(void) |
2846 | { | |
2847 | int rc = -1; | |
1da177e4 | 2848 | |
b4fff5f8 | 2849 | BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 LT |
2850 | |
2851 | rc = proto_register(&unix_proto, 1); | |
ac7bfa62 | 2852 | if (rc != 0) { |
5cc208be | 2853 | pr_crit("%s: Cannot create unix_sock SLAB cache!\n", __func__); |
1da177e4 LT |
2854 | goto out; |
2855 | } | |
2856 | ||
2857 | sock_register(&unix_family_ops); | |
097e66c5 | 2858 | register_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2859 | out: |
2860 | return rc; | |
2861 | } | |
2862 | ||
2863 | static void __exit af_unix_exit(void) | |
2864 | { | |
2865 | sock_unregister(PF_UNIX); | |
1da177e4 | 2866 | proto_unregister(&unix_proto); |
097e66c5 | 2867 | unregister_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2868 | } |
2869 | ||
3d366960 DW |
2870 | /* Earlier than device_initcall() so that other drivers invoking |
2871 | request_module() don't end up in a loop when modprobe tries | |
2872 | to use a UNIX socket. But later than subsys_initcall() because | |
2873 | we depend on stuff initialised there */ | |
2874 | fs_initcall(af_unix_init); | |
1da177e4 LT |
2875 | module_exit(af_unix_exit); |
2876 | ||
2877 | MODULE_LICENSE("GPL"); | |
2878 | MODULE_ALIAS_NETPROTO(PF_UNIX); |