]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NET4: Implementation of BSD Unix domain sockets. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
5 | * |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License | |
8 | * as published by the Free Software Foundation; either version | |
9 | * 2 of the License, or (at your option) any later version. | |
10 | * | |
1da177e4 LT |
11 | * Fixes: |
12 | * Linus Torvalds : Assorted bug cures. | |
13 | * Niibe Yutaka : async I/O support. | |
14 | * Carsten Paeth : PF_UNIX check, address fixes. | |
15 | * Alan Cox : Limit size of allocated blocks. | |
16 | * Alan Cox : Fixed the stupid socketpair bug. | |
17 | * Alan Cox : BSD compatibility fine tuning. | |
18 | * Alan Cox : Fixed a bug in connect when interrupted. | |
19 | * Alan Cox : Sorted out a proper draft version of | |
20 | * file descriptor passing hacked up from | |
21 | * Mike Shaver's work. | |
22 | * Marty Leisner : Fixes to fd passing | |
23 | * Nick Nevin : recvmsg bugfix. | |
24 | * Alan Cox : Started proper garbage collector | |
25 | * Heiko EiBfeldt : Missing verify_area check | |
26 | * Alan Cox : Started POSIXisms | |
27 | * Andreas Schwab : Replace inode by dentry for proper | |
28 | * reference counting | |
29 | * Kirk Petersen : Made this a module | |
30 | * Christoph Rohland : Elegant non-blocking accept/connect algorithm. | |
31 | * Lots of bug fixes. | |
32 | * Alexey Kuznetosv : Repaired (I hope) bugs introduces | |
33 | * by above two patches. | |
34 | * Andrea Arcangeli : If possible we block in connect(2) | |
35 | * if the max backlog of the listen socket | |
36 | * is been reached. This won't break | |
37 | * old apps and it will avoid huge amount | |
38 | * of socks hashed (this for unix_gc() | |
39 | * performances reasons). | |
40 | * Security fix that limits the max | |
41 | * number of socks to 2*max_files and | |
42 | * the number of skb queueable in the | |
43 | * dgram receiver. | |
44 | * Artur Skawina : Hash function optimizations | |
45 | * Alexey Kuznetsov : Full scale SMP. Lot of bugs are introduced 8) | |
46 | * Malcolm Beattie : Set peercred for socketpair | |
47 | * Michal Ostrowski : Module initialization cleanup. | |
48 | * Arnaldo C. Melo : Remove MOD_{INC,DEC}_USE_COUNT, | |
49 | * the core infrastructure is doing that | |
50 | * for all net proto families now (2.5.69+) | |
51 | * | |
52 | * | |
53 | * Known differences from reference BSD that was tested: | |
54 | * | |
55 | * [TO FIX] | |
56 | * ECONNREFUSED is not returned from one end of a connected() socket to the | |
57 | * other the moment one end closes. | |
58 | * fstat() doesn't return st_dev=0, and give the blksize as high water mark | |
59 | * and a fake inode identifier (nor the BSD first socket fstat twice bug). | |
60 | * [NOT TO FIX] | |
61 | * accept() returns a path name even if the connecting socket has closed | |
62 | * in the meantime (BSD loses the path and gives up). | |
63 | * accept() returns 0 length path for an unbound connector. BSD returns 16 | |
64 | * and a null first byte in the path (but not for gethost/peername - BSD bug ??) | |
65 | * socketpair(...SOCK_RAW..) doesn't panic the kernel. | |
66 | * BSD af_unix apparently has connect forgetting to block properly. | |
67 | * (need to check this with the POSIX spec in detail) | |
68 | * | |
69 | * Differences from 2.0.0-11-... (ANK) | |
70 | * Bug fixes and improvements. | |
71 | * - client shutdown killed server socket. | |
72 | * - removed all useless cli/sti pairs. | |
73 | * | |
74 | * Semantic changes/extensions. | |
75 | * - generic control message passing. | |
76 | * - SCM_CREDENTIALS control message. | |
77 | * - "Abstract" (not FS based) socket bindings. | |
78 | * Abstract names are sequences of bytes (not zero terminated) | |
79 | * started by 0, so that this name space does not intersect | |
80 | * with BSD names. | |
81 | */ | |
82 | ||
5cc208be | 83 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
84 | ||
1da177e4 | 85 | #include <linux/module.h> |
1da177e4 | 86 | #include <linux/kernel.h> |
1da177e4 | 87 | #include <linux/signal.h> |
3f07c014 | 88 | #include <linux/sched/signal.h> |
1da177e4 LT |
89 | #include <linux/errno.h> |
90 | #include <linux/string.h> | |
91 | #include <linux/stat.h> | |
92 | #include <linux/dcache.h> | |
93 | #include <linux/namei.h> | |
94 | #include <linux/socket.h> | |
95 | #include <linux/un.h> | |
96 | #include <linux/fcntl.h> | |
97 | #include <linux/termios.h> | |
98 | #include <linux/sockios.h> | |
99 | #include <linux/net.h> | |
100 | #include <linux/in.h> | |
101 | #include <linux/fs.h> | |
102 | #include <linux/slab.h> | |
7c0f6ba6 | 103 | #include <linux/uaccess.h> |
1da177e4 LT |
104 | #include <linux/skbuff.h> |
105 | #include <linux/netdevice.h> | |
457c4cbc | 106 | #include <net/net_namespace.h> |
1da177e4 | 107 | #include <net/sock.h> |
c752f073 | 108 | #include <net/tcp_states.h> |
1da177e4 LT |
109 | #include <net/af_unix.h> |
110 | #include <linux/proc_fs.h> | |
111 | #include <linux/seq_file.h> | |
112 | #include <net/scm.h> | |
113 | #include <linux/init.h> | |
114 | #include <linux/poll.h> | |
1da177e4 LT |
115 | #include <linux/rtnetlink.h> |
116 | #include <linux/mount.h> | |
117 | #include <net/checksum.h> | |
118 | #include <linux/security.h> | |
2b15af6f | 119 | #include <linux/freezer.h> |
ba94f308 | 120 | #include <linux/file.h> |
1da177e4 | 121 | |
7123aaa3 | 122 | struct hlist_head unix_socket_table[2 * UNIX_HASH_SIZE]; |
fa7ff56f PE |
123 | EXPORT_SYMBOL_GPL(unix_socket_table); |
124 | DEFINE_SPINLOCK(unix_table_lock); | |
125 | EXPORT_SYMBOL_GPL(unix_table_lock); | |
518de9b3 | 126 | static atomic_long_t unix_nr_socks; |
1da177e4 | 127 | |
1da177e4 | 128 | |
7123aaa3 ED |
129 | static struct hlist_head *unix_sockets_unbound(void *addr) |
130 | { | |
131 | unsigned long hash = (unsigned long)addr; | |
132 | ||
133 | hash ^= hash >> 16; | |
134 | hash ^= hash >> 8; | |
135 | hash %= UNIX_HASH_SIZE; | |
136 | return &unix_socket_table[UNIX_HASH_SIZE + hash]; | |
137 | } | |
138 | ||
139 | #define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash < UNIX_HASH_SIZE) | |
1da177e4 | 140 | |
877ce7c1 | 141 | #ifdef CONFIG_SECURITY_NETWORK |
dc49c1f9 | 142 | static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 | 143 | { |
37a9a8df | 144 | UNIXCB(skb).secid = scm->secid; |
877ce7c1 CZ |
145 | } |
146 | ||
147 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
148 | { | |
37a9a8df SS |
149 | scm->secid = UNIXCB(skb).secid; |
150 | } | |
151 | ||
152 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
153 | { | |
154 | return (scm->secid == UNIXCB(skb).secid); | |
877ce7c1 CZ |
155 | } |
156 | #else | |
dc49c1f9 | 157 | static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 CZ |
158 | { } |
159 | ||
160 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
161 | { } | |
37a9a8df SS |
162 | |
163 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
164 | { | |
165 | return true; | |
166 | } | |
877ce7c1 CZ |
167 | #endif /* CONFIG_SECURITY_NETWORK */ |
168 | ||
1da177e4 LT |
169 | /* |
170 | * SMP locking strategy: | |
fbe9cc4a | 171 | * hash table is protected with spinlock unix_table_lock |
663717f6 | 172 | * each socket state is protected by separate spin lock. |
1da177e4 LT |
173 | */ |
174 | ||
95c96174 | 175 | static inline unsigned int unix_hash_fold(__wsum n) |
1da177e4 | 176 | { |
0a13404d | 177 | unsigned int hash = (__force unsigned int)csum_fold(n); |
95c96174 | 178 | |
1da177e4 LT |
179 | hash ^= hash>>8; |
180 | return hash&(UNIX_HASH_SIZE-1); | |
181 | } | |
182 | ||
183 | #define unix_peer(sk) (unix_sk(sk)->peer) | |
184 | ||
185 | static inline int unix_our_peer(struct sock *sk, struct sock *osk) | |
186 | { | |
187 | return unix_peer(osk) == sk; | |
188 | } | |
189 | ||
190 | static inline int unix_may_send(struct sock *sk, struct sock *osk) | |
191 | { | |
6eba6a37 | 192 | return unix_peer(osk) == NULL || unix_our_peer(sk, osk); |
1da177e4 LT |
193 | } |
194 | ||
3c73419c RW |
195 | static inline int unix_recvq_full(struct sock const *sk) |
196 | { | |
197 | return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog; | |
198 | } | |
199 | ||
fa7ff56f | 200 | struct sock *unix_peer_get(struct sock *s) |
1da177e4 LT |
201 | { |
202 | struct sock *peer; | |
203 | ||
1c92b4e5 | 204 | unix_state_lock(s); |
1da177e4 LT |
205 | peer = unix_peer(s); |
206 | if (peer) | |
207 | sock_hold(peer); | |
1c92b4e5 | 208 | unix_state_unlock(s); |
1da177e4 LT |
209 | return peer; |
210 | } | |
fa7ff56f | 211 | EXPORT_SYMBOL_GPL(unix_peer_get); |
1da177e4 LT |
212 | |
213 | static inline void unix_release_addr(struct unix_address *addr) | |
214 | { | |
8c9814b9 | 215 | if (refcount_dec_and_test(&addr->refcnt)) |
1da177e4 LT |
216 | kfree(addr); |
217 | } | |
218 | ||
219 | /* | |
220 | * Check unix socket name: | |
221 | * - should be not zero length. | |
222 | * - if started by not zero, should be NULL terminated (FS object) | |
223 | * - if started by zero, it is abstract name. | |
224 | */ | |
ac7bfa62 | 225 | |
95c96174 | 226 | static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp) |
1da177e4 | 227 | { |
33c4368e KK |
228 | *hashp = 0; |
229 | ||
1da177e4 LT |
230 | if (len <= sizeof(short) || len > sizeof(*sunaddr)) |
231 | return -EINVAL; | |
232 | if (!sunaddr || sunaddr->sun_family != AF_UNIX) | |
233 | return -EINVAL; | |
234 | if (sunaddr->sun_path[0]) { | |
235 | /* | |
236 | * This may look like an off by one error but it is a bit more | |
237 | * subtle. 108 is the longest valid AF_UNIX path for a binding. | |
25985edc | 238 | * sun_path[108] doesn't as such exist. However in kernel space |
1da177e4 LT |
239 | * we are guaranteed that it is a valid memory location in our |
240 | * kernel address buffer. | |
241 | */ | |
e27dfcea | 242 | ((char *)sunaddr)[len] = 0; |
1da177e4 LT |
243 | len = strlen(sunaddr->sun_path)+1+sizeof(short); |
244 | return len; | |
245 | } | |
246 | ||
07f0757a | 247 | *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0)); |
1da177e4 LT |
248 | return len; |
249 | } | |
250 | ||
251 | static void __unix_remove_socket(struct sock *sk) | |
252 | { | |
253 | sk_del_node_init(sk); | |
254 | } | |
255 | ||
256 | static void __unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
257 | { | |
547b792c | 258 | WARN_ON(!sk_unhashed(sk)); |
1da177e4 LT |
259 | sk_add_node(sk, list); |
260 | } | |
261 | ||
262 | static inline void unix_remove_socket(struct sock *sk) | |
263 | { | |
fbe9cc4a | 264 | spin_lock(&unix_table_lock); |
1da177e4 | 265 | __unix_remove_socket(sk); |
fbe9cc4a | 266 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
267 | } |
268 | ||
269 | static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
270 | { | |
fbe9cc4a | 271 | spin_lock(&unix_table_lock); |
1da177e4 | 272 | __unix_insert_socket(list, sk); |
fbe9cc4a | 273 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
274 | } |
275 | ||
097e66c5 DL |
276 | static struct sock *__unix_find_socket_byname(struct net *net, |
277 | struct sockaddr_un *sunname, | |
95c96174 | 278 | int len, int type, unsigned int hash) |
1da177e4 LT |
279 | { |
280 | struct sock *s; | |
1da177e4 | 281 | |
b67bfe0d | 282 | sk_for_each(s, &unix_socket_table[hash ^ type]) { |
1da177e4 LT |
283 | struct unix_sock *u = unix_sk(s); |
284 | ||
878628fb | 285 | if (!net_eq(sock_net(s), net)) |
097e66c5 DL |
286 | continue; |
287 | ||
1da177e4 LT |
288 | if (u->addr->len == len && |
289 | !memcmp(u->addr->name, sunname, len)) | |
290 | goto found; | |
291 | } | |
292 | s = NULL; | |
293 | found: | |
294 | return s; | |
295 | } | |
296 | ||
097e66c5 DL |
297 | static inline struct sock *unix_find_socket_byname(struct net *net, |
298 | struct sockaddr_un *sunname, | |
1da177e4 | 299 | int len, int type, |
95c96174 | 300 | unsigned int hash) |
1da177e4 LT |
301 | { |
302 | struct sock *s; | |
303 | ||
fbe9cc4a | 304 | spin_lock(&unix_table_lock); |
097e66c5 | 305 | s = __unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
306 | if (s) |
307 | sock_hold(s); | |
fbe9cc4a | 308 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
309 | return s; |
310 | } | |
311 | ||
6616f788 | 312 | static struct sock *unix_find_socket_byinode(struct inode *i) |
1da177e4 LT |
313 | { |
314 | struct sock *s; | |
1da177e4 | 315 | |
fbe9cc4a | 316 | spin_lock(&unix_table_lock); |
b67bfe0d | 317 | sk_for_each(s, |
1da177e4 | 318 | &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) { |
40ffe67d | 319 | struct dentry *dentry = unix_sk(s)->path.dentry; |
1da177e4 | 320 | |
beef5121 | 321 | if (dentry && d_backing_inode(dentry) == i) { |
1da177e4 LT |
322 | sock_hold(s); |
323 | goto found; | |
324 | } | |
325 | } | |
326 | s = NULL; | |
327 | found: | |
fbe9cc4a | 328 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
329 | return s; |
330 | } | |
331 | ||
7d267278 RW |
332 | /* Support code for asymmetrically connected dgram sockets |
333 | * | |
334 | * If a datagram socket is connected to a socket not itself connected | |
335 | * to the first socket (eg, /dev/log), clients may only enqueue more | |
336 | * messages if the present receive queue of the server socket is not | |
337 | * "too large". This means there's a second writeability condition | |
338 | * poll and sendmsg need to test. The dgram recv code will do a wake | |
339 | * up on the peer_wait wait queue of a socket upon reception of a | |
340 | * datagram which needs to be propagated to sleeping would-be writers | |
341 | * since these might not have sent anything so far. This can't be | |
342 | * accomplished via poll_wait because the lifetime of the server | |
343 | * socket might be less than that of its clients if these break their | |
344 | * association with it or if the server socket is closed while clients | |
345 | * are still connected to it and there's no way to inform "a polling | |
346 | * implementation" that it should let go of a certain wait queue | |
347 | * | |
ac6424b9 | 348 | * In order to propagate a wake up, a wait_queue_entry_t of the client |
7d267278 RW |
349 | * socket is enqueued on the peer_wait queue of the server socket |
350 | * whose wake function does a wake_up on the ordinary client socket | |
351 | * wait queue. This connection is established whenever a write (or | |
352 | * poll for write) hit the flow control condition and broken when the | |
353 | * association to the server socket is dissolved or after a wake up | |
354 | * was relayed. | |
355 | */ | |
356 | ||
ac6424b9 | 357 | static int unix_dgram_peer_wake_relay(wait_queue_entry_t *q, unsigned mode, int flags, |
7d267278 RW |
358 | void *key) |
359 | { | |
360 | struct unix_sock *u; | |
361 | wait_queue_head_t *u_sleep; | |
362 | ||
363 | u = container_of(q, struct unix_sock, peer_wake); | |
364 | ||
365 | __remove_wait_queue(&unix_sk(u->peer_wake.private)->peer_wait, | |
366 | q); | |
367 | u->peer_wake.private = NULL; | |
368 | ||
369 | /* relaying can only happen while the wq still exists */ | |
370 | u_sleep = sk_sleep(&u->sk); | |
371 | if (u_sleep) | |
3ad6f93e | 372 | wake_up_interruptible_poll(u_sleep, key_to_poll(key)); |
7d267278 RW |
373 | |
374 | return 0; | |
375 | } | |
376 | ||
377 | static int unix_dgram_peer_wake_connect(struct sock *sk, struct sock *other) | |
378 | { | |
379 | struct unix_sock *u, *u_other; | |
380 | int rc; | |
381 | ||
382 | u = unix_sk(sk); | |
383 | u_other = unix_sk(other); | |
384 | rc = 0; | |
385 | spin_lock(&u_other->peer_wait.lock); | |
386 | ||
387 | if (!u->peer_wake.private) { | |
388 | u->peer_wake.private = other; | |
389 | __add_wait_queue(&u_other->peer_wait, &u->peer_wake); | |
390 | ||
391 | rc = 1; | |
392 | } | |
393 | ||
394 | spin_unlock(&u_other->peer_wait.lock); | |
395 | return rc; | |
396 | } | |
397 | ||
398 | static void unix_dgram_peer_wake_disconnect(struct sock *sk, | |
399 | struct sock *other) | |
400 | { | |
401 | struct unix_sock *u, *u_other; | |
402 | ||
403 | u = unix_sk(sk); | |
404 | u_other = unix_sk(other); | |
405 | spin_lock(&u_other->peer_wait.lock); | |
406 | ||
407 | if (u->peer_wake.private == other) { | |
408 | __remove_wait_queue(&u_other->peer_wait, &u->peer_wake); | |
409 | u->peer_wake.private = NULL; | |
410 | } | |
411 | ||
412 | spin_unlock(&u_other->peer_wait.lock); | |
413 | } | |
414 | ||
415 | static void unix_dgram_peer_wake_disconnect_wakeup(struct sock *sk, | |
416 | struct sock *other) | |
417 | { | |
418 | unix_dgram_peer_wake_disconnect(sk, other); | |
419 | wake_up_interruptible_poll(sk_sleep(sk), | |
a9a08845 LT |
420 | EPOLLOUT | |
421 | EPOLLWRNORM | | |
422 | EPOLLWRBAND); | |
7d267278 RW |
423 | } |
424 | ||
425 | /* preconditions: | |
426 | * - unix_peer(sk) == other | |
427 | * - association is stable | |
428 | */ | |
429 | static int unix_dgram_peer_wake_me(struct sock *sk, struct sock *other) | |
430 | { | |
431 | int connected; | |
432 | ||
433 | connected = unix_dgram_peer_wake_connect(sk, other); | |
434 | ||
51f7e951 JB |
435 | /* If other is SOCK_DEAD, we want to make sure we signal |
436 | * POLLOUT, such that a subsequent write() can get a | |
437 | * -ECONNREFUSED. Otherwise, if we haven't queued any skbs | |
438 | * to other and its full, we will hang waiting for POLLOUT. | |
439 | */ | |
440 | if (unix_recvq_full(other) && !sock_flag(other, SOCK_DEAD)) | |
7d267278 RW |
441 | return 1; |
442 | ||
443 | if (connected) | |
444 | unix_dgram_peer_wake_disconnect(sk, other); | |
445 | ||
446 | return 0; | |
447 | } | |
448 | ||
1586a587 | 449 | static int unix_writable(const struct sock *sk) |
1da177e4 | 450 | { |
1586a587 | 451 | return sk->sk_state != TCP_LISTEN && |
14afee4b | 452 | (refcount_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf; |
1da177e4 LT |
453 | } |
454 | ||
455 | static void unix_write_space(struct sock *sk) | |
456 | { | |
43815482 ED |
457 | struct socket_wq *wq; |
458 | ||
459 | rcu_read_lock(); | |
1da177e4 | 460 | if (unix_writable(sk)) { |
43815482 | 461 | wq = rcu_dereference(sk->sk_wq); |
1ce0bf50 | 462 | if (skwq_has_sleeper(wq)) |
67426b75 | 463 | wake_up_interruptible_sync_poll(&wq->wait, |
a9a08845 | 464 | EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND); |
8d8ad9d7 | 465 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
1da177e4 | 466 | } |
43815482 | 467 | rcu_read_unlock(); |
1da177e4 LT |
468 | } |
469 | ||
470 | /* When dgram socket disconnects (or changes its peer), we clear its receive | |
471 | * queue of packets arrived from previous peer. First, it allows to do | |
472 | * flow control based only on wmem_alloc; second, sk connected to peer | |
473 | * may receive messages only from that peer. */ | |
474 | static void unix_dgram_disconnected(struct sock *sk, struct sock *other) | |
475 | { | |
b03efcfb | 476 | if (!skb_queue_empty(&sk->sk_receive_queue)) { |
1da177e4 LT |
477 | skb_queue_purge(&sk->sk_receive_queue); |
478 | wake_up_interruptible_all(&unix_sk(sk)->peer_wait); | |
479 | ||
480 | /* If one link of bidirectional dgram pipe is disconnected, | |
481 | * we signal error. Messages are lost. Do not make this, | |
482 | * when peer was not connected to us. | |
483 | */ | |
484 | if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) { | |
485 | other->sk_err = ECONNRESET; | |
486 | other->sk_error_report(other); | |
487 | } | |
488 | } | |
489 | } | |
490 | ||
491 | static void unix_sock_destructor(struct sock *sk) | |
492 | { | |
493 | struct unix_sock *u = unix_sk(sk); | |
494 | ||
495 | skb_queue_purge(&sk->sk_receive_queue); | |
496 | ||
14afee4b | 497 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
547b792c IJ |
498 | WARN_ON(!sk_unhashed(sk)); |
499 | WARN_ON(sk->sk_socket); | |
1da177e4 | 500 | if (!sock_flag(sk, SOCK_DEAD)) { |
5cc208be | 501 | pr_info("Attempt to release alive unix socket: %p\n", sk); |
1da177e4 LT |
502 | return; |
503 | } | |
504 | ||
505 | if (u->addr) | |
506 | unix_release_addr(u->addr); | |
507 | ||
518de9b3 | 508 | atomic_long_dec(&unix_nr_socks); |
6f756a8c | 509 | local_bh_disable(); |
a8076d8d | 510 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
6f756a8c | 511 | local_bh_enable(); |
1da177e4 | 512 | #ifdef UNIX_REFCNT_DEBUG |
5cc208be | 513 | pr_debug("UNIX %p is destroyed, %ld are still alive.\n", sk, |
518de9b3 | 514 | atomic_long_read(&unix_nr_socks)); |
1da177e4 LT |
515 | #endif |
516 | } | |
517 | ||
ded34e0f | 518 | static void unix_release_sock(struct sock *sk, int embrion) |
1da177e4 LT |
519 | { |
520 | struct unix_sock *u = unix_sk(sk); | |
40ffe67d | 521 | struct path path; |
1da177e4 LT |
522 | struct sock *skpair; |
523 | struct sk_buff *skb; | |
524 | int state; | |
525 | ||
526 | unix_remove_socket(sk); | |
527 | ||
528 | /* Clear state */ | |
1c92b4e5 | 529 | unix_state_lock(sk); |
1da177e4 LT |
530 | sock_orphan(sk); |
531 | sk->sk_shutdown = SHUTDOWN_MASK; | |
40ffe67d AV |
532 | path = u->path; |
533 | u->path.dentry = NULL; | |
534 | u->path.mnt = NULL; | |
1da177e4 LT |
535 | state = sk->sk_state; |
536 | sk->sk_state = TCP_CLOSE; | |
1c92b4e5 | 537 | unix_state_unlock(sk); |
1da177e4 LT |
538 | |
539 | wake_up_interruptible_all(&u->peer_wait); | |
540 | ||
e27dfcea | 541 | skpair = unix_peer(sk); |
1da177e4 | 542 | |
e27dfcea | 543 | if (skpair != NULL) { |
1da177e4 | 544 | if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) { |
1c92b4e5 | 545 | unix_state_lock(skpair); |
1da177e4 LT |
546 | /* No more writes */ |
547 | skpair->sk_shutdown = SHUTDOWN_MASK; | |
548 | if (!skb_queue_empty(&sk->sk_receive_queue) || embrion) | |
549 | skpair->sk_err = ECONNRESET; | |
1c92b4e5 | 550 | unix_state_unlock(skpair); |
1da177e4 | 551 | skpair->sk_state_change(skpair); |
8d8ad9d7 | 552 | sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP); |
1da177e4 | 553 | } |
7d267278 RW |
554 | |
555 | unix_dgram_peer_wake_disconnect(sk, skpair); | |
1da177e4 LT |
556 | sock_put(skpair); /* It may now die */ |
557 | unix_peer(sk) = NULL; | |
558 | } | |
559 | ||
560 | /* Try to flush out this socket. Throw out buffers at least */ | |
561 | ||
562 | while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) { | |
e27dfcea | 563 | if (state == TCP_LISTEN) |
1da177e4 LT |
564 | unix_release_sock(skb->sk, 1); |
565 | /* passed fds are erased in the kfree_skb hook */ | |
73ed5d25 | 566 | UNIXCB(skb).consumed = skb->len; |
1da177e4 LT |
567 | kfree_skb(skb); |
568 | } | |
569 | ||
40ffe67d AV |
570 | if (path.dentry) |
571 | path_put(&path); | |
1da177e4 LT |
572 | |
573 | sock_put(sk); | |
574 | ||
575 | /* ---- Socket is dead now and most probably destroyed ---- */ | |
576 | ||
577 | /* | |
e04dae84 | 578 | * Fixme: BSD difference: In BSD all sockets connected to us get |
1da177e4 LT |
579 | * ECONNRESET and we die on the spot. In Linux we behave |
580 | * like files and pipes do and wait for the last | |
581 | * dereference. | |
582 | * | |
583 | * Can't we simply set sock->err? | |
584 | * | |
585 | * What the above comment does talk about? --ANK(980817) | |
586 | */ | |
587 | ||
9305cfa4 | 588 | if (unix_tot_inflight) |
ac7bfa62 | 589 | unix_gc(); /* Garbage collect fds */ |
1da177e4 LT |
590 | } |
591 | ||
109f6e39 EB |
592 | static void init_peercred(struct sock *sk) |
593 | { | |
594 | put_pid(sk->sk_peer_pid); | |
595 | if (sk->sk_peer_cred) | |
596 | put_cred(sk->sk_peer_cred); | |
597 | sk->sk_peer_pid = get_pid(task_tgid(current)); | |
598 | sk->sk_peer_cred = get_current_cred(); | |
599 | } | |
600 | ||
601 | static void copy_peercred(struct sock *sk, struct sock *peersk) | |
602 | { | |
603 | put_pid(sk->sk_peer_pid); | |
604 | if (sk->sk_peer_cred) | |
605 | put_cred(sk->sk_peer_cred); | |
606 | sk->sk_peer_pid = get_pid(peersk->sk_peer_pid); | |
607 | sk->sk_peer_cred = get_cred(peersk->sk_peer_cred); | |
608 | } | |
609 | ||
1da177e4 LT |
610 | static int unix_listen(struct socket *sock, int backlog) |
611 | { | |
612 | int err; | |
613 | struct sock *sk = sock->sk; | |
614 | struct unix_sock *u = unix_sk(sk); | |
109f6e39 | 615 | struct pid *old_pid = NULL; |
1da177e4 LT |
616 | |
617 | err = -EOPNOTSUPP; | |
6eba6a37 ED |
618 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
619 | goto out; /* Only stream/seqpacket sockets accept */ | |
1da177e4 LT |
620 | err = -EINVAL; |
621 | if (!u->addr) | |
6eba6a37 | 622 | goto out; /* No listens on an unbound socket */ |
1c92b4e5 | 623 | unix_state_lock(sk); |
1da177e4 LT |
624 | if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN) |
625 | goto out_unlock; | |
626 | if (backlog > sk->sk_max_ack_backlog) | |
627 | wake_up_interruptible_all(&u->peer_wait); | |
628 | sk->sk_max_ack_backlog = backlog; | |
629 | sk->sk_state = TCP_LISTEN; | |
630 | /* set credentials so connect can copy them */ | |
109f6e39 | 631 | init_peercred(sk); |
1da177e4 LT |
632 | err = 0; |
633 | ||
634 | out_unlock: | |
1c92b4e5 | 635 | unix_state_unlock(sk); |
109f6e39 | 636 | put_pid(old_pid); |
1da177e4 LT |
637 | out: |
638 | return err; | |
639 | } | |
640 | ||
641 | static int unix_release(struct socket *); | |
642 | static int unix_bind(struct socket *, struct sockaddr *, int); | |
643 | static int unix_stream_connect(struct socket *, struct sockaddr *, | |
644 | int addr_len, int flags); | |
645 | static int unix_socketpair(struct socket *, struct socket *); | |
cdfbabfb | 646 | static int unix_accept(struct socket *, struct socket *, int, bool); |
9b2c45d4 | 647 | static int unix_getname(struct socket *, struct sockaddr *, int); |
a11e1d43 LT |
648 | static __poll_t unix_poll(struct file *, struct socket *, poll_table *); |
649 | static __poll_t unix_dgram_poll(struct file *, struct socket *, | |
650 | poll_table *); | |
1da177e4 LT |
651 | static int unix_ioctl(struct socket *, unsigned int, unsigned long); |
652 | static int unix_shutdown(struct socket *, int); | |
1b784140 YX |
653 | static int unix_stream_sendmsg(struct socket *, struct msghdr *, size_t); |
654 | static int unix_stream_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
869e7c62 HFS |
655 | static ssize_t unix_stream_sendpage(struct socket *, struct page *, int offset, |
656 | size_t size, int flags); | |
2b514574 HFS |
657 | static ssize_t unix_stream_splice_read(struct socket *, loff_t *ppos, |
658 | struct pipe_inode_info *, size_t size, | |
659 | unsigned int flags); | |
1b784140 YX |
660 | static int unix_dgram_sendmsg(struct socket *, struct msghdr *, size_t); |
661 | static int unix_dgram_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
1da177e4 LT |
662 | static int unix_dgram_connect(struct socket *, struct sockaddr *, |
663 | int, int); | |
1b784140 YX |
664 | static int unix_seqpacket_sendmsg(struct socket *, struct msghdr *, size_t); |
665 | static int unix_seqpacket_recvmsg(struct socket *, struct msghdr *, size_t, | |
666 | int); | |
1da177e4 | 667 | |
12663bfc | 668 | static int unix_set_peek_off(struct sock *sk, int val) |
f55bb7f9 PE |
669 | { |
670 | struct unix_sock *u = unix_sk(sk); | |
671 | ||
6e1ce3c3 | 672 | if (mutex_lock_interruptible(&u->iolock)) |
12663bfc SL |
673 | return -EINTR; |
674 | ||
f55bb7f9 | 675 | sk->sk_peek_off = val; |
6e1ce3c3 | 676 | mutex_unlock(&u->iolock); |
12663bfc SL |
677 | |
678 | return 0; | |
f55bb7f9 PE |
679 | } |
680 | ||
681 | ||
90ddc4f0 | 682 | static const struct proto_ops unix_stream_ops = { |
1da177e4 LT |
683 | .family = PF_UNIX, |
684 | .owner = THIS_MODULE, | |
685 | .release = unix_release, | |
686 | .bind = unix_bind, | |
687 | .connect = unix_stream_connect, | |
688 | .socketpair = unix_socketpair, | |
689 | .accept = unix_accept, | |
690 | .getname = unix_getname, | |
a11e1d43 | 691 | .poll = unix_poll, |
1da177e4 LT |
692 | .ioctl = unix_ioctl, |
693 | .listen = unix_listen, | |
694 | .shutdown = unix_shutdown, | |
695 | .setsockopt = sock_no_setsockopt, | |
696 | .getsockopt = sock_no_getsockopt, | |
697 | .sendmsg = unix_stream_sendmsg, | |
698 | .recvmsg = unix_stream_recvmsg, | |
699 | .mmap = sock_no_mmap, | |
869e7c62 | 700 | .sendpage = unix_stream_sendpage, |
2b514574 | 701 | .splice_read = unix_stream_splice_read, |
fc0d7536 | 702 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
703 | }; |
704 | ||
90ddc4f0 | 705 | static const struct proto_ops unix_dgram_ops = { |
1da177e4 LT |
706 | .family = PF_UNIX, |
707 | .owner = THIS_MODULE, | |
708 | .release = unix_release, | |
709 | .bind = unix_bind, | |
710 | .connect = unix_dgram_connect, | |
711 | .socketpair = unix_socketpair, | |
712 | .accept = sock_no_accept, | |
713 | .getname = unix_getname, | |
a11e1d43 | 714 | .poll = unix_dgram_poll, |
1da177e4 LT |
715 | .ioctl = unix_ioctl, |
716 | .listen = sock_no_listen, | |
717 | .shutdown = unix_shutdown, | |
718 | .setsockopt = sock_no_setsockopt, | |
719 | .getsockopt = sock_no_getsockopt, | |
720 | .sendmsg = unix_dgram_sendmsg, | |
721 | .recvmsg = unix_dgram_recvmsg, | |
722 | .mmap = sock_no_mmap, | |
723 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 724 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
725 | }; |
726 | ||
90ddc4f0 | 727 | static const struct proto_ops unix_seqpacket_ops = { |
1da177e4 LT |
728 | .family = PF_UNIX, |
729 | .owner = THIS_MODULE, | |
730 | .release = unix_release, | |
731 | .bind = unix_bind, | |
732 | .connect = unix_stream_connect, | |
733 | .socketpair = unix_socketpair, | |
734 | .accept = unix_accept, | |
735 | .getname = unix_getname, | |
a11e1d43 | 736 | .poll = unix_dgram_poll, |
1da177e4 LT |
737 | .ioctl = unix_ioctl, |
738 | .listen = unix_listen, | |
739 | .shutdown = unix_shutdown, | |
740 | .setsockopt = sock_no_setsockopt, | |
741 | .getsockopt = sock_no_getsockopt, | |
742 | .sendmsg = unix_seqpacket_sendmsg, | |
a05d2ad1 | 743 | .recvmsg = unix_seqpacket_recvmsg, |
1da177e4 LT |
744 | .mmap = sock_no_mmap, |
745 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 746 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
747 | }; |
748 | ||
749 | static struct proto unix_proto = { | |
248969ae ED |
750 | .name = "UNIX", |
751 | .owner = THIS_MODULE, | |
248969ae | 752 | .obj_size = sizeof(struct unix_sock), |
1da177e4 LT |
753 | }; |
754 | ||
11aa9c28 | 755 | static struct sock *unix_create1(struct net *net, struct socket *sock, int kern) |
1da177e4 LT |
756 | { |
757 | struct sock *sk = NULL; | |
758 | struct unix_sock *u; | |
759 | ||
518de9b3 ED |
760 | atomic_long_inc(&unix_nr_socks); |
761 | if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files()) | |
1da177e4 LT |
762 | goto out; |
763 | ||
11aa9c28 | 764 | sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto, kern); |
1da177e4 LT |
765 | if (!sk) |
766 | goto out; | |
767 | ||
6eba6a37 | 768 | sock_init_data(sock, sk); |
1da177e4 | 769 | |
3aa9799e | 770 | sk->sk_allocation = GFP_KERNEL_ACCOUNT; |
1da177e4 | 771 | sk->sk_write_space = unix_write_space; |
a0a53c8b | 772 | sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen; |
1da177e4 LT |
773 | sk->sk_destruct = unix_sock_destructor; |
774 | u = unix_sk(sk); | |
40ffe67d AV |
775 | u->path.dentry = NULL; |
776 | u->path.mnt = NULL; | |
fd19f329 | 777 | spin_lock_init(&u->lock); |
516e0cc5 | 778 | atomic_long_set(&u->inflight, 0); |
1fd05ba5 | 779 | INIT_LIST_HEAD(&u->link); |
6e1ce3c3 LT |
780 | mutex_init(&u->iolock); /* single task reading lock */ |
781 | mutex_init(&u->bindlock); /* single task binding lock */ | |
1da177e4 | 782 | init_waitqueue_head(&u->peer_wait); |
7d267278 | 783 | init_waitqueue_func_entry(&u->peer_wake, unix_dgram_peer_wake_relay); |
7123aaa3 | 784 | unix_insert_socket(unix_sockets_unbound(sk), sk); |
1da177e4 | 785 | out: |
284b327b | 786 | if (sk == NULL) |
518de9b3 | 787 | atomic_long_dec(&unix_nr_socks); |
920de804 ED |
788 | else { |
789 | local_bh_disable(); | |
a8076d8d | 790 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
920de804 ED |
791 | local_bh_enable(); |
792 | } | |
1da177e4 LT |
793 | return sk; |
794 | } | |
795 | ||
3f378b68 EP |
796 | static int unix_create(struct net *net, struct socket *sock, int protocol, |
797 | int kern) | |
1da177e4 LT |
798 | { |
799 | if (protocol && protocol != PF_UNIX) | |
800 | return -EPROTONOSUPPORT; | |
801 | ||
802 | sock->state = SS_UNCONNECTED; | |
803 | ||
804 | switch (sock->type) { | |
805 | case SOCK_STREAM: | |
806 | sock->ops = &unix_stream_ops; | |
807 | break; | |
808 | /* | |
809 | * Believe it or not BSD has AF_UNIX, SOCK_RAW though | |
810 | * nothing uses it. | |
811 | */ | |
812 | case SOCK_RAW: | |
e27dfcea | 813 | sock->type = SOCK_DGRAM; |
110af3ac | 814 | /* fall through */ |
1da177e4 LT |
815 | case SOCK_DGRAM: |
816 | sock->ops = &unix_dgram_ops; | |
817 | break; | |
818 | case SOCK_SEQPACKET: | |
819 | sock->ops = &unix_seqpacket_ops; | |
820 | break; | |
821 | default: | |
822 | return -ESOCKTNOSUPPORT; | |
823 | } | |
824 | ||
11aa9c28 | 825 | return unix_create1(net, sock, kern) ? 0 : -ENOMEM; |
1da177e4 LT |
826 | } |
827 | ||
828 | static int unix_release(struct socket *sock) | |
829 | { | |
830 | struct sock *sk = sock->sk; | |
831 | ||
832 | if (!sk) | |
833 | return 0; | |
834 | ||
ded34e0f | 835 | unix_release_sock(sk, 0); |
1da177e4 LT |
836 | sock->sk = NULL; |
837 | ||
ded34e0f | 838 | return 0; |
1da177e4 LT |
839 | } |
840 | ||
841 | static int unix_autobind(struct socket *sock) | |
842 | { | |
843 | struct sock *sk = sock->sk; | |
3b1e0a65 | 844 | struct net *net = sock_net(sk); |
1da177e4 LT |
845 | struct unix_sock *u = unix_sk(sk); |
846 | static u32 ordernum = 1; | |
6eba6a37 | 847 | struct unix_address *addr; |
1da177e4 | 848 | int err; |
8df73ff9 | 849 | unsigned int retries = 0; |
1da177e4 | 850 | |
6e1ce3c3 | 851 | err = mutex_lock_interruptible(&u->bindlock); |
37ab4fa7 SL |
852 | if (err) |
853 | return err; | |
1da177e4 LT |
854 | |
855 | err = 0; | |
856 | if (u->addr) | |
857 | goto out; | |
858 | ||
859 | err = -ENOMEM; | |
0da974f4 | 860 | addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL); |
1da177e4 LT |
861 | if (!addr) |
862 | goto out; | |
863 | ||
1da177e4 | 864 | addr->name->sun_family = AF_UNIX; |
8c9814b9 | 865 | refcount_set(&addr->refcnt, 1); |
1da177e4 LT |
866 | |
867 | retry: | |
868 | addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short); | |
07f0757a | 869 | addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0)); |
1da177e4 | 870 | |
fbe9cc4a | 871 | spin_lock(&unix_table_lock); |
1da177e4 LT |
872 | ordernum = (ordernum+1)&0xFFFFF; |
873 | ||
097e66c5 | 874 | if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type, |
1da177e4 | 875 | addr->hash)) { |
fbe9cc4a | 876 | spin_unlock(&unix_table_lock); |
8df73ff9 TH |
877 | /* |
878 | * __unix_find_socket_byname() may take long time if many names | |
879 | * are already in use. | |
880 | */ | |
881 | cond_resched(); | |
882 | /* Give up if all names seems to be in use. */ | |
883 | if (retries++ == 0xFFFFF) { | |
884 | err = -ENOSPC; | |
885 | kfree(addr); | |
886 | goto out; | |
887 | } | |
1da177e4 LT |
888 | goto retry; |
889 | } | |
890 | addr->hash ^= sk->sk_type; | |
891 | ||
892 | __unix_remove_socket(sk); | |
893 | u->addr = addr; | |
894 | __unix_insert_socket(&unix_socket_table[addr->hash], sk); | |
fbe9cc4a | 895 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
896 | err = 0; |
897 | ||
6e1ce3c3 | 898 | out: mutex_unlock(&u->bindlock); |
1da177e4 LT |
899 | return err; |
900 | } | |
901 | ||
097e66c5 DL |
902 | static struct sock *unix_find_other(struct net *net, |
903 | struct sockaddr_un *sunname, int len, | |
95c96174 | 904 | int type, unsigned int hash, int *error) |
1da177e4 LT |
905 | { |
906 | struct sock *u; | |
421748ec | 907 | struct path path; |
1da177e4 | 908 | int err = 0; |
ac7bfa62 | 909 | |
1da177e4 | 910 | if (sunname->sun_path[0]) { |
421748ec AV |
911 | struct inode *inode; |
912 | err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path); | |
1da177e4 LT |
913 | if (err) |
914 | goto fail; | |
beef5121 | 915 | inode = d_backing_inode(path.dentry); |
421748ec | 916 | err = inode_permission(inode, MAY_WRITE); |
1da177e4 LT |
917 | if (err) |
918 | goto put_fail; | |
919 | ||
920 | err = -ECONNREFUSED; | |
421748ec | 921 | if (!S_ISSOCK(inode->i_mode)) |
1da177e4 | 922 | goto put_fail; |
6616f788 | 923 | u = unix_find_socket_byinode(inode); |
1da177e4 LT |
924 | if (!u) |
925 | goto put_fail; | |
926 | ||
927 | if (u->sk_type == type) | |
68ac1234 | 928 | touch_atime(&path); |
1da177e4 | 929 | |
421748ec | 930 | path_put(&path); |
1da177e4 | 931 | |
e27dfcea | 932 | err = -EPROTOTYPE; |
1da177e4 LT |
933 | if (u->sk_type != type) { |
934 | sock_put(u); | |
935 | goto fail; | |
936 | } | |
937 | } else { | |
938 | err = -ECONNREFUSED; | |
e27dfcea | 939 | u = unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
940 | if (u) { |
941 | struct dentry *dentry; | |
40ffe67d | 942 | dentry = unix_sk(u)->path.dentry; |
1da177e4 | 943 | if (dentry) |
68ac1234 | 944 | touch_atime(&unix_sk(u)->path); |
1da177e4 LT |
945 | } else |
946 | goto fail; | |
947 | } | |
948 | return u; | |
949 | ||
950 | put_fail: | |
421748ec | 951 | path_put(&path); |
1da177e4 | 952 | fail: |
e27dfcea | 953 | *error = err; |
1da177e4 LT |
954 | return NULL; |
955 | } | |
956 | ||
38f7bd94 | 957 | static int unix_mknod(const char *sun_path, umode_t mode, struct path *res) |
faf02010 | 958 | { |
38f7bd94 LT |
959 | struct dentry *dentry; |
960 | struct path path; | |
961 | int err = 0; | |
962 | /* | |
963 | * Get the parent directory, calculate the hash for last | |
964 | * component. | |
965 | */ | |
966 | dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0); | |
967 | err = PTR_ERR(dentry); | |
968 | if (IS_ERR(dentry)) | |
969 | return err; | |
faf02010 | 970 | |
38f7bd94 LT |
971 | /* |
972 | * All right, let's create it. | |
973 | */ | |
974 | err = security_path_mknod(&path, dentry, mode, 0); | |
faf02010 | 975 | if (!err) { |
38f7bd94 | 976 | err = vfs_mknod(d_inode(path.dentry), dentry, mode, 0); |
faf02010 | 977 | if (!err) { |
38f7bd94 | 978 | res->mnt = mntget(path.mnt); |
faf02010 AV |
979 | res->dentry = dget(dentry); |
980 | } | |
981 | } | |
38f7bd94 | 982 | done_path_create(&path, dentry); |
faf02010 AV |
983 | return err; |
984 | } | |
1da177e4 LT |
985 | |
986 | static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
987 | { | |
988 | struct sock *sk = sock->sk; | |
3b1e0a65 | 989 | struct net *net = sock_net(sk); |
1da177e4 | 990 | struct unix_sock *u = unix_sk(sk); |
e27dfcea | 991 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
dae6ad8f | 992 | char *sun_path = sunaddr->sun_path; |
38f7bd94 | 993 | int err; |
95c96174 | 994 | unsigned int hash; |
1da177e4 LT |
995 | struct unix_address *addr; |
996 | struct hlist_head *list; | |
82fe0d2b | 997 | struct path path = { }; |
1da177e4 LT |
998 | |
999 | err = -EINVAL; | |
defbcf2d MJ |
1000 | if (addr_len < offsetofend(struct sockaddr_un, sun_family) || |
1001 | sunaddr->sun_family != AF_UNIX) | |
1da177e4 LT |
1002 | goto out; |
1003 | ||
e27dfcea | 1004 | if (addr_len == sizeof(short)) { |
1da177e4 LT |
1005 | err = unix_autobind(sock); |
1006 | goto out; | |
1007 | } | |
1008 | ||
1009 | err = unix_mkname(sunaddr, addr_len, &hash); | |
1010 | if (err < 0) | |
1011 | goto out; | |
1012 | addr_len = err; | |
1013 | ||
0fb44559 WC |
1014 | if (sun_path[0]) { |
1015 | umode_t mode = S_IFSOCK | | |
1016 | (SOCK_INODE(sock)->i_mode & ~current_umask()); | |
1017 | err = unix_mknod(sun_path, mode, &path); | |
1018 | if (err) { | |
1019 | if (err == -EEXIST) | |
1020 | err = -EADDRINUSE; | |
1021 | goto out; | |
1022 | } | |
1023 | } | |
1024 | ||
6e1ce3c3 | 1025 | err = mutex_lock_interruptible(&u->bindlock); |
37ab4fa7 | 1026 | if (err) |
0fb44559 | 1027 | goto out_put; |
1da177e4 LT |
1028 | |
1029 | err = -EINVAL; | |
1030 | if (u->addr) | |
1031 | goto out_up; | |
1032 | ||
1033 | err = -ENOMEM; | |
1034 | addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL); | |
1035 | if (!addr) | |
1036 | goto out_up; | |
1037 | ||
1038 | memcpy(addr->name, sunaddr, addr_len); | |
1039 | addr->len = addr_len; | |
1040 | addr->hash = hash ^ sk->sk_type; | |
8c9814b9 | 1041 | refcount_set(&addr->refcnt, 1); |
1da177e4 | 1042 | |
38f7bd94 | 1043 | if (sun_path[0]) { |
1da177e4 | 1044 | addr->hash = UNIX_HASH_SIZE; |
beef5121 | 1045 | hash = d_backing_inode(path.dentry)->i_ino & (UNIX_HASH_SIZE - 1); |
faf02010 | 1046 | spin_lock(&unix_table_lock); |
38f7bd94 | 1047 | u->path = path; |
faf02010 AV |
1048 | list = &unix_socket_table[hash]; |
1049 | } else { | |
1050 | spin_lock(&unix_table_lock); | |
1da177e4 | 1051 | err = -EADDRINUSE; |
097e66c5 | 1052 | if (__unix_find_socket_byname(net, sunaddr, addr_len, |
1da177e4 LT |
1053 | sk->sk_type, hash)) { |
1054 | unix_release_addr(addr); | |
1055 | goto out_unlock; | |
1056 | } | |
1057 | ||
1058 | list = &unix_socket_table[addr->hash]; | |
1da177e4 LT |
1059 | } |
1060 | ||
1061 | err = 0; | |
1062 | __unix_remove_socket(sk); | |
1063 | u->addr = addr; | |
1064 | __unix_insert_socket(list, sk); | |
1065 | ||
1066 | out_unlock: | |
fbe9cc4a | 1067 | spin_unlock(&unix_table_lock); |
1da177e4 | 1068 | out_up: |
6e1ce3c3 | 1069 | mutex_unlock(&u->bindlock); |
0fb44559 WC |
1070 | out_put: |
1071 | if (err) | |
1072 | path_put(&path); | |
1da177e4 LT |
1073 | out: |
1074 | return err; | |
1da177e4 LT |
1075 | } |
1076 | ||
278a3de5 DM |
1077 | static void unix_state_double_lock(struct sock *sk1, struct sock *sk2) |
1078 | { | |
1079 | if (unlikely(sk1 == sk2) || !sk2) { | |
1080 | unix_state_lock(sk1); | |
1081 | return; | |
1082 | } | |
1083 | if (sk1 < sk2) { | |
1084 | unix_state_lock(sk1); | |
1085 | unix_state_lock_nested(sk2); | |
1086 | } else { | |
1087 | unix_state_lock(sk2); | |
1088 | unix_state_lock_nested(sk1); | |
1089 | } | |
1090 | } | |
1091 | ||
1092 | static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2) | |
1093 | { | |
1094 | if (unlikely(sk1 == sk2) || !sk2) { | |
1095 | unix_state_unlock(sk1); | |
1096 | return; | |
1097 | } | |
1098 | unix_state_unlock(sk1); | |
1099 | unix_state_unlock(sk2); | |
1100 | } | |
1101 | ||
1da177e4 LT |
1102 | static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr, |
1103 | int alen, int flags) | |
1104 | { | |
1105 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1106 | struct net *net = sock_net(sk); |
e27dfcea | 1107 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr; |
1da177e4 | 1108 | struct sock *other; |
95c96174 | 1109 | unsigned int hash; |
1da177e4 LT |
1110 | int err; |
1111 | ||
defbcf2d MJ |
1112 | err = -EINVAL; |
1113 | if (alen < offsetofend(struct sockaddr, sa_family)) | |
1114 | goto out; | |
1115 | ||
1da177e4 LT |
1116 | if (addr->sa_family != AF_UNSPEC) { |
1117 | err = unix_mkname(sunaddr, alen, &hash); | |
1118 | if (err < 0) | |
1119 | goto out; | |
1120 | alen = err; | |
1121 | ||
1122 | if (test_bit(SOCK_PASSCRED, &sock->flags) && | |
1123 | !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0) | |
1124 | goto out; | |
1125 | ||
278a3de5 | 1126 | restart: |
e27dfcea | 1127 | other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err); |
1da177e4 LT |
1128 | if (!other) |
1129 | goto out; | |
1130 | ||
278a3de5 DM |
1131 | unix_state_double_lock(sk, other); |
1132 | ||
1133 | /* Apparently VFS overslept socket death. Retry. */ | |
1134 | if (sock_flag(other, SOCK_DEAD)) { | |
1135 | unix_state_double_unlock(sk, other); | |
1136 | sock_put(other); | |
1137 | goto restart; | |
1138 | } | |
1da177e4 LT |
1139 | |
1140 | err = -EPERM; | |
1141 | if (!unix_may_send(sk, other)) | |
1142 | goto out_unlock; | |
1143 | ||
1144 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1145 | if (err) | |
1146 | goto out_unlock; | |
1147 | ||
1148 | } else { | |
1149 | /* | |
1150 | * 1003.1g breaking connected state with AF_UNSPEC | |
1151 | */ | |
1152 | other = NULL; | |
278a3de5 | 1153 | unix_state_double_lock(sk, other); |
1da177e4 LT |
1154 | } |
1155 | ||
1156 | /* | |
1157 | * If it was connected, reconnect. | |
1158 | */ | |
1159 | if (unix_peer(sk)) { | |
1160 | struct sock *old_peer = unix_peer(sk); | |
e27dfcea | 1161 | unix_peer(sk) = other; |
7d267278 RW |
1162 | unix_dgram_peer_wake_disconnect_wakeup(sk, old_peer); |
1163 | ||
278a3de5 | 1164 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1165 | |
1166 | if (other != old_peer) | |
1167 | unix_dgram_disconnected(sk, old_peer); | |
1168 | sock_put(old_peer); | |
1169 | } else { | |
e27dfcea | 1170 | unix_peer(sk) = other; |
278a3de5 | 1171 | unix_state_double_unlock(sk, other); |
1da177e4 | 1172 | } |
ac7bfa62 | 1173 | return 0; |
1da177e4 LT |
1174 | |
1175 | out_unlock: | |
278a3de5 | 1176 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1177 | sock_put(other); |
1178 | out: | |
1179 | return err; | |
1180 | } | |
1181 | ||
1182 | static long unix_wait_for_peer(struct sock *other, long timeo) | |
1183 | { | |
1184 | struct unix_sock *u = unix_sk(other); | |
1185 | int sched; | |
1186 | DEFINE_WAIT(wait); | |
1187 | ||
1188 | prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE); | |
1189 | ||
1190 | sched = !sock_flag(other, SOCK_DEAD) && | |
1191 | !(other->sk_shutdown & RCV_SHUTDOWN) && | |
3c73419c | 1192 | unix_recvq_full(other); |
1da177e4 | 1193 | |
1c92b4e5 | 1194 | unix_state_unlock(other); |
1da177e4 LT |
1195 | |
1196 | if (sched) | |
1197 | timeo = schedule_timeout(timeo); | |
1198 | ||
1199 | finish_wait(&u->peer_wait, &wait); | |
1200 | return timeo; | |
1201 | } | |
1202 | ||
1203 | static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
1204 | int addr_len, int flags) | |
1205 | { | |
e27dfcea | 1206 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
1da177e4 | 1207 | struct sock *sk = sock->sk; |
3b1e0a65 | 1208 | struct net *net = sock_net(sk); |
1da177e4 LT |
1209 | struct unix_sock *u = unix_sk(sk), *newu, *otheru; |
1210 | struct sock *newsk = NULL; | |
1211 | struct sock *other = NULL; | |
1212 | struct sk_buff *skb = NULL; | |
95c96174 | 1213 | unsigned int hash; |
1da177e4 LT |
1214 | int st; |
1215 | int err; | |
1216 | long timeo; | |
1217 | ||
1218 | err = unix_mkname(sunaddr, addr_len, &hash); | |
1219 | if (err < 0) | |
1220 | goto out; | |
1221 | addr_len = err; | |
1222 | ||
f64f9e71 JP |
1223 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr && |
1224 | (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1225 | goto out; |
1226 | ||
1227 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
1228 | ||
1229 | /* First of all allocate resources. | |
1230 | If we will make it after state is locked, | |
1231 | we will have to recheck all again in any case. | |
1232 | */ | |
1233 | ||
1234 | err = -ENOMEM; | |
1235 | ||
1236 | /* create new sock for complete connection */ | |
11aa9c28 | 1237 | newsk = unix_create1(sock_net(sk), NULL, 0); |
1da177e4 LT |
1238 | if (newsk == NULL) |
1239 | goto out; | |
1240 | ||
1241 | /* Allocate skb for sending to listening sock */ | |
1242 | skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL); | |
1243 | if (skb == NULL) | |
1244 | goto out; | |
1245 | ||
1246 | restart: | |
1247 | /* Find listening sock. */ | |
097e66c5 | 1248 | other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err); |
1da177e4 LT |
1249 | if (!other) |
1250 | goto out; | |
1251 | ||
1252 | /* Latch state of peer */ | |
1c92b4e5 | 1253 | unix_state_lock(other); |
1da177e4 LT |
1254 | |
1255 | /* Apparently VFS overslept socket death. Retry. */ | |
1256 | if (sock_flag(other, SOCK_DEAD)) { | |
1c92b4e5 | 1257 | unix_state_unlock(other); |
1da177e4 LT |
1258 | sock_put(other); |
1259 | goto restart; | |
1260 | } | |
1261 | ||
1262 | err = -ECONNREFUSED; | |
1263 | if (other->sk_state != TCP_LISTEN) | |
1264 | goto out_unlock; | |
77238f2b TS |
1265 | if (other->sk_shutdown & RCV_SHUTDOWN) |
1266 | goto out_unlock; | |
1da177e4 | 1267 | |
3c73419c | 1268 | if (unix_recvq_full(other)) { |
1da177e4 LT |
1269 | err = -EAGAIN; |
1270 | if (!timeo) | |
1271 | goto out_unlock; | |
1272 | ||
1273 | timeo = unix_wait_for_peer(other, timeo); | |
1274 | ||
1275 | err = sock_intr_errno(timeo); | |
1276 | if (signal_pending(current)) | |
1277 | goto out; | |
1278 | sock_put(other); | |
1279 | goto restart; | |
ac7bfa62 | 1280 | } |
1da177e4 LT |
1281 | |
1282 | /* Latch our state. | |
1283 | ||
e5537bfc | 1284 | It is tricky place. We need to grab our state lock and cannot |
1da177e4 LT |
1285 | drop lock on peer. It is dangerous because deadlock is |
1286 | possible. Connect to self case and simultaneous | |
1287 | attempt to connect are eliminated by checking socket | |
1288 | state. other is TCP_LISTEN, if sk is TCP_LISTEN we | |
1289 | check this before attempt to grab lock. | |
1290 | ||
1291 | Well, and we have to recheck the state after socket locked. | |
1292 | */ | |
1293 | st = sk->sk_state; | |
1294 | ||
1295 | switch (st) { | |
1296 | case TCP_CLOSE: | |
1297 | /* This is ok... continue with connect */ | |
1298 | break; | |
1299 | case TCP_ESTABLISHED: | |
1300 | /* Socket is already connected */ | |
1301 | err = -EISCONN; | |
1302 | goto out_unlock; | |
1303 | default: | |
1304 | err = -EINVAL; | |
1305 | goto out_unlock; | |
1306 | } | |
1307 | ||
1c92b4e5 | 1308 | unix_state_lock_nested(sk); |
1da177e4 LT |
1309 | |
1310 | if (sk->sk_state != st) { | |
1c92b4e5 DM |
1311 | unix_state_unlock(sk); |
1312 | unix_state_unlock(other); | |
1da177e4 LT |
1313 | sock_put(other); |
1314 | goto restart; | |
1315 | } | |
1316 | ||
3610cda5 | 1317 | err = security_unix_stream_connect(sk, other, newsk); |
1da177e4 | 1318 | if (err) { |
1c92b4e5 | 1319 | unix_state_unlock(sk); |
1da177e4 LT |
1320 | goto out_unlock; |
1321 | } | |
1322 | ||
1323 | /* The way is open! Fastly set all the necessary fields... */ | |
1324 | ||
1325 | sock_hold(sk); | |
1326 | unix_peer(newsk) = sk; | |
1327 | newsk->sk_state = TCP_ESTABLISHED; | |
1328 | newsk->sk_type = sk->sk_type; | |
109f6e39 | 1329 | init_peercred(newsk); |
1da177e4 | 1330 | newu = unix_sk(newsk); |
eaefd110 | 1331 | RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq); |
1da177e4 LT |
1332 | otheru = unix_sk(other); |
1333 | ||
1334 | /* copy address information from listening to new sock*/ | |
1335 | if (otheru->addr) { | |
8c9814b9 | 1336 | refcount_inc(&otheru->addr->refcnt); |
1da177e4 LT |
1337 | newu->addr = otheru->addr; |
1338 | } | |
40ffe67d AV |
1339 | if (otheru->path.dentry) { |
1340 | path_get(&otheru->path); | |
1341 | newu->path = otheru->path; | |
1da177e4 LT |
1342 | } |
1343 | ||
1344 | /* Set credentials */ | |
109f6e39 | 1345 | copy_peercred(sk, other); |
1da177e4 | 1346 | |
1da177e4 LT |
1347 | sock->state = SS_CONNECTED; |
1348 | sk->sk_state = TCP_ESTABLISHED; | |
830a1e5c BL |
1349 | sock_hold(newsk); |
1350 | ||
4e857c58 | 1351 | smp_mb__after_atomic(); /* sock_hold() does an atomic_inc() */ |
830a1e5c | 1352 | unix_peer(sk) = newsk; |
1da177e4 | 1353 | |
1c92b4e5 | 1354 | unix_state_unlock(sk); |
1da177e4 LT |
1355 | |
1356 | /* take ten and and send info to listening sock */ | |
1357 | spin_lock(&other->sk_receive_queue.lock); | |
1358 | __skb_queue_tail(&other->sk_receive_queue, skb); | |
1da177e4 | 1359 | spin_unlock(&other->sk_receive_queue.lock); |
1c92b4e5 | 1360 | unix_state_unlock(other); |
676d2369 | 1361 | other->sk_data_ready(other); |
1da177e4 LT |
1362 | sock_put(other); |
1363 | return 0; | |
1364 | ||
1365 | out_unlock: | |
1366 | if (other) | |
1c92b4e5 | 1367 | unix_state_unlock(other); |
1da177e4 LT |
1368 | |
1369 | out: | |
40d44446 | 1370 | kfree_skb(skb); |
1da177e4 LT |
1371 | if (newsk) |
1372 | unix_release_sock(newsk, 0); | |
1373 | if (other) | |
1374 | sock_put(other); | |
1375 | return err; | |
1376 | } | |
1377 | ||
1378 | static int unix_socketpair(struct socket *socka, struct socket *sockb) | |
1379 | { | |
e27dfcea | 1380 | struct sock *ska = socka->sk, *skb = sockb->sk; |
1da177e4 LT |
1381 | |
1382 | /* Join our sockets back to back */ | |
1383 | sock_hold(ska); | |
1384 | sock_hold(skb); | |
e27dfcea JK |
1385 | unix_peer(ska) = skb; |
1386 | unix_peer(skb) = ska; | |
109f6e39 EB |
1387 | init_peercred(ska); |
1388 | init_peercred(skb); | |
1da177e4 LT |
1389 | |
1390 | if (ska->sk_type != SOCK_DGRAM) { | |
1391 | ska->sk_state = TCP_ESTABLISHED; | |
1392 | skb->sk_state = TCP_ESTABLISHED; | |
1393 | socka->state = SS_CONNECTED; | |
1394 | sockb->state = SS_CONNECTED; | |
1395 | } | |
1396 | return 0; | |
1397 | } | |
1398 | ||
90c6bd34 DB |
1399 | static void unix_sock_inherit_flags(const struct socket *old, |
1400 | struct socket *new) | |
1401 | { | |
1402 | if (test_bit(SOCK_PASSCRED, &old->flags)) | |
1403 | set_bit(SOCK_PASSCRED, &new->flags); | |
1404 | if (test_bit(SOCK_PASSSEC, &old->flags)) | |
1405 | set_bit(SOCK_PASSSEC, &new->flags); | |
1406 | } | |
1407 | ||
cdfbabfb DH |
1408 | static int unix_accept(struct socket *sock, struct socket *newsock, int flags, |
1409 | bool kern) | |
1da177e4 LT |
1410 | { |
1411 | struct sock *sk = sock->sk; | |
1412 | struct sock *tsk; | |
1413 | struct sk_buff *skb; | |
1414 | int err; | |
1415 | ||
1416 | err = -EOPNOTSUPP; | |
6eba6a37 | 1417 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
1da177e4 LT |
1418 | goto out; |
1419 | ||
1420 | err = -EINVAL; | |
1421 | if (sk->sk_state != TCP_LISTEN) | |
1422 | goto out; | |
1423 | ||
1424 | /* If socket state is TCP_LISTEN it cannot change (for now...), | |
1425 | * so that no locks are necessary. | |
1426 | */ | |
1427 | ||
1428 | skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err); | |
1429 | if (!skb) { | |
1430 | /* This means receive shutdown. */ | |
1431 | if (err == 0) | |
1432 | err = -EINVAL; | |
1433 | goto out; | |
1434 | } | |
1435 | ||
1436 | tsk = skb->sk; | |
1437 | skb_free_datagram(sk, skb); | |
1438 | wake_up_interruptible(&unix_sk(sk)->peer_wait); | |
1439 | ||
1440 | /* attach accepted sock to socket */ | |
1c92b4e5 | 1441 | unix_state_lock(tsk); |
1da177e4 | 1442 | newsock->state = SS_CONNECTED; |
90c6bd34 | 1443 | unix_sock_inherit_flags(sock, newsock); |
1da177e4 | 1444 | sock_graft(tsk, newsock); |
1c92b4e5 | 1445 | unix_state_unlock(tsk); |
1da177e4 LT |
1446 | return 0; |
1447 | ||
1448 | out: | |
1449 | return err; | |
1450 | } | |
1451 | ||
1452 | ||
9b2c45d4 | 1453 | static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int peer) |
1da177e4 LT |
1454 | { |
1455 | struct sock *sk = sock->sk; | |
1456 | struct unix_sock *u; | |
13cfa97b | 1457 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr); |
1da177e4 LT |
1458 | int err = 0; |
1459 | ||
1460 | if (peer) { | |
1461 | sk = unix_peer_get(sk); | |
1462 | ||
1463 | err = -ENOTCONN; | |
1464 | if (!sk) | |
1465 | goto out; | |
1466 | err = 0; | |
1467 | } else { | |
1468 | sock_hold(sk); | |
1469 | } | |
1470 | ||
1471 | u = unix_sk(sk); | |
1c92b4e5 | 1472 | unix_state_lock(sk); |
1da177e4 LT |
1473 | if (!u->addr) { |
1474 | sunaddr->sun_family = AF_UNIX; | |
1475 | sunaddr->sun_path[0] = 0; | |
9b2c45d4 | 1476 | err = sizeof(short); |
1da177e4 LT |
1477 | } else { |
1478 | struct unix_address *addr = u->addr; | |
1479 | ||
9b2c45d4 DV |
1480 | err = addr->len; |
1481 | memcpy(sunaddr, addr->name, addr->len); | |
1da177e4 | 1482 | } |
1c92b4e5 | 1483 | unix_state_unlock(sk); |
1da177e4 LT |
1484 | sock_put(sk); |
1485 | out: | |
1486 | return err; | |
1487 | } | |
1488 | ||
1489 | static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb) | |
1490 | { | |
1491 | int i; | |
1492 | ||
1493 | scm->fp = UNIXCB(skb).fp; | |
1da177e4 LT |
1494 | UNIXCB(skb).fp = NULL; |
1495 | ||
6eba6a37 | 1496 | for (i = scm->fp->count-1; i >= 0; i--) |
415e3d3e | 1497 | unix_notinflight(scm->fp->user, scm->fp->fp[i]); |
1da177e4 LT |
1498 | } |
1499 | ||
7361c36c | 1500 | static void unix_destruct_scm(struct sk_buff *skb) |
1da177e4 LT |
1501 | { |
1502 | struct scm_cookie scm; | |
1503 | memset(&scm, 0, sizeof(scm)); | |
7361c36c | 1504 | scm.pid = UNIXCB(skb).pid; |
7361c36c EB |
1505 | if (UNIXCB(skb).fp) |
1506 | unix_detach_fds(&scm, skb); | |
1da177e4 LT |
1507 | |
1508 | /* Alas, it calls VFS */ | |
1509 | /* So fscking what? fput() had been SMP-safe since the last Summer */ | |
1510 | scm_destroy(&scm); | |
1511 | sock_wfree(skb); | |
1512 | } | |
1513 | ||
712f4aad | 1514 | /* |
1515 | * The "user->unix_inflight" variable is protected by the garbage | |
1516 | * collection lock, and we just read it locklessly here. If you go | |
1517 | * over the limit, there might be a tiny race in actually noticing | |
1518 | * it across threads. Tough. | |
1519 | */ | |
1520 | static inline bool too_many_unix_fds(struct task_struct *p) | |
1521 | { | |
1522 | struct user_struct *user = current_user(); | |
1523 | ||
1524 | if (unlikely(user->unix_inflight > task_rlimit(p, RLIMIT_NOFILE))) | |
1525 | return !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN); | |
1526 | return false; | |
1527 | } | |
1528 | ||
6209344f | 1529 | static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb) |
1da177e4 LT |
1530 | { |
1531 | int i; | |
25888e30 | 1532 | |
712f4aad | 1533 | if (too_many_unix_fds(current)) |
1534 | return -ETOOMANYREFS; | |
1535 | ||
6209344f MS |
1536 | /* |
1537 | * Need to duplicate file references for the sake of garbage | |
1538 | * collection. Otherwise a socket in the fps might become a | |
1539 | * candidate for GC while the skb is not yet queued. | |
1540 | */ | |
1541 | UNIXCB(skb).fp = scm_fp_dup(scm->fp); | |
1542 | if (!UNIXCB(skb).fp) | |
1543 | return -ENOMEM; | |
1544 | ||
712f4aad | 1545 | for (i = scm->fp->count - 1; i >= 0; i--) |
415e3d3e | 1546 | unix_inflight(scm->fp->user, scm->fp->fp[i]); |
27eac47b | 1547 | return 0; |
1da177e4 LT |
1548 | } |
1549 | ||
f78a5fda | 1550 | static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds) |
7361c36c EB |
1551 | { |
1552 | int err = 0; | |
16e57262 | 1553 | |
f78a5fda | 1554 | UNIXCB(skb).pid = get_pid(scm->pid); |
6b0ee8c0 EB |
1555 | UNIXCB(skb).uid = scm->creds.uid; |
1556 | UNIXCB(skb).gid = scm->creds.gid; | |
7361c36c | 1557 | UNIXCB(skb).fp = NULL; |
37a9a8df | 1558 | unix_get_secdata(scm, skb); |
7361c36c EB |
1559 | if (scm->fp && send_fds) |
1560 | err = unix_attach_fds(scm, skb); | |
1561 | ||
1562 | skb->destructor = unix_destruct_scm; | |
1563 | return err; | |
1564 | } | |
1565 | ||
9490f886 HFS |
1566 | static bool unix_passcred_enabled(const struct socket *sock, |
1567 | const struct sock *other) | |
1568 | { | |
1569 | return test_bit(SOCK_PASSCRED, &sock->flags) || | |
1570 | !other->sk_socket || | |
1571 | test_bit(SOCK_PASSCRED, &other->sk_socket->flags); | |
1572 | } | |
1573 | ||
16e57262 ED |
1574 | /* |
1575 | * Some apps rely on write() giving SCM_CREDENTIALS | |
1576 | * We include credentials if source or destination socket | |
1577 | * asserted SOCK_PASSCRED. | |
1578 | */ | |
1579 | static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock, | |
1580 | const struct sock *other) | |
1581 | { | |
6b0ee8c0 | 1582 | if (UNIXCB(skb).pid) |
16e57262 | 1583 | return; |
9490f886 | 1584 | if (unix_passcred_enabled(sock, other)) { |
16e57262 | 1585 | UNIXCB(skb).pid = get_pid(task_tgid(current)); |
6e0895c2 | 1586 | current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid); |
16e57262 ED |
1587 | } |
1588 | } | |
1589 | ||
9490f886 HFS |
1590 | static int maybe_init_creds(struct scm_cookie *scm, |
1591 | struct socket *socket, | |
1592 | const struct sock *other) | |
1593 | { | |
1594 | int err; | |
1595 | struct msghdr msg = { .msg_controllen = 0 }; | |
1596 | ||
1597 | err = scm_send(socket, &msg, scm, false); | |
1598 | if (err) | |
1599 | return err; | |
1600 | ||
1601 | if (unix_passcred_enabled(socket, other)) { | |
1602 | scm->pid = get_pid(task_tgid(current)); | |
1603 | current_uid_gid(&scm->creds.uid, &scm->creds.gid); | |
1604 | } | |
1605 | return err; | |
1606 | } | |
1607 | ||
1608 | static bool unix_skb_scm_eq(struct sk_buff *skb, | |
1609 | struct scm_cookie *scm) | |
1610 | { | |
1611 | const struct unix_skb_parms *u = &UNIXCB(skb); | |
1612 | ||
1613 | return u->pid == scm->pid && | |
1614 | uid_eq(u->uid, scm->creds.uid) && | |
1615 | gid_eq(u->gid, scm->creds.gid) && | |
1616 | unix_secdata_eq(scm, skb); | |
1617 | } | |
1618 | ||
1da177e4 LT |
1619 | /* |
1620 | * Send AF_UNIX data. | |
1621 | */ | |
1622 | ||
1b784140 YX |
1623 | static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg, |
1624 | size_t len) | |
1da177e4 | 1625 | { |
1da177e4 | 1626 | struct sock *sk = sock->sk; |
3b1e0a65 | 1627 | struct net *net = sock_net(sk); |
1da177e4 | 1628 | struct unix_sock *u = unix_sk(sk); |
342dfc30 | 1629 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name); |
1da177e4 LT |
1630 | struct sock *other = NULL; |
1631 | int namelen = 0; /* fake GCC */ | |
1632 | int err; | |
95c96174 | 1633 | unsigned int hash; |
f78a5fda | 1634 | struct sk_buff *skb; |
1da177e4 | 1635 | long timeo; |
7cc05662 | 1636 | struct scm_cookie scm; |
eb6a2481 | 1637 | int data_len = 0; |
7d267278 | 1638 | int sk_locked; |
1da177e4 | 1639 | |
5f23b734 | 1640 | wait_for_unix_gc(); |
7cc05662 | 1641 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1642 | if (err < 0) |
1643 | return err; | |
1644 | ||
1645 | err = -EOPNOTSUPP; | |
1646 | if (msg->msg_flags&MSG_OOB) | |
1647 | goto out; | |
1648 | ||
1649 | if (msg->msg_namelen) { | |
1650 | err = unix_mkname(sunaddr, msg->msg_namelen, &hash); | |
1651 | if (err < 0) | |
1652 | goto out; | |
1653 | namelen = err; | |
1654 | } else { | |
1655 | sunaddr = NULL; | |
1656 | err = -ENOTCONN; | |
1657 | other = unix_peer_get(sk); | |
1658 | if (!other) | |
1659 | goto out; | |
1660 | } | |
1661 | ||
f64f9e71 JP |
1662 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr |
1663 | && (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1664 | goto out; |
1665 | ||
1666 | err = -EMSGSIZE; | |
1667 | if (len > sk->sk_sndbuf - 32) | |
1668 | goto out; | |
1669 | ||
31ff6aa5 | 1670 | if (len > SKB_MAX_ALLOC) { |
eb6a2481 ED |
1671 | data_len = min_t(size_t, |
1672 | len - SKB_MAX_ALLOC, | |
1673 | MAX_SKB_FRAGS * PAGE_SIZE); | |
31ff6aa5 KT |
1674 | data_len = PAGE_ALIGN(data_len); |
1675 | ||
1676 | BUILD_BUG_ON(SKB_MAX_ALLOC < PAGE_SIZE); | |
1677 | } | |
eb6a2481 ED |
1678 | |
1679 | skb = sock_alloc_send_pskb(sk, len - data_len, data_len, | |
28d64271 ED |
1680 | msg->msg_flags & MSG_DONTWAIT, &err, |
1681 | PAGE_ALLOC_COSTLY_ORDER); | |
e27dfcea | 1682 | if (skb == NULL) |
1da177e4 LT |
1683 | goto out; |
1684 | ||
7cc05662 | 1685 | err = unix_scm_to_skb(&scm, skb, true); |
25888e30 | 1686 | if (err < 0) |
7361c36c | 1687 | goto out_free; |
877ce7c1 | 1688 | |
eb6a2481 ED |
1689 | skb_put(skb, len - data_len); |
1690 | skb->data_len = data_len; | |
1691 | skb->len = len; | |
c0371da6 | 1692 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len); |
1da177e4 LT |
1693 | if (err) |
1694 | goto out_free; | |
1695 | ||
1696 | timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); | |
1697 | ||
1698 | restart: | |
1699 | if (!other) { | |
1700 | err = -ECONNRESET; | |
1701 | if (sunaddr == NULL) | |
1702 | goto out_free; | |
1703 | ||
097e66c5 | 1704 | other = unix_find_other(net, sunaddr, namelen, sk->sk_type, |
1da177e4 | 1705 | hash, &err); |
e27dfcea | 1706 | if (other == NULL) |
1da177e4 LT |
1707 | goto out_free; |
1708 | } | |
1709 | ||
d6ae3bae AC |
1710 | if (sk_filter(other, skb) < 0) { |
1711 | /* Toss the packet but do not return any error to the sender */ | |
1712 | err = len; | |
1713 | goto out_free; | |
1714 | } | |
1715 | ||
7d267278 | 1716 | sk_locked = 0; |
1c92b4e5 | 1717 | unix_state_lock(other); |
7d267278 | 1718 | restart_locked: |
1da177e4 LT |
1719 | err = -EPERM; |
1720 | if (!unix_may_send(sk, other)) | |
1721 | goto out_unlock; | |
1722 | ||
7d267278 | 1723 | if (unlikely(sock_flag(other, SOCK_DEAD))) { |
1da177e4 LT |
1724 | /* |
1725 | * Check with 1003.1g - what should | |
1726 | * datagram error | |
1727 | */ | |
1c92b4e5 | 1728 | unix_state_unlock(other); |
1da177e4 LT |
1729 | sock_put(other); |
1730 | ||
7d267278 RW |
1731 | if (!sk_locked) |
1732 | unix_state_lock(sk); | |
1733 | ||
1da177e4 | 1734 | err = 0; |
1da177e4 | 1735 | if (unix_peer(sk) == other) { |
e27dfcea | 1736 | unix_peer(sk) = NULL; |
7d267278 RW |
1737 | unix_dgram_peer_wake_disconnect_wakeup(sk, other); |
1738 | ||
1c92b4e5 | 1739 | unix_state_unlock(sk); |
1da177e4 LT |
1740 | |
1741 | unix_dgram_disconnected(sk, other); | |
1742 | sock_put(other); | |
1743 | err = -ECONNREFUSED; | |
1744 | } else { | |
1c92b4e5 | 1745 | unix_state_unlock(sk); |
1da177e4 LT |
1746 | } |
1747 | ||
1748 | other = NULL; | |
1749 | if (err) | |
1750 | goto out_free; | |
1751 | goto restart; | |
1752 | } | |
1753 | ||
1754 | err = -EPIPE; | |
1755 | if (other->sk_shutdown & RCV_SHUTDOWN) | |
1756 | goto out_unlock; | |
1757 | ||
1758 | if (sk->sk_type != SOCK_SEQPACKET) { | |
1759 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1760 | if (err) | |
1761 | goto out_unlock; | |
1762 | } | |
1763 | ||
a5527dda RW |
1764 | /* other == sk && unix_peer(other) != sk if |
1765 | * - unix_peer(sk) == NULL, destination address bound to sk | |
1766 | * - unix_peer(sk) == sk by time of get but disconnected before lock | |
1767 | */ | |
1768 | if (other != sk && | |
1769 | unlikely(unix_peer(other) != sk && unix_recvq_full(other))) { | |
7d267278 RW |
1770 | if (timeo) { |
1771 | timeo = unix_wait_for_peer(other, timeo); | |
1772 | ||
1773 | err = sock_intr_errno(timeo); | |
1774 | if (signal_pending(current)) | |
1775 | goto out_free; | |
1776 | ||
1777 | goto restart; | |
1da177e4 LT |
1778 | } |
1779 | ||
7d267278 RW |
1780 | if (!sk_locked) { |
1781 | unix_state_unlock(other); | |
1782 | unix_state_double_lock(sk, other); | |
1783 | } | |
1da177e4 | 1784 | |
7d267278 RW |
1785 | if (unix_peer(sk) != other || |
1786 | unix_dgram_peer_wake_me(sk, other)) { | |
1787 | err = -EAGAIN; | |
1788 | sk_locked = 1; | |
1789 | goto out_unlock; | |
1790 | } | |
1da177e4 | 1791 | |
7d267278 RW |
1792 | if (!sk_locked) { |
1793 | sk_locked = 1; | |
1794 | goto restart_locked; | |
1795 | } | |
1da177e4 LT |
1796 | } |
1797 | ||
7d267278 RW |
1798 | if (unlikely(sk_locked)) |
1799 | unix_state_unlock(sk); | |
1800 | ||
3f66116e AC |
1801 | if (sock_flag(other, SOCK_RCVTSTAMP)) |
1802 | __net_timestamp(skb); | |
16e57262 | 1803 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1804 | skb_queue_tail(&other->sk_receive_queue, skb); |
1c92b4e5 | 1805 | unix_state_unlock(other); |
676d2369 | 1806 | other->sk_data_ready(other); |
1da177e4 | 1807 | sock_put(other); |
7cc05662 | 1808 | scm_destroy(&scm); |
1da177e4 LT |
1809 | return len; |
1810 | ||
1811 | out_unlock: | |
7d267278 RW |
1812 | if (sk_locked) |
1813 | unix_state_unlock(sk); | |
1c92b4e5 | 1814 | unix_state_unlock(other); |
1da177e4 LT |
1815 | out_free: |
1816 | kfree_skb(skb); | |
1817 | out: | |
1818 | if (other) | |
1819 | sock_put(other); | |
7cc05662 | 1820 | scm_destroy(&scm); |
1da177e4 LT |
1821 | return err; |
1822 | } | |
1823 | ||
e370a723 | 1824 | /* We use paged skbs for stream sockets, and limit occupancy to 32768 |
d4e9a408 | 1825 | * bytes, and a minimum of a full page. |
e370a723 ED |
1826 | */ |
1827 | #define UNIX_SKB_FRAGS_SZ (PAGE_SIZE << get_order(32768)) | |
ac7bfa62 | 1828 | |
1b784140 YX |
1829 | static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg, |
1830 | size_t len) | |
1da177e4 | 1831 | { |
1da177e4 LT |
1832 | struct sock *sk = sock->sk; |
1833 | struct sock *other = NULL; | |
6eba6a37 | 1834 | int err, size; |
f78a5fda | 1835 | struct sk_buff *skb; |
e27dfcea | 1836 | int sent = 0; |
7cc05662 | 1837 | struct scm_cookie scm; |
8ba69ba6 | 1838 | bool fds_sent = false; |
e370a723 | 1839 | int data_len; |
1da177e4 | 1840 | |
5f23b734 | 1841 | wait_for_unix_gc(); |
7cc05662 | 1842 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1843 | if (err < 0) |
1844 | return err; | |
1845 | ||
1846 | err = -EOPNOTSUPP; | |
1847 | if (msg->msg_flags&MSG_OOB) | |
1848 | goto out_err; | |
1849 | ||
1850 | if (msg->msg_namelen) { | |
1851 | err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; | |
1852 | goto out_err; | |
1853 | } else { | |
1da177e4 | 1854 | err = -ENOTCONN; |
830a1e5c | 1855 | other = unix_peer(sk); |
1da177e4 LT |
1856 | if (!other) |
1857 | goto out_err; | |
1858 | } | |
1859 | ||
1860 | if (sk->sk_shutdown & SEND_SHUTDOWN) | |
1861 | goto pipe_err; | |
1862 | ||
6eba6a37 | 1863 | while (sent < len) { |
e370a723 | 1864 | size = len - sent; |
1da177e4 LT |
1865 | |
1866 | /* Keep two messages in the pipe so it schedules better */ | |
e370a723 | 1867 | size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64); |
1da177e4 | 1868 | |
e370a723 ED |
1869 | /* allow fallback to order-0 allocations */ |
1870 | size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ); | |
ac7bfa62 | 1871 | |
e370a723 | 1872 | data_len = max_t(int, 0, size - SKB_MAX_HEAD(0)); |
1da177e4 | 1873 | |
31ff6aa5 KT |
1874 | data_len = min_t(size_t, size, PAGE_ALIGN(data_len)); |
1875 | ||
e370a723 | 1876 | skb = sock_alloc_send_pskb(sk, size - data_len, data_len, |
28d64271 ED |
1877 | msg->msg_flags & MSG_DONTWAIT, &err, |
1878 | get_order(UNIX_SKB_FRAGS_SZ)); | |
e370a723 | 1879 | if (!skb) |
1da177e4 LT |
1880 | goto out_err; |
1881 | ||
f78a5fda | 1882 | /* Only send the fds in the first buffer */ |
7cc05662 | 1883 | err = unix_scm_to_skb(&scm, skb, !fds_sent); |
25888e30 | 1884 | if (err < 0) { |
7361c36c | 1885 | kfree_skb(skb); |
f78a5fda | 1886 | goto out_err; |
6209344f | 1887 | } |
7361c36c | 1888 | fds_sent = true; |
1da177e4 | 1889 | |
e370a723 ED |
1890 | skb_put(skb, size - data_len); |
1891 | skb->data_len = data_len; | |
1892 | skb->len = size; | |
c0371da6 | 1893 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size); |
6eba6a37 | 1894 | if (err) { |
1da177e4 | 1895 | kfree_skb(skb); |
f78a5fda | 1896 | goto out_err; |
1da177e4 LT |
1897 | } |
1898 | ||
1c92b4e5 | 1899 | unix_state_lock(other); |
1da177e4 LT |
1900 | |
1901 | if (sock_flag(other, SOCK_DEAD) || | |
1902 | (other->sk_shutdown & RCV_SHUTDOWN)) | |
1903 | goto pipe_err_free; | |
1904 | ||
16e57262 | 1905 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1906 | skb_queue_tail(&other->sk_receive_queue, skb); |
1c92b4e5 | 1907 | unix_state_unlock(other); |
676d2369 | 1908 | other->sk_data_ready(other); |
e27dfcea | 1909 | sent += size; |
1da177e4 | 1910 | } |
1da177e4 | 1911 | |
7cc05662 | 1912 | scm_destroy(&scm); |
1da177e4 LT |
1913 | |
1914 | return sent; | |
1915 | ||
1916 | pipe_err_free: | |
1c92b4e5 | 1917 | unix_state_unlock(other); |
1da177e4 LT |
1918 | kfree_skb(skb); |
1919 | pipe_err: | |
6eba6a37 ED |
1920 | if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL)) |
1921 | send_sig(SIGPIPE, current, 0); | |
1da177e4 LT |
1922 | err = -EPIPE; |
1923 | out_err: | |
7cc05662 | 1924 | scm_destroy(&scm); |
1da177e4 LT |
1925 | return sent ? : err; |
1926 | } | |
1927 | ||
869e7c62 HFS |
1928 | static ssize_t unix_stream_sendpage(struct socket *socket, struct page *page, |
1929 | int offset, size_t size, int flags) | |
1930 | { | |
9490f886 HFS |
1931 | int err; |
1932 | bool send_sigpipe = false; | |
1933 | bool init_scm = true; | |
1934 | struct scm_cookie scm; | |
869e7c62 HFS |
1935 | struct sock *other, *sk = socket->sk; |
1936 | struct sk_buff *skb, *newskb = NULL, *tail = NULL; | |
1937 | ||
1938 | if (flags & MSG_OOB) | |
1939 | return -EOPNOTSUPP; | |
1940 | ||
1941 | other = unix_peer(sk); | |
1942 | if (!other || sk->sk_state != TCP_ESTABLISHED) | |
1943 | return -ENOTCONN; | |
1944 | ||
1945 | if (false) { | |
1946 | alloc_skb: | |
1947 | unix_state_unlock(other); | |
6e1ce3c3 | 1948 | mutex_unlock(&unix_sk(other)->iolock); |
869e7c62 HFS |
1949 | newskb = sock_alloc_send_pskb(sk, 0, 0, flags & MSG_DONTWAIT, |
1950 | &err, 0); | |
1951 | if (!newskb) | |
9490f886 | 1952 | goto err; |
869e7c62 HFS |
1953 | } |
1954 | ||
6e1ce3c3 | 1955 | /* we must acquire iolock as we modify already present |
869e7c62 HFS |
1956 | * skbs in the sk_receive_queue and mess with skb->len |
1957 | */ | |
6e1ce3c3 | 1958 | err = mutex_lock_interruptible(&unix_sk(other)->iolock); |
869e7c62 HFS |
1959 | if (err) { |
1960 | err = flags & MSG_DONTWAIT ? -EAGAIN : -ERESTARTSYS; | |
869e7c62 HFS |
1961 | goto err; |
1962 | } | |
1963 | ||
1964 | if (sk->sk_shutdown & SEND_SHUTDOWN) { | |
1965 | err = -EPIPE; | |
9490f886 | 1966 | send_sigpipe = true; |
869e7c62 HFS |
1967 | goto err_unlock; |
1968 | } | |
1969 | ||
1970 | unix_state_lock(other); | |
1971 | ||
1972 | if (sock_flag(other, SOCK_DEAD) || | |
1973 | other->sk_shutdown & RCV_SHUTDOWN) { | |
1974 | err = -EPIPE; | |
9490f886 | 1975 | send_sigpipe = true; |
869e7c62 HFS |
1976 | goto err_state_unlock; |
1977 | } | |
1978 | ||
9490f886 HFS |
1979 | if (init_scm) { |
1980 | err = maybe_init_creds(&scm, socket, other); | |
1981 | if (err) | |
1982 | goto err_state_unlock; | |
1983 | init_scm = false; | |
1984 | } | |
1985 | ||
869e7c62 HFS |
1986 | skb = skb_peek_tail(&other->sk_receive_queue); |
1987 | if (tail && tail == skb) { | |
1988 | skb = newskb; | |
9490f886 HFS |
1989 | } else if (!skb || !unix_skb_scm_eq(skb, &scm)) { |
1990 | if (newskb) { | |
869e7c62 | 1991 | skb = newskb; |
9490f886 HFS |
1992 | } else { |
1993 | tail = skb; | |
869e7c62 | 1994 | goto alloc_skb; |
9490f886 | 1995 | } |
869e7c62 HFS |
1996 | } else if (newskb) { |
1997 | /* this is fast path, we don't necessarily need to | |
1998 | * call to kfree_skb even though with newskb == NULL | |
1999 | * this - does no harm | |
2000 | */ | |
2001 | consume_skb(newskb); | |
8844f972 | 2002 | newskb = NULL; |
869e7c62 HFS |
2003 | } |
2004 | ||
2005 | if (skb_append_pagefrags(skb, page, offset, size)) { | |
2006 | tail = skb; | |
2007 | goto alloc_skb; | |
2008 | } | |
2009 | ||
2010 | skb->len += size; | |
2011 | skb->data_len += size; | |
2012 | skb->truesize += size; | |
14afee4b | 2013 | refcount_add(size, &sk->sk_wmem_alloc); |
869e7c62 | 2014 | |
a3a116e0 | 2015 | if (newskb) { |
9490f886 HFS |
2016 | err = unix_scm_to_skb(&scm, skb, false); |
2017 | if (err) | |
2018 | goto err_state_unlock; | |
a3a116e0 | 2019 | spin_lock(&other->sk_receive_queue.lock); |
869e7c62 | 2020 | __skb_queue_tail(&other->sk_receive_queue, newskb); |
a3a116e0 HFS |
2021 | spin_unlock(&other->sk_receive_queue.lock); |
2022 | } | |
869e7c62 HFS |
2023 | |
2024 | unix_state_unlock(other); | |
6e1ce3c3 | 2025 | mutex_unlock(&unix_sk(other)->iolock); |
869e7c62 HFS |
2026 | |
2027 | other->sk_data_ready(other); | |
9490f886 | 2028 | scm_destroy(&scm); |
869e7c62 HFS |
2029 | return size; |
2030 | ||
2031 | err_state_unlock: | |
2032 | unix_state_unlock(other); | |
2033 | err_unlock: | |
6e1ce3c3 | 2034 | mutex_unlock(&unix_sk(other)->iolock); |
869e7c62 HFS |
2035 | err: |
2036 | kfree_skb(newskb); | |
2037 | if (send_sigpipe && !(flags & MSG_NOSIGNAL)) | |
2038 | send_sig(SIGPIPE, current, 0); | |
9490f886 HFS |
2039 | if (!init_scm) |
2040 | scm_destroy(&scm); | |
869e7c62 HFS |
2041 | return err; |
2042 | } | |
2043 | ||
1b784140 YX |
2044 | static int unix_seqpacket_sendmsg(struct socket *sock, struct msghdr *msg, |
2045 | size_t len) | |
1da177e4 LT |
2046 | { |
2047 | int err; | |
2048 | struct sock *sk = sock->sk; | |
ac7bfa62 | 2049 | |
1da177e4 LT |
2050 | err = sock_error(sk); |
2051 | if (err) | |
2052 | return err; | |
2053 | ||
2054 | if (sk->sk_state != TCP_ESTABLISHED) | |
2055 | return -ENOTCONN; | |
2056 | ||
2057 | if (msg->msg_namelen) | |
2058 | msg->msg_namelen = 0; | |
2059 | ||
1b784140 | 2060 | return unix_dgram_sendmsg(sock, msg, len); |
1da177e4 | 2061 | } |
ac7bfa62 | 2062 | |
1b784140 YX |
2063 | static int unix_seqpacket_recvmsg(struct socket *sock, struct msghdr *msg, |
2064 | size_t size, int flags) | |
a05d2ad1 EB |
2065 | { |
2066 | struct sock *sk = sock->sk; | |
2067 | ||
2068 | if (sk->sk_state != TCP_ESTABLISHED) | |
2069 | return -ENOTCONN; | |
2070 | ||
1b784140 | 2071 | return unix_dgram_recvmsg(sock, msg, size, flags); |
a05d2ad1 EB |
2072 | } |
2073 | ||
1da177e4 LT |
2074 | static void unix_copy_addr(struct msghdr *msg, struct sock *sk) |
2075 | { | |
2076 | struct unix_sock *u = unix_sk(sk); | |
2077 | ||
1da177e4 LT |
2078 | if (u->addr) { |
2079 | msg->msg_namelen = u->addr->len; | |
2080 | memcpy(msg->msg_name, u->addr->name, u->addr->len); | |
2081 | } | |
2082 | } | |
2083 | ||
1b784140 YX |
2084 | static int unix_dgram_recvmsg(struct socket *sock, struct msghdr *msg, |
2085 | size_t size, int flags) | |
1da177e4 | 2086 | { |
7cc05662 | 2087 | struct scm_cookie scm; |
1da177e4 LT |
2088 | struct sock *sk = sock->sk; |
2089 | struct unix_sock *u = unix_sk(sk); | |
64874280 RW |
2090 | struct sk_buff *skb, *last; |
2091 | long timeo; | |
1da177e4 | 2092 | int err; |
f55bb7f9 | 2093 | int peeked, skip; |
1da177e4 LT |
2094 | |
2095 | err = -EOPNOTSUPP; | |
2096 | if (flags&MSG_OOB) | |
2097 | goto out; | |
2098 | ||
64874280 | 2099 | timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); |
1da177e4 | 2100 | |
64874280 | 2101 | do { |
6e1ce3c3 | 2102 | mutex_lock(&u->iolock); |
f55bb7f9 | 2103 | |
64874280 | 2104 | skip = sk_peek_offset(sk, flags); |
7c13f97f PA |
2105 | skb = __skb_try_recv_datagram(sk, flags, NULL, &peeked, &skip, |
2106 | &err, &last); | |
64874280 RW |
2107 | if (skb) |
2108 | break; | |
2109 | ||
6e1ce3c3 | 2110 | mutex_unlock(&u->iolock); |
64874280 RW |
2111 | |
2112 | if (err != -EAGAIN) | |
2113 | break; | |
2114 | } while (timeo && | |
2115 | !__skb_wait_for_more_packets(sk, &err, &timeo, last)); | |
2116 | ||
6e1ce3c3 | 2117 | if (!skb) { /* implies iolock unlocked */ |
0a112258 FZ |
2118 | unix_state_lock(sk); |
2119 | /* Signal EOF on disconnected non-blocking SEQPACKET socket. */ | |
2120 | if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN && | |
2121 | (sk->sk_shutdown & RCV_SHUTDOWN)) | |
2122 | err = 0; | |
2123 | unix_state_unlock(sk); | |
64874280 | 2124 | goto out; |
0a112258 | 2125 | } |
1da177e4 | 2126 | |
77b75f4d RW |
2127 | if (wq_has_sleeper(&u->peer_wait)) |
2128 | wake_up_interruptible_sync_poll(&u->peer_wait, | |
a9a08845 LT |
2129 | EPOLLOUT | EPOLLWRNORM | |
2130 | EPOLLWRBAND); | |
1da177e4 LT |
2131 | |
2132 | if (msg->msg_name) | |
2133 | unix_copy_addr(msg, skb->sk); | |
2134 | ||
f55bb7f9 PE |
2135 | if (size > skb->len - skip) |
2136 | size = skb->len - skip; | |
2137 | else if (size < skb->len - skip) | |
1da177e4 LT |
2138 | msg->msg_flags |= MSG_TRUNC; |
2139 | ||
51f3d02b | 2140 | err = skb_copy_datagram_msg(skb, skip, msg, size); |
1da177e4 LT |
2141 | if (err) |
2142 | goto out_free; | |
2143 | ||
3f66116e AC |
2144 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
2145 | __sock_recv_timestamp(msg, sk, skb); | |
2146 | ||
7cc05662 CH |
2147 | memset(&scm, 0, sizeof(scm)); |
2148 | ||
2149 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); | |
2150 | unix_set_secdata(&scm, skb); | |
1da177e4 | 2151 | |
6eba6a37 | 2152 | if (!(flags & MSG_PEEK)) { |
1da177e4 | 2153 | if (UNIXCB(skb).fp) |
7cc05662 | 2154 | unix_detach_fds(&scm, skb); |
f55bb7f9 PE |
2155 | |
2156 | sk_peek_offset_bwd(sk, skb->len); | |
6eba6a37 | 2157 | } else { |
1da177e4 LT |
2158 | /* It is questionable: on PEEK we could: |
2159 | - do not return fds - good, but too simple 8) | |
2160 | - return fds, and do not return them on read (old strategy, | |
2161 | apparently wrong) | |
2162 | - clone fds (I chose it for now, it is the most universal | |
2163 | solution) | |
ac7bfa62 YH |
2164 | |
2165 | POSIX 1003.1g does not actually define this clearly | |
2166 | at all. POSIX 1003.1g doesn't define a lot of things | |
2167 | clearly however! | |
2168 | ||
1da177e4 | 2169 | */ |
f55bb7f9 PE |
2170 | |
2171 | sk_peek_offset_fwd(sk, size); | |
2172 | ||
1da177e4 | 2173 | if (UNIXCB(skb).fp) |
7cc05662 | 2174 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 2175 | } |
9f6f9af7 | 2176 | err = (flags & MSG_TRUNC) ? skb->len - skip : size; |
1da177e4 | 2177 | |
7cc05662 | 2178 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2179 | |
2180 | out_free: | |
6eba6a37 | 2181 | skb_free_datagram(sk, skb); |
6e1ce3c3 | 2182 | mutex_unlock(&u->iolock); |
1da177e4 LT |
2183 | out: |
2184 | return err; | |
2185 | } | |
2186 | ||
2187 | /* | |
79f632c7 | 2188 | * Sleep until more data has arrived. But check for races.. |
1da177e4 | 2189 | */ |
79f632c7 | 2190 | static long unix_stream_data_wait(struct sock *sk, long timeo, |
06a77b07 WC |
2191 | struct sk_buff *last, unsigned int last_len, |
2192 | bool freezable) | |
1da177e4 | 2193 | { |
2b514574 | 2194 | struct sk_buff *tail; |
1da177e4 LT |
2195 | DEFINE_WAIT(wait); |
2196 | ||
1c92b4e5 | 2197 | unix_state_lock(sk); |
1da177e4 LT |
2198 | |
2199 | for (;;) { | |
aa395145 | 2200 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 2201 | |
2b514574 HFS |
2202 | tail = skb_peek_tail(&sk->sk_receive_queue); |
2203 | if (tail != last || | |
2204 | (tail && tail->len != last_len) || | |
1da177e4 LT |
2205 | sk->sk_err || |
2206 | (sk->sk_shutdown & RCV_SHUTDOWN) || | |
2207 | signal_pending(current) || | |
2208 | !timeo) | |
2209 | break; | |
2210 | ||
9cd3e072 | 2211 | sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk); |
1c92b4e5 | 2212 | unix_state_unlock(sk); |
06a77b07 WC |
2213 | if (freezable) |
2214 | timeo = freezable_schedule_timeout(timeo); | |
2215 | else | |
2216 | timeo = schedule_timeout(timeo); | |
1c92b4e5 | 2217 | unix_state_lock(sk); |
b48732e4 MS |
2218 | |
2219 | if (sock_flag(sk, SOCK_DEAD)) | |
2220 | break; | |
2221 | ||
9cd3e072 | 2222 | sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk); |
1da177e4 LT |
2223 | } |
2224 | ||
aa395145 | 2225 | finish_wait(sk_sleep(sk), &wait); |
1c92b4e5 | 2226 | unix_state_unlock(sk); |
1da177e4 LT |
2227 | return timeo; |
2228 | } | |
2229 | ||
e370a723 ED |
2230 | static unsigned int unix_skb_len(const struct sk_buff *skb) |
2231 | { | |
2232 | return skb->len - UNIXCB(skb).consumed; | |
2233 | } | |
2234 | ||
2b514574 HFS |
2235 | struct unix_stream_read_state { |
2236 | int (*recv_actor)(struct sk_buff *, int, int, | |
2237 | struct unix_stream_read_state *); | |
2238 | struct socket *socket; | |
2239 | struct msghdr *msg; | |
2240 | struct pipe_inode_info *pipe; | |
2241 | size_t size; | |
2242 | int flags; | |
2243 | unsigned int splice_flags; | |
2244 | }; | |
2245 | ||
06a77b07 WC |
2246 | static int unix_stream_read_generic(struct unix_stream_read_state *state, |
2247 | bool freezable) | |
1da177e4 | 2248 | { |
7cc05662 | 2249 | struct scm_cookie scm; |
2b514574 | 2250 | struct socket *sock = state->socket; |
1da177e4 LT |
2251 | struct sock *sk = sock->sk; |
2252 | struct unix_sock *u = unix_sk(sk); | |
1da177e4 | 2253 | int copied = 0; |
2b514574 | 2254 | int flags = state->flags; |
de144391 | 2255 | int noblock = flags & MSG_DONTWAIT; |
2b514574 | 2256 | bool check_creds = false; |
1da177e4 LT |
2257 | int target; |
2258 | int err = 0; | |
2259 | long timeo; | |
fc0d7536 | 2260 | int skip; |
2b514574 HFS |
2261 | size_t size = state->size; |
2262 | unsigned int last_len; | |
1da177e4 | 2263 | |
1b92ee3d RW |
2264 | if (unlikely(sk->sk_state != TCP_ESTABLISHED)) { |
2265 | err = -EINVAL; | |
1da177e4 | 2266 | goto out; |
1b92ee3d | 2267 | } |
1da177e4 | 2268 | |
1b92ee3d RW |
2269 | if (unlikely(flags & MSG_OOB)) { |
2270 | err = -EOPNOTSUPP; | |
1da177e4 | 2271 | goto out; |
1b92ee3d | 2272 | } |
1da177e4 | 2273 | |
2b514574 | 2274 | target = sock_rcvlowat(sk, flags & MSG_WAITALL, size); |
de144391 | 2275 | timeo = sock_rcvtimeo(sk, noblock); |
1da177e4 | 2276 | |
2b514574 HFS |
2277 | memset(&scm, 0, sizeof(scm)); |
2278 | ||
1da177e4 LT |
2279 | /* Lock the socket to prevent queue disordering |
2280 | * while sleeps in memcpy_tomsg | |
2281 | */ | |
6e1ce3c3 | 2282 | mutex_lock(&u->iolock); |
1da177e4 | 2283 | |
a0917e0b | 2284 | skip = max(sk_peek_offset(sk, flags), 0); |
e9193d60 | 2285 | |
6eba6a37 | 2286 | do { |
1da177e4 | 2287 | int chunk; |
73ed5d25 | 2288 | bool drop_skb; |
79f632c7 | 2289 | struct sk_buff *skb, *last; |
1da177e4 | 2290 | |
18eceb81 | 2291 | redo: |
3c0d2f37 | 2292 | unix_state_lock(sk); |
b48732e4 MS |
2293 | if (sock_flag(sk, SOCK_DEAD)) { |
2294 | err = -ECONNRESET; | |
2295 | goto unlock; | |
2296 | } | |
79f632c7 | 2297 | last = skb = skb_peek(&sk->sk_receive_queue); |
2b514574 | 2298 | last_len = last ? last->len : 0; |
fc0d7536 | 2299 | again: |
6eba6a37 | 2300 | if (skb == NULL) { |
1da177e4 | 2301 | if (copied >= target) |
3c0d2f37 | 2302 | goto unlock; |
1da177e4 LT |
2303 | |
2304 | /* | |
2305 | * POSIX 1003.1g mandates this order. | |
2306 | */ | |
ac7bfa62 | 2307 | |
6eba6a37 ED |
2308 | err = sock_error(sk); |
2309 | if (err) | |
3c0d2f37 | 2310 | goto unlock; |
1da177e4 | 2311 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
3c0d2f37 MS |
2312 | goto unlock; |
2313 | ||
2314 | unix_state_unlock(sk); | |
1b92ee3d RW |
2315 | if (!timeo) { |
2316 | err = -EAGAIN; | |
1da177e4 | 2317 | break; |
1b92ee3d RW |
2318 | } |
2319 | ||
6e1ce3c3 | 2320 | mutex_unlock(&u->iolock); |
1da177e4 | 2321 | |
2b514574 | 2322 | timeo = unix_stream_data_wait(sk, timeo, last, |
06a77b07 | 2323 | last_len, freezable); |
1da177e4 | 2324 | |
3822b5c2 | 2325 | if (signal_pending(current)) { |
1da177e4 | 2326 | err = sock_intr_errno(timeo); |
fa0dc04d | 2327 | scm_destroy(&scm); |
1da177e4 LT |
2328 | goto out; |
2329 | } | |
b3ca9b02 | 2330 | |
6e1ce3c3 | 2331 | mutex_lock(&u->iolock); |
18eceb81 | 2332 | goto redo; |
2b514574 | 2333 | unlock: |
3c0d2f37 MS |
2334 | unix_state_unlock(sk); |
2335 | break; | |
1da177e4 | 2336 | } |
fc0d7536 | 2337 | |
e370a723 ED |
2338 | while (skip >= unix_skb_len(skb)) { |
2339 | skip -= unix_skb_len(skb); | |
79f632c7 | 2340 | last = skb; |
2b514574 | 2341 | last_len = skb->len; |
fc0d7536 | 2342 | skb = skb_peek_next(skb, &sk->sk_receive_queue); |
79f632c7 BP |
2343 | if (!skb) |
2344 | goto again; | |
fc0d7536 PE |
2345 | } |
2346 | ||
3c0d2f37 | 2347 | unix_state_unlock(sk); |
1da177e4 LT |
2348 | |
2349 | if (check_creds) { | |
2350 | /* Never glue messages from different writers */ | |
9490f886 | 2351 | if (!unix_skb_scm_eq(skb, &scm)) |
1da177e4 | 2352 | break; |
0e82e7f6 | 2353 | } else if (test_bit(SOCK_PASSCRED, &sock->flags)) { |
1da177e4 | 2354 | /* Copy credentials */ |
7cc05662 | 2355 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); |
37a9a8df | 2356 | unix_set_secdata(&scm, skb); |
2b514574 | 2357 | check_creds = true; |
1da177e4 LT |
2358 | } |
2359 | ||
2360 | /* Copy address just once */ | |
2b514574 HFS |
2361 | if (state->msg && state->msg->msg_name) { |
2362 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, | |
2363 | state->msg->msg_name); | |
2364 | unix_copy_addr(state->msg, skb->sk); | |
1da177e4 LT |
2365 | sunaddr = NULL; |
2366 | } | |
2367 | ||
e370a723 | 2368 | chunk = min_t(unsigned int, unix_skb_len(skb) - skip, size); |
73ed5d25 | 2369 | skb_get(skb); |
2b514574 | 2370 | chunk = state->recv_actor(skb, skip, chunk, state); |
73ed5d25 HFS |
2371 | drop_skb = !unix_skb_len(skb); |
2372 | /* skb is only safe to use if !drop_skb */ | |
2373 | consume_skb(skb); | |
2b514574 | 2374 | if (chunk < 0) { |
1da177e4 LT |
2375 | if (copied == 0) |
2376 | copied = -EFAULT; | |
2377 | break; | |
2378 | } | |
2379 | copied += chunk; | |
2380 | size -= chunk; | |
2381 | ||
73ed5d25 HFS |
2382 | if (drop_skb) { |
2383 | /* the skb was touched by a concurrent reader; | |
2384 | * we should not expect anything from this skb | |
2385 | * anymore and assume it invalid - we can be | |
2386 | * sure it was dropped from the socket queue | |
2387 | * | |
2388 | * let's report a short read | |
2389 | */ | |
2390 | err = 0; | |
2391 | break; | |
2392 | } | |
2393 | ||
1da177e4 | 2394 | /* Mark read part of skb as used */ |
6eba6a37 | 2395 | if (!(flags & MSG_PEEK)) { |
e370a723 | 2396 | UNIXCB(skb).consumed += chunk; |
1da177e4 | 2397 | |
fc0d7536 PE |
2398 | sk_peek_offset_bwd(sk, chunk); |
2399 | ||
1da177e4 | 2400 | if (UNIXCB(skb).fp) |
7cc05662 | 2401 | unix_detach_fds(&scm, skb); |
1da177e4 | 2402 | |
e370a723 | 2403 | if (unix_skb_len(skb)) |
1da177e4 | 2404 | break; |
1da177e4 | 2405 | |
6f01fd6e | 2406 | skb_unlink(skb, &sk->sk_receive_queue); |
70d4bf6d | 2407 | consume_skb(skb); |
1da177e4 | 2408 | |
7cc05662 | 2409 | if (scm.fp) |
1da177e4 | 2410 | break; |
6eba6a37 | 2411 | } else { |
1da177e4 LT |
2412 | /* It is questionable, see note in unix_dgram_recvmsg. |
2413 | */ | |
2414 | if (UNIXCB(skb).fp) | |
7cc05662 | 2415 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 2416 | |
e9193d60 | 2417 | sk_peek_offset_fwd(sk, chunk); |
fc0d7536 | 2418 | |
9f389e35 AC |
2419 | if (UNIXCB(skb).fp) |
2420 | break; | |
2421 | ||
e9193d60 | 2422 | skip = 0; |
9f389e35 AC |
2423 | last = skb; |
2424 | last_len = skb->len; | |
2425 | unix_state_lock(sk); | |
2426 | skb = skb_peek_next(skb, &sk->sk_receive_queue); | |
2427 | if (skb) | |
2428 | goto again; | |
2429 | unix_state_unlock(sk); | |
1da177e4 LT |
2430 | break; |
2431 | } | |
2432 | } while (size); | |
2433 | ||
6e1ce3c3 | 2434 | mutex_unlock(&u->iolock); |
2b514574 HFS |
2435 | if (state->msg) |
2436 | scm_recv(sock, state->msg, &scm, flags); | |
2437 | else | |
2438 | scm_destroy(&scm); | |
1da177e4 LT |
2439 | out: |
2440 | return copied ? : err; | |
2441 | } | |
2442 | ||
2b514574 HFS |
2443 | static int unix_stream_read_actor(struct sk_buff *skb, |
2444 | int skip, int chunk, | |
2445 | struct unix_stream_read_state *state) | |
2446 | { | |
2447 | int ret; | |
2448 | ||
2449 | ret = skb_copy_datagram_msg(skb, UNIXCB(skb).consumed + skip, | |
2450 | state->msg, chunk); | |
2451 | return ret ?: chunk; | |
2452 | } | |
2453 | ||
2454 | static int unix_stream_recvmsg(struct socket *sock, struct msghdr *msg, | |
2455 | size_t size, int flags) | |
2456 | { | |
2457 | struct unix_stream_read_state state = { | |
2458 | .recv_actor = unix_stream_read_actor, | |
2459 | .socket = sock, | |
2460 | .msg = msg, | |
2461 | .size = size, | |
2462 | .flags = flags | |
2463 | }; | |
2464 | ||
06a77b07 | 2465 | return unix_stream_read_generic(&state, true); |
2b514574 HFS |
2466 | } |
2467 | ||
2b514574 HFS |
2468 | static int unix_stream_splice_actor(struct sk_buff *skb, |
2469 | int skip, int chunk, | |
2470 | struct unix_stream_read_state *state) | |
2471 | { | |
2472 | return skb_splice_bits(skb, state->socket->sk, | |
2473 | UNIXCB(skb).consumed + skip, | |
25869262 | 2474 | state->pipe, chunk, state->splice_flags); |
2b514574 HFS |
2475 | } |
2476 | ||
2477 | static ssize_t unix_stream_splice_read(struct socket *sock, loff_t *ppos, | |
2478 | struct pipe_inode_info *pipe, | |
2479 | size_t size, unsigned int flags) | |
2480 | { | |
2481 | struct unix_stream_read_state state = { | |
2482 | .recv_actor = unix_stream_splice_actor, | |
2483 | .socket = sock, | |
2484 | .pipe = pipe, | |
2485 | .size = size, | |
2486 | .splice_flags = flags, | |
2487 | }; | |
2488 | ||
2489 | if (unlikely(*ppos)) | |
2490 | return -ESPIPE; | |
2491 | ||
2492 | if (sock->file->f_flags & O_NONBLOCK || | |
2493 | flags & SPLICE_F_NONBLOCK) | |
2494 | state.flags = MSG_DONTWAIT; | |
2495 | ||
06a77b07 | 2496 | return unix_stream_read_generic(&state, false); |
2b514574 HFS |
2497 | } |
2498 | ||
1da177e4 LT |
2499 | static int unix_shutdown(struct socket *sock, int mode) |
2500 | { | |
2501 | struct sock *sk = sock->sk; | |
2502 | struct sock *other; | |
2503 | ||
fc61b928 XW |
2504 | if (mode < SHUT_RD || mode > SHUT_RDWR) |
2505 | return -EINVAL; | |
2506 | /* This maps: | |
2507 | * SHUT_RD (0) -> RCV_SHUTDOWN (1) | |
2508 | * SHUT_WR (1) -> SEND_SHUTDOWN (2) | |
2509 | * SHUT_RDWR (2) -> SHUTDOWN_MASK (3) | |
2510 | */ | |
2511 | ++mode; | |
7180a031 AC |
2512 | |
2513 | unix_state_lock(sk); | |
2514 | sk->sk_shutdown |= mode; | |
2515 | other = unix_peer(sk); | |
2516 | if (other) | |
2517 | sock_hold(other); | |
2518 | unix_state_unlock(sk); | |
2519 | sk->sk_state_change(sk); | |
2520 | ||
2521 | if (other && | |
2522 | (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) { | |
2523 | ||
2524 | int peer_mode = 0; | |
2525 | ||
2526 | if (mode&RCV_SHUTDOWN) | |
2527 | peer_mode |= SEND_SHUTDOWN; | |
2528 | if (mode&SEND_SHUTDOWN) | |
2529 | peer_mode |= RCV_SHUTDOWN; | |
2530 | unix_state_lock(other); | |
2531 | other->sk_shutdown |= peer_mode; | |
2532 | unix_state_unlock(other); | |
2533 | other->sk_state_change(other); | |
2534 | if (peer_mode == SHUTDOWN_MASK) | |
2535 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP); | |
2536 | else if (peer_mode & RCV_SHUTDOWN) | |
2537 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN); | |
1da177e4 | 2538 | } |
7180a031 AC |
2539 | if (other) |
2540 | sock_put(other); | |
2541 | ||
1da177e4 LT |
2542 | return 0; |
2543 | } | |
2544 | ||
885ee74d PE |
2545 | long unix_inq_len(struct sock *sk) |
2546 | { | |
2547 | struct sk_buff *skb; | |
2548 | long amount = 0; | |
2549 | ||
2550 | if (sk->sk_state == TCP_LISTEN) | |
2551 | return -EINVAL; | |
2552 | ||
2553 | spin_lock(&sk->sk_receive_queue.lock); | |
2554 | if (sk->sk_type == SOCK_STREAM || | |
2555 | sk->sk_type == SOCK_SEQPACKET) { | |
2556 | skb_queue_walk(&sk->sk_receive_queue, skb) | |
e370a723 | 2557 | amount += unix_skb_len(skb); |
885ee74d PE |
2558 | } else { |
2559 | skb = skb_peek(&sk->sk_receive_queue); | |
2560 | if (skb) | |
2561 | amount = skb->len; | |
2562 | } | |
2563 | spin_unlock(&sk->sk_receive_queue.lock); | |
2564 | ||
2565 | return amount; | |
2566 | } | |
2567 | EXPORT_SYMBOL_GPL(unix_inq_len); | |
2568 | ||
2569 | long unix_outq_len(struct sock *sk) | |
2570 | { | |
2571 | return sk_wmem_alloc_get(sk); | |
2572 | } | |
2573 | EXPORT_SYMBOL_GPL(unix_outq_len); | |
2574 | ||
ba94f308 AV |
2575 | static int unix_open_file(struct sock *sk) |
2576 | { | |
2577 | struct path path; | |
2578 | struct file *f; | |
2579 | int fd; | |
2580 | ||
2581 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) | |
2582 | return -EPERM; | |
2583 | ||
2584 | unix_state_lock(sk); | |
2585 | path = unix_sk(sk)->path; | |
2586 | if (!path.dentry) { | |
2587 | unix_state_unlock(sk); | |
2588 | return -ENOENT; | |
2589 | } | |
2590 | ||
2591 | path_get(&path); | |
2592 | unix_state_unlock(sk); | |
2593 | ||
2594 | fd = get_unused_fd_flags(O_CLOEXEC); | |
2595 | if (fd < 0) | |
2596 | goto out; | |
2597 | ||
2598 | f = dentry_open(&path, O_PATH, current_cred()); | |
2599 | if (IS_ERR(f)) { | |
2600 | put_unused_fd(fd); | |
2601 | fd = PTR_ERR(f); | |
2602 | goto out; | |
2603 | } | |
2604 | ||
2605 | fd_install(fd, f); | |
2606 | out: | |
2607 | path_put(&path); | |
2608 | ||
2609 | return fd; | |
2610 | } | |
2611 | ||
1da177e4 LT |
2612 | static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
2613 | { | |
2614 | struct sock *sk = sock->sk; | |
e27dfcea | 2615 | long amount = 0; |
1da177e4 LT |
2616 | int err; |
2617 | ||
6eba6a37 ED |
2618 | switch (cmd) { |
2619 | case SIOCOUTQ: | |
885ee74d | 2620 | amount = unix_outq_len(sk); |
6eba6a37 ED |
2621 | err = put_user(amount, (int __user *)arg); |
2622 | break; | |
2623 | case SIOCINQ: | |
885ee74d PE |
2624 | amount = unix_inq_len(sk); |
2625 | if (amount < 0) | |
2626 | err = amount; | |
2627 | else | |
1da177e4 | 2628 | err = put_user(amount, (int __user *)arg); |
885ee74d | 2629 | break; |
ba94f308 AV |
2630 | case SIOCUNIXFILE: |
2631 | err = unix_open_file(sk); | |
2632 | break; | |
6eba6a37 ED |
2633 | default: |
2634 | err = -ENOIOCTLCMD; | |
2635 | break; | |
1da177e4 LT |
2636 | } |
2637 | return err; | |
2638 | } | |
2639 | ||
a11e1d43 | 2640 | static __poll_t unix_poll(struct file *file, struct socket *sock, poll_table *wait) |
1da177e4 LT |
2641 | { |
2642 | struct sock *sk = sock->sk; | |
a11e1d43 LT |
2643 | __poll_t mask; |
2644 | ||
89ab066d | 2645 | sock_poll_wait(file, sock, wait); |
a11e1d43 | 2646 | mask = 0; |
1da177e4 LT |
2647 | |
2648 | /* exceptional events? */ | |
2649 | if (sk->sk_err) | |
a9a08845 | 2650 | mask |= EPOLLERR; |
1da177e4 | 2651 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
a9a08845 | 2652 | mask |= EPOLLHUP; |
f348d70a | 2653 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
a9a08845 | 2654 | mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; |
1da177e4 LT |
2655 | |
2656 | /* readable? */ | |
db40980f | 2657 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
a9a08845 | 2658 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 LT |
2659 | |
2660 | /* Connection-based need to check for termination and startup */ | |
6eba6a37 ED |
2661 | if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) && |
2662 | sk->sk_state == TCP_CLOSE) | |
a9a08845 | 2663 | mask |= EPOLLHUP; |
1da177e4 LT |
2664 | |
2665 | /* | |
2666 | * we set writable also when the other side has shut down the | |
2667 | * connection. This prevents stuck sockets. | |
2668 | */ | |
2669 | if (unix_writable(sk)) | |
a9a08845 | 2670 | mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; |
1da177e4 LT |
2671 | |
2672 | return mask; | |
2673 | } | |
2674 | ||
a11e1d43 LT |
2675 | static __poll_t unix_dgram_poll(struct file *file, struct socket *sock, |
2676 | poll_table *wait) | |
3c73419c | 2677 | { |
ec0d215f | 2678 | struct sock *sk = sock->sk, *other; |
a11e1d43 LT |
2679 | unsigned int writable; |
2680 | __poll_t mask; | |
2681 | ||
89ab066d | 2682 | sock_poll_wait(file, sock, wait); |
a11e1d43 | 2683 | mask = 0; |
3c73419c RW |
2684 | |
2685 | /* exceptional events? */ | |
2686 | if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue)) | |
a9a08845 LT |
2687 | mask |= EPOLLERR | |
2688 | (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0); | |
7d4c04fc | 2689 | |
3c73419c | 2690 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
a9a08845 | 2691 | mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; |
3c73419c | 2692 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
a9a08845 | 2693 | mask |= EPOLLHUP; |
3c73419c RW |
2694 | |
2695 | /* readable? */ | |
5456f09a | 2696 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
a9a08845 | 2697 | mask |= EPOLLIN | EPOLLRDNORM; |
3c73419c RW |
2698 | |
2699 | /* Connection-based need to check for termination and startup */ | |
2700 | if (sk->sk_type == SOCK_SEQPACKET) { | |
2701 | if (sk->sk_state == TCP_CLOSE) | |
a9a08845 | 2702 | mask |= EPOLLHUP; |
3c73419c RW |
2703 | /* connection hasn't started yet? */ |
2704 | if (sk->sk_state == TCP_SYN_SENT) | |
2705 | return mask; | |
2706 | } | |
2707 | ||
973a34aa | 2708 | /* No write status requested, avoid expensive OUT tests. */ |
a11e1d43 | 2709 | if (!(poll_requested_events(wait) & (EPOLLWRBAND|EPOLLWRNORM|EPOLLOUT))) |
973a34aa ED |
2710 | return mask; |
2711 | ||
ec0d215f | 2712 | writable = unix_writable(sk); |
7d267278 RW |
2713 | if (writable) { |
2714 | unix_state_lock(sk); | |
2715 | ||
2716 | other = unix_peer(sk); | |
2717 | if (other && unix_peer(other) != sk && | |
2718 | unix_recvq_full(other) && | |
2719 | unix_dgram_peer_wake_me(sk, other)) | |
2720 | writable = 0; | |
2721 | ||
2722 | unix_state_unlock(sk); | |
ec0d215f RW |
2723 | } |
2724 | ||
2725 | if (writable) | |
a9a08845 | 2726 | mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; |
3c73419c | 2727 | else |
9cd3e072 | 2728 | sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk); |
3c73419c | 2729 | |
3c73419c RW |
2730 | return mask; |
2731 | } | |
1da177e4 LT |
2732 | |
2733 | #ifdef CONFIG_PROC_FS | |
a53eb3fe | 2734 | |
7123aaa3 ED |
2735 | #define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1) |
2736 | ||
2737 | #define get_bucket(x) ((x) >> BUCKET_SPACE) | |
2738 | #define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1)) | |
2739 | #define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o)) | |
a53eb3fe | 2740 | |
7123aaa3 | 2741 | static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos) |
1da177e4 | 2742 | { |
7123aaa3 ED |
2743 | unsigned long offset = get_offset(*pos); |
2744 | unsigned long bucket = get_bucket(*pos); | |
2745 | struct sock *sk; | |
2746 | unsigned long count = 0; | |
1da177e4 | 2747 | |
7123aaa3 ED |
2748 | for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) { |
2749 | if (sock_net(sk) != seq_file_net(seq)) | |
097e66c5 | 2750 | continue; |
7123aaa3 ED |
2751 | if (++count == offset) |
2752 | break; | |
2753 | } | |
2754 | ||
2755 | return sk; | |
2756 | } | |
2757 | ||
2758 | static struct sock *unix_next_socket(struct seq_file *seq, | |
2759 | struct sock *sk, | |
2760 | loff_t *pos) | |
2761 | { | |
2762 | unsigned long bucket; | |
2763 | ||
2764 | while (sk > (struct sock *)SEQ_START_TOKEN) { | |
2765 | sk = sk_next(sk); | |
2766 | if (!sk) | |
2767 | goto next_bucket; | |
2768 | if (sock_net(sk) == seq_file_net(seq)) | |
2769 | return sk; | |
1da177e4 | 2770 | } |
7123aaa3 ED |
2771 | |
2772 | do { | |
2773 | sk = unix_from_bucket(seq, pos); | |
2774 | if (sk) | |
2775 | return sk; | |
2776 | ||
2777 | next_bucket: | |
2778 | bucket = get_bucket(*pos) + 1; | |
2779 | *pos = set_bucket_offset(bucket, 1); | |
2780 | } while (bucket < ARRAY_SIZE(unix_socket_table)); | |
2781 | ||
1da177e4 LT |
2782 | return NULL; |
2783 | } | |
2784 | ||
1da177e4 | 2785 | static void *unix_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 2786 | __acquires(unix_table_lock) |
1da177e4 | 2787 | { |
fbe9cc4a | 2788 | spin_lock(&unix_table_lock); |
7123aaa3 ED |
2789 | |
2790 | if (!*pos) | |
2791 | return SEQ_START_TOKEN; | |
2792 | ||
2793 | if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table)) | |
2794 | return NULL; | |
2795 | ||
2796 | return unix_next_socket(seq, NULL, pos); | |
1da177e4 LT |
2797 | } |
2798 | ||
2799 | static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2800 | { | |
2801 | ++*pos; | |
7123aaa3 | 2802 | return unix_next_socket(seq, v, pos); |
1da177e4 LT |
2803 | } |
2804 | ||
2805 | static void unix_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2806 | __releases(unix_table_lock) |
1da177e4 | 2807 | { |
fbe9cc4a | 2808 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
2809 | } |
2810 | ||
2811 | static int unix_seq_show(struct seq_file *seq, void *v) | |
2812 | { | |
ac7bfa62 | 2813 | |
b9f3124f | 2814 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2815 | seq_puts(seq, "Num RefCount Protocol Flags Type St " |
2816 | "Inode Path\n"); | |
2817 | else { | |
2818 | struct sock *s = v; | |
2819 | struct unix_sock *u = unix_sk(s); | |
1c92b4e5 | 2820 | unix_state_lock(s); |
1da177e4 | 2821 | |
71338aa7 | 2822 | seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu", |
1da177e4 | 2823 | s, |
41c6d650 | 2824 | refcount_read(&s->sk_refcnt), |
1da177e4 LT |
2825 | 0, |
2826 | s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0, | |
2827 | s->sk_type, | |
2828 | s->sk_socket ? | |
2829 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) : | |
2830 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING), | |
2831 | sock_i_ino(s)); | |
2832 | ||
2833 | if (u->addr) { | |
2834 | int i, len; | |
2835 | seq_putc(seq, ' '); | |
2836 | ||
2837 | i = 0; | |
2838 | len = u->addr->len - sizeof(short); | |
2839 | if (!UNIX_ABSTRACT(s)) | |
2840 | len--; | |
2841 | else { | |
2842 | seq_putc(seq, '@'); | |
2843 | i++; | |
2844 | } | |
2845 | for ( ; i < len; i++) | |
e7947ea7 IB |
2846 | seq_putc(seq, u->addr->name->sun_path[i] ?: |
2847 | '@'); | |
1da177e4 | 2848 | } |
1c92b4e5 | 2849 | unix_state_unlock(s); |
1da177e4 LT |
2850 | seq_putc(seq, '\n'); |
2851 | } | |
2852 | ||
2853 | return 0; | |
2854 | } | |
2855 | ||
56b3d975 | 2856 | static const struct seq_operations unix_seq_ops = { |
1da177e4 LT |
2857 | .start = unix_seq_start, |
2858 | .next = unix_seq_next, | |
2859 | .stop = unix_seq_stop, | |
2860 | .show = unix_seq_show, | |
2861 | }; | |
1da177e4 LT |
2862 | #endif |
2863 | ||
ec1b4cf7 | 2864 | static const struct net_proto_family unix_family_ops = { |
1da177e4 LT |
2865 | .family = PF_UNIX, |
2866 | .create = unix_create, | |
2867 | .owner = THIS_MODULE, | |
2868 | }; | |
2869 | ||
097e66c5 | 2870 | |
2c8c1e72 | 2871 | static int __net_init unix_net_init(struct net *net) |
097e66c5 DL |
2872 | { |
2873 | int error = -ENOMEM; | |
2874 | ||
a0a53c8b | 2875 | net->unx.sysctl_max_dgram_qlen = 10; |
1597fbc0 PE |
2876 | if (unix_sysctl_register(net)) |
2877 | goto out; | |
d392e497 | 2878 | |
097e66c5 | 2879 | #ifdef CONFIG_PROC_FS |
c3506372 CH |
2880 | if (!proc_create_net("unix", 0, net->proc_net, &unix_seq_ops, |
2881 | sizeof(struct seq_net_private))) { | |
1597fbc0 | 2882 | unix_sysctl_unregister(net); |
097e66c5 | 2883 | goto out; |
1597fbc0 | 2884 | } |
097e66c5 DL |
2885 | #endif |
2886 | error = 0; | |
2887 | out: | |
48dcc33e | 2888 | return error; |
097e66c5 DL |
2889 | } |
2890 | ||
2c8c1e72 | 2891 | static void __net_exit unix_net_exit(struct net *net) |
097e66c5 | 2892 | { |
1597fbc0 | 2893 | unix_sysctl_unregister(net); |
ece31ffd | 2894 | remove_proc_entry("unix", net->proc_net); |
097e66c5 DL |
2895 | } |
2896 | ||
2897 | static struct pernet_operations unix_net_ops = { | |
2898 | .init = unix_net_init, | |
2899 | .exit = unix_net_exit, | |
2900 | }; | |
2901 | ||
1da177e4 LT |
2902 | static int __init af_unix_init(void) |
2903 | { | |
2904 | int rc = -1; | |
1da177e4 | 2905 | |
b4fff5f8 | 2906 | BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 LT |
2907 | |
2908 | rc = proto_register(&unix_proto, 1); | |
ac7bfa62 | 2909 | if (rc != 0) { |
5cc208be | 2910 | pr_crit("%s: Cannot create unix_sock SLAB cache!\n", __func__); |
1da177e4 LT |
2911 | goto out; |
2912 | } | |
2913 | ||
2914 | sock_register(&unix_family_ops); | |
097e66c5 | 2915 | register_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2916 | out: |
2917 | return rc; | |
2918 | } | |
2919 | ||
2920 | static void __exit af_unix_exit(void) | |
2921 | { | |
2922 | sock_unregister(PF_UNIX); | |
1da177e4 | 2923 | proto_unregister(&unix_proto); |
097e66c5 | 2924 | unregister_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2925 | } |
2926 | ||
3d366960 DW |
2927 | /* Earlier than device_initcall() so that other drivers invoking |
2928 | request_module() don't end up in a loop when modprobe tries | |
2929 | to use a UNIX socket. But later than subsys_initcall() because | |
2930 | we depend on stuff initialised there */ | |
2931 | fs_initcall(af_unix_init); | |
1da177e4 LT |
2932 | module_exit(af_unix_exit); |
2933 | ||
2934 | MODULE_LICENSE("GPL"); | |
2935 | MODULE_ALIAS_NETPROTO(PF_UNIX); |