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 LT |
87 | #include <linux/signal.h> |
88 | #include <linux/sched.h> | |
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> | |
103 | #include <asm/uaccess.h> | |
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> |
1da177e4 | 120 | |
7123aaa3 | 121 | struct hlist_head unix_socket_table[2 * UNIX_HASH_SIZE]; |
fa7ff56f PE |
122 | EXPORT_SYMBOL_GPL(unix_socket_table); |
123 | DEFINE_SPINLOCK(unix_table_lock); | |
124 | EXPORT_SYMBOL_GPL(unix_table_lock); | |
518de9b3 | 125 | static atomic_long_t unix_nr_socks; |
1da177e4 | 126 | |
1da177e4 | 127 | |
7123aaa3 ED |
128 | static struct hlist_head *unix_sockets_unbound(void *addr) |
129 | { | |
130 | unsigned long hash = (unsigned long)addr; | |
131 | ||
132 | hash ^= hash >> 16; | |
133 | hash ^= hash >> 8; | |
134 | hash %= UNIX_HASH_SIZE; | |
135 | return &unix_socket_table[UNIX_HASH_SIZE + hash]; | |
136 | } | |
137 | ||
138 | #define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash < UNIX_HASH_SIZE) | |
1da177e4 | 139 | |
877ce7c1 | 140 | #ifdef CONFIG_SECURITY_NETWORK |
dc49c1f9 | 141 | static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 | 142 | { |
37a9a8df | 143 | UNIXCB(skb).secid = scm->secid; |
877ce7c1 CZ |
144 | } |
145 | ||
146 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
147 | { | |
37a9a8df SS |
148 | scm->secid = UNIXCB(skb).secid; |
149 | } | |
150 | ||
151 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
152 | { | |
153 | return (scm->secid == UNIXCB(skb).secid); | |
877ce7c1 CZ |
154 | } |
155 | #else | |
dc49c1f9 | 156 | static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 CZ |
157 | { } |
158 | ||
159 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
160 | { } | |
37a9a8df SS |
161 | |
162 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
163 | { | |
164 | return true; | |
165 | } | |
877ce7c1 CZ |
166 | #endif /* CONFIG_SECURITY_NETWORK */ |
167 | ||
1da177e4 LT |
168 | /* |
169 | * SMP locking strategy: | |
fbe9cc4a | 170 | * hash table is protected with spinlock unix_table_lock |
663717f6 | 171 | * each socket state is protected by separate spin lock. |
1da177e4 LT |
172 | */ |
173 | ||
95c96174 | 174 | static inline unsigned int unix_hash_fold(__wsum n) |
1da177e4 | 175 | { |
0a13404d | 176 | unsigned int hash = (__force unsigned int)csum_fold(n); |
95c96174 | 177 | |
1da177e4 LT |
178 | hash ^= hash>>8; |
179 | return hash&(UNIX_HASH_SIZE-1); | |
180 | } | |
181 | ||
182 | #define unix_peer(sk) (unix_sk(sk)->peer) | |
183 | ||
184 | static inline int unix_our_peer(struct sock *sk, struct sock *osk) | |
185 | { | |
186 | return unix_peer(osk) == sk; | |
187 | } | |
188 | ||
189 | static inline int unix_may_send(struct sock *sk, struct sock *osk) | |
190 | { | |
6eba6a37 | 191 | return unix_peer(osk) == NULL || unix_our_peer(sk, osk); |
1da177e4 LT |
192 | } |
193 | ||
3c73419c RW |
194 | static inline int unix_recvq_full(struct sock const *sk) |
195 | { | |
196 | return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog; | |
197 | } | |
198 | ||
fa7ff56f | 199 | struct sock *unix_peer_get(struct sock *s) |
1da177e4 LT |
200 | { |
201 | struct sock *peer; | |
202 | ||
1c92b4e5 | 203 | unix_state_lock(s); |
1da177e4 LT |
204 | peer = unix_peer(s); |
205 | if (peer) | |
206 | sock_hold(peer); | |
1c92b4e5 | 207 | unix_state_unlock(s); |
1da177e4 LT |
208 | return peer; |
209 | } | |
fa7ff56f | 210 | EXPORT_SYMBOL_GPL(unix_peer_get); |
1da177e4 LT |
211 | |
212 | static inline void unix_release_addr(struct unix_address *addr) | |
213 | { | |
214 | if (atomic_dec_and_test(&addr->refcnt)) | |
215 | kfree(addr); | |
216 | } | |
217 | ||
218 | /* | |
219 | * Check unix socket name: | |
220 | * - should be not zero length. | |
221 | * - if started by not zero, should be NULL terminated (FS object) | |
222 | * - if started by zero, it is abstract name. | |
223 | */ | |
ac7bfa62 | 224 | |
95c96174 | 225 | static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp) |
1da177e4 LT |
226 | { |
227 | if (len <= sizeof(short) || len > sizeof(*sunaddr)) | |
228 | return -EINVAL; | |
229 | if (!sunaddr || sunaddr->sun_family != AF_UNIX) | |
230 | return -EINVAL; | |
231 | if (sunaddr->sun_path[0]) { | |
232 | /* | |
233 | * This may look like an off by one error but it is a bit more | |
234 | * subtle. 108 is the longest valid AF_UNIX path for a binding. | |
25985edc | 235 | * sun_path[108] doesn't as such exist. However in kernel space |
1da177e4 LT |
236 | * we are guaranteed that it is a valid memory location in our |
237 | * kernel address buffer. | |
238 | */ | |
e27dfcea | 239 | ((char *)sunaddr)[len] = 0; |
1da177e4 LT |
240 | len = strlen(sunaddr->sun_path)+1+sizeof(short); |
241 | return len; | |
242 | } | |
243 | ||
07f0757a | 244 | *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0)); |
1da177e4 LT |
245 | return len; |
246 | } | |
247 | ||
248 | static void __unix_remove_socket(struct sock *sk) | |
249 | { | |
250 | sk_del_node_init(sk); | |
251 | } | |
252 | ||
253 | static void __unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
254 | { | |
547b792c | 255 | WARN_ON(!sk_unhashed(sk)); |
1da177e4 LT |
256 | sk_add_node(sk, list); |
257 | } | |
258 | ||
259 | static inline void unix_remove_socket(struct sock *sk) | |
260 | { | |
fbe9cc4a | 261 | spin_lock(&unix_table_lock); |
1da177e4 | 262 | __unix_remove_socket(sk); |
fbe9cc4a | 263 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
264 | } |
265 | ||
266 | static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
267 | { | |
fbe9cc4a | 268 | spin_lock(&unix_table_lock); |
1da177e4 | 269 | __unix_insert_socket(list, sk); |
fbe9cc4a | 270 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
271 | } |
272 | ||
097e66c5 DL |
273 | static struct sock *__unix_find_socket_byname(struct net *net, |
274 | struct sockaddr_un *sunname, | |
95c96174 | 275 | int len, int type, unsigned int hash) |
1da177e4 LT |
276 | { |
277 | struct sock *s; | |
1da177e4 | 278 | |
b67bfe0d | 279 | sk_for_each(s, &unix_socket_table[hash ^ type]) { |
1da177e4 LT |
280 | struct unix_sock *u = unix_sk(s); |
281 | ||
878628fb | 282 | if (!net_eq(sock_net(s), net)) |
097e66c5 DL |
283 | continue; |
284 | ||
1da177e4 LT |
285 | if (u->addr->len == len && |
286 | !memcmp(u->addr->name, sunname, len)) | |
287 | goto found; | |
288 | } | |
289 | s = NULL; | |
290 | found: | |
291 | return s; | |
292 | } | |
293 | ||
097e66c5 DL |
294 | static inline struct sock *unix_find_socket_byname(struct net *net, |
295 | struct sockaddr_un *sunname, | |
1da177e4 | 296 | int len, int type, |
95c96174 | 297 | unsigned int hash) |
1da177e4 LT |
298 | { |
299 | struct sock *s; | |
300 | ||
fbe9cc4a | 301 | spin_lock(&unix_table_lock); |
097e66c5 | 302 | s = __unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
303 | if (s) |
304 | sock_hold(s); | |
fbe9cc4a | 305 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
306 | return s; |
307 | } | |
308 | ||
6616f788 | 309 | static struct sock *unix_find_socket_byinode(struct inode *i) |
1da177e4 LT |
310 | { |
311 | struct sock *s; | |
1da177e4 | 312 | |
fbe9cc4a | 313 | spin_lock(&unix_table_lock); |
b67bfe0d | 314 | sk_for_each(s, |
1da177e4 | 315 | &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) { |
40ffe67d | 316 | struct dentry *dentry = unix_sk(s)->path.dentry; |
1da177e4 | 317 | |
eb0a4a47 | 318 | if (dentry && d_real_inode(dentry) == i) { |
1da177e4 LT |
319 | sock_hold(s); |
320 | goto found; | |
321 | } | |
322 | } | |
323 | s = NULL; | |
324 | found: | |
fbe9cc4a | 325 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
326 | return s; |
327 | } | |
328 | ||
7d267278 RW |
329 | /* Support code for asymmetrically connected dgram sockets |
330 | * | |
331 | * If a datagram socket is connected to a socket not itself connected | |
332 | * to the first socket (eg, /dev/log), clients may only enqueue more | |
333 | * messages if the present receive queue of the server socket is not | |
334 | * "too large". This means there's a second writeability condition | |
335 | * poll and sendmsg need to test. The dgram recv code will do a wake | |
336 | * up on the peer_wait wait queue of a socket upon reception of a | |
337 | * datagram which needs to be propagated to sleeping would-be writers | |
338 | * since these might not have sent anything so far. This can't be | |
339 | * accomplished via poll_wait because the lifetime of the server | |
340 | * socket might be less than that of its clients if these break their | |
341 | * association with it or if the server socket is closed while clients | |
342 | * are still connected to it and there's no way to inform "a polling | |
343 | * implementation" that it should let go of a certain wait queue | |
344 | * | |
345 | * In order to propagate a wake up, a wait_queue_t of the client | |
346 | * socket is enqueued on the peer_wait queue of the server socket | |
347 | * whose wake function does a wake_up on the ordinary client socket | |
348 | * wait queue. This connection is established whenever a write (or | |
349 | * poll for write) hit the flow control condition and broken when the | |
350 | * association to the server socket is dissolved or after a wake up | |
351 | * was relayed. | |
352 | */ | |
353 | ||
354 | static int unix_dgram_peer_wake_relay(wait_queue_t *q, unsigned mode, int flags, | |
355 | void *key) | |
356 | { | |
357 | struct unix_sock *u; | |
358 | wait_queue_head_t *u_sleep; | |
359 | ||
360 | u = container_of(q, struct unix_sock, peer_wake); | |
361 | ||
362 | __remove_wait_queue(&unix_sk(u->peer_wake.private)->peer_wait, | |
363 | q); | |
364 | u->peer_wake.private = NULL; | |
365 | ||
366 | /* relaying can only happen while the wq still exists */ | |
367 | u_sleep = sk_sleep(&u->sk); | |
368 | if (u_sleep) | |
369 | wake_up_interruptible_poll(u_sleep, key); | |
370 | ||
371 | return 0; | |
372 | } | |
373 | ||
374 | static int unix_dgram_peer_wake_connect(struct sock *sk, struct sock *other) | |
375 | { | |
376 | struct unix_sock *u, *u_other; | |
377 | int rc; | |
378 | ||
379 | u = unix_sk(sk); | |
380 | u_other = unix_sk(other); | |
381 | rc = 0; | |
382 | spin_lock(&u_other->peer_wait.lock); | |
383 | ||
384 | if (!u->peer_wake.private) { | |
385 | u->peer_wake.private = other; | |
386 | __add_wait_queue(&u_other->peer_wait, &u->peer_wake); | |
387 | ||
388 | rc = 1; | |
389 | } | |
390 | ||
391 | spin_unlock(&u_other->peer_wait.lock); | |
392 | return rc; | |
393 | } | |
394 | ||
395 | static void unix_dgram_peer_wake_disconnect(struct sock *sk, | |
396 | struct sock *other) | |
397 | { | |
398 | struct unix_sock *u, *u_other; | |
399 | ||
400 | u = unix_sk(sk); | |
401 | u_other = unix_sk(other); | |
402 | spin_lock(&u_other->peer_wait.lock); | |
403 | ||
404 | if (u->peer_wake.private == other) { | |
405 | __remove_wait_queue(&u_other->peer_wait, &u->peer_wake); | |
406 | u->peer_wake.private = NULL; | |
407 | } | |
408 | ||
409 | spin_unlock(&u_other->peer_wait.lock); | |
410 | } | |
411 | ||
412 | static void unix_dgram_peer_wake_disconnect_wakeup(struct sock *sk, | |
413 | struct sock *other) | |
414 | { | |
415 | unix_dgram_peer_wake_disconnect(sk, other); | |
416 | wake_up_interruptible_poll(sk_sleep(sk), | |
417 | POLLOUT | | |
418 | POLLWRNORM | | |
419 | POLLWRBAND); | |
420 | } | |
421 | ||
422 | /* preconditions: | |
423 | * - unix_peer(sk) == other | |
424 | * - association is stable | |
425 | */ | |
426 | static int unix_dgram_peer_wake_me(struct sock *sk, struct sock *other) | |
427 | { | |
428 | int connected; | |
429 | ||
430 | connected = unix_dgram_peer_wake_connect(sk, other); | |
431 | ||
432 | if (unix_recvq_full(other)) | |
433 | return 1; | |
434 | ||
435 | if (connected) | |
436 | unix_dgram_peer_wake_disconnect(sk, other); | |
437 | ||
438 | return 0; | |
439 | } | |
440 | ||
1586a587 | 441 | static int unix_writable(const struct sock *sk) |
1da177e4 | 442 | { |
1586a587 ED |
443 | return sk->sk_state != TCP_LISTEN && |
444 | (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf; | |
1da177e4 LT |
445 | } |
446 | ||
447 | static void unix_write_space(struct sock *sk) | |
448 | { | |
43815482 ED |
449 | struct socket_wq *wq; |
450 | ||
451 | rcu_read_lock(); | |
1da177e4 | 452 | if (unix_writable(sk)) { |
43815482 | 453 | wq = rcu_dereference(sk->sk_wq); |
1ce0bf50 | 454 | if (skwq_has_sleeper(wq)) |
67426b75 ED |
455 | wake_up_interruptible_sync_poll(&wq->wait, |
456 | POLLOUT | POLLWRNORM | POLLWRBAND); | |
8d8ad9d7 | 457 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
1da177e4 | 458 | } |
43815482 | 459 | rcu_read_unlock(); |
1da177e4 LT |
460 | } |
461 | ||
462 | /* When dgram socket disconnects (or changes its peer), we clear its receive | |
463 | * queue of packets arrived from previous peer. First, it allows to do | |
464 | * flow control based only on wmem_alloc; second, sk connected to peer | |
465 | * may receive messages only from that peer. */ | |
466 | static void unix_dgram_disconnected(struct sock *sk, struct sock *other) | |
467 | { | |
b03efcfb | 468 | if (!skb_queue_empty(&sk->sk_receive_queue)) { |
1da177e4 LT |
469 | skb_queue_purge(&sk->sk_receive_queue); |
470 | wake_up_interruptible_all(&unix_sk(sk)->peer_wait); | |
471 | ||
472 | /* If one link of bidirectional dgram pipe is disconnected, | |
473 | * we signal error. Messages are lost. Do not make this, | |
474 | * when peer was not connected to us. | |
475 | */ | |
476 | if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) { | |
477 | other->sk_err = ECONNRESET; | |
478 | other->sk_error_report(other); | |
479 | } | |
480 | } | |
481 | } | |
482 | ||
483 | static void unix_sock_destructor(struct sock *sk) | |
484 | { | |
485 | struct unix_sock *u = unix_sk(sk); | |
486 | ||
487 | skb_queue_purge(&sk->sk_receive_queue); | |
488 | ||
547b792c IJ |
489 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); |
490 | WARN_ON(!sk_unhashed(sk)); | |
491 | WARN_ON(sk->sk_socket); | |
1da177e4 | 492 | if (!sock_flag(sk, SOCK_DEAD)) { |
5cc208be | 493 | pr_info("Attempt to release alive unix socket: %p\n", sk); |
1da177e4 LT |
494 | return; |
495 | } | |
496 | ||
497 | if (u->addr) | |
498 | unix_release_addr(u->addr); | |
499 | ||
518de9b3 | 500 | atomic_long_dec(&unix_nr_socks); |
6f756a8c | 501 | local_bh_disable(); |
a8076d8d | 502 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
6f756a8c | 503 | local_bh_enable(); |
1da177e4 | 504 | #ifdef UNIX_REFCNT_DEBUG |
5cc208be | 505 | pr_debug("UNIX %p is destroyed, %ld are still alive.\n", sk, |
518de9b3 | 506 | atomic_long_read(&unix_nr_socks)); |
1da177e4 LT |
507 | #endif |
508 | } | |
509 | ||
ded34e0f | 510 | static void unix_release_sock(struct sock *sk, int embrion) |
1da177e4 LT |
511 | { |
512 | struct unix_sock *u = unix_sk(sk); | |
40ffe67d | 513 | struct path path; |
1da177e4 LT |
514 | struct sock *skpair; |
515 | struct sk_buff *skb; | |
516 | int state; | |
517 | ||
518 | unix_remove_socket(sk); | |
519 | ||
520 | /* Clear state */ | |
1c92b4e5 | 521 | unix_state_lock(sk); |
1da177e4 LT |
522 | sock_orphan(sk); |
523 | sk->sk_shutdown = SHUTDOWN_MASK; | |
40ffe67d AV |
524 | path = u->path; |
525 | u->path.dentry = NULL; | |
526 | u->path.mnt = NULL; | |
1da177e4 LT |
527 | state = sk->sk_state; |
528 | sk->sk_state = TCP_CLOSE; | |
1c92b4e5 | 529 | unix_state_unlock(sk); |
1da177e4 LT |
530 | |
531 | wake_up_interruptible_all(&u->peer_wait); | |
532 | ||
e27dfcea | 533 | skpair = unix_peer(sk); |
1da177e4 | 534 | |
e27dfcea | 535 | if (skpair != NULL) { |
1da177e4 | 536 | if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) { |
1c92b4e5 | 537 | unix_state_lock(skpair); |
1da177e4 LT |
538 | /* No more writes */ |
539 | skpair->sk_shutdown = SHUTDOWN_MASK; | |
540 | if (!skb_queue_empty(&sk->sk_receive_queue) || embrion) | |
541 | skpair->sk_err = ECONNRESET; | |
1c92b4e5 | 542 | unix_state_unlock(skpair); |
1da177e4 | 543 | skpair->sk_state_change(skpair); |
8d8ad9d7 | 544 | sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP); |
1da177e4 | 545 | } |
7d267278 RW |
546 | |
547 | unix_dgram_peer_wake_disconnect(sk, skpair); | |
1da177e4 LT |
548 | sock_put(skpair); /* It may now die */ |
549 | unix_peer(sk) = NULL; | |
550 | } | |
551 | ||
552 | /* Try to flush out this socket. Throw out buffers at least */ | |
553 | ||
554 | while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) { | |
e27dfcea | 555 | if (state == TCP_LISTEN) |
1da177e4 LT |
556 | unix_release_sock(skb->sk, 1); |
557 | /* passed fds are erased in the kfree_skb hook */ | |
73ed5d25 | 558 | UNIXCB(skb).consumed = skb->len; |
1da177e4 LT |
559 | kfree_skb(skb); |
560 | } | |
561 | ||
40ffe67d AV |
562 | if (path.dentry) |
563 | path_put(&path); | |
1da177e4 LT |
564 | |
565 | sock_put(sk); | |
566 | ||
567 | /* ---- Socket is dead now and most probably destroyed ---- */ | |
568 | ||
569 | /* | |
e04dae84 | 570 | * Fixme: BSD difference: In BSD all sockets connected to us get |
1da177e4 LT |
571 | * ECONNRESET and we die on the spot. In Linux we behave |
572 | * like files and pipes do and wait for the last | |
573 | * dereference. | |
574 | * | |
575 | * Can't we simply set sock->err? | |
576 | * | |
577 | * What the above comment does talk about? --ANK(980817) | |
578 | */ | |
579 | ||
9305cfa4 | 580 | if (unix_tot_inflight) |
ac7bfa62 | 581 | unix_gc(); /* Garbage collect fds */ |
1da177e4 LT |
582 | } |
583 | ||
109f6e39 EB |
584 | static void init_peercred(struct sock *sk) |
585 | { | |
586 | put_pid(sk->sk_peer_pid); | |
587 | if (sk->sk_peer_cred) | |
588 | put_cred(sk->sk_peer_cred); | |
589 | sk->sk_peer_pid = get_pid(task_tgid(current)); | |
590 | sk->sk_peer_cred = get_current_cred(); | |
591 | } | |
592 | ||
593 | static void copy_peercred(struct sock *sk, struct sock *peersk) | |
594 | { | |
595 | put_pid(sk->sk_peer_pid); | |
596 | if (sk->sk_peer_cred) | |
597 | put_cred(sk->sk_peer_cred); | |
598 | sk->sk_peer_pid = get_pid(peersk->sk_peer_pid); | |
599 | sk->sk_peer_cred = get_cred(peersk->sk_peer_cred); | |
600 | } | |
601 | ||
1da177e4 LT |
602 | static int unix_listen(struct socket *sock, int backlog) |
603 | { | |
604 | int err; | |
605 | struct sock *sk = sock->sk; | |
606 | struct unix_sock *u = unix_sk(sk); | |
109f6e39 | 607 | struct pid *old_pid = NULL; |
1da177e4 LT |
608 | |
609 | err = -EOPNOTSUPP; | |
6eba6a37 ED |
610 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
611 | goto out; /* Only stream/seqpacket sockets accept */ | |
1da177e4 LT |
612 | err = -EINVAL; |
613 | if (!u->addr) | |
6eba6a37 | 614 | goto out; /* No listens on an unbound socket */ |
1c92b4e5 | 615 | unix_state_lock(sk); |
1da177e4 LT |
616 | if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN) |
617 | goto out_unlock; | |
618 | if (backlog > sk->sk_max_ack_backlog) | |
619 | wake_up_interruptible_all(&u->peer_wait); | |
620 | sk->sk_max_ack_backlog = backlog; | |
621 | sk->sk_state = TCP_LISTEN; | |
622 | /* set credentials so connect can copy them */ | |
109f6e39 | 623 | init_peercred(sk); |
1da177e4 LT |
624 | err = 0; |
625 | ||
626 | out_unlock: | |
1c92b4e5 | 627 | unix_state_unlock(sk); |
109f6e39 | 628 | put_pid(old_pid); |
1da177e4 LT |
629 | out: |
630 | return err; | |
631 | } | |
632 | ||
633 | static int unix_release(struct socket *); | |
634 | static int unix_bind(struct socket *, struct sockaddr *, int); | |
635 | static int unix_stream_connect(struct socket *, struct sockaddr *, | |
636 | int addr_len, int flags); | |
637 | static int unix_socketpair(struct socket *, struct socket *); | |
638 | static int unix_accept(struct socket *, struct socket *, int); | |
639 | static int unix_getname(struct socket *, struct sockaddr *, int *, int); | |
640 | static unsigned int unix_poll(struct file *, struct socket *, poll_table *); | |
ec0d215f RW |
641 | static unsigned int unix_dgram_poll(struct file *, struct socket *, |
642 | poll_table *); | |
1da177e4 LT |
643 | static int unix_ioctl(struct socket *, unsigned int, unsigned long); |
644 | static int unix_shutdown(struct socket *, int); | |
1b784140 YX |
645 | static int unix_stream_sendmsg(struct socket *, struct msghdr *, size_t); |
646 | static int unix_stream_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
869e7c62 HFS |
647 | static ssize_t unix_stream_sendpage(struct socket *, struct page *, int offset, |
648 | size_t size, int flags); | |
2b514574 HFS |
649 | static ssize_t unix_stream_splice_read(struct socket *, loff_t *ppos, |
650 | struct pipe_inode_info *, size_t size, | |
651 | unsigned int flags); | |
1b784140 YX |
652 | static int unix_dgram_sendmsg(struct socket *, struct msghdr *, size_t); |
653 | static int unix_dgram_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
1da177e4 LT |
654 | static int unix_dgram_connect(struct socket *, struct sockaddr *, |
655 | int, int); | |
1b784140 YX |
656 | static int unix_seqpacket_sendmsg(struct socket *, struct msghdr *, size_t); |
657 | static int unix_seqpacket_recvmsg(struct socket *, struct msghdr *, size_t, | |
658 | int); | |
1da177e4 | 659 | |
12663bfc | 660 | static int unix_set_peek_off(struct sock *sk, int val) |
f55bb7f9 PE |
661 | { |
662 | struct unix_sock *u = unix_sk(sk); | |
663 | ||
12663bfc SL |
664 | if (mutex_lock_interruptible(&u->readlock)) |
665 | return -EINTR; | |
666 | ||
f55bb7f9 PE |
667 | sk->sk_peek_off = val; |
668 | mutex_unlock(&u->readlock); | |
12663bfc SL |
669 | |
670 | return 0; | |
f55bb7f9 PE |
671 | } |
672 | ||
673 | ||
90ddc4f0 | 674 | static const struct proto_ops unix_stream_ops = { |
1da177e4 LT |
675 | .family = PF_UNIX, |
676 | .owner = THIS_MODULE, | |
677 | .release = unix_release, | |
678 | .bind = unix_bind, | |
679 | .connect = unix_stream_connect, | |
680 | .socketpair = unix_socketpair, | |
681 | .accept = unix_accept, | |
682 | .getname = unix_getname, | |
683 | .poll = unix_poll, | |
684 | .ioctl = unix_ioctl, | |
685 | .listen = unix_listen, | |
686 | .shutdown = unix_shutdown, | |
687 | .setsockopt = sock_no_setsockopt, | |
688 | .getsockopt = sock_no_getsockopt, | |
689 | .sendmsg = unix_stream_sendmsg, | |
690 | .recvmsg = unix_stream_recvmsg, | |
691 | .mmap = sock_no_mmap, | |
869e7c62 | 692 | .sendpage = unix_stream_sendpage, |
2b514574 | 693 | .splice_read = unix_stream_splice_read, |
fc0d7536 | 694 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
695 | }; |
696 | ||
90ddc4f0 | 697 | static const struct proto_ops unix_dgram_ops = { |
1da177e4 LT |
698 | .family = PF_UNIX, |
699 | .owner = THIS_MODULE, | |
700 | .release = unix_release, | |
701 | .bind = unix_bind, | |
702 | .connect = unix_dgram_connect, | |
703 | .socketpair = unix_socketpair, | |
704 | .accept = sock_no_accept, | |
705 | .getname = unix_getname, | |
ec0d215f | 706 | .poll = unix_dgram_poll, |
1da177e4 LT |
707 | .ioctl = unix_ioctl, |
708 | .listen = sock_no_listen, | |
709 | .shutdown = unix_shutdown, | |
710 | .setsockopt = sock_no_setsockopt, | |
711 | .getsockopt = sock_no_getsockopt, | |
712 | .sendmsg = unix_dgram_sendmsg, | |
713 | .recvmsg = unix_dgram_recvmsg, | |
714 | .mmap = sock_no_mmap, | |
715 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 716 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
717 | }; |
718 | ||
90ddc4f0 | 719 | static const struct proto_ops unix_seqpacket_ops = { |
1da177e4 LT |
720 | .family = PF_UNIX, |
721 | .owner = THIS_MODULE, | |
722 | .release = unix_release, | |
723 | .bind = unix_bind, | |
724 | .connect = unix_stream_connect, | |
725 | .socketpair = unix_socketpair, | |
726 | .accept = unix_accept, | |
727 | .getname = unix_getname, | |
ec0d215f | 728 | .poll = unix_dgram_poll, |
1da177e4 LT |
729 | .ioctl = unix_ioctl, |
730 | .listen = unix_listen, | |
731 | .shutdown = unix_shutdown, | |
732 | .setsockopt = sock_no_setsockopt, | |
733 | .getsockopt = sock_no_getsockopt, | |
734 | .sendmsg = unix_seqpacket_sendmsg, | |
a05d2ad1 | 735 | .recvmsg = unix_seqpacket_recvmsg, |
1da177e4 LT |
736 | .mmap = sock_no_mmap, |
737 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 738 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
739 | }; |
740 | ||
741 | static struct proto unix_proto = { | |
248969ae ED |
742 | .name = "UNIX", |
743 | .owner = THIS_MODULE, | |
248969ae | 744 | .obj_size = sizeof(struct unix_sock), |
1da177e4 LT |
745 | }; |
746 | ||
a09785a2 IM |
747 | /* |
748 | * AF_UNIX sockets do not interact with hardware, hence they | |
749 | * dont trigger interrupts - so it's safe for them to have | |
750 | * bh-unsafe locking for their sk_receive_queue.lock. Split off | |
751 | * this special lock-class by reinitializing the spinlock key: | |
752 | */ | |
753 | static struct lock_class_key af_unix_sk_receive_queue_lock_key; | |
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); |
a09785a2 IM |
769 | lockdep_set_class(&sk->sk_receive_queue.lock, |
770 | &af_unix_sk_receive_queue_lock_key); | |
1da177e4 | 771 | |
3aa9799e | 772 | sk->sk_allocation = GFP_KERNEL_ACCOUNT; |
1da177e4 | 773 | sk->sk_write_space = unix_write_space; |
a0a53c8b | 774 | sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen; |
1da177e4 LT |
775 | sk->sk_destruct = unix_sock_destructor; |
776 | u = unix_sk(sk); | |
40ffe67d AV |
777 | u->path.dentry = NULL; |
778 | u->path.mnt = NULL; | |
fd19f329 | 779 | spin_lock_init(&u->lock); |
516e0cc5 | 780 | atomic_long_set(&u->inflight, 0); |
1fd05ba5 | 781 | INIT_LIST_HEAD(&u->link); |
57b47a53 | 782 | mutex_init(&u->readlock); /* single task reading lock */ |
1da177e4 | 783 | init_waitqueue_head(&u->peer_wait); |
7d267278 | 784 | init_waitqueue_func_entry(&u->peer_wake, unix_dgram_peer_wake_relay); |
7123aaa3 | 785 | unix_insert_socket(unix_sockets_unbound(sk), sk); |
1da177e4 | 786 | out: |
284b327b | 787 | if (sk == NULL) |
518de9b3 | 788 | atomic_long_dec(&unix_nr_socks); |
920de804 ED |
789 | else { |
790 | local_bh_disable(); | |
a8076d8d | 791 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
920de804 ED |
792 | local_bh_enable(); |
793 | } | |
1da177e4 LT |
794 | return sk; |
795 | } | |
796 | ||
3f378b68 EP |
797 | static int unix_create(struct net *net, struct socket *sock, int protocol, |
798 | int kern) | |
1da177e4 LT |
799 | { |
800 | if (protocol && protocol != PF_UNIX) | |
801 | return -EPROTONOSUPPORT; | |
802 | ||
803 | sock->state = SS_UNCONNECTED; | |
804 | ||
805 | switch (sock->type) { | |
806 | case SOCK_STREAM: | |
807 | sock->ops = &unix_stream_ops; | |
808 | break; | |
809 | /* | |
810 | * Believe it or not BSD has AF_UNIX, SOCK_RAW though | |
811 | * nothing uses it. | |
812 | */ | |
813 | case SOCK_RAW: | |
e27dfcea | 814 | sock->type = SOCK_DGRAM; |
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 | |
37ab4fa7 SL |
851 | err = mutex_lock_interruptible(&u->readlock); |
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 LT |
864 | addr->name->sun_family = AF_UNIX; |
865 | atomic_set(&addr->refcnt, 1); | |
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 | ||
57b47a53 | 898 | out: mutex_unlock(&u->readlock); |
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; | |
eb0a4a47 | 915 | inode = d_real_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 | ||
d3607752 | 957 | static int unix_mknod(struct dentry *dentry, const struct path *path, umode_t mode, |
c845acb3 | 958 | struct path *res) |
faf02010 | 959 | { |
c845acb3 | 960 | int err; |
faf02010 | 961 | |
c845acb3 | 962 | err = security_path_mknod(path, dentry, mode, 0); |
faf02010 | 963 | if (!err) { |
c845acb3 | 964 | err = vfs_mknod(d_inode(path->dentry), dentry, mode, 0); |
faf02010 | 965 | if (!err) { |
c845acb3 | 966 | res->mnt = mntget(path->mnt); |
faf02010 AV |
967 | res->dentry = dget(dentry); |
968 | } | |
969 | } | |
c845acb3 | 970 | |
faf02010 AV |
971 | return err; |
972 | } | |
1da177e4 LT |
973 | |
974 | static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
975 | { | |
976 | struct sock *sk = sock->sk; | |
3b1e0a65 | 977 | struct net *net = sock_net(sk); |
1da177e4 | 978 | struct unix_sock *u = unix_sk(sk); |
e27dfcea | 979 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
dae6ad8f | 980 | char *sun_path = sunaddr->sun_path; |
c845acb3 | 981 | int err, name_err; |
95c96174 | 982 | unsigned int hash; |
1da177e4 LT |
983 | struct unix_address *addr; |
984 | struct hlist_head *list; | |
c845acb3 RW |
985 | struct path path; |
986 | struct dentry *dentry; | |
1da177e4 LT |
987 | |
988 | err = -EINVAL; | |
989 | if (sunaddr->sun_family != AF_UNIX) | |
990 | goto out; | |
991 | ||
e27dfcea | 992 | if (addr_len == sizeof(short)) { |
1da177e4 LT |
993 | err = unix_autobind(sock); |
994 | goto out; | |
995 | } | |
996 | ||
997 | err = unix_mkname(sunaddr, addr_len, &hash); | |
998 | if (err < 0) | |
999 | goto out; | |
1000 | addr_len = err; | |
1001 | ||
c845acb3 RW |
1002 | name_err = 0; |
1003 | dentry = NULL; | |
1004 | if (sun_path[0]) { | |
1005 | /* Get the parent directory, calculate the hash for last | |
1006 | * component. | |
1007 | */ | |
1008 | dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0); | |
1009 | ||
1010 | if (IS_ERR(dentry)) { | |
1011 | /* delay report until after 'already bound' check */ | |
1012 | name_err = PTR_ERR(dentry); | |
1013 | dentry = NULL; | |
1014 | } | |
1015 | } | |
1016 | ||
37ab4fa7 SL |
1017 | err = mutex_lock_interruptible(&u->readlock); |
1018 | if (err) | |
c845acb3 | 1019 | goto out_path; |
1da177e4 LT |
1020 | |
1021 | err = -EINVAL; | |
1022 | if (u->addr) | |
1023 | goto out_up; | |
1024 | ||
c845acb3 RW |
1025 | if (name_err) { |
1026 | err = name_err == -EEXIST ? -EADDRINUSE : name_err; | |
1027 | goto out_up; | |
1028 | } | |
1029 | ||
1da177e4 LT |
1030 | err = -ENOMEM; |
1031 | addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL); | |
1032 | if (!addr) | |
1033 | goto out_up; | |
1034 | ||
1035 | memcpy(addr->name, sunaddr, addr_len); | |
1036 | addr->len = addr_len; | |
1037 | addr->hash = hash ^ sk->sk_type; | |
1038 | atomic_set(&addr->refcnt, 1); | |
1039 | ||
c845acb3 RW |
1040 | if (dentry) { |
1041 | struct path u_path; | |
faf02010 | 1042 | umode_t mode = S_IFSOCK | |
ce3b0f8d | 1043 | (SOCK_INODE(sock)->i_mode & ~current_umask()); |
c845acb3 | 1044 | err = unix_mknod(dentry, &path, mode, &u_path); |
faf02010 AV |
1045 | if (err) { |
1046 | if (err == -EEXIST) | |
1047 | err = -EADDRINUSE; | |
1048 | unix_release_addr(addr); | |
1049 | goto out_up; | |
1050 | } | |
1da177e4 | 1051 | addr->hash = UNIX_HASH_SIZE; |
eb0a4a47 | 1052 | hash = d_real_inode(dentry)->i_ino & (UNIX_HASH_SIZE - 1); |
faf02010 | 1053 | spin_lock(&unix_table_lock); |
c845acb3 | 1054 | u->path = u_path; |
faf02010 AV |
1055 | list = &unix_socket_table[hash]; |
1056 | } else { | |
1057 | spin_lock(&unix_table_lock); | |
1da177e4 | 1058 | err = -EADDRINUSE; |
097e66c5 | 1059 | if (__unix_find_socket_byname(net, sunaddr, addr_len, |
1da177e4 LT |
1060 | sk->sk_type, hash)) { |
1061 | unix_release_addr(addr); | |
1062 | goto out_unlock; | |
1063 | } | |
1064 | ||
1065 | list = &unix_socket_table[addr->hash]; | |
1da177e4 LT |
1066 | } |
1067 | ||
1068 | err = 0; | |
1069 | __unix_remove_socket(sk); | |
1070 | u->addr = addr; | |
1071 | __unix_insert_socket(list, sk); | |
1072 | ||
1073 | out_unlock: | |
fbe9cc4a | 1074 | spin_unlock(&unix_table_lock); |
1da177e4 | 1075 | out_up: |
57b47a53 | 1076 | mutex_unlock(&u->readlock); |
c845acb3 RW |
1077 | out_path: |
1078 | if (dentry) | |
1079 | done_path_create(&path, dentry); | |
1080 | ||
1da177e4 LT |
1081 | out: |
1082 | return err; | |
1da177e4 LT |
1083 | } |
1084 | ||
278a3de5 DM |
1085 | static void unix_state_double_lock(struct sock *sk1, struct sock *sk2) |
1086 | { | |
1087 | if (unlikely(sk1 == sk2) || !sk2) { | |
1088 | unix_state_lock(sk1); | |
1089 | return; | |
1090 | } | |
1091 | if (sk1 < sk2) { | |
1092 | unix_state_lock(sk1); | |
1093 | unix_state_lock_nested(sk2); | |
1094 | } else { | |
1095 | unix_state_lock(sk2); | |
1096 | unix_state_lock_nested(sk1); | |
1097 | } | |
1098 | } | |
1099 | ||
1100 | static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2) | |
1101 | { | |
1102 | if (unlikely(sk1 == sk2) || !sk2) { | |
1103 | unix_state_unlock(sk1); | |
1104 | return; | |
1105 | } | |
1106 | unix_state_unlock(sk1); | |
1107 | unix_state_unlock(sk2); | |
1108 | } | |
1109 | ||
1da177e4 LT |
1110 | static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr, |
1111 | int alen, int flags) | |
1112 | { | |
1113 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1114 | struct net *net = sock_net(sk); |
e27dfcea | 1115 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr; |
1da177e4 | 1116 | struct sock *other; |
95c96174 | 1117 | unsigned int hash; |
1da177e4 LT |
1118 | int err; |
1119 | ||
1120 | if (addr->sa_family != AF_UNSPEC) { | |
1121 | err = unix_mkname(sunaddr, alen, &hash); | |
1122 | if (err < 0) | |
1123 | goto out; | |
1124 | alen = err; | |
1125 | ||
1126 | if (test_bit(SOCK_PASSCRED, &sock->flags) && | |
1127 | !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0) | |
1128 | goto out; | |
1129 | ||
278a3de5 | 1130 | restart: |
e27dfcea | 1131 | other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err); |
1da177e4 LT |
1132 | if (!other) |
1133 | goto out; | |
1134 | ||
278a3de5 DM |
1135 | unix_state_double_lock(sk, other); |
1136 | ||
1137 | /* Apparently VFS overslept socket death. Retry. */ | |
1138 | if (sock_flag(other, SOCK_DEAD)) { | |
1139 | unix_state_double_unlock(sk, other); | |
1140 | sock_put(other); | |
1141 | goto restart; | |
1142 | } | |
1da177e4 LT |
1143 | |
1144 | err = -EPERM; | |
1145 | if (!unix_may_send(sk, other)) | |
1146 | goto out_unlock; | |
1147 | ||
1148 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1149 | if (err) | |
1150 | goto out_unlock; | |
1151 | ||
1152 | } else { | |
1153 | /* | |
1154 | * 1003.1g breaking connected state with AF_UNSPEC | |
1155 | */ | |
1156 | other = NULL; | |
278a3de5 | 1157 | unix_state_double_lock(sk, other); |
1da177e4 LT |
1158 | } |
1159 | ||
1160 | /* | |
1161 | * If it was connected, reconnect. | |
1162 | */ | |
1163 | if (unix_peer(sk)) { | |
1164 | struct sock *old_peer = unix_peer(sk); | |
e27dfcea | 1165 | unix_peer(sk) = other; |
7d267278 RW |
1166 | unix_dgram_peer_wake_disconnect_wakeup(sk, old_peer); |
1167 | ||
278a3de5 | 1168 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1169 | |
1170 | if (other != old_peer) | |
1171 | unix_dgram_disconnected(sk, old_peer); | |
1172 | sock_put(old_peer); | |
1173 | } else { | |
e27dfcea | 1174 | unix_peer(sk) = other; |
278a3de5 | 1175 | unix_state_double_unlock(sk, other); |
1da177e4 | 1176 | } |
ac7bfa62 | 1177 | return 0; |
1da177e4 LT |
1178 | |
1179 | out_unlock: | |
278a3de5 | 1180 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1181 | sock_put(other); |
1182 | out: | |
1183 | return err; | |
1184 | } | |
1185 | ||
1186 | static long unix_wait_for_peer(struct sock *other, long timeo) | |
1187 | { | |
1188 | struct unix_sock *u = unix_sk(other); | |
1189 | int sched; | |
1190 | DEFINE_WAIT(wait); | |
1191 | ||
1192 | prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE); | |
1193 | ||
1194 | sched = !sock_flag(other, SOCK_DEAD) && | |
1195 | !(other->sk_shutdown & RCV_SHUTDOWN) && | |
3c73419c | 1196 | unix_recvq_full(other); |
1da177e4 | 1197 | |
1c92b4e5 | 1198 | unix_state_unlock(other); |
1da177e4 LT |
1199 | |
1200 | if (sched) | |
1201 | timeo = schedule_timeout(timeo); | |
1202 | ||
1203 | finish_wait(&u->peer_wait, &wait); | |
1204 | return timeo; | |
1205 | } | |
1206 | ||
1207 | static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
1208 | int addr_len, int flags) | |
1209 | { | |
e27dfcea | 1210 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
1da177e4 | 1211 | struct sock *sk = sock->sk; |
3b1e0a65 | 1212 | struct net *net = sock_net(sk); |
1da177e4 LT |
1213 | struct unix_sock *u = unix_sk(sk), *newu, *otheru; |
1214 | struct sock *newsk = NULL; | |
1215 | struct sock *other = NULL; | |
1216 | struct sk_buff *skb = NULL; | |
95c96174 | 1217 | unsigned int hash; |
1da177e4 LT |
1218 | int st; |
1219 | int err; | |
1220 | long timeo; | |
1221 | ||
1222 | err = unix_mkname(sunaddr, addr_len, &hash); | |
1223 | if (err < 0) | |
1224 | goto out; | |
1225 | addr_len = err; | |
1226 | ||
f64f9e71 JP |
1227 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr && |
1228 | (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1229 | goto out; |
1230 | ||
1231 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
1232 | ||
1233 | /* First of all allocate resources. | |
1234 | If we will make it after state is locked, | |
1235 | we will have to recheck all again in any case. | |
1236 | */ | |
1237 | ||
1238 | err = -ENOMEM; | |
1239 | ||
1240 | /* create new sock for complete connection */ | |
11aa9c28 | 1241 | newsk = unix_create1(sock_net(sk), NULL, 0); |
1da177e4 LT |
1242 | if (newsk == NULL) |
1243 | goto out; | |
1244 | ||
1245 | /* Allocate skb for sending to listening sock */ | |
1246 | skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL); | |
1247 | if (skb == NULL) | |
1248 | goto out; | |
1249 | ||
1250 | restart: | |
1251 | /* Find listening sock. */ | |
097e66c5 | 1252 | other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err); |
1da177e4 LT |
1253 | if (!other) |
1254 | goto out; | |
1255 | ||
1256 | /* Latch state of peer */ | |
1c92b4e5 | 1257 | unix_state_lock(other); |
1da177e4 LT |
1258 | |
1259 | /* Apparently VFS overslept socket death. Retry. */ | |
1260 | if (sock_flag(other, SOCK_DEAD)) { | |
1c92b4e5 | 1261 | unix_state_unlock(other); |
1da177e4 LT |
1262 | sock_put(other); |
1263 | goto restart; | |
1264 | } | |
1265 | ||
1266 | err = -ECONNREFUSED; | |
1267 | if (other->sk_state != TCP_LISTEN) | |
1268 | goto out_unlock; | |
77238f2b TS |
1269 | if (other->sk_shutdown & RCV_SHUTDOWN) |
1270 | goto out_unlock; | |
1da177e4 | 1271 | |
3c73419c | 1272 | if (unix_recvq_full(other)) { |
1da177e4 LT |
1273 | err = -EAGAIN; |
1274 | if (!timeo) | |
1275 | goto out_unlock; | |
1276 | ||
1277 | timeo = unix_wait_for_peer(other, timeo); | |
1278 | ||
1279 | err = sock_intr_errno(timeo); | |
1280 | if (signal_pending(current)) | |
1281 | goto out; | |
1282 | sock_put(other); | |
1283 | goto restart; | |
ac7bfa62 | 1284 | } |
1da177e4 LT |
1285 | |
1286 | /* Latch our state. | |
1287 | ||
e5537bfc | 1288 | It is tricky place. We need to grab our state lock and cannot |
1da177e4 LT |
1289 | drop lock on peer. It is dangerous because deadlock is |
1290 | possible. Connect to self case and simultaneous | |
1291 | attempt to connect are eliminated by checking socket | |
1292 | state. other is TCP_LISTEN, if sk is TCP_LISTEN we | |
1293 | check this before attempt to grab lock. | |
1294 | ||
1295 | Well, and we have to recheck the state after socket locked. | |
1296 | */ | |
1297 | st = sk->sk_state; | |
1298 | ||
1299 | switch (st) { | |
1300 | case TCP_CLOSE: | |
1301 | /* This is ok... continue with connect */ | |
1302 | break; | |
1303 | case TCP_ESTABLISHED: | |
1304 | /* Socket is already connected */ | |
1305 | err = -EISCONN; | |
1306 | goto out_unlock; | |
1307 | default: | |
1308 | err = -EINVAL; | |
1309 | goto out_unlock; | |
1310 | } | |
1311 | ||
1c92b4e5 | 1312 | unix_state_lock_nested(sk); |
1da177e4 LT |
1313 | |
1314 | if (sk->sk_state != st) { | |
1c92b4e5 DM |
1315 | unix_state_unlock(sk); |
1316 | unix_state_unlock(other); | |
1da177e4 LT |
1317 | sock_put(other); |
1318 | goto restart; | |
1319 | } | |
1320 | ||
3610cda5 | 1321 | err = security_unix_stream_connect(sk, other, newsk); |
1da177e4 | 1322 | if (err) { |
1c92b4e5 | 1323 | unix_state_unlock(sk); |
1da177e4 LT |
1324 | goto out_unlock; |
1325 | } | |
1326 | ||
1327 | /* The way is open! Fastly set all the necessary fields... */ | |
1328 | ||
1329 | sock_hold(sk); | |
1330 | unix_peer(newsk) = sk; | |
1331 | newsk->sk_state = TCP_ESTABLISHED; | |
1332 | newsk->sk_type = sk->sk_type; | |
109f6e39 | 1333 | init_peercred(newsk); |
1da177e4 | 1334 | newu = unix_sk(newsk); |
eaefd110 | 1335 | RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq); |
1da177e4 LT |
1336 | otheru = unix_sk(other); |
1337 | ||
1338 | /* copy address information from listening to new sock*/ | |
1339 | if (otheru->addr) { | |
1340 | atomic_inc(&otheru->addr->refcnt); | |
1341 | newu->addr = otheru->addr; | |
1342 | } | |
40ffe67d AV |
1343 | if (otheru->path.dentry) { |
1344 | path_get(&otheru->path); | |
1345 | newu->path = otheru->path; | |
1da177e4 LT |
1346 | } |
1347 | ||
1348 | /* Set credentials */ | |
109f6e39 | 1349 | copy_peercred(sk, other); |
1da177e4 | 1350 | |
1da177e4 LT |
1351 | sock->state = SS_CONNECTED; |
1352 | sk->sk_state = TCP_ESTABLISHED; | |
830a1e5c BL |
1353 | sock_hold(newsk); |
1354 | ||
4e857c58 | 1355 | smp_mb__after_atomic(); /* sock_hold() does an atomic_inc() */ |
830a1e5c | 1356 | unix_peer(sk) = newsk; |
1da177e4 | 1357 | |
1c92b4e5 | 1358 | unix_state_unlock(sk); |
1da177e4 LT |
1359 | |
1360 | /* take ten and and send info to listening sock */ | |
1361 | spin_lock(&other->sk_receive_queue.lock); | |
1362 | __skb_queue_tail(&other->sk_receive_queue, skb); | |
1da177e4 | 1363 | spin_unlock(&other->sk_receive_queue.lock); |
1c92b4e5 | 1364 | unix_state_unlock(other); |
676d2369 | 1365 | other->sk_data_ready(other); |
1da177e4 LT |
1366 | sock_put(other); |
1367 | return 0; | |
1368 | ||
1369 | out_unlock: | |
1370 | if (other) | |
1c92b4e5 | 1371 | unix_state_unlock(other); |
1da177e4 LT |
1372 | |
1373 | out: | |
40d44446 | 1374 | kfree_skb(skb); |
1da177e4 LT |
1375 | if (newsk) |
1376 | unix_release_sock(newsk, 0); | |
1377 | if (other) | |
1378 | sock_put(other); | |
1379 | return err; | |
1380 | } | |
1381 | ||
1382 | static int unix_socketpair(struct socket *socka, struct socket *sockb) | |
1383 | { | |
e27dfcea | 1384 | struct sock *ska = socka->sk, *skb = sockb->sk; |
1da177e4 LT |
1385 | |
1386 | /* Join our sockets back to back */ | |
1387 | sock_hold(ska); | |
1388 | sock_hold(skb); | |
e27dfcea JK |
1389 | unix_peer(ska) = skb; |
1390 | unix_peer(skb) = ska; | |
109f6e39 EB |
1391 | init_peercred(ska); |
1392 | init_peercred(skb); | |
1da177e4 LT |
1393 | |
1394 | if (ska->sk_type != SOCK_DGRAM) { | |
1395 | ska->sk_state = TCP_ESTABLISHED; | |
1396 | skb->sk_state = TCP_ESTABLISHED; | |
1397 | socka->state = SS_CONNECTED; | |
1398 | sockb->state = SS_CONNECTED; | |
1399 | } | |
1400 | return 0; | |
1401 | } | |
1402 | ||
90c6bd34 DB |
1403 | static void unix_sock_inherit_flags(const struct socket *old, |
1404 | struct socket *new) | |
1405 | { | |
1406 | if (test_bit(SOCK_PASSCRED, &old->flags)) | |
1407 | set_bit(SOCK_PASSCRED, &new->flags); | |
1408 | if (test_bit(SOCK_PASSSEC, &old->flags)) | |
1409 | set_bit(SOCK_PASSSEC, &new->flags); | |
1410 | } | |
1411 | ||
1da177e4 LT |
1412 | static int unix_accept(struct socket *sock, struct socket *newsock, int flags) |
1413 | { | |
1414 | struct sock *sk = sock->sk; | |
1415 | struct sock *tsk; | |
1416 | struct sk_buff *skb; | |
1417 | int err; | |
1418 | ||
1419 | err = -EOPNOTSUPP; | |
6eba6a37 | 1420 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
1da177e4 LT |
1421 | goto out; |
1422 | ||
1423 | err = -EINVAL; | |
1424 | if (sk->sk_state != TCP_LISTEN) | |
1425 | goto out; | |
1426 | ||
1427 | /* If socket state is TCP_LISTEN it cannot change (for now...), | |
1428 | * so that no locks are necessary. | |
1429 | */ | |
1430 | ||
1431 | skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err); | |
1432 | if (!skb) { | |
1433 | /* This means receive shutdown. */ | |
1434 | if (err == 0) | |
1435 | err = -EINVAL; | |
1436 | goto out; | |
1437 | } | |
1438 | ||
1439 | tsk = skb->sk; | |
1440 | skb_free_datagram(sk, skb); | |
1441 | wake_up_interruptible(&unix_sk(sk)->peer_wait); | |
1442 | ||
1443 | /* attach accepted sock to socket */ | |
1c92b4e5 | 1444 | unix_state_lock(tsk); |
1da177e4 | 1445 | newsock->state = SS_CONNECTED; |
90c6bd34 | 1446 | unix_sock_inherit_flags(sock, newsock); |
1da177e4 | 1447 | sock_graft(tsk, newsock); |
1c92b4e5 | 1448 | unix_state_unlock(tsk); |
1da177e4 LT |
1449 | return 0; |
1450 | ||
1451 | out: | |
1452 | return err; | |
1453 | } | |
1454 | ||
1455 | ||
1456 | static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer) | |
1457 | { | |
1458 | struct sock *sk = sock->sk; | |
1459 | struct unix_sock *u; | |
13cfa97b | 1460 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr); |
1da177e4 LT |
1461 | int err = 0; |
1462 | ||
1463 | if (peer) { | |
1464 | sk = unix_peer_get(sk); | |
1465 | ||
1466 | err = -ENOTCONN; | |
1467 | if (!sk) | |
1468 | goto out; | |
1469 | err = 0; | |
1470 | } else { | |
1471 | sock_hold(sk); | |
1472 | } | |
1473 | ||
1474 | u = unix_sk(sk); | |
1c92b4e5 | 1475 | unix_state_lock(sk); |
1da177e4 LT |
1476 | if (!u->addr) { |
1477 | sunaddr->sun_family = AF_UNIX; | |
1478 | sunaddr->sun_path[0] = 0; | |
1479 | *uaddr_len = sizeof(short); | |
1480 | } else { | |
1481 | struct unix_address *addr = u->addr; | |
1482 | ||
1483 | *uaddr_len = addr->len; | |
1484 | memcpy(sunaddr, addr->name, *uaddr_len); | |
1485 | } | |
1c92b4e5 | 1486 | unix_state_unlock(sk); |
1da177e4 LT |
1487 | sock_put(sk); |
1488 | out: | |
1489 | return err; | |
1490 | } | |
1491 | ||
1492 | static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb) | |
1493 | { | |
1494 | int i; | |
1495 | ||
1496 | scm->fp = UNIXCB(skb).fp; | |
1da177e4 LT |
1497 | UNIXCB(skb).fp = NULL; |
1498 | ||
6eba6a37 | 1499 | for (i = scm->fp->count-1; i >= 0; i--) |
415e3d3e | 1500 | unix_notinflight(scm->fp->user, scm->fp->fp[i]); |
1da177e4 LT |
1501 | } |
1502 | ||
7361c36c | 1503 | static void unix_destruct_scm(struct sk_buff *skb) |
1da177e4 LT |
1504 | { |
1505 | struct scm_cookie scm; | |
1506 | memset(&scm, 0, sizeof(scm)); | |
7361c36c | 1507 | scm.pid = UNIXCB(skb).pid; |
7361c36c EB |
1508 | if (UNIXCB(skb).fp) |
1509 | unix_detach_fds(&scm, skb); | |
1da177e4 LT |
1510 | |
1511 | /* Alas, it calls VFS */ | |
1512 | /* So fscking what? fput() had been SMP-safe since the last Summer */ | |
1513 | scm_destroy(&scm); | |
1514 | sock_wfree(skb); | |
1515 | } | |
1516 | ||
712f4aad | 1517 | /* |
1518 | * The "user->unix_inflight" variable is protected by the garbage | |
1519 | * collection lock, and we just read it locklessly here. If you go | |
1520 | * over the limit, there might be a tiny race in actually noticing | |
1521 | * it across threads. Tough. | |
1522 | */ | |
1523 | static inline bool too_many_unix_fds(struct task_struct *p) | |
1524 | { | |
1525 | struct user_struct *user = current_user(); | |
1526 | ||
1527 | if (unlikely(user->unix_inflight > task_rlimit(p, RLIMIT_NOFILE))) | |
1528 | return !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN); | |
1529 | return false; | |
1530 | } | |
1531 | ||
25888e30 ED |
1532 | #define MAX_RECURSION_LEVEL 4 |
1533 | ||
6209344f | 1534 | static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb) |
1da177e4 LT |
1535 | { |
1536 | int i; | |
25888e30 | 1537 | unsigned char max_level = 0; |
25888e30 | 1538 | |
712f4aad | 1539 | if (too_many_unix_fds(current)) |
1540 | return -ETOOMANYREFS; | |
1541 | ||
25888e30 ED |
1542 | for (i = scm->fp->count - 1; i >= 0; i--) { |
1543 | struct sock *sk = unix_get_socket(scm->fp->fp[i]); | |
1544 | ||
3575dbf2 | 1545 | if (sk) |
25888e30 ED |
1546 | max_level = max(max_level, |
1547 | unix_sk(sk)->recursion_level); | |
25888e30 ED |
1548 | } |
1549 | if (unlikely(max_level > MAX_RECURSION_LEVEL)) | |
1550 | return -ETOOMANYREFS; | |
6209344f MS |
1551 | |
1552 | /* | |
1553 | * Need to duplicate file references for the sake of garbage | |
1554 | * collection. Otherwise a socket in the fps might become a | |
1555 | * candidate for GC while the skb is not yet queued. | |
1556 | */ | |
1557 | UNIXCB(skb).fp = scm_fp_dup(scm->fp); | |
1558 | if (!UNIXCB(skb).fp) | |
1559 | return -ENOMEM; | |
1560 | ||
712f4aad | 1561 | for (i = scm->fp->count - 1; i >= 0; i--) |
415e3d3e | 1562 | unix_inflight(scm->fp->user, scm->fp->fp[i]); |
25888e30 | 1563 | return max_level; |
1da177e4 LT |
1564 | } |
1565 | ||
f78a5fda | 1566 | static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds) |
7361c36c EB |
1567 | { |
1568 | int err = 0; | |
16e57262 | 1569 | |
f78a5fda | 1570 | UNIXCB(skb).pid = get_pid(scm->pid); |
6b0ee8c0 EB |
1571 | UNIXCB(skb).uid = scm->creds.uid; |
1572 | UNIXCB(skb).gid = scm->creds.gid; | |
7361c36c | 1573 | UNIXCB(skb).fp = NULL; |
37a9a8df | 1574 | unix_get_secdata(scm, skb); |
7361c36c EB |
1575 | if (scm->fp && send_fds) |
1576 | err = unix_attach_fds(scm, skb); | |
1577 | ||
1578 | skb->destructor = unix_destruct_scm; | |
1579 | return err; | |
1580 | } | |
1581 | ||
9490f886 HFS |
1582 | static bool unix_passcred_enabled(const struct socket *sock, |
1583 | const struct sock *other) | |
1584 | { | |
1585 | return test_bit(SOCK_PASSCRED, &sock->flags) || | |
1586 | !other->sk_socket || | |
1587 | test_bit(SOCK_PASSCRED, &other->sk_socket->flags); | |
1588 | } | |
1589 | ||
16e57262 ED |
1590 | /* |
1591 | * Some apps rely on write() giving SCM_CREDENTIALS | |
1592 | * We include credentials if source or destination socket | |
1593 | * asserted SOCK_PASSCRED. | |
1594 | */ | |
1595 | static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock, | |
1596 | const struct sock *other) | |
1597 | { | |
6b0ee8c0 | 1598 | if (UNIXCB(skb).pid) |
16e57262 | 1599 | return; |
9490f886 | 1600 | if (unix_passcred_enabled(sock, other)) { |
16e57262 | 1601 | UNIXCB(skb).pid = get_pid(task_tgid(current)); |
6e0895c2 | 1602 | current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid); |
16e57262 ED |
1603 | } |
1604 | } | |
1605 | ||
9490f886 HFS |
1606 | static int maybe_init_creds(struct scm_cookie *scm, |
1607 | struct socket *socket, | |
1608 | const struct sock *other) | |
1609 | { | |
1610 | int err; | |
1611 | struct msghdr msg = { .msg_controllen = 0 }; | |
1612 | ||
1613 | err = scm_send(socket, &msg, scm, false); | |
1614 | if (err) | |
1615 | return err; | |
1616 | ||
1617 | if (unix_passcred_enabled(socket, other)) { | |
1618 | scm->pid = get_pid(task_tgid(current)); | |
1619 | current_uid_gid(&scm->creds.uid, &scm->creds.gid); | |
1620 | } | |
1621 | return err; | |
1622 | } | |
1623 | ||
1624 | static bool unix_skb_scm_eq(struct sk_buff *skb, | |
1625 | struct scm_cookie *scm) | |
1626 | { | |
1627 | const struct unix_skb_parms *u = &UNIXCB(skb); | |
1628 | ||
1629 | return u->pid == scm->pid && | |
1630 | uid_eq(u->uid, scm->creds.uid) && | |
1631 | gid_eq(u->gid, scm->creds.gid) && | |
1632 | unix_secdata_eq(scm, skb); | |
1633 | } | |
1634 | ||
1da177e4 LT |
1635 | /* |
1636 | * Send AF_UNIX data. | |
1637 | */ | |
1638 | ||
1b784140 YX |
1639 | static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg, |
1640 | size_t len) | |
1da177e4 | 1641 | { |
1da177e4 | 1642 | struct sock *sk = sock->sk; |
3b1e0a65 | 1643 | struct net *net = sock_net(sk); |
1da177e4 | 1644 | struct unix_sock *u = unix_sk(sk); |
342dfc30 | 1645 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name); |
1da177e4 LT |
1646 | struct sock *other = NULL; |
1647 | int namelen = 0; /* fake GCC */ | |
1648 | int err; | |
95c96174 | 1649 | unsigned int hash; |
f78a5fda | 1650 | struct sk_buff *skb; |
1da177e4 | 1651 | long timeo; |
7cc05662 | 1652 | struct scm_cookie scm; |
25888e30 | 1653 | int max_level; |
eb6a2481 | 1654 | int data_len = 0; |
7d267278 | 1655 | int sk_locked; |
1da177e4 | 1656 | |
5f23b734 | 1657 | wait_for_unix_gc(); |
7cc05662 | 1658 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1659 | if (err < 0) |
1660 | return err; | |
1661 | ||
1662 | err = -EOPNOTSUPP; | |
1663 | if (msg->msg_flags&MSG_OOB) | |
1664 | goto out; | |
1665 | ||
1666 | if (msg->msg_namelen) { | |
1667 | err = unix_mkname(sunaddr, msg->msg_namelen, &hash); | |
1668 | if (err < 0) | |
1669 | goto out; | |
1670 | namelen = err; | |
1671 | } else { | |
1672 | sunaddr = NULL; | |
1673 | err = -ENOTCONN; | |
1674 | other = unix_peer_get(sk); | |
1675 | if (!other) | |
1676 | goto out; | |
1677 | } | |
1678 | ||
f64f9e71 JP |
1679 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr |
1680 | && (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1681 | goto out; |
1682 | ||
1683 | err = -EMSGSIZE; | |
1684 | if (len > sk->sk_sndbuf - 32) | |
1685 | goto out; | |
1686 | ||
31ff6aa5 | 1687 | if (len > SKB_MAX_ALLOC) { |
eb6a2481 ED |
1688 | data_len = min_t(size_t, |
1689 | len - SKB_MAX_ALLOC, | |
1690 | MAX_SKB_FRAGS * PAGE_SIZE); | |
31ff6aa5 KT |
1691 | data_len = PAGE_ALIGN(data_len); |
1692 | ||
1693 | BUILD_BUG_ON(SKB_MAX_ALLOC < PAGE_SIZE); | |
1694 | } | |
eb6a2481 ED |
1695 | |
1696 | skb = sock_alloc_send_pskb(sk, len - data_len, data_len, | |
28d64271 ED |
1697 | msg->msg_flags & MSG_DONTWAIT, &err, |
1698 | PAGE_ALLOC_COSTLY_ORDER); | |
e27dfcea | 1699 | if (skb == NULL) |
1da177e4 LT |
1700 | goto out; |
1701 | ||
7cc05662 | 1702 | err = unix_scm_to_skb(&scm, skb, true); |
25888e30 | 1703 | if (err < 0) |
7361c36c | 1704 | goto out_free; |
25888e30 | 1705 | max_level = err + 1; |
877ce7c1 | 1706 | |
eb6a2481 ED |
1707 | skb_put(skb, len - data_len); |
1708 | skb->data_len = data_len; | |
1709 | skb->len = len; | |
c0371da6 | 1710 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len); |
1da177e4 LT |
1711 | if (err) |
1712 | goto out_free; | |
1713 | ||
1714 | timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); | |
1715 | ||
1716 | restart: | |
1717 | if (!other) { | |
1718 | err = -ECONNRESET; | |
1719 | if (sunaddr == NULL) | |
1720 | goto out_free; | |
1721 | ||
097e66c5 | 1722 | other = unix_find_other(net, sunaddr, namelen, sk->sk_type, |
1da177e4 | 1723 | hash, &err); |
e27dfcea | 1724 | if (other == NULL) |
1da177e4 LT |
1725 | goto out_free; |
1726 | } | |
1727 | ||
d6ae3bae AC |
1728 | if (sk_filter(other, skb) < 0) { |
1729 | /* Toss the packet but do not return any error to the sender */ | |
1730 | err = len; | |
1731 | goto out_free; | |
1732 | } | |
1733 | ||
7d267278 | 1734 | sk_locked = 0; |
1c92b4e5 | 1735 | unix_state_lock(other); |
7d267278 | 1736 | restart_locked: |
1da177e4 LT |
1737 | err = -EPERM; |
1738 | if (!unix_may_send(sk, other)) | |
1739 | goto out_unlock; | |
1740 | ||
7d267278 | 1741 | if (unlikely(sock_flag(other, SOCK_DEAD))) { |
1da177e4 LT |
1742 | /* |
1743 | * Check with 1003.1g - what should | |
1744 | * datagram error | |
1745 | */ | |
1c92b4e5 | 1746 | unix_state_unlock(other); |
1da177e4 LT |
1747 | sock_put(other); |
1748 | ||
7d267278 RW |
1749 | if (!sk_locked) |
1750 | unix_state_lock(sk); | |
1751 | ||
1da177e4 | 1752 | err = 0; |
1da177e4 | 1753 | if (unix_peer(sk) == other) { |
e27dfcea | 1754 | unix_peer(sk) = NULL; |
7d267278 RW |
1755 | unix_dgram_peer_wake_disconnect_wakeup(sk, other); |
1756 | ||
1c92b4e5 | 1757 | unix_state_unlock(sk); |
1da177e4 LT |
1758 | |
1759 | unix_dgram_disconnected(sk, other); | |
1760 | sock_put(other); | |
1761 | err = -ECONNREFUSED; | |
1762 | } else { | |
1c92b4e5 | 1763 | unix_state_unlock(sk); |
1da177e4 LT |
1764 | } |
1765 | ||
1766 | other = NULL; | |
1767 | if (err) | |
1768 | goto out_free; | |
1769 | goto restart; | |
1770 | } | |
1771 | ||
1772 | err = -EPIPE; | |
1773 | if (other->sk_shutdown & RCV_SHUTDOWN) | |
1774 | goto out_unlock; | |
1775 | ||
1776 | if (sk->sk_type != SOCK_SEQPACKET) { | |
1777 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1778 | if (err) | |
1779 | goto out_unlock; | |
1780 | } | |
1781 | ||
a5527dda RW |
1782 | /* other == sk && unix_peer(other) != sk if |
1783 | * - unix_peer(sk) == NULL, destination address bound to sk | |
1784 | * - unix_peer(sk) == sk by time of get but disconnected before lock | |
1785 | */ | |
1786 | if (other != sk && | |
1787 | unlikely(unix_peer(other) != sk && unix_recvq_full(other))) { | |
7d267278 RW |
1788 | if (timeo) { |
1789 | timeo = unix_wait_for_peer(other, timeo); | |
1790 | ||
1791 | err = sock_intr_errno(timeo); | |
1792 | if (signal_pending(current)) | |
1793 | goto out_free; | |
1794 | ||
1795 | goto restart; | |
1da177e4 LT |
1796 | } |
1797 | ||
7d267278 RW |
1798 | if (!sk_locked) { |
1799 | unix_state_unlock(other); | |
1800 | unix_state_double_lock(sk, other); | |
1801 | } | |
1da177e4 | 1802 | |
7d267278 RW |
1803 | if (unix_peer(sk) != other || |
1804 | unix_dgram_peer_wake_me(sk, other)) { | |
1805 | err = -EAGAIN; | |
1806 | sk_locked = 1; | |
1807 | goto out_unlock; | |
1808 | } | |
1da177e4 | 1809 | |
7d267278 RW |
1810 | if (!sk_locked) { |
1811 | sk_locked = 1; | |
1812 | goto restart_locked; | |
1813 | } | |
1da177e4 LT |
1814 | } |
1815 | ||
7d267278 RW |
1816 | if (unlikely(sk_locked)) |
1817 | unix_state_unlock(sk); | |
1818 | ||
3f66116e AC |
1819 | if (sock_flag(other, SOCK_RCVTSTAMP)) |
1820 | __net_timestamp(skb); | |
16e57262 | 1821 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1822 | skb_queue_tail(&other->sk_receive_queue, skb); |
25888e30 ED |
1823 | if (max_level > unix_sk(other)->recursion_level) |
1824 | unix_sk(other)->recursion_level = max_level; | |
1c92b4e5 | 1825 | unix_state_unlock(other); |
676d2369 | 1826 | other->sk_data_ready(other); |
1da177e4 | 1827 | sock_put(other); |
7cc05662 | 1828 | scm_destroy(&scm); |
1da177e4 LT |
1829 | return len; |
1830 | ||
1831 | out_unlock: | |
7d267278 RW |
1832 | if (sk_locked) |
1833 | unix_state_unlock(sk); | |
1c92b4e5 | 1834 | unix_state_unlock(other); |
1da177e4 LT |
1835 | out_free: |
1836 | kfree_skb(skb); | |
1837 | out: | |
1838 | if (other) | |
1839 | sock_put(other); | |
7cc05662 | 1840 | scm_destroy(&scm); |
1da177e4 LT |
1841 | return err; |
1842 | } | |
1843 | ||
e370a723 ED |
1844 | /* We use paged skbs for stream sockets, and limit occupancy to 32768 |
1845 | * bytes, and a minimun of a full page. | |
1846 | */ | |
1847 | #define UNIX_SKB_FRAGS_SZ (PAGE_SIZE << get_order(32768)) | |
ac7bfa62 | 1848 | |
1b784140 YX |
1849 | static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg, |
1850 | size_t len) | |
1da177e4 | 1851 | { |
1da177e4 LT |
1852 | struct sock *sk = sock->sk; |
1853 | struct sock *other = NULL; | |
6eba6a37 | 1854 | int err, size; |
f78a5fda | 1855 | struct sk_buff *skb; |
e27dfcea | 1856 | int sent = 0; |
7cc05662 | 1857 | struct scm_cookie scm; |
8ba69ba6 | 1858 | bool fds_sent = false; |
25888e30 | 1859 | int max_level; |
e370a723 | 1860 | int data_len; |
1da177e4 | 1861 | |
5f23b734 | 1862 | wait_for_unix_gc(); |
7cc05662 | 1863 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1864 | if (err < 0) |
1865 | return err; | |
1866 | ||
1867 | err = -EOPNOTSUPP; | |
1868 | if (msg->msg_flags&MSG_OOB) | |
1869 | goto out_err; | |
1870 | ||
1871 | if (msg->msg_namelen) { | |
1872 | err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; | |
1873 | goto out_err; | |
1874 | } else { | |
1da177e4 | 1875 | err = -ENOTCONN; |
830a1e5c | 1876 | other = unix_peer(sk); |
1da177e4 LT |
1877 | if (!other) |
1878 | goto out_err; | |
1879 | } | |
1880 | ||
1881 | if (sk->sk_shutdown & SEND_SHUTDOWN) | |
1882 | goto pipe_err; | |
1883 | ||
6eba6a37 | 1884 | while (sent < len) { |
e370a723 | 1885 | size = len - sent; |
1da177e4 LT |
1886 | |
1887 | /* Keep two messages in the pipe so it schedules better */ | |
e370a723 | 1888 | size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64); |
1da177e4 | 1889 | |
e370a723 ED |
1890 | /* allow fallback to order-0 allocations */ |
1891 | size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ); | |
ac7bfa62 | 1892 | |
e370a723 | 1893 | data_len = max_t(int, 0, size - SKB_MAX_HEAD(0)); |
1da177e4 | 1894 | |
31ff6aa5 KT |
1895 | data_len = min_t(size_t, size, PAGE_ALIGN(data_len)); |
1896 | ||
e370a723 | 1897 | skb = sock_alloc_send_pskb(sk, size - data_len, data_len, |
28d64271 ED |
1898 | msg->msg_flags & MSG_DONTWAIT, &err, |
1899 | get_order(UNIX_SKB_FRAGS_SZ)); | |
e370a723 | 1900 | if (!skb) |
1da177e4 LT |
1901 | goto out_err; |
1902 | ||
f78a5fda | 1903 | /* Only send the fds in the first buffer */ |
7cc05662 | 1904 | err = unix_scm_to_skb(&scm, skb, !fds_sent); |
25888e30 | 1905 | if (err < 0) { |
7361c36c | 1906 | kfree_skb(skb); |
f78a5fda | 1907 | goto out_err; |
6209344f | 1908 | } |
25888e30 | 1909 | max_level = err + 1; |
7361c36c | 1910 | fds_sent = true; |
1da177e4 | 1911 | |
e370a723 ED |
1912 | skb_put(skb, size - data_len); |
1913 | skb->data_len = data_len; | |
1914 | skb->len = size; | |
c0371da6 | 1915 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size); |
6eba6a37 | 1916 | if (err) { |
1da177e4 | 1917 | kfree_skb(skb); |
f78a5fda | 1918 | goto out_err; |
1da177e4 LT |
1919 | } |
1920 | ||
1c92b4e5 | 1921 | unix_state_lock(other); |
1da177e4 LT |
1922 | |
1923 | if (sock_flag(other, SOCK_DEAD) || | |
1924 | (other->sk_shutdown & RCV_SHUTDOWN)) | |
1925 | goto pipe_err_free; | |
1926 | ||
16e57262 | 1927 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1928 | skb_queue_tail(&other->sk_receive_queue, skb); |
25888e30 ED |
1929 | if (max_level > unix_sk(other)->recursion_level) |
1930 | unix_sk(other)->recursion_level = max_level; | |
1c92b4e5 | 1931 | unix_state_unlock(other); |
676d2369 | 1932 | other->sk_data_ready(other); |
e27dfcea | 1933 | sent += size; |
1da177e4 | 1934 | } |
1da177e4 | 1935 | |
7cc05662 | 1936 | scm_destroy(&scm); |
1da177e4 LT |
1937 | |
1938 | return sent; | |
1939 | ||
1940 | pipe_err_free: | |
1c92b4e5 | 1941 | unix_state_unlock(other); |
1da177e4 LT |
1942 | kfree_skb(skb); |
1943 | pipe_err: | |
6eba6a37 ED |
1944 | if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL)) |
1945 | send_sig(SIGPIPE, current, 0); | |
1da177e4 LT |
1946 | err = -EPIPE; |
1947 | out_err: | |
7cc05662 | 1948 | scm_destroy(&scm); |
1da177e4 LT |
1949 | return sent ? : err; |
1950 | } | |
1951 | ||
869e7c62 HFS |
1952 | static ssize_t unix_stream_sendpage(struct socket *socket, struct page *page, |
1953 | int offset, size_t size, int flags) | |
1954 | { | |
9490f886 HFS |
1955 | int err; |
1956 | bool send_sigpipe = false; | |
1957 | bool init_scm = true; | |
1958 | struct scm_cookie scm; | |
869e7c62 HFS |
1959 | struct sock *other, *sk = socket->sk; |
1960 | struct sk_buff *skb, *newskb = NULL, *tail = NULL; | |
1961 | ||
1962 | if (flags & MSG_OOB) | |
1963 | return -EOPNOTSUPP; | |
1964 | ||
1965 | other = unix_peer(sk); | |
1966 | if (!other || sk->sk_state != TCP_ESTABLISHED) | |
1967 | return -ENOTCONN; | |
1968 | ||
1969 | if (false) { | |
1970 | alloc_skb: | |
1971 | unix_state_unlock(other); | |
1972 | mutex_unlock(&unix_sk(other)->readlock); | |
1973 | newskb = sock_alloc_send_pskb(sk, 0, 0, flags & MSG_DONTWAIT, | |
1974 | &err, 0); | |
1975 | if (!newskb) | |
9490f886 | 1976 | goto err; |
869e7c62 HFS |
1977 | } |
1978 | ||
1979 | /* we must acquire readlock as we modify already present | |
1980 | * skbs in the sk_receive_queue and mess with skb->len | |
1981 | */ | |
1982 | err = mutex_lock_interruptible(&unix_sk(other)->readlock); | |
1983 | if (err) { | |
1984 | err = flags & MSG_DONTWAIT ? -EAGAIN : -ERESTARTSYS; | |
869e7c62 HFS |
1985 | goto err; |
1986 | } | |
1987 | ||
1988 | if (sk->sk_shutdown & SEND_SHUTDOWN) { | |
1989 | err = -EPIPE; | |
9490f886 | 1990 | send_sigpipe = true; |
869e7c62 HFS |
1991 | goto err_unlock; |
1992 | } | |
1993 | ||
1994 | unix_state_lock(other); | |
1995 | ||
1996 | if (sock_flag(other, SOCK_DEAD) || | |
1997 | other->sk_shutdown & RCV_SHUTDOWN) { | |
1998 | err = -EPIPE; | |
9490f886 | 1999 | send_sigpipe = true; |
869e7c62 HFS |
2000 | goto err_state_unlock; |
2001 | } | |
2002 | ||
9490f886 HFS |
2003 | if (init_scm) { |
2004 | err = maybe_init_creds(&scm, socket, other); | |
2005 | if (err) | |
2006 | goto err_state_unlock; | |
2007 | init_scm = false; | |
2008 | } | |
2009 | ||
869e7c62 HFS |
2010 | skb = skb_peek_tail(&other->sk_receive_queue); |
2011 | if (tail && tail == skb) { | |
2012 | skb = newskb; | |
9490f886 HFS |
2013 | } else if (!skb || !unix_skb_scm_eq(skb, &scm)) { |
2014 | if (newskb) { | |
869e7c62 | 2015 | skb = newskb; |
9490f886 HFS |
2016 | } else { |
2017 | tail = skb; | |
869e7c62 | 2018 | goto alloc_skb; |
9490f886 | 2019 | } |
869e7c62 HFS |
2020 | } else if (newskb) { |
2021 | /* this is fast path, we don't necessarily need to | |
2022 | * call to kfree_skb even though with newskb == NULL | |
2023 | * this - does no harm | |
2024 | */ | |
2025 | consume_skb(newskb); | |
8844f972 | 2026 | newskb = NULL; |
869e7c62 HFS |
2027 | } |
2028 | ||
2029 | if (skb_append_pagefrags(skb, page, offset, size)) { | |
2030 | tail = skb; | |
2031 | goto alloc_skb; | |
2032 | } | |
2033 | ||
2034 | skb->len += size; | |
2035 | skb->data_len += size; | |
2036 | skb->truesize += size; | |
2037 | atomic_add(size, &sk->sk_wmem_alloc); | |
2038 | ||
a3a116e0 | 2039 | if (newskb) { |
9490f886 HFS |
2040 | err = unix_scm_to_skb(&scm, skb, false); |
2041 | if (err) | |
2042 | goto err_state_unlock; | |
a3a116e0 | 2043 | spin_lock(&other->sk_receive_queue.lock); |
869e7c62 | 2044 | __skb_queue_tail(&other->sk_receive_queue, newskb); |
a3a116e0 HFS |
2045 | spin_unlock(&other->sk_receive_queue.lock); |
2046 | } | |
869e7c62 HFS |
2047 | |
2048 | unix_state_unlock(other); | |
2049 | mutex_unlock(&unix_sk(other)->readlock); | |
2050 | ||
2051 | other->sk_data_ready(other); | |
9490f886 | 2052 | scm_destroy(&scm); |
869e7c62 HFS |
2053 | return size; |
2054 | ||
2055 | err_state_unlock: | |
2056 | unix_state_unlock(other); | |
2057 | err_unlock: | |
2058 | mutex_unlock(&unix_sk(other)->readlock); | |
2059 | err: | |
2060 | kfree_skb(newskb); | |
2061 | if (send_sigpipe && !(flags & MSG_NOSIGNAL)) | |
2062 | send_sig(SIGPIPE, current, 0); | |
9490f886 HFS |
2063 | if (!init_scm) |
2064 | scm_destroy(&scm); | |
869e7c62 HFS |
2065 | return err; |
2066 | } | |
2067 | ||
1b784140 YX |
2068 | static int unix_seqpacket_sendmsg(struct socket *sock, struct msghdr *msg, |
2069 | size_t len) | |
1da177e4 LT |
2070 | { |
2071 | int err; | |
2072 | struct sock *sk = sock->sk; | |
ac7bfa62 | 2073 | |
1da177e4 LT |
2074 | err = sock_error(sk); |
2075 | if (err) | |
2076 | return err; | |
2077 | ||
2078 | if (sk->sk_state != TCP_ESTABLISHED) | |
2079 | return -ENOTCONN; | |
2080 | ||
2081 | if (msg->msg_namelen) | |
2082 | msg->msg_namelen = 0; | |
2083 | ||
1b784140 | 2084 | return unix_dgram_sendmsg(sock, msg, len); |
1da177e4 | 2085 | } |
ac7bfa62 | 2086 | |
1b784140 YX |
2087 | static int unix_seqpacket_recvmsg(struct socket *sock, struct msghdr *msg, |
2088 | size_t size, int flags) | |
a05d2ad1 EB |
2089 | { |
2090 | struct sock *sk = sock->sk; | |
2091 | ||
2092 | if (sk->sk_state != TCP_ESTABLISHED) | |
2093 | return -ENOTCONN; | |
2094 | ||
1b784140 | 2095 | return unix_dgram_recvmsg(sock, msg, size, flags); |
a05d2ad1 EB |
2096 | } |
2097 | ||
1da177e4 LT |
2098 | static void unix_copy_addr(struct msghdr *msg, struct sock *sk) |
2099 | { | |
2100 | struct unix_sock *u = unix_sk(sk); | |
2101 | ||
1da177e4 LT |
2102 | if (u->addr) { |
2103 | msg->msg_namelen = u->addr->len; | |
2104 | memcpy(msg->msg_name, u->addr->name, u->addr->len); | |
2105 | } | |
2106 | } | |
2107 | ||
1b784140 YX |
2108 | static int unix_dgram_recvmsg(struct socket *sock, struct msghdr *msg, |
2109 | size_t size, int flags) | |
1da177e4 | 2110 | { |
7cc05662 | 2111 | struct scm_cookie scm; |
1da177e4 LT |
2112 | struct sock *sk = sock->sk; |
2113 | struct unix_sock *u = unix_sk(sk); | |
64874280 RW |
2114 | struct sk_buff *skb, *last; |
2115 | long timeo; | |
1da177e4 | 2116 | int err; |
f55bb7f9 | 2117 | int peeked, skip; |
1da177e4 LT |
2118 | |
2119 | err = -EOPNOTSUPP; | |
2120 | if (flags&MSG_OOB) | |
2121 | goto out; | |
2122 | ||
64874280 | 2123 | timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); |
1da177e4 | 2124 | |
64874280 RW |
2125 | do { |
2126 | mutex_lock(&u->readlock); | |
f55bb7f9 | 2127 | |
64874280 RW |
2128 | skip = sk_peek_offset(sk, flags); |
2129 | skb = __skb_try_recv_datagram(sk, flags, &peeked, &skip, &err, | |
2130 | &last); | |
2131 | if (skb) | |
2132 | break; | |
2133 | ||
2134 | mutex_unlock(&u->readlock); | |
2135 | ||
2136 | if (err != -EAGAIN) | |
2137 | break; | |
2138 | } while (timeo && | |
2139 | !__skb_wait_for_more_packets(sk, &err, &timeo, last)); | |
2140 | ||
2141 | if (!skb) { /* implies readlock unlocked */ | |
0a112258 FZ |
2142 | unix_state_lock(sk); |
2143 | /* Signal EOF on disconnected non-blocking SEQPACKET socket. */ | |
2144 | if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN && | |
2145 | (sk->sk_shutdown & RCV_SHUTDOWN)) | |
2146 | err = 0; | |
2147 | unix_state_unlock(sk); | |
64874280 | 2148 | goto out; |
0a112258 | 2149 | } |
1da177e4 | 2150 | |
77b75f4d RW |
2151 | if (wq_has_sleeper(&u->peer_wait)) |
2152 | wake_up_interruptible_sync_poll(&u->peer_wait, | |
2153 | POLLOUT | POLLWRNORM | | |
2154 | POLLWRBAND); | |
1da177e4 LT |
2155 | |
2156 | if (msg->msg_name) | |
2157 | unix_copy_addr(msg, skb->sk); | |
2158 | ||
f55bb7f9 PE |
2159 | if (size > skb->len - skip) |
2160 | size = skb->len - skip; | |
2161 | else if (size < skb->len - skip) | |
1da177e4 LT |
2162 | msg->msg_flags |= MSG_TRUNC; |
2163 | ||
51f3d02b | 2164 | err = skb_copy_datagram_msg(skb, skip, msg, size); |
1da177e4 LT |
2165 | if (err) |
2166 | goto out_free; | |
2167 | ||
3f66116e AC |
2168 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
2169 | __sock_recv_timestamp(msg, sk, skb); | |
2170 | ||
7cc05662 CH |
2171 | memset(&scm, 0, sizeof(scm)); |
2172 | ||
2173 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); | |
2174 | unix_set_secdata(&scm, skb); | |
1da177e4 | 2175 | |
6eba6a37 | 2176 | if (!(flags & MSG_PEEK)) { |
1da177e4 | 2177 | if (UNIXCB(skb).fp) |
7cc05662 | 2178 | unix_detach_fds(&scm, skb); |
f55bb7f9 PE |
2179 | |
2180 | sk_peek_offset_bwd(sk, skb->len); | |
6eba6a37 | 2181 | } else { |
1da177e4 LT |
2182 | /* It is questionable: on PEEK we could: |
2183 | - do not return fds - good, but too simple 8) | |
2184 | - return fds, and do not return them on read (old strategy, | |
2185 | apparently wrong) | |
2186 | - clone fds (I chose it for now, it is the most universal | |
2187 | solution) | |
ac7bfa62 YH |
2188 | |
2189 | POSIX 1003.1g does not actually define this clearly | |
2190 | at all. POSIX 1003.1g doesn't define a lot of things | |
2191 | clearly however! | |
2192 | ||
1da177e4 | 2193 | */ |
f55bb7f9 PE |
2194 | |
2195 | sk_peek_offset_fwd(sk, size); | |
2196 | ||
1da177e4 | 2197 | if (UNIXCB(skb).fp) |
7cc05662 | 2198 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 2199 | } |
9f6f9af7 | 2200 | err = (flags & MSG_TRUNC) ? skb->len - skip : size; |
1da177e4 | 2201 | |
7cc05662 | 2202 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2203 | |
2204 | out_free: | |
6eba6a37 | 2205 | skb_free_datagram(sk, skb); |
57b47a53 | 2206 | mutex_unlock(&u->readlock); |
1da177e4 LT |
2207 | out: |
2208 | return err; | |
2209 | } | |
2210 | ||
2211 | /* | |
79f632c7 | 2212 | * Sleep until more data has arrived. But check for races.. |
1da177e4 | 2213 | */ |
79f632c7 | 2214 | static long unix_stream_data_wait(struct sock *sk, long timeo, |
2b514574 | 2215 | struct sk_buff *last, unsigned int last_len) |
1da177e4 | 2216 | { |
2b514574 | 2217 | struct sk_buff *tail; |
1da177e4 LT |
2218 | DEFINE_WAIT(wait); |
2219 | ||
1c92b4e5 | 2220 | unix_state_lock(sk); |
1da177e4 LT |
2221 | |
2222 | for (;;) { | |
aa395145 | 2223 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 2224 | |
2b514574 HFS |
2225 | tail = skb_peek_tail(&sk->sk_receive_queue); |
2226 | if (tail != last || | |
2227 | (tail && tail->len != last_len) || | |
1da177e4 LT |
2228 | sk->sk_err || |
2229 | (sk->sk_shutdown & RCV_SHUTDOWN) || | |
2230 | signal_pending(current) || | |
2231 | !timeo) | |
2232 | break; | |
2233 | ||
9cd3e072 | 2234 | sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk); |
1c92b4e5 | 2235 | unix_state_unlock(sk); |
2b15af6f | 2236 | timeo = freezable_schedule_timeout(timeo); |
1c92b4e5 | 2237 | unix_state_lock(sk); |
b48732e4 MS |
2238 | |
2239 | if (sock_flag(sk, SOCK_DEAD)) | |
2240 | break; | |
2241 | ||
9cd3e072 | 2242 | sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk); |
1da177e4 LT |
2243 | } |
2244 | ||
aa395145 | 2245 | finish_wait(sk_sleep(sk), &wait); |
1c92b4e5 | 2246 | unix_state_unlock(sk); |
1da177e4 LT |
2247 | return timeo; |
2248 | } | |
2249 | ||
e370a723 ED |
2250 | static unsigned int unix_skb_len(const struct sk_buff *skb) |
2251 | { | |
2252 | return skb->len - UNIXCB(skb).consumed; | |
2253 | } | |
2254 | ||
2b514574 HFS |
2255 | struct unix_stream_read_state { |
2256 | int (*recv_actor)(struct sk_buff *, int, int, | |
2257 | struct unix_stream_read_state *); | |
2258 | struct socket *socket; | |
2259 | struct msghdr *msg; | |
2260 | struct pipe_inode_info *pipe; | |
2261 | size_t size; | |
2262 | int flags; | |
2263 | unsigned int splice_flags; | |
2264 | }; | |
2265 | ||
2266 | static int unix_stream_read_generic(struct unix_stream_read_state *state) | |
1da177e4 | 2267 | { |
7cc05662 | 2268 | struct scm_cookie scm; |
2b514574 | 2269 | struct socket *sock = state->socket; |
1da177e4 LT |
2270 | struct sock *sk = sock->sk; |
2271 | struct unix_sock *u = unix_sk(sk); | |
1da177e4 | 2272 | int copied = 0; |
2b514574 | 2273 | int flags = state->flags; |
de144391 | 2274 | int noblock = flags & MSG_DONTWAIT; |
2b514574 | 2275 | bool check_creds = false; |
1da177e4 LT |
2276 | int target; |
2277 | int err = 0; | |
2278 | long timeo; | |
fc0d7536 | 2279 | int skip; |
2b514574 HFS |
2280 | size_t size = state->size; |
2281 | unsigned int last_len; | |
1da177e4 | 2282 | |
1b92ee3d RW |
2283 | if (unlikely(sk->sk_state != TCP_ESTABLISHED)) { |
2284 | err = -EINVAL; | |
1da177e4 | 2285 | goto out; |
1b92ee3d | 2286 | } |
1da177e4 | 2287 | |
1b92ee3d RW |
2288 | if (unlikely(flags & MSG_OOB)) { |
2289 | err = -EOPNOTSUPP; | |
1da177e4 | 2290 | goto out; |
1b92ee3d | 2291 | } |
1da177e4 | 2292 | |
2b514574 | 2293 | target = sock_rcvlowat(sk, flags & MSG_WAITALL, size); |
de144391 | 2294 | timeo = sock_rcvtimeo(sk, noblock); |
1da177e4 | 2295 | |
2b514574 HFS |
2296 | memset(&scm, 0, sizeof(scm)); |
2297 | ||
1da177e4 LT |
2298 | /* Lock the socket to prevent queue disordering |
2299 | * while sleeps in memcpy_tomsg | |
2300 | */ | |
3822b5c2 | 2301 | mutex_lock(&u->readlock); |
1da177e4 | 2302 | |
e9193d60 AV |
2303 | if (flags & MSG_PEEK) |
2304 | skip = sk_peek_offset(sk, flags); | |
2305 | else | |
2306 | skip = 0; | |
2307 | ||
6eba6a37 | 2308 | do { |
1da177e4 | 2309 | int chunk; |
73ed5d25 | 2310 | bool drop_skb; |
79f632c7 | 2311 | struct sk_buff *skb, *last; |
1da177e4 | 2312 | |
18eceb81 | 2313 | redo: |
3c0d2f37 | 2314 | unix_state_lock(sk); |
b48732e4 MS |
2315 | if (sock_flag(sk, SOCK_DEAD)) { |
2316 | err = -ECONNRESET; | |
2317 | goto unlock; | |
2318 | } | |
79f632c7 | 2319 | last = skb = skb_peek(&sk->sk_receive_queue); |
2b514574 | 2320 | last_len = last ? last->len : 0; |
fc0d7536 | 2321 | again: |
6eba6a37 | 2322 | if (skb == NULL) { |
25888e30 | 2323 | unix_sk(sk)->recursion_level = 0; |
1da177e4 | 2324 | if (copied >= target) |
3c0d2f37 | 2325 | goto unlock; |
1da177e4 LT |
2326 | |
2327 | /* | |
2328 | * POSIX 1003.1g mandates this order. | |
2329 | */ | |
ac7bfa62 | 2330 | |
6eba6a37 ED |
2331 | err = sock_error(sk); |
2332 | if (err) | |
3c0d2f37 | 2333 | goto unlock; |
1da177e4 | 2334 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
3c0d2f37 MS |
2335 | goto unlock; |
2336 | ||
2337 | unix_state_unlock(sk); | |
1b92ee3d RW |
2338 | if (!timeo) { |
2339 | err = -EAGAIN; | |
1da177e4 | 2340 | break; |
1b92ee3d RW |
2341 | } |
2342 | ||
57b47a53 | 2343 | mutex_unlock(&u->readlock); |
1da177e4 | 2344 | |
2b514574 HFS |
2345 | timeo = unix_stream_data_wait(sk, timeo, last, |
2346 | last_len); | |
1da177e4 | 2347 | |
3822b5c2 | 2348 | if (signal_pending(current)) { |
1da177e4 | 2349 | err = sock_intr_errno(timeo); |
fa0dc04d | 2350 | scm_destroy(&scm); |
1da177e4 LT |
2351 | goto out; |
2352 | } | |
b3ca9b02 | 2353 | |
3822b5c2 | 2354 | mutex_lock(&u->readlock); |
18eceb81 | 2355 | goto redo; |
2b514574 | 2356 | unlock: |
3c0d2f37 MS |
2357 | unix_state_unlock(sk); |
2358 | break; | |
1da177e4 | 2359 | } |
fc0d7536 | 2360 | |
e370a723 ED |
2361 | while (skip >= unix_skb_len(skb)) { |
2362 | skip -= unix_skb_len(skb); | |
79f632c7 | 2363 | last = skb; |
2b514574 | 2364 | last_len = skb->len; |
fc0d7536 | 2365 | skb = skb_peek_next(skb, &sk->sk_receive_queue); |
79f632c7 BP |
2366 | if (!skb) |
2367 | goto again; | |
fc0d7536 PE |
2368 | } |
2369 | ||
3c0d2f37 | 2370 | unix_state_unlock(sk); |
1da177e4 LT |
2371 | |
2372 | if (check_creds) { | |
2373 | /* Never glue messages from different writers */ | |
9490f886 | 2374 | if (!unix_skb_scm_eq(skb, &scm)) |
1da177e4 | 2375 | break; |
0e82e7f6 | 2376 | } else if (test_bit(SOCK_PASSCRED, &sock->flags)) { |
1da177e4 | 2377 | /* Copy credentials */ |
7cc05662 | 2378 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); |
37a9a8df | 2379 | unix_set_secdata(&scm, skb); |
2b514574 | 2380 | check_creds = true; |
1da177e4 LT |
2381 | } |
2382 | ||
2383 | /* Copy address just once */ | |
2b514574 HFS |
2384 | if (state->msg && state->msg->msg_name) { |
2385 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, | |
2386 | state->msg->msg_name); | |
2387 | unix_copy_addr(state->msg, skb->sk); | |
1da177e4 LT |
2388 | sunaddr = NULL; |
2389 | } | |
2390 | ||
e370a723 | 2391 | chunk = min_t(unsigned int, unix_skb_len(skb) - skip, size); |
73ed5d25 | 2392 | skb_get(skb); |
2b514574 | 2393 | chunk = state->recv_actor(skb, skip, chunk, state); |
73ed5d25 HFS |
2394 | drop_skb = !unix_skb_len(skb); |
2395 | /* skb is only safe to use if !drop_skb */ | |
2396 | consume_skb(skb); | |
2b514574 | 2397 | if (chunk < 0) { |
1da177e4 LT |
2398 | if (copied == 0) |
2399 | copied = -EFAULT; | |
2400 | break; | |
2401 | } | |
2402 | copied += chunk; | |
2403 | size -= chunk; | |
2404 | ||
73ed5d25 HFS |
2405 | if (drop_skb) { |
2406 | /* the skb was touched by a concurrent reader; | |
2407 | * we should not expect anything from this skb | |
2408 | * anymore and assume it invalid - we can be | |
2409 | * sure it was dropped from the socket queue | |
2410 | * | |
2411 | * let's report a short read | |
2412 | */ | |
2413 | err = 0; | |
2414 | break; | |
2415 | } | |
2416 | ||
1da177e4 | 2417 | /* Mark read part of skb as used */ |
6eba6a37 | 2418 | if (!(flags & MSG_PEEK)) { |
e370a723 | 2419 | UNIXCB(skb).consumed += chunk; |
1da177e4 | 2420 | |
fc0d7536 PE |
2421 | sk_peek_offset_bwd(sk, chunk); |
2422 | ||
1da177e4 | 2423 | if (UNIXCB(skb).fp) |
7cc05662 | 2424 | unix_detach_fds(&scm, skb); |
1da177e4 | 2425 | |
e370a723 | 2426 | if (unix_skb_len(skb)) |
1da177e4 | 2427 | break; |
1da177e4 | 2428 | |
6f01fd6e | 2429 | skb_unlink(skb, &sk->sk_receive_queue); |
70d4bf6d | 2430 | consume_skb(skb); |
1da177e4 | 2431 | |
7cc05662 | 2432 | if (scm.fp) |
1da177e4 | 2433 | break; |
6eba6a37 | 2434 | } else { |
1da177e4 LT |
2435 | /* It is questionable, see note in unix_dgram_recvmsg. |
2436 | */ | |
2437 | if (UNIXCB(skb).fp) | |
7cc05662 | 2438 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 2439 | |
e9193d60 | 2440 | sk_peek_offset_fwd(sk, chunk); |
fc0d7536 | 2441 | |
9f389e35 AC |
2442 | if (UNIXCB(skb).fp) |
2443 | break; | |
2444 | ||
e9193d60 | 2445 | skip = 0; |
9f389e35 AC |
2446 | last = skb; |
2447 | last_len = skb->len; | |
2448 | unix_state_lock(sk); | |
2449 | skb = skb_peek_next(skb, &sk->sk_receive_queue); | |
2450 | if (skb) | |
2451 | goto again; | |
2452 | unix_state_unlock(sk); | |
1da177e4 LT |
2453 | break; |
2454 | } | |
2455 | } while (size); | |
2456 | ||
57b47a53 | 2457 | mutex_unlock(&u->readlock); |
2b514574 HFS |
2458 | if (state->msg) |
2459 | scm_recv(sock, state->msg, &scm, flags); | |
2460 | else | |
2461 | scm_destroy(&scm); | |
1da177e4 LT |
2462 | out: |
2463 | return copied ? : err; | |
2464 | } | |
2465 | ||
2b514574 HFS |
2466 | static int unix_stream_read_actor(struct sk_buff *skb, |
2467 | int skip, int chunk, | |
2468 | struct unix_stream_read_state *state) | |
2469 | { | |
2470 | int ret; | |
2471 | ||
2472 | ret = skb_copy_datagram_msg(skb, UNIXCB(skb).consumed + skip, | |
2473 | state->msg, chunk); | |
2474 | return ret ?: chunk; | |
2475 | } | |
2476 | ||
2477 | static int unix_stream_recvmsg(struct socket *sock, struct msghdr *msg, | |
2478 | size_t size, int flags) | |
2479 | { | |
2480 | struct unix_stream_read_state state = { | |
2481 | .recv_actor = unix_stream_read_actor, | |
2482 | .socket = sock, | |
2483 | .msg = msg, | |
2484 | .size = size, | |
2485 | .flags = flags | |
2486 | }; | |
2487 | ||
2488 | return unix_stream_read_generic(&state); | |
2489 | } | |
2490 | ||
2491 | static ssize_t skb_unix_socket_splice(struct sock *sk, | |
2492 | struct pipe_inode_info *pipe, | |
2493 | struct splice_pipe_desc *spd) | |
2494 | { | |
2495 | int ret; | |
2496 | struct unix_sock *u = unix_sk(sk); | |
2497 | ||
2498 | mutex_unlock(&u->readlock); | |
2499 | ret = splice_to_pipe(pipe, spd); | |
2500 | mutex_lock(&u->readlock); | |
2501 | ||
2502 | return ret; | |
2503 | } | |
2504 | ||
2505 | static int unix_stream_splice_actor(struct sk_buff *skb, | |
2506 | int skip, int chunk, | |
2507 | struct unix_stream_read_state *state) | |
2508 | { | |
2509 | return skb_splice_bits(skb, state->socket->sk, | |
2510 | UNIXCB(skb).consumed + skip, | |
2511 | state->pipe, chunk, state->splice_flags, | |
2512 | skb_unix_socket_splice); | |
2513 | } | |
2514 | ||
2515 | static ssize_t unix_stream_splice_read(struct socket *sock, loff_t *ppos, | |
2516 | struct pipe_inode_info *pipe, | |
2517 | size_t size, unsigned int flags) | |
2518 | { | |
2519 | struct unix_stream_read_state state = { | |
2520 | .recv_actor = unix_stream_splice_actor, | |
2521 | .socket = sock, | |
2522 | .pipe = pipe, | |
2523 | .size = size, | |
2524 | .splice_flags = flags, | |
2525 | }; | |
2526 | ||
2527 | if (unlikely(*ppos)) | |
2528 | return -ESPIPE; | |
2529 | ||
2530 | if (sock->file->f_flags & O_NONBLOCK || | |
2531 | flags & SPLICE_F_NONBLOCK) | |
2532 | state.flags = MSG_DONTWAIT; | |
2533 | ||
2534 | return unix_stream_read_generic(&state); | |
2535 | } | |
2536 | ||
1da177e4 LT |
2537 | static int unix_shutdown(struct socket *sock, int mode) |
2538 | { | |
2539 | struct sock *sk = sock->sk; | |
2540 | struct sock *other; | |
2541 | ||
fc61b928 XW |
2542 | if (mode < SHUT_RD || mode > SHUT_RDWR) |
2543 | return -EINVAL; | |
2544 | /* This maps: | |
2545 | * SHUT_RD (0) -> RCV_SHUTDOWN (1) | |
2546 | * SHUT_WR (1) -> SEND_SHUTDOWN (2) | |
2547 | * SHUT_RDWR (2) -> SHUTDOWN_MASK (3) | |
2548 | */ | |
2549 | ++mode; | |
7180a031 AC |
2550 | |
2551 | unix_state_lock(sk); | |
2552 | sk->sk_shutdown |= mode; | |
2553 | other = unix_peer(sk); | |
2554 | if (other) | |
2555 | sock_hold(other); | |
2556 | unix_state_unlock(sk); | |
2557 | sk->sk_state_change(sk); | |
2558 | ||
2559 | if (other && | |
2560 | (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) { | |
2561 | ||
2562 | int peer_mode = 0; | |
2563 | ||
2564 | if (mode&RCV_SHUTDOWN) | |
2565 | peer_mode |= SEND_SHUTDOWN; | |
2566 | if (mode&SEND_SHUTDOWN) | |
2567 | peer_mode |= RCV_SHUTDOWN; | |
2568 | unix_state_lock(other); | |
2569 | other->sk_shutdown |= peer_mode; | |
2570 | unix_state_unlock(other); | |
2571 | other->sk_state_change(other); | |
2572 | if (peer_mode == SHUTDOWN_MASK) | |
2573 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP); | |
2574 | else if (peer_mode & RCV_SHUTDOWN) | |
2575 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN); | |
1da177e4 | 2576 | } |
7180a031 AC |
2577 | if (other) |
2578 | sock_put(other); | |
2579 | ||
1da177e4 LT |
2580 | return 0; |
2581 | } | |
2582 | ||
885ee74d PE |
2583 | long unix_inq_len(struct sock *sk) |
2584 | { | |
2585 | struct sk_buff *skb; | |
2586 | long amount = 0; | |
2587 | ||
2588 | if (sk->sk_state == TCP_LISTEN) | |
2589 | return -EINVAL; | |
2590 | ||
2591 | spin_lock(&sk->sk_receive_queue.lock); | |
2592 | if (sk->sk_type == SOCK_STREAM || | |
2593 | sk->sk_type == SOCK_SEQPACKET) { | |
2594 | skb_queue_walk(&sk->sk_receive_queue, skb) | |
e370a723 | 2595 | amount += unix_skb_len(skb); |
885ee74d PE |
2596 | } else { |
2597 | skb = skb_peek(&sk->sk_receive_queue); | |
2598 | if (skb) | |
2599 | amount = skb->len; | |
2600 | } | |
2601 | spin_unlock(&sk->sk_receive_queue.lock); | |
2602 | ||
2603 | return amount; | |
2604 | } | |
2605 | EXPORT_SYMBOL_GPL(unix_inq_len); | |
2606 | ||
2607 | long unix_outq_len(struct sock *sk) | |
2608 | { | |
2609 | return sk_wmem_alloc_get(sk); | |
2610 | } | |
2611 | EXPORT_SYMBOL_GPL(unix_outq_len); | |
2612 | ||
1da177e4 LT |
2613 | static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
2614 | { | |
2615 | struct sock *sk = sock->sk; | |
e27dfcea | 2616 | long amount = 0; |
1da177e4 LT |
2617 | int err; |
2618 | ||
6eba6a37 ED |
2619 | switch (cmd) { |
2620 | case SIOCOUTQ: | |
885ee74d | 2621 | amount = unix_outq_len(sk); |
6eba6a37 ED |
2622 | err = put_user(amount, (int __user *)arg); |
2623 | break; | |
2624 | case SIOCINQ: | |
885ee74d PE |
2625 | amount = unix_inq_len(sk); |
2626 | if (amount < 0) | |
2627 | err = amount; | |
2628 | else | |
1da177e4 | 2629 | err = put_user(amount, (int __user *)arg); |
885ee74d | 2630 | break; |
6eba6a37 ED |
2631 | default: |
2632 | err = -ENOIOCTLCMD; | |
2633 | break; | |
1da177e4 LT |
2634 | } |
2635 | return err; | |
2636 | } | |
2637 | ||
6eba6a37 | 2638 | static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait) |
1da177e4 LT |
2639 | { |
2640 | struct sock *sk = sock->sk; | |
2641 | unsigned int mask; | |
2642 | ||
aa395145 | 2643 | sock_poll_wait(file, sk_sleep(sk), wait); |
1da177e4 LT |
2644 | mask = 0; |
2645 | ||
2646 | /* exceptional events? */ | |
2647 | if (sk->sk_err) | |
2648 | mask |= POLLERR; | |
2649 | if (sk->sk_shutdown == SHUTDOWN_MASK) | |
2650 | mask |= POLLHUP; | |
f348d70a | 2651 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
db40980f | 2652 | mask |= POLLRDHUP | POLLIN | POLLRDNORM; |
1da177e4 LT |
2653 | |
2654 | /* readable? */ | |
db40980f | 2655 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
2656 | mask |= POLLIN | POLLRDNORM; |
2657 | ||
2658 | /* Connection-based need to check for termination and startup */ | |
6eba6a37 ED |
2659 | if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) && |
2660 | sk->sk_state == TCP_CLOSE) | |
1da177e4 LT |
2661 | mask |= POLLHUP; |
2662 | ||
2663 | /* | |
2664 | * we set writable also when the other side has shut down the | |
2665 | * connection. This prevents stuck sockets. | |
2666 | */ | |
2667 | if (unix_writable(sk)) | |
2668 | mask |= POLLOUT | POLLWRNORM | POLLWRBAND; | |
2669 | ||
2670 | return mask; | |
2671 | } | |
2672 | ||
ec0d215f RW |
2673 | static unsigned int unix_dgram_poll(struct file *file, struct socket *sock, |
2674 | poll_table *wait) | |
3c73419c | 2675 | { |
ec0d215f RW |
2676 | struct sock *sk = sock->sk, *other; |
2677 | unsigned int mask, writable; | |
3c73419c | 2678 | |
aa395145 | 2679 | sock_poll_wait(file, sk_sleep(sk), wait); |
3c73419c RW |
2680 | mask = 0; |
2681 | ||
2682 | /* exceptional events? */ | |
2683 | if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue)) | |
7d4c04fc | 2684 | mask |= POLLERR | |
8facd5fb | 2685 | (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0); |
7d4c04fc | 2686 | |
3c73419c | 2687 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
5456f09a | 2688 | mask |= POLLRDHUP | POLLIN | POLLRDNORM; |
3c73419c RW |
2689 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
2690 | mask |= POLLHUP; | |
2691 | ||
2692 | /* readable? */ | |
5456f09a | 2693 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
3c73419c RW |
2694 | mask |= POLLIN | POLLRDNORM; |
2695 | ||
2696 | /* Connection-based need to check for termination and startup */ | |
2697 | if (sk->sk_type == SOCK_SEQPACKET) { | |
2698 | if (sk->sk_state == TCP_CLOSE) | |
2699 | mask |= POLLHUP; | |
2700 | /* connection hasn't started yet? */ | |
2701 | if (sk->sk_state == TCP_SYN_SENT) | |
2702 | return mask; | |
2703 | } | |
2704 | ||
973a34aa | 2705 | /* No write status requested, avoid expensive OUT tests. */ |
626cf236 | 2706 | if (!(poll_requested_events(wait) & (POLLWRBAND|POLLWRNORM|POLLOUT))) |
973a34aa ED |
2707 | return mask; |
2708 | ||
ec0d215f | 2709 | writable = unix_writable(sk); |
7d267278 RW |
2710 | if (writable) { |
2711 | unix_state_lock(sk); | |
2712 | ||
2713 | other = unix_peer(sk); | |
2714 | if (other && unix_peer(other) != sk && | |
2715 | unix_recvq_full(other) && | |
2716 | unix_dgram_peer_wake_me(sk, other)) | |
2717 | writable = 0; | |
2718 | ||
2719 | unix_state_unlock(sk); | |
ec0d215f RW |
2720 | } |
2721 | ||
2722 | if (writable) | |
3c73419c RW |
2723 | mask |= POLLOUT | POLLWRNORM | POLLWRBAND; |
2724 | else | |
9cd3e072 | 2725 | sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk); |
3c73419c | 2726 | |
3c73419c RW |
2727 | return mask; |
2728 | } | |
1da177e4 LT |
2729 | |
2730 | #ifdef CONFIG_PROC_FS | |
a53eb3fe | 2731 | |
7123aaa3 ED |
2732 | #define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1) |
2733 | ||
2734 | #define get_bucket(x) ((x) >> BUCKET_SPACE) | |
2735 | #define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1)) | |
2736 | #define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o)) | |
a53eb3fe | 2737 | |
7123aaa3 | 2738 | static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos) |
1da177e4 | 2739 | { |
7123aaa3 ED |
2740 | unsigned long offset = get_offset(*pos); |
2741 | unsigned long bucket = get_bucket(*pos); | |
2742 | struct sock *sk; | |
2743 | unsigned long count = 0; | |
1da177e4 | 2744 | |
7123aaa3 ED |
2745 | for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) { |
2746 | if (sock_net(sk) != seq_file_net(seq)) | |
097e66c5 | 2747 | continue; |
7123aaa3 ED |
2748 | if (++count == offset) |
2749 | break; | |
2750 | } | |
2751 | ||
2752 | return sk; | |
2753 | } | |
2754 | ||
2755 | static struct sock *unix_next_socket(struct seq_file *seq, | |
2756 | struct sock *sk, | |
2757 | loff_t *pos) | |
2758 | { | |
2759 | unsigned long bucket; | |
2760 | ||
2761 | while (sk > (struct sock *)SEQ_START_TOKEN) { | |
2762 | sk = sk_next(sk); | |
2763 | if (!sk) | |
2764 | goto next_bucket; | |
2765 | if (sock_net(sk) == seq_file_net(seq)) | |
2766 | return sk; | |
1da177e4 | 2767 | } |
7123aaa3 ED |
2768 | |
2769 | do { | |
2770 | sk = unix_from_bucket(seq, pos); | |
2771 | if (sk) | |
2772 | return sk; | |
2773 | ||
2774 | next_bucket: | |
2775 | bucket = get_bucket(*pos) + 1; | |
2776 | *pos = set_bucket_offset(bucket, 1); | |
2777 | } while (bucket < ARRAY_SIZE(unix_socket_table)); | |
2778 | ||
1da177e4 LT |
2779 | return NULL; |
2780 | } | |
2781 | ||
1da177e4 | 2782 | static void *unix_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 2783 | __acquires(unix_table_lock) |
1da177e4 | 2784 | { |
fbe9cc4a | 2785 | spin_lock(&unix_table_lock); |
7123aaa3 ED |
2786 | |
2787 | if (!*pos) | |
2788 | return SEQ_START_TOKEN; | |
2789 | ||
2790 | if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table)) | |
2791 | return NULL; | |
2792 | ||
2793 | return unix_next_socket(seq, NULL, pos); | |
1da177e4 LT |
2794 | } |
2795 | ||
2796 | static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2797 | { | |
2798 | ++*pos; | |
7123aaa3 | 2799 | return unix_next_socket(seq, v, pos); |
1da177e4 LT |
2800 | } |
2801 | ||
2802 | static void unix_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2803 | __releases(unix_table_lock) |
1da177e4 | 2804 | { |
fbe9cc4a | 2805 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
2806 | } |
2807 | ||
2808 | static int unix_seq_show(struct seq_file *seq, void *v) | |
2809 | { | |
ac7bfa62 | 2810 | |
b9f3124f | 2811 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2812 | seq_puts(seq, "Num RefCount Protocol Flags Type St " |
2813 | "Inode Path\n"); | |
2814 | else { | |
2815 | struct sock *s = v; | |
2816 | struct unix_sock *u = unix_sk(s); | |
1c92b4e5 | 2817 | unix_state_lock(s); |
1da177e4 | 2818 | |
71338aa7 | 2819 | seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu", |
1da177e4 LT |
2820 | s, |
2821 | atomic_read(&s->sk_refcnt), | |
2822 | 0, | |
2823 | s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0, | |
2824 | s->sk_type, | |
2825 | s->sk_socket ? | |
2826 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) : | |
2827 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING), | |
2828 | sock_i_ino(s)); | |
2829 | ||
2830 | if (u->addr) { | |
2831 | int i, len; | |
2832 | seq_putc(seq, ' '); | |
2833 | ||
2834 | i = 0; | |
2835 | len = u->addr->len - sizeof(short); | |
2836 | if (!UNIX_ABSTRACT(s)) | |
2837 | len--; | |
2838 | else { | |
2839 | seq_putc(seq, '@'); | |
2840 | i++; | |
2841 | } | |
2842 | for ( ; i < len; i++) | |
2843 | seq_putc(seq, u->addr->name->sun_path[i]); | |
2844 | } | |
1c92b4e5 | 2845 | unix_state_unlock(s); |
1da177e4 LT |
2846 | seq_putc(seq, '\n'); |
2847 | } | |
2848 | ||
2849 | return 0; | |
2850 | } | |
2851 | ||
56b3d975 | 2852 | static const struct seq_operations unix_seq_ops = { |
1da177e4 LT |
2853 | .start = unix_seq_start, |
2854 | .next = unix_seq_next, | |
2855 | .stop = unix_seq_stop, | |
2856 | .show = unix_seq_show, | |
2857 | }; | |
2858 | ||
1da177e4 LT |
2859 | static int unix_seq_open(struct inode *inode, struct file *file) |
2860 | { | |
e372c414 | 2861 | return seq_open_net(inode, file, &unix_seq_ops, |
8b51b064 | 2862 | sizeof(struct seq_net_private)); |
1da177e4 LT |
2863 | } |
2864 | ||
da7071d7 | 2865 | static const struct file_operations unix_seq_fops = { |
1da177e4 LT |
2866 | .owner = THIS_MODULE, |
2867 | .open = unix_seq_open, | |
2868 | .read = seq_read, | |
2869 | .llseek = seq_lseek, | |
e372c414 | 2870 | .release = seq_release_net, |
1da177e4 LT |
2871 | }; |
2872 | ||
2873 | #endif | |
2874 | ||
ec1b4cf7 | 2875 | static const struct net_proto_family unix_family_ops = { |
1da177e4 LT |
2876 | .family = PF_UNIX, |
2877 | .create = unix_create, | |
2878 | .owner = THIS_MODULE, | |
2879 | }; | |
2880 | ||
097e66c5 | 2881 | |
2c8c1e72 | 2882 | static int __net_init unix_net_init(struct net *net) |
097e66c5 DL |
2883 | { |
2884 | int error = -ENOMEM; | |
2885 | ||
a0a53c8b | 2886 | net->unx.sysctl_max_dgram_qlen = 10; |
1597fbc0 PE |
2887 | if (unix_sysctl_register(net)) |
2888 | goto out; | |
d392e497 | 2889 | |
097e66c5 | 2890 | #ifdef CONFIG_PROC_FS |
d4beaa66 | 2891 | if (!proc_create("unix", 0, net->proc_net, &unix_seq_fops)) { |
1597fbc0 | 2892 | unix_sysctl_unregister(net); |
097e66c5 | 2893 | goto out; |
1597fbc0 | 2894 | } |
097e66c5 DL |
2895 | #endif |
2896 | error = 0; | |
2897 | out: | |
48dcc33e | 2898 | return error; |
097e66c5 DL |
2899 | } |
2900 | ||
2c8c1e72 | 2901 | static void __net_exit unix_net_exit(struct net *net) |
097e66c5 | 2902 | { |
1597fbc0 | 2903 | unix_sysctl_unregister(net); |
ece31ffd | 2904 | remove_proc_entry("unix", net->proc_net); |
097e66c5 DL |
2905 | } |
2906 | ||
2907 | static struct pernet_operations unix_net_ops = { | |
2908 | .init = unix_net_init, | |
2909 | .exit = unix_net_exit, | |
2910 | }; | |
2911 | ||
1da177e4 LT |
2912 | static int __init af_unix_init(void) |
2913 | { | |
2914 | int rc = -1; | |
1da177e4 | 2915 | |
b4fff5f8 | 2916 | BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 LT |
2917 | |
2918 | rc = proto_register(&unix_proto, 1); | |
ac7bfa62 | 2919 | if (rc != 0) { |
5cc208be | 2920 | pr_crit("%s: Cannot create unix_sock SLAB cache!\n", __func__); |
1da177e4 LT |
2921 | goto out; |
2922 | } | |
2923 | ||
2924 | sock_register(&unix_family_ops); | |
097e66c5 | 2925 | register_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2926 | out: |
2927 | return rc; | |
2928 | } | |
2929 | ||
2930 | static void __exit af_unix_exit(void) | |
2931 | { | |
2932 | sock_unregister(PF_UNIX); | |
1da177e4 | 2933 | proto_unregister(&unix_proto); |
097e66c5 | 2934 | unregister_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2935 | } |
2936 | ||
3d366960 DW |
2937 | /* Earlier than device_initcall() so that other drivers invoking |
2938 | request_module() don't end up in a loop when modprobe tries | |
2939 | to use a UNIX socket. But later than subsys_initcall() because | |
2940 | we depend on stuff initialised there */ | |
2941 | fs_initcall(af_unix_init); | |
1da177e4 LT |
2942 | module_exit(af_unix_exit); |
2943 | ||
2944 | MODULE_LICENSE("GPL"); | |
2945 | MODULE_ALIAS_NETPROTO(PF_UNIX); |