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 | ||
83 | #include <linux/module.h> | |
1da177e4 | 84 | #include <linux/kernel.h> |
1da177e4 LT |
85 | #include <linux/signal.h> |
86 | #include <linux/sched.h> | |
87 | #include <linux/errno.h> | |
88 | #include <linux/string.h> | |
89 | #include <linux/stat.h> | |
90 | #include <linux/dcache.h> | |
91 | #include <linux/namei.h> | |
92 | #include <linux/socket.h> | |
93 | #include <linux/un.h> | |
94 | #include <linux/fcntl.h> | |
95 | #include <linux/termios.h> | |
96 | #include <linux/sockios.h> | |
97 | #include <linux/net.h> | |
98 | #include <linux/in.h> | |
99 | #include <linux/fs.h> | |
100 | #include <linux/slab.h> | |
101 | #include <asm/uaccess.h> | |
102 | #include <linux/skbuff.h> | |
103 | #include <linux/netdevice.h> | |
457c4cbc | 104 | #include <net/net_namespace.h> |
1da177e4 | 105 | #include <net/sock.h> |
c752f073 | 106 | #include <net/tcp_states.h> |
1da177e4 LT |
107 | #include <net/af_unix.h> |
108 | #include <linux/proc_fs.h> | |
109 | #include <linux/seq_file.h> | |
110 | #include <net/scm.h> | |
111 | #include <linux/init.h> | |
112 | #include <linux/poll.h> | |
1da177e4 LT |
113 | #include <linux/rtnetlink.h> |
114 | #include <linux/mount.h> | |
115 | #include <net/checksum.h> | |
116 | #include <linux/security.h> | |
117 | ||
fa7ff56f PE |
118 | struct hlist_head unix_socket_table[UNIX_HASH_SIZE + 1]; |
119 | EXPORT_SYMBOL_GPL(unix_socket_table); | |
120 | DEFINE_SPINLOCK(unix_table_lock); | |
121 | EXPORT_SYMBOL_GPL(unix_table_lock); | |
518de9b3 | 122 | static atomic_long_t unix_nr_socks; |
1da177e4 LT |
123 | |
124 | #define unix_sockets_unbound (&unix_socket_table[UNIX_HASH_SIZE]) | |
125 | ||
126 | #define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash != UNIX_HASH_SIZE) | |
127 | ||
877ce7c1 | 128 | #ifdef CONFIG_SECURITY_NETWORK |
dc49c1f9 | 129 | static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 | 130 | { |
dc49c1f9 | 131 | memcpy(UNIXSID(skb), &scm->secid, sizeof(u32)); |
877ce7c1 CZ |
132 | } |
133 | ||
134 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
135 | { | |
dc49c1f9 | 136 | scm->secid = *UNIXSID(skb); |
877ce7c1 CZ |
137 | } |
138 | #else | |
dc49c1f9 | 139 | static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 CZ |
140 | { } |
141 | ||
142 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
143 | { } | |
144 | #endif /* CONFIG_SECURITY_NETWORK */ | |
145 | ||
1da177e4 LT |
146 | /* |
147 | * SMP locking strategy: | |
fbe9cc4a | 148 | * hash table is protected with spinlock unix_table_lock |
663717f6 | 149 | * each socket state is protected by separate spin lock. |
1da177e4 LT |
150 | */ |
151 | ||
44bb9363 | 152 | static inline unsigned unix_hash_fold(__wsum n) |
1da177e4 | 153 | { |
44bb9363 | 154 | unsigned hash = (__force unsigned)n; |
1da177e4 LT |
155 | hash ^= hash>>16; |
156 | hash ^= hash>>8; | |
157 | return hash&(UNIX_HASH_SIZE-1); | |
158 | } | |
159 | ||
160 | #define unix_peer(sk) (unix_sk(sk)->peer) | |
161 | ||
162 | static inline int unix_our_peer(struct sock *sk, struct sock *osk) | |
163 | { | |
164 | return unix_peer(osk) == sk; | |
165 | } | |
166 | ||
167 | static inline int unix_may_send(struct sock *sk, struct sock *osk) | |
168 | { | |
6eba6a37 | 169 | return unix_peer(osk) == NULL || unix_our_peer(sk, osk); |
1da177e4 LT |
170 | } |
171 | ||
3c73419c RW |
172 | static inline int unix_recvq_full(struct sock const *sk) |
173 | { | |
174 | return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog; | |
175 | } | |
176 | ||
fa7ff56f | 177 | struct sock *unix_peer_get(struct sock *s) |
1da177e4 LT |
178 | { |
179 | struct sock *peer; | |
180 | ||
1c92b4e5 | 181 | unix_state_lock(s); |
1da177e4 LT |
182 | peer = unix_peer(s); |
183 | if (peer) | |
184 | sock_hold(peer); | |
1c92b4e5 | 185 | unix_state_unlock(s); |
1da177e4 LT |
186 | return peer; |
187 | } | |
fa7ff56f | 188 | EXPORT_SYMBOL_GPL(unix_peer_get); |
1da177e4 LT |
189 | |
190 | static inline void unix_release_addr(struct unix_address *addr) | |
191 | { | |
192 | if (atomic_dec_and_test(&addr->refcnt)) | |
193 | kfree(addr); | |
194 | } | |
195 | ||
196 | /* | |
197 | * Check unix socket name: | |
198 | * - should be not zero length. | |
199 | * - if started by not zero, should be NULL terminated (FS object) | |
200 | * - if started by zero, it is abstract name. | |
201 | */ | |
ac7bfa62 | 202 | |
6eba6a37 | 203 | static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned *hashp) |
1da177e4 LT |
204 | { |
205 | if (len <= sizeof(short) || len > sizeof(*sunaddr)) | |
206 | return -EINVAL; | |
207 | if (!sunaddr || sunaddr->sun_family != AF_UNIX) | |
208 | return -EINVAL; | |
209 | if (sunaddr->sun_path[0]) { | |
210 | /* | |
211 | * This may look like an off by one error but it is a bit more | |
212 | * subtle. 108 is the longest valid AF_UNIX path for a binding. | |
25985edc | 213 | * sun_path[108] doesn't as such exist. However in kernel space |
1da177e4 LT |
214 | * we are guaranteed that it is a valid memory location in our |
215 | * kernel address buffer. | |
216 | */ | |
e27dfcea | 217 | ((char *)sunaddr)[len] = 0; |
1da177e4 LT |
218 | len = strlen(sunaddr->sun_path)+1+sizeof(short); |
219 | return len; | |
220 | } | |
221 | ||
07f0757a | 222 | *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0)); |
1da177e4 LT |
223 | return len; |
224 | } | |
225 | ||
226 | static void __unix_remove_socket(struct sock *sk) | |
227 | { | |
228 | sk_del_node_init(sk); | |
229 | } | |
230 | ||
231 | static void __unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
232 | { | |
547b792c | 233 | WARN_ON(!sk_unhashed(sk)); |
1da177e4 LT |
234 | sk_add_node(sk, list); |
235 | } | |
236 | ||
237 | static inline void unix_remove_socket(struct sock *sk) | |
238 | { | |
fbe9cc4a | 239 | spin_lock(&unix_table_lock); |
1da177e4 | 240 | __unix_remove_socket(sk); |
fbe9cc4a | 241 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
242 | } |
243 | ||
244 | static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
245 | { | |
fbe9cc4a | 246 | spin_lock(&unix_table_lock); |
1da177e4 | 247 | __unix_insert_socket(list, sk); |
fbe9cc4a | 248 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
249 | } |
250 | ||
097e66c5 DL |
251 | static struct sock *__unix_find_socket_byname(struct net *net, |
252 | struct sockaddr_un *sunname, | |
1da177e4 LT |
253 | int len, int type, unsigned hash) |
254 | { | |
255 | struct sock *s; | |
256 | struct hlist_node *node; | |
257 | ||
258 | sk_for_each(s, node, &unix_socket_table[hash ^ type]) { | |
259 | struct unix_sock *u = unix_sk(s); | |
260 | ||
878628fb | 261 | if (!net_eq(sock_net(s), net)) |
097e66c5 DL |
262 | continue; |
263 | ||
1da177e4 LT |
264 | if (u->addr->len == len && |
265 | !memcmp(u->addr->name, sunname, len)) | |
266 | goto found; | |
267 | } | |
268 | s = NULL; | |
269 | found: | |
270 | return s; | |
271 | } | |
272 | ||
097e66c5 DL |
273 | static inline struct sock *unix_find_socket_byname(struct net *net, |
274 | struct sockaddr_un *sunname, | |
1da177e4 LT |
275 | int len, int type, |
276 | unsigned hash) | |
277 | { | |
278 | struct sock *s; | |
279 | ||
fbe9cc4a | 280 | spin_lock(&unix_table_lock); |
097e66c5 | 281 | s = __unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
282 | if (s) |
283 | sock_hold(s); | |
fbe9cc4a | 284 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
285 | return s; |
286 | } | |
287 | ||
6616f788 | 288 | static struct sock *unix_find_socket_byinode(struct inode *i) |
1da177e4 LT |
289 | { |
290 | struct sock *s; | |
291 | struct hlist_node *node; | |
292 | ||
fbe9cc4a | 293 | spin_lock(&unix_table_lock); |
1da177e4 LT |
294 | sk_for_each(s, node, |
295 | &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) { | |
296 | struct dentry *dentry = unix_sk(s)->dentry; | |
297 | ||
6eba6a37 | 298 | if (dentry && dentry->d_inode == i) { |
1da177e4 LT |
299 | sock_hold(s); |
300 | goto found; | |
301 | } | |
302 | } | |
303 | s = NULL; | |
304 | found: | |
fbe9cc4a | 305 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
306 | return s; |
307 | } | |
308 | ||
309 | static inline int unix_writable(struct sock *sk) | |
310 | { | |
311 | return (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf; | |
312 | } | |
313 | ||
314 | static void unix_write_space(struct sock *sk) | |
315 | { | |
43815482 ED |
316 | struct socket_wq *wq; |
317 | ||
318 | rcu_read_lock(); | |
1da177e4 | 319 | if (unix_writable(sk)) { |
43815482 ED |
320 | wq = rcu_dereference(sk->sk_wq); |
321 | if (wq_has_sleeper(wq)) | |
67426b75 ED |
322 | wake_up_interruptible_sync_poll(&wq->wait, |
323 | POLLOUT | POLLWRNORM | POLLWRBAND); | |
8d8ad9d7 | 324 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
1da177e4 | 325 | } |
43815482 | 326 | rcu_read_unlock(); |
1da177e4 LT |
327 | } |
328 | ||
329 | /* When dgram socket disconnects (or changes its peer), we clear its receive | |
330 | * queue of packets arrived from previous peer. First, it allows to do | |
331 | * flow control based only on wmem_alloc; second, sk connected to peer | |
332 | * may receive messages only from that peer. */ | |
333 | static void unix_dgram_disconnected(struct sock *sk, struct sock *other) | |
334 | { | |
b03efcfb | 335 | if (!skb_queue_empty(&sk->sk_receive_queue)) { |
1da177e4 LT |
336 | skb_queue_purge(&sk->sk_receive_queue); |
337 | wake_up_interruptible_all(&unix_sk(sk)->peer_wait); | |
338 | ||
339 | /* If one link of bidirectional dgram pipe is disconnected, | |
340 | * we signal error. Messages are lost. Do not make this, | |
341 | * when peer was not connected to us. | |
342 | */ | |
343 | if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) { | |
344 | other->sk_err = ECONNRESET; | |
345 | other->sk_error_report(other); | |
346 | } | |
347 | } | |
348 | } | |
349 | ||
350 | static void unix_sock_destructor(struct sock *sk) | |
351 | { | |
352 | struct unix_sock *u = unix_sk(sk); | |
353 | ||
354 | skb_queue_purge(&sk->sk_receive_queue); | |
355 | ||
547b792c IJ |
356 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); |
357 | WARN_ON(!sk_unhashed(sk)); | |
358 | WARN_ON(sk->sk_socket); | |
1da177e4 | 359 | if (!sock_flag(sk, SOCK_DEAD)) { |
6b41e7dd | 360 | printk(KERN_INFO "Attempt to release alive unix socket: %p\n", sk); |
1da177e4 LT |
361 | return; |
362 | } | |
363 | ||
364 | if (u->addr) | |
365 | unix_release_addr(u->addr); | |
366 | ||
518de9b3 | 367 | atomic_long_dec(&unix_nr_socks); |
6f756a8c | 368 | local_bh_disable(); |
a8076d8d | 369 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
6f756a8c | 370 | local_bh_enable(); |
1da177e4 | 371 | #ifdef UNIX_REFCNT_DEBUG |
518de9b3 ED |
372 | printk(KERN_DEBUG "UNIX %p is destroyed, %ld are still alive.\n", sk, |
373 | atomic_long_read(&unix_nr_socks)); | |
1da177e4 LT |
374 | #endif |
375 | } | |
376 | ||
6eba6a37 | 377 | static int unix_release_sock(struct sock *sk, int embrion) |
1da177e4 LT |
378 | { |
379 | struct unix_sock *u = unix_sk(sk); | |
380 | struct dentry *dentry; | |
381 | struct vfsmount *mnt; | |
382 | struct sock *skpair; | |
383 | struct sk_buff *skb; | |
384 | int state; | |
385 | ||
386 | unix_remove_socket(sk); | |
387 | ||
388 | /* Clear state */ | |
1c92b4e5 | 389 | unix_state_lock(sk); |
1da177e4 LT |
390 | sock_orphan(sk); |
391 | sk->sk_shutdown = SHUTDOWN_MASK; | |
392 | dentry = u->dentry; | |
393 | u->dentry = NULL; | |
394 | mnt = u->mnt; | |
395 | u->mnt = NULL; | |
396 | state = sk->sk_state; | |
397 | sk->sk_state = TCP_CLOSE; | |
1c92b4e5 | 398 | unix_state_unlock(sk); |
1da177e4 LT |
399 | |
400 | wake_up_interruptible_all(&u->peer_wait); | |
401 | ||
e27dfcea | 402 | skpair = unix_peer(sk); |
1da177e4 | 403 | |
e27dfcea | 404 | if (skpair != NULL) { |
1da177e4 | 405 | if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) { |
1c92b4e5 | 406 | unix_state_lock(skpair); |
1da177e4 LT |
407 | /* No more writes */ |
408 | skpair->sk_shutdown = SHUTDOWN_MASK; | |
409 | if (!skb_queue_empty(&sk->sk_receive_queue) || embrion) | |
410 | skpair->sk_err = ECONNRESET; | |
1c92b4e5 | 411 | unix_state_unlock(skpair); |
1da177e4 | 412 | skpair->sk_state_change(skpair); |
8d8ad9d7 | 413 | sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP); |
1da177e4 LT |
414 | } |
415 | sock_put(skpair); /* It may now die */ | |
416 | unix_peer(sk) = NULL; | |
417 | } | |
418 | ||
419 | /* Try to flush out this socket. Throw out buffers at least */ | |
420 | ||
421 | while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) { | |
e27dfcea | 422 | if (state == TCP_LISTEN) |
1da177e4 LT |
423 | unix_release_sock(skb->sk, 1); |
424 | /* passed fds are erased in the kfree_skb hook */ | |
425 | kfree_skb(skb); | |
426 | } | |
427 | ||
428 | if (dentry) { | |
429 | dput(dentry); | |
430 | mntput(mnt); | |
431 | } | |
432 | ||
433 | sock_put(sk); | |
434 | ||
435 | /* ---- Socket is dead now and most probably destroyed ---- */ | |
436 | ||
437 | /* | |
438 | * Fixme: BSD difference: In BSD all sockets connected to use get | |
439 | * ECONNRESET and we die on the spot. In Linux we behave | |
440 | * like files and pipes do and wait for the last | |
441 | * dereference. | |
442 | * | |
443 | * Can't we simply set sock->err? | |
444 | * | |
445 | * What the above comment does talk about? --ANK(980817) | |
446 | */ | |
447 | ||
9305cfa4 | 448 | if (unix_tot_inflight) |
ac7bfa62 | 449 | unix_gc(); /* Garbage collect fds */ |
1da177e4 LT |
450 | |
451 | return 0; | |
452 | } | |
453 | ||
109f6e39 EB |
454 | static void init_peercred(struct sock *sk) |
455 | { | |
456 | put_pid(sk->sk_peer_pid); | |
457 | if (sk->sk_peer_cred) | |
458 | put_cred(sk->sk_peer_cred); | |
459 | sk->sk_peer_pid = get_pid(task_tgid(current)); | |
460 | sk->sk_peer_cred = get_current_cred(); | |
461 | } | |
462 | ||
463 | static void copy_peercred(struct sock *sk, struct sock *peersk) | |
464 | { | |
465 | put_pid(sk->sk_peer_pid); | |
466 | if (sk->sk_peer_cred) | |
467 | put_cred(sk->sk_peer_cred); | |
468 | sk->sk_peer_pid = get_pid(peersk->sk_peer_pid); | |
469 | sk->sk_peer_cred = get_cred(peersk->sk_peer_cred); | |
470 | } | |
471 | ||
1da177e4 LT |
472 | static int unix_listen(struct socket *sock, int backlog) |
473 | { | |
474 | int err; | |
475 | struct sock *sk = sock->sk; | |
476 | struct unix_sock *u = unix_sk(sk); | |
109f6e39 EB |
477 | struct pid *old_pid = NULL; |
478 | const struct cred *old_cred = NULL; | |
1da177e4 LT |
479 | |
480 | err = -EOPNOTSUPP; | |
6eba6a37 ED |
481 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
482 | goto out; /* Only stream/seqpacket sockets accept */ | |
1da177e4 LT |
483 | err = -EINVAL; |
484 | if (!u->addr) | |
6eba6a37 | 485 | goto out; /* No listens on an unbound socket */ |
1c92b4e5 | 486 | unix_state_lock(sk); |
1da177e4 LT |
487 | if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN) |
488 | goto out_unlock; | |
489 | if (backlog > sk->sk_max_ack_backlog) | |
490 | wake_up_interruptible_all(&u->peer_wait); | |
491 | sk->sk_max_ack_backlog = backlog; | |
492 | sk->sk_state = TCP_LISTEN; | |
493 | /* set credentials so connect can copy them */ | |
109f6e39 | 494 | init_peercred(sk); |
1da177e4 LT |
495 | err = 0; |
496 | ||
497 | out_unlock: | |
1c92b4e5 | 498 | unix_state_unlock(sk); |
109f6e39 EB |
499 | put_pid(old_pid); |
500 | if (old_cred) | |
501 | put_cred(old_cred); | |
1da177e4 LT |
502 | out: |
503 | return err; | |
504 | } | |
505 | ||
506 | static int unix_release(struct socket *); | |
507 | static int unix_bind(struct socket *, struct sockaddr *, int); | |
508 | static int unix_stream_connect(struct socket *, struct sockaddr *, | |
509 | int addr_len, int flags); | |
510 | static int unix_socketpair(struct socket *, struct socket *); | |
511 | static int unix_accept(struct socket *, struct socket *, int); | |
512 | static int unix_getname(struct socket *, struct sockaddr *, int *, int); | |
513 | static unsigned int unix_poll(struct file *, struct socket *, poll_table *); | |
ec0d215f RW |
514 | static unsigned int unix_dgram_poll(struct file *, struct socket *, |
515 | poll_table *); | |
1da177e4 LT |
516 | static int unix_ioctl(struct socket *, unsigned int, unsigned long); |
517 | static int unix_shutdown(struct socket *, int); | |
518 | static int unix_stream_sendmsg(struct kiocb *, struct socket *, | |
519 | struct msghdr *, size_t); | |
520 | static int unix_stream_recvmsg(struct kiocb *, struct socket *, | |
521 | struct msghdr *, size_t, int); | |
522 | static int unix_dgram_sendmsg(struct kiocb *, struct socket *, | |
523 | struct msghdr *, size_t); | |
524 | static int unix_dgram_recvmsg(struct kiocb *, struct socket *, | |
525 | struct msghdr *, size_t, int); | |
526 | static int unix_dgram_connect(struct socket *, struct sockaddr *, | |
527 | int, int); | |
528 | static int unix_seqpacket_sendmsg(struct kiocb *, struct socket *, | |
529 | struct msghdr *, size_t); | |
a05d2ad1 EB |
530 | static int unix_seqpacket_recvmsg(struct kiocb *, struct socket *, |
531 | struct msghdr *, size_t, int); | |
1da177e4 | 532 | |
f55bb7f9 PE |
533 | static void unix_set_peek_off(struct sock *sk, int val) |
534 | { | |
535 | struct unix_sock *u = unix_sk(sk); | |
536 | ||
537 | mutex_lock(&u->readlock); | |
538 | sk->sk_peek_off = val; | |
539 | mutex_unlock(&u->readlock); | |
540 | } | |
541 | ||
542 | ||
90ddc4f0 | 543 | static const struct proto_ops unix_stream_ops = { |
1da177e4 LT |
544 | .family = PF_UNIX, |
545 | .owner = THIS_MODULE, | |
546 | .release = unix_release, | |
547 | .bind = unix_bind, | |
548 | .connect = unix_stream_connect, | |
549 | .socketpair = unix_socketpair, | |
550 | .accept = unix_accept, | |
551 | .getname = unix_getname, | |
552 | .poll = unix_poll, | |
553 | .ioctl = unix_ioctl, | |
554 | .listen = unix_listen, | |
555 | .shutdown = unix_shutdown, | |
556 | .setsockopt = sock_no_setsockopt, | |
557 | .getsockopt = sock_no_getsockopt, | |
558 | .sendmsg = unix_stream_sendmsg, | |
559 | .recvmsg = unix_stream_recvmsg, | |
560 | .mmap = sock_no_mmap, | |
561 | .sendpage = sock_no_sendpage, | |
562 | }; | |
563 | ||
90ddc4f0 | 564 | static const struct proto_ops unix_dgram_ops = { |
1da177e4 LT |
565 | .family = PF_UNIX, |
566 | .owner = THIS_MODULE, | |
567 | .release = unix_release, | |
568 | .bind = unix_bind, | |
569 | .connect = unix_dgram_connect, | |
570 | .socketpair = unix_socketpair, | |
571 | .accept = sock_no_accept, | |
572 | .getname = unix_getname, | |
ec0d215f | 573 | .poll = unix_dgram_poll, |
1da177e4 LT |
574 | .ioctl = unix_ioctl, |
575 | .listen = sock_no_listen, | |
576 | .shutdown = unix_shutdown, | |
577 | .setsockopt = sock_no_setsockopt, | |
578 | .getsockopt = sock_no_getsockopt, | |
579 | .sendmsg = unix_dgram_sendmsg, | |
580 | .recvmsg = unix_dgram_recvmsg, | |
581 | .mmap = sock_no_mmap, | |
582 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 583 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
584 | }; |
585 | ||
90ddc4f0 | 586 | static const struct proto_ops unix_seqpacket_ops = { |
1da177e4 LT |
587 | .family = PF_UNIX, |
588 | .owner = THIS_MODULE, | |
589 | .release = unix_release, | |
590 | .bind = unix_bind, | |
591 | .connect = unix_stream_connect, | |
592 | .socketpair = unix_socketpair, | |
593 | .accept = unix_accept, | |
594 | .getname = unix_getname, | |
ec0d215f | 595 | .poll = unix_dgram_poll, |
1da177e4 LT |
596 | .ioctl = unix_ioctl, |
597 | .listen = unix_listen, | |
598 | .shutdown = unix_shutdown, | |
599 | .setsockopt = sock_no_setsockopt, | |
600 | .getsockopt = sock_no_getsockopt, | |
601 | .sendmsg = unix_seqpacket_sendmsg, | |
a05d2ad1 | 602 | .recvmsg = unix_seqpacket_recvmsg, |
1da177e4 LT |
603 | .mmap = sock_no_mmap, |
604 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 605 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
606 | }; |
607 | ||
608 | static struct proto unix_proto = { | |
248969ae ED |
609 | .name = "UNIX", |
610 | .owner = THIS_MODULE, | |
248969ae | 611 | .obj_size = sizeof(struct unix_sock), |
1da177e4 LT |
612 | }; |
613 | ||
a09785a2 IM |
614 | /* |
615 | * AF_UNIX sockets do not interact with hardware, hence they | |
616 | * dont trigger interrupts - so it's safe for them to have | |
617 | * bh-unsafe locking for their sk_receive_queue.lock. Split off | |
618 | * this special lock-class by reinitializing the spinlock key: | |
619 | */ | |
620 | static struct lock_class_key af_unix_sk_receive_queue_lock_key; | |
621 | ||
6eba6a37 | 622 | static struct sock *unix_create1(struct net *net, struct socket *sock) |
1da177e4 LT |
623 | { |
624 | struct sock *sk = NULL; | |
625 | struct unix_sock *u; | |
626 | ||
518de9b3 ED |
627 | atomic_long_inc(&unix_nr_socks); |
628 | if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files()) | |
1da177e4 LT |
629 | goto out; |
630 | ||
6257ff21 | 631 | sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto); |
1da177e4 LT |
632 | if (!sk) |
633 | goto out; | |
634 | ||
6eba6a37 | 635 | sock_init_data(sock, sk); |
a09785a2 IM |
636 | lockdep_set_class(&sk->sk_receive_queue.lock, |
637 | &af_unix_sk_receive_queue_lock_key); | |
1da177e4 LT |
638 | |
639 | sk->sk_write_space = unix_write_space; | |
a0a53c8b | 640 | sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen; |
1da177e4 LT |
641 | sk->sk_destruct = unix_sock_destructor; |
642 | u = unix_sk(sk); | |
643 | u->dentry = NULL; | |
644 | u->mnt = NULL; | |
fd19f329 | 645 | spin_lock_init(&u->lock); |
516e0cc5 | 646 | atomic_long_set(&u->inflight, 0); |
1fd05ba5 | 647 | INIT_LIST_HEAD(&u->link); |
57b47a53 | 648 | mutex_init(&u->readlock); /* single task reading lock */ |
1da177e4 LT |
649 | init_waitqueue_head(&u->peer_wait); |
650 | unix_insert_socket(unix_sockets_unbound, sk); | |
651 | out: | |
284b327b | 652 | if (sk == NULL) |
518de9b3 | 653 | atomic_long_dec(&unix_nr_socks); |
920de804 ED |
654 | else { |
655 | local_bh_disable(); | |
a8076d8d | 656 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
920de804 ED |
657 | local_bh_enable(); |
658 | } | |
1da177e4 LT |
659 | return sk; |
660 | } | |
661 | ||
3f378b68 EP |
662 | static int unix_create(struct net *net, struct socket *sock, int protocol, |
663 | int kern) | |
1da177e4 LT |
664 | { |
665 | if (protocol && protocol != PF_UNIX) | |
666 | return -EPROTONOSUPPORT; | |
667 | ||
668 | sock->state = SS_UNCONNECTED; | |
669 | ||
670 | switch (sock->type) { | |
671 | case SOCK_STREAM: | |
672 | sock->ops = &unix_stream_ops; | |
673 | break; | |
674 | /* | |
675 | * Believe it or not BSD has AF_UNIX, SOCK_RAW though | |
676 | * nothing uses it. | |
677 | */ | |
678 | case SOCK_RAW: | |
e27dfcea | 679 | sock->type = SOCK_DGRAM; |
1da177e4 LT |
680 | case SOCK_DGRAM: |
681 | sock->ops = &unix_dgram_ops; | |
682 | break; | |
683 | case SOCK_SEQPACKET: | |
684 | sock->ops = &unix_seqpacket_ops; | |
685 | break; | |
686 | default: | |
687 | return -ESOCKTNOSUPPORT; | |
688 | } | |
689 | ||
1b8d7ae4 | 690 | return unix_create1(net, sock) ? 0 : -ENOMEM; |
1da177e4 LT |
691 | } |
692 | ||
693 | static int unix_release(struct socket *sock) | |
694 | { | |
695 | struct sock *sk = sock->sk; | |
696 | ||
697 | if (!sk) | |
698 | return 0; | |
699 | ||
700 | sock->sk = NULL; | |
701 | ||
6eba6a37 | 702 | return unix_release_sock(sk, 0); |
1da177e4 LT |
703 | } |
704 | ||
705 | static int unix_autobind(struct socket *sock) | |
706 | { | |
707 | struct sock *sk = sock->sk; | |
3b1e0a65 | 708 | struct net *net = sock_net(sk); |
1da177e4 LT |
709 | struct unix_sock *u = unix_sk(sk); |
710 | static u32 ordernum = 1; | |
6eba6a37 | 711 | struct unix_address *addr; |
1da177e4 | 712 | int err; |
8df73ff9 | 713 | unsigned int retries = 0; |
1da177e4 | 714 | |
57b47a53 | 715 | mutex_lock(&u->readlock); |
1da177e4 LT |
716 | |
717 | err = 0; | |
718 | if (u->addr) | |
719 | goto out; | |
720 | ||
721 | err = -ENOMEM; | |
0da974f4 | 722 | addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL); |
1da177e4 LT |
723 | if (!addr) |
724 | goto out; | |
725 | ||
1da177e4 LT |
726 | addr->name->sun_family = AF_UNIX; |
727 | atomic_set(&addr->refcnt, 1); | |
728 | ||
729 | retry: | |
730 | addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short); | |
07f0757a | 731 | addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0)); |
1da177e4 | 732 | |
fbe9cc4a | 733 | spin_lock(&unix_table_lock); |
1da177e4 LT |
734 | ordernum = (ordernum+1)&0xFFFFF; |
735 | ||
097e66c5 | 736 | if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type, |
1da177e4 | 737 | addr->hash)) { |
fbe9cc4a | 738 | spin_unlock(&unix_table_lock); |
8df73ff9 TH |
739 | /* |
740 | * __unix_find_socket_byname() may take long time if many names | |
741 | * are already in use. | |
742 | */ | |
743 | cond_resched(); | |
744 | /* Give up if all names seems to be in use. */ | |
745 | if (retries++ == 0xFFFFF) { | |
746 | err = -ENOSPC; | |
747 | kfree(addr); | |
748 | goto out; | |
749 | } | |
1da177e4 LT |
750 | goto retry; |
751 | } | |
752 | addr->hash ^= sk->sk_type; | |
753 | ||
754 | __unix_remove_socket(sk); | |
755 | u->addr = addr; | |
756 | __unix_insert_socket(&unix_socket_table[addr->hash], sk); | |
fbe9cc4a | 757 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
758 | err = 0; |
759 | ||
57b47a53 | 760 | out: mutex_unlock(&u->readlock); |
1da177e4 LT |
761 | return err; |
762 | } | |
763 | ||
097e66c5 DL |
764 | static struct sock *unix_find_other(struct net *net, |
765 | struct sockaddr_un *sunname, int len, | |
1da177e4 LT |
766 | int type, unsigned hash, int *error) |
767 | { | |
768 | struct sock *u; | |
421748ec | 769 | struct path path; |
1da177e4 | 770 | int err = 0; |
ac7bfa62 | 771 | |
1da177e4 | 772 | if (sunname->sun_path[0]) { |
421748ec AV |
773 | struct inode *inode; |
774 | err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path); | |
1da177e4 LT |
775 | if (err) |
776 | goto fail; | |
421748ec AV |
777 | inode = path.dentry->d_inode; |
778 | err = inode_permission(inode, MAY_WRITE); | |
1da177e4 LT |
779 | if (err) |
780 | goto put_fail; | |
781 | ||
782 | err = -ECONNREFUSED; | |
421748ec | 783 | if (!S_ISSOCK(inode->i_mode)) |
1da177e4 | 784 | goto put_fail; |
6616f788 | 785 | u = unix_find_socket_byinode(inode); |
1da177e4 LT |
786 | if (!u) |
787 | goto put_fail; | |
788 | ||
789 | if (u->sk_type == type) | |
421748ec | 790 | touch_atime(path.mnt, path.dentry); |
1da177e4 | 791 | |
421748ec | 792 | path_put(&path); |
1da177e4 | 793 | |
e27dfcea | 794 | err = -EPROTOTYPE; |
1da177e4 LT |
795 | if (u->sk_type != type) { |
796 | sock_put(u); | |
797 | goto fail; | |
798 | } | |
799 | } else { | |
800 | err = -ECONNREFUSED; | |
e27dfcea | 801 | u = unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
802 | if (u) { |
803 | struct dentry *dentry; | |
804 | dentry = unix_sk(u)->dentry; | |
805 | if (dentry) | |
806 | touch_atime(unix_sk(u)->mnt, dentry); | |
807 | } else | |
808 | goto fail; | |
809 | } | |
810 | return u; | |
811 | ||
812 | put_fail: | |
421748ec | 813 | path_put(&path); |
1da177e4 | 814 | fail: |
e27dfcea | 815 | *error = err; |
1da177e4 LT |
816 | return NULL; |
817 | } | |
818 | ||
819 | ||
820 | static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
821 | { | |
822 | struct sock *sk = sock->sk; | |
3b1e0a65 | 823 | struct net *net = sock_net(sk); |
1da177e4 | 824 | struct unix_sock *u = unix_sk(sk); |
e27dfcea | 825 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
dae6ad8f | 826 | char *sun_path = sunaddr->sun_path; |
6eba6a37 | 827 | struct dentry *dentry = NULL; |
dae6ad8f | 828 | struct path path; |
1da177e4 LT |
829 | int err; |
830 | unsigned hash; | |
831 | struct unix_address *addr; | |
832 | struct hlist_head *list; | |
833 | ||
834 | err = -EINVAL; | |
835 | if (sunaddr->sun_family != AF_UNIX) | |
836 | goto out; | |
837 | ||
e27dfcea | 838 | if (addr_len == sizeof(short)) { |
1da177e4 LT |
839 | err = unix_autobind(sock); |
840 | goto out; | |
841 | } | |
842 | ||
843 | err = unix_mkname(sunaddr, addr_len, &hash); | |
844 | if (err < 0) | |
845 | goto out; | |
846 | addr_len = err; | |
847 | ||
57b47a53 | 848 | mutex_lock(&u->readlock); |
1da177e4 LT |
849 | |
850 | err = -EINVAL; | |
851 | if (u->addr) | |
852 | goto out_up; | |
853 | ||
854 | err = -ENOMEM; | |
855 | addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL); | |
856 | if (!addr) | |
857 | goto out_up; | |
858 | ||
859 | memcpy(addr->name, sunaddr, addr_len); | |
860 | addr->len = addr_len; | |
861 | addr->hash = hash ^ sk->sk_type; | |
862 | atomic_set(&addr->refcnt, 1); | |
863 | ||
dae6ad8f | 864 | if (sun_path[0]) { |
04fc66e7 | 865 | umode_t mode; |
1da177e4 LT |
866 | err = 0; |
867 | /* | |
868 | * Get the parent directory, calculate the hash for last | |
869 | * component. | |
870 | */ | |
dae6ad8f | 871 | dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0); |
1da177e4 LT |
872 | err = PTR_ERR(dentry); |
873 | if (IS_ERR(dentry)) | |
dae6ad8f | 874 | goto out_mknod_parent; |
f81a0bff | 875 | |
1da177e4 LT |
876 | /* |
877 | * All right, let's create it. | |
878 | */ | |
879 | mode = S_IFSOCK | | |
ce3b0f8d | 880 | (SOCK_INODE(sock)->i_mode & ~current_umask()); |
dae6ad8f | 881 | err = mnt_want_write(path.mnt); |
463c3197 DH |
882 | if (err) |
883 | goto out_mknod_dput; | |
dae6ad8f | 884 | err = security_path_mknod(&path, dentry, mode, 0); |
be6d3e56 KT |
885 | if (err) |
886 | goto out_mknod_drop_write; | |
dae6ad8f | 887 | err = vfs_mknod(path.dentry->d_inode, dentry, mode, 0); |
be6d3e56 | 888 | out_mknod_drop_write: |
dae6ad8f | 889 | mnt_drop_write(path.mnt); |
1da177e4 LT |
890 | if (err) |
891 | goto out_mknod_dput; | |
dae6ad8f AV |
892 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
893 | dput(path.dentry); | |
894 | path.dentry = dentry; | |
1da177e4 LT |
895 | |
896 | addr->hash = UNIX_HASH_SIZE; | |
897 | } | |
898 | ||
fbe9cc4a | 899 | spin_lock(&unix_table_lock); |
1da177e4 | 900 | |
dae6ad8f | 901 | if (!sun_path[0]) { |
1da177e4 | 902 | err = -EADDRINUSE; |
097e66c5 | 903 | if (__unix_find_socket_byname(net, sunaddr, addr_len, |
1da177e4 LT |
904 | sk->sk_type, hash)) { |
905 | unix_release_addr(addr); | |
906 | goto out_unlock; | |
907 | } | |
908 | ||
909 | list = &unix_socket_table[addr->hash]; | |
910 | } else { | |
911 | list = &unix_socket_table[dentry->d_inode->i_ino & (UNIX_HASH_SIZE-1)]; | |
dae6ad8f AV |
912 | u->dentry = path.dentry; |
913 | u->mnt = path.mnt; | |
1da177e4 LT |
914 | } |
915 | ||
916 | err = 0; | |
917 | __unix_remove_socket(sk); | |
918 | u->addr = addr; | |
919 | __unix_insert_socket(list, sk); | |
920 | ||
921 | out_unlock: | |
fbe9cc4a | 922 | spin_unlock(&unix_table_lock); |
1da177e4 | 923 | out_up: |
57b47a53 | 924 | mutex_unlock(&u->readlock); |
1da177e4 LT |
925 | out: |
926 | return err; | |
927 | ||
928 | out_mknod_dput: | |
929 | dput(dentry); | |
dae6ad8f AV |
930 | mutex_unlock(&path.dentry->d_inode->i_mutex); |
931 | path_put(&path); | |
1da177e4 | 932 | out_mknod_parent: |
e27dfcea JK |
933 | if (err == -EEXIST) |
934 | err = -EADDRINUSE; | |
1da177e4 LT |
935 | unix_release_addr(addr); |
936 | goto out_up; | |
937 | } | |
938 | ||
278a3de5 DM |
939 | static void unix_state_double_lock(struct sock *sk1, struct sock *sk2) |
940 | { | |
941 | if (unlikely(sk1 == sk2) || !sk2) { | |
942 | unix_state_lock(sk1); | |
943 | return; | |
944 | } | |
945 | if (sk1 < sk2) { | |
946 | unix_state_lock(sk1); | |
947 | unix_state_lock_nested(sk2); | |
948 | } else { | |
949 | unix_state_lock(sk2); | |
950 | unix_state_lock_nested(sk1); | |
951 | } | |
952 | } | |
953 | ||
954 | static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2) | |
955 | { | |
956 | if (unlikely(sk1 == sk2) || !sk2) { | |
957 | unix_state_unlock(sk1); | |
958 | return; | |
959 | } | |
960 | unix_state_unlock(sk1); | |
961 | unix_state_unlock(sk2); | |
962 | } | |
963 | ||
1da177e4 LT |
964 | static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr, |
965 | int alen, int flags) | |
966 | { | |
967 | struct sock *sk = sock->sk; | |
3b1e0a65 | 968 | struct net *net = sock_net(sk); |
e27dfcea | 969 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr; |
1da177e4 LT |
970 | struct sock *other; |
971 | unsigned hash; | |
972 | int err; | |
973 | ||
974 | if (addr->sa_family != AF_UNSPEC) { | |
975 | err = unix_mkname(sunaddr, alen, &hash); | |
976 | if (err < 0) | |
977 | goto out; | |
978 | alen = err; | |
979 | ||
980 | if (test_bit(SOCK_PASSCRED, &sock->flags) && | |
981 | !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0) | |
982 | goto out; | |
983 | ||
278a3de5 | 984 | restart: |
e27dfcea | 985 | other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err); |
1da177e4 LT |
986 | if (!other) |
987 | goto out; | |
988 | ||
278a3de5 DM |
989 | unix_state_double_lock(sk, other); |
990 | ||
991 | /* Apparently VFS overslept socket death. Retry. */ | |
992 | if (sock_flag(other, SOCK_DEAD)) { | |
993 | unix_state_double_unlock(sk, other); | |
994 | sock_put(other); | |
995 | goto restart; | |
996 | } | |
1da177e4 LT |
997 | |
998 | err = -EPERM; | |
999 | if (!unix_may_send(sk, other)) | |
1000 | goto out_unlock; | |
1001 | ||
1002 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1003 | if (err) | |
1004 | goto out_unlock; | |
1005 | ||
1006 | } else { | |
1007 | /* | |
1008 | * 1003.1g breaking connected state with AF_UNSPEC | |
1009 | */ | |
1010 | other = NULL; | |
278a3de5 | 1011 | unix_state_double_lock(sk, other); |
1da177e4 LT |
1012 | } |
1013 | ||
1014 | /* | |
1015 | * If it was connected, reconnect. | |
1016 | */ | |
1017 | if (unix_peer(sk)) { | |
1018 | struct sock *old_peer = unix_peer(sk); | |
e27dfcea | 1019 | unix_peer(sk) = other; |
278a3de5 | 1020 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1021 | |
1022 | if (other != old_peer) | |
1023 | unix_dgram_disconnected(sk, old_peer); | |
1024 | sock_put(old_peer); | |
1025 | } else { | |
e27dfcea | 1026 | unix_peer(sk) = other; |
278a3de5 | 1027 | unix_state_double_unlock(sk, other); |
1da177e4 | 1028 | } |
ac7bfa62 | 1029 | return 0; |
1da177e4 LT |
1030 | |
1031 | out_unlock: | |
278a3de5 | 1032 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1033 | sock_put(other); |
1034 | out: | |
1035 | return err; | |
1036 | } | |
1037 | ||
1038 | static long unix_wait_for_peer(struct sock *other, long timeo) | |
1039 | { | |
1040 | struct unix_sock *u = unix_sk(other); | |
1041 | int sched; | |
1042 | DEFINE_WAIT(wait); | |
1043 | ||
1044 | prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE); | |
1045 | ||
1046 | sched = !sock_flag(other, SOCK_DEAD) && | |
1047 | !(other->sk_shutdown & RCV_SHUTDOWN) && | |
3c73419c | 1048 | unix_recvq_full(other); |
1da177e4 | 1049 | |
1c92b4e5 | 1050 | unix_state_unlock(other); |
1da177e4 LT |
1051 | |
1052 | if (sched) | |
1053 | timeo = schedule_timeout(timeo); | |
1054 | ||
1055 | finish_wait(&u->peer_wait, &wait); | |
1056 | return timeo; | |
1057 | } | |
1058 | ||
1059 | static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
1060 | int addr_len, int flags) | |
1061 | { | |
e27dfcea | 1062 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
1da177e4 | 1063 | struct sock *sk = sock->sk; |
3b1e0a65 | 1064 | struct net *net = sock_net(sk); |
1da177e4 LT |
1065 | struct unix_sock *u = unix_sk(sk), *newu, *otheru; |
1066 | struct sock *newsk = NULL; | |
1067 | struct sock *other = NULL; | |
1068 | struct sk_buff *skb = NULL; | |
1069 | unsigned hash; | |
1070 | int st; | |
1071 | int err; | |
1072 | long timeo; | |
1073 | ||
1074 | err = unix_mkname(sunaddr, addr_len, &hash); | |
1075 | if (err < 0) | |
1076 | goto out; | |
1077 | addr_len = err; | |
1078 | ||
f64f9e71 JP |
1079 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr && |
1080 | (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1081 | goto out; |
1082 | ||
1083 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
1084 | ||
1085 | /* First of all allocate resources. | |
1086 | If we will make it after state is locked, | |
1087 | we will have to recheck all again in any case. | |
1088 | */ | |
1089 | ||
1090 | err = -ENOMEM; | |
1091 | ||
1092 | /* create new sock for complete connection */ | |
3b1e0a65 | 1093 | newsk = unix_create1(sock_net(sk), NULL); |
1da177e4 LT |
1094 | if (newsk == NULL) |
1095 | goto out; | |
1096 | ||
1097 | /* Allocate skb for sending to listening sock */ | |
1098 | skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL); | |
1099 | if (skb == NULL) | |
1100 | goto out; | |
1101 | ||
1102 | restart: | |
1103 | /* Find listening sock. */ | |
097e66c5 | 1104 | other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err); |
1da177e4 LT |
1105 | if (!other) |
1106 | goto out; | |
1107 | ||
1108 | /* Latch state of peer */ | |
1c92b4e5 | 1109 | unix_state_lock(other); |
1da177e4 LT |
1110 | |
1111 | /* Apparently VFS overslept socket death. Retry. */ | |
1112 | if (sock_flag(other, SOCK_DEAD)) { | |
1c92b4e5 | 1113 | unix_state_unlock(other); |
1da177e4 LT |
1114 | sock_put(other); |
1115 | goto restart; | |
1116 | } | |
1117 | ||
1118 | err = -ECONNREFUSED; | |
1119 | if (other->sk_state != TCP_LISTEN) | |
1120 | goto out_unlock; | |
77238f2b TS |
1121 | if (other->sk_shutdown & RCV_SHUTDOWN) |
1122 | goto out_unlock; | |
1da177e4 | 1123 | |
3c73419c | 1124 | if (unix_recvq_full(other)) { |
1da177e4 LT |
1125 | err = -EAGAIN; |
1126 | if (!timeo) | |
1127 | goto out_unlock; | |
1128 | ||
1129 | timeo = unix_wait_for_peer(other, timeo); | |
1130 | ||
1131 | err = sock_intr_errno(timeo); | |
1132 | if (signal_pending(current)) | |
1133 | goto out; | |
1134 | sock_put(other); | |
1135 | goto restart; | |
ac7bfa62 | 1136 | } |
1da177e4 LT |
1137 | |
1138 | /* Latch our state. | |
1139 | ||
e5537bfc | 1140 | It is tricky place. We need to grab our state lock and cannot |
1da177e4 LT |
1141 | drop lock on peer. It is dangerous because deadlock is |
1142 | possible. Connect to self case and simultaneous | |
1143 | attempt to connect are eliminated by checking socket | |
1144 | state. other is TCP_LISTEN, if sk is TCP_LISTEN we | |
1145 | check this before attempt to grab lock. | |
1146 | ||
1147 | Well, and we have to recheck the state after socket locked. | |
1148 | */ | |
1149 | st = sk->sk_state; | |
1150 | ||
1151 | switch (st) { | |
1152 | case TCP_CLOSE: | |
1153 | /* This is ok... continue with connect */ | |
1154 | break; | |
1155 | case TCP_ESTABLISHED: | |
1156 | /* Socket is already connected */ | |
1157 | err = -EISCONN; | |
1158 | goto out_unlock; | |
1159 | default: | |
1160 | err = -EINVAL; | |
1161 | goto out_unlock; | |
1162 | } | |
1163 | ||
1c92b4e5 | 1164 | unix_state_lock_nested(sk); |
1da177e4 LT |
1165 | |
1166 | if (sk->sk_state != st) { | |
1c92b4e5 DM |
1167 | unix_state_unlock(sk); |
1168 | unix_state_unlock(other); | |
1da177e4 LT |
1169 | sock_put(other); |
1170 | goto restart; | |
1171 | } | |
1172 | ||
3610cda5 | 1173 | err = security_unix_stream_connect(sk, other, newsk); |
1da177e4 | 1174 | if (err) { |
1c92b4e5 | 1175 | unix_state_unlock(sk); |
1da177e4 LT |
1176 | goto out_unlock; |
1177 | } | |
1178 | ||
1179 | /* The way is open! Fastly set all the necessary fields... */ | |
1180 | ||
1181 | sock_hold(sk); | |
1182 | unix_peer(newsk) = sk; | |
1183 | newsk->sk_state = TCP_ESTABLISHED; | |
1184 | newsk->sk_type = sk->sk_type; | |
109f6e39 | 1185 | init_peercred(newsk); |
1da177e4 | 1186 | newu = unix_sk(newsk); |
eaefd110 | 1187 | RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq); |
1da177e4 LT |
1188 | otheru = unix_sk(other); |
1189 | ||
1190 | /* copy address information from listening to new sock*/ | |
1191 | if (otheru->addr) { | |
1192 | atomic_inc(&otheru->addr->refcnt); | |
1193 | newu->addr = otheru->addr; | |
1194 | } | |
1195 | if (otheru->dentry) { | |
1196 | newu->dentry = dget(otheru->dentry); | |
1197 | newu->mnt = mntget(otheru->mnt); | |
1198 | } | |
1199 | ||
1200 | /* Set credentials */ | |
109f6e39 | 1201 | copy_peercred(sk, other); |
1da177e4 | 1202 | |
1da177e4 LT |
1203 | sock->state = SS_CONNECTED; |
1204 | sk->sk_state = TCP_ESTABLISHED; | |
830a1e5c BL |
1205 | sock_hold(newsk); |
1206 | ||
1207 | smp_mb__after_atomic_inc(); /* sock_hold() does an atomic_inc() */ | |
1208 | unix_peer(sk) = newsk; | |
1da177e4 | 1209 | |
1c92b4e5 | 1210 | unix_state_unlock(sk); |
1da177e4 LT |
1211 | |
1212 | /* take ten and and send info to listening sock */ | |
1213 | spin_lock(&other->sk_receive_queue.lock); | |
1214 | __skb_queue_tail(&other->sk_receive_queue, skb); | |
1da177e4 | 1215 | spin_unlock(&other->sk_receive_queue.lock); |
1c92b4e5 | 1216 | unix_state_unlock(other); |
1da177e4 LT |
1217 | other->sk_data_ready(other, 0); |
1218 | sock_put(other); | |
1219 | return 0; | |
1220 | ||
1221 | out_unlock: | |
1222 | if (other) | |
1c92b4e5 | 1223 | unix_state_unlock(other); |
1da177e4 LT |
1224 | |
1225 | out: | |
40d44446 | 1226 | kfree_skb(skb); |
1da177e4 LT |
1227 | if (newsk) |
1228 | unix_release_sock(newsk, 0); | |
1229 | if (other) | |
1230 | sock_put(other); | |
1231 | return err; | |
1232 | } | |
1233 | ||
1234 | static int unix_socketpair(struct socket *socka, struct socket *sockb) | |
1235 | { | |
e27dfcea | 1236 | struct sock *ska = socka->sk, *skb = sockb->sk; |
1da177e4 LT |
1237 | |
1238 | /* Join our sockets back to back */ | |
1239 | sock_hold(ska); | |
1240 | sock_hold(skb); | |
e27dfcea JK |
1241 | unix_peer(ska) = skb; |
1242 | unix_peer(skb) = ska; | |
109f6e39 EB |
1243 | init_peercred(ska); |
1244 | init_peercred(skb); | |
1da177e4 LT |
1245 | |
1246 | if (ska->sk_type != SOCK_DGRAM) { | |
1247 | ska->sk_state = TCP_ESTABLISHED; | |
1248 | skb->sk_state = TCP_ESTABLISHED; | |
1249 | socka->state = SS_CONNECTED; | |
1250 | sockb->state = SS_CONNECTED; | |
1251 | } | |
1252 | return 0; | |
1253 | } | |
1254 | ||
1255 | static int unix_accept(struct socket *sock, struct socket *newsock, int flags) | |
1256 | { | |
1257 | struct sock *sk = sock->sk; | |
1258 | struct sock *tsk; | |
1259 | struct sk_buff *skb; | |
1260 | int err; | |
1261 | ||
1262 | err = -EOPNOTSUPP; | |
6eba6a37 | 1263 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
1da177e4 LT |
1264 | goto out; |
1265 | ||
1266 | err = -EINVAL; | |
1267 | if (sk->sk_state != TCP_LISTEN) | |
1268 | goto out; | |
1269 | ||
1270 | /* If socket state is TCP_LISTEN it cannot change (for now...), | |
1271 | * so that no locks are necessary. | |
1272 | */ | |
1273 | ||
1274 | skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err); | |
1275 | if (!skb) { | |
1276 | /* This means receive shutdown. */ | |
1277 | if (err == 0) | |
1278 | err = -EINVAL; | |
1279 | goto out; | |
1280 | } | |
1281 | ||
1282 | tsk = skb->sk; | |
1283 | skb_free_datagram(sk, skb); | |
1284 | wake_up_interruptible(&unix_sk(sk)->peer_wait); | |
1285 | ||
1286 | /* attach accepted sock to socket */ | |
1c92b4e5 | 1287 | unix_state_lock(tsk); |
1da177e4 LT |
1288 | newsock->state = SS_CONNECTED; |
1289 | sock_graft(tsk, newsock); | |
1c92b4e5 | 1290 | unix_state_unlock(tsk); |
1da177e4 LT |
1291 | return 0; |
1292 | ||
1293 | out: | |
1294 | return err; | |
1295 | } | |
1296 | ||
1297 | ||
1298 | static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer) | |
1299 | { | |
1300 | struct sock *sk = sock->sk; | |
1301 | struct unix_sock *u; | |
13cfa97b | 1302 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr); |
1da177e4 LT |
1303 | int err = 0; |
1304 | ||
1305 | if (peer) { | |
1306 | sk = unix_peer_get(sk); | |
1307 | ||
1308 | err = -ENOTCONN; | |
1309 | if (!sk) | |
1310 | goto out; | |
1311 | err = 0; | |
1312 | } else { | |
1313 | sock_hold(sk); | |
1314 | } | |
1315 | ||
1316 | u = unix_sk(sk); | |
1c92b4e5 | 1317 | unix_state_lock(sk); |
1da177e4 LT |
1318 | if (!u->addr) { |
1319 | sunaddr->sun_family = AF_UNIX; | |
1320 | sunaddr->sun_path[0] = 0; | |
1321 | *uaddr_len = sizeof(short); | |
1322 | } else { | |
1323 | struct unix_address *addr = u->addr; | |
1324 | ||
1325 | *uaddr_len = addr->len; | |
1326 | memcpy(sunaddr, addr->name, *uaddr_len); | |
1327 | } | |
1c92b4e5 | 1328 | unix_state_unlock(sk); |
1da177e4 LT |
1329 | sock_put(sk); |
1330 | out: | |
1331 | return err; | |
1332 | } | |
1333 | ||
1334 | static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb) | |
1335 | { | |
1336 | int i; | |
1337 | ||
1338 | scm->fp = UNIXCB(skb).fp; | |
1da177e4 LT |
1339 | UNIXCB(skb).fp = NULL; |
1340 | ||
6eba6a37 | 1341 | for (i = scm->fp->count-1; i >= 0; i--) |
1da177e4 LT |
1342 | unix_notinflight(scm->fp->fp[i]); |
1343 | } | |
1344 | ||
7361c36c | 1345 | static void unix_destruct_scm(struct sk_buff *skb) |
1da177e4 LT |
1346 | { |
1347 | struct scm_cookie scm; | |
1348 | memset(&scm, 0, sizeof(scm)); | |
7361c36c EB |
1349 | scm.pid = UNIXCB(skb).pid; |
1350 | scm.cred = UNIXCB(skb).cred; | |
1351 | if (UNIXCB(skb).fp) | |
1352 | unix_detach_fds(&scm, skb); | |
1da177e4 LT |
1353 | |
1354 | /* Alas, it calls VFS */ | |
1355 | /* So fscking what? fput() had been SMP-safe since the last Summer */ | |
1356 | scm_destroy(&scm); | |
1357 | sock_wfree(skb); | |
1358 | } | |
1359 | ||
25888e30 ED |
1360 | #define MAX_RECURSION_LEVEL 4 |
1361 | ||
6209344f | 1362 | static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb) |
1da177e4 LT |
1363 | { |
1364 | int i; | |
25888e30 ED |
1365 | unsigned char max_level = 0; |
1366 | int unix_sock_count = 0; | |
1367 | ||
1368 | for (i = scm->fp->count - 1; i >= 0; i--) { | |
1369 | struct sock *sk = unix_get_socket(scm->fp->fp[i]); | |
1370 | ||
1371 | if (sk) { | |
1372 | unix_sock_count++; | |
1373 | max_level = max(max_level, | |
1374 | unix_sk(sk)->recursion_level); | |
1375 | } | |
1376 | } | |
1377 | if (unlikely(max_level > MAX_RECURSION_LEVEL)) | |
1378 | return -ETOOMANYREFS; | |
6209344f MS |
1379 | |
1380 | /* | |
1381 | * Need to duplicate file references for the sake of garbage | |
1382 | * collection. Otherwise a socket in the fps might become a | |
1383 | * candidate for GC while the skb is not yet queued. | |
1384 | */ | |
1385 | UNIXCB(skb).fp = scm_fp_dup(scm->fp); | |
1386 | if (!UNIXCB(skb).fp) | |
1387 | return -ENOMEM; | |
1388 | ||
25888e30 ED |
1389 | if (unix_sock_count) { |
1390 | for (i = scm->fp->count - 1; i >= 0; i--) | |
1391 | unix_inflight(scm->fp->fp[i]); | |
1392 | } | |
1393 | return max_level; | |
1da177e4 LT |
1394 | } |
1395 | ||
f78a5fda | 1396 | static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds) |
7361c36c EB |
1397 | { |
1398 | int err = 0; | |
16e57262 | 1399 | |
f78a5fda | 1400 | UNIXCB(skb).pid = get_pid(scm->pid); |
16e57262 ED |
1401 | if (scm->cred) |
1402 | UNIXCB(skb).cred = get_cred(scm->cred); | |
7361c36c EB |
1403 | UNIXCB(skb).fp = NULL; |
1404 | if (scm->fp && send_fds) | |
1405 | err = unix_attach_fds(scm, skb); | |
1406 | ||
1407 | skb->destructor = unix_destruct_scm; | |
1408 | return err; | |
1409 | } | |
1410 | ||
16e57262 ED |
1411 | /* |
1412 | * Some apps rely on write() giving SCM_CREDENTIALS | |
1413 | * We include credentials if source or destination socket | |
1414 | * asserted SOCK_PASSCRED. | |
1415 | */ | |
1416 | static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock, | |
1417 | const struct sock *other) | |
1418 | { | |
1419 | if (UNIXCB(skb).cred) | |
1420 | return; | |
1421 | if (test_bit(SOCK_PASSCRED, &sock->flags) || | |
1422 | !other->sk_socket || | |
1423 | test_bit(SOCK_PASSCRED, &other->sk_socket->flags)) { | |
1424 | UNIXCB(skb).pid = get_pid(task_tgid(current)); | |
1425 | UNIXCB(skb).cred = get_current_cred(); | |
1426 | } | |
1427 | } | |
1428 | ||
1da177e4 LT |
1429 | /* |
1430 | * Send AF_UNIX data. | |
1431 | */ | |
1432 | ||
1433 | static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock, | |
1434 | struct msghdr *msg, size_t len) | |
1435 | { | |
1436 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
1437 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1438 | struct net *net = sock_net(sk); |
1da177e4 | 1439 | struct unix_sock *u = unix_sk(sk); |
e27dfcea | 1440 | struct sockaddr_un *sunaddr = msg->msg_name; |
1da177e4 LT |
1441 | struct sock *other = NULL; |
1442 | int namelen = 0; /* fake GCC */ | |
1443 | int err; | |
1444 | unsigned hash; | |
f78a5fda | 1445 | struct sk_buff *skb; |
1da177e4 LT |
1446 | long timeo; |
1447 | struct scm_cookie tmp_scm; | |
25888e30 | 1448 | int max_level; |
1da177e4 LT |
1449 | |
1450 | if (NULL == siocb->scm) | |
1451 | siocb->scm = &tmp_scm; | |
5f23b734 | 1452 | wait_for_unix_gc(); |
1da177e4 LT |
1453 | err = scm_send(sock, msg, siocb->scm); |
1454 | if (err < 0) | |
1455 | return err; | |
1456 | ||
1457 | err = -EOPNOTSUPP; | |
1458 | if (msg->msg_flags&MSG_OOB) | |
1459 | goto out; | |
1460 | ||
1461 | if (msg->msg_namelen) { | |
1462 | err = unix_mkname(sunaddr, msg->msg_namelen, &hash); | |
1463 | if (err < 0) | |
1464 | goto out; | |
1465 | namelen = err; | |
1466 | } else { | |
1467 | sunaddr = NULL; | |
1468 | err = -ENOTCONN; | |
1469 | other = unix_peer_get(sk); | |
1470 | if (!other) | |
1471 | goto out; | |
1472 | } | |
1473 | ||
f64f9e71 JP |
1474 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr |
1475 | && (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1476 | goto out; |
1477 | ||
1478 | err = -EMSGSIZE; | |
1479 | if (len > sk->sk_sndbuf - 32) | |
1480 | goto out; | |
1481 | ||
1482 | skb = sock_alloc_send_skb(sk, len, msg->msg_flags&MSG_DONTWAIT, &err); | |
e27dfcea | 1483 | if (skb == NULL) |
1da177e4 LT |
1484 | goto out; |
1485 | ||
f78a5fda | 1486 | err = unix_scm_to_skb(siocb->scm, skb, true); |
25888e30 | 1487 | if (err < 0) |
7361c36c | 1488 | goto out_free; |
25888e30 | 1489 | max_level = err + 1; |
dc49c1f9 | 1490 | unix_get_secdata(siocb->scm, skb); |
877ce7c1 | 1491 | |
badff6d0 | 1492 | skb_reset_transport_header(skb); |
6eba6a37 | 1493 | err = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len); |
1da177e4 LT |
1494 | if (err) |
1495 | goto out_free; | |
1496 | ||
1497 | timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); | |
1498 | ||
1499 | restart: | |
1500 | if (!other) { | |
1501 | err = -ECONNRESET; | |
1502 | if (sunaddr == NULL) | |
1503 | goto out_free; | |
1504 | ||
097e66c5 | 1505 | other = unix_find_other(net, sunaddr, namelen, sk->sk_type, |
1da177e4 | 1506 | hash, &err); |
e27dfcea | 1507 | if (other == NULL) |
1da177e4 LT |
1508 | goto out_free; |
1509 | } | |
1510 | ||
d6ae3bae AC |
1511 | if (sk_filter(other, skb) < 0) { |
1512 | /* Toss the packet but do not return any error to the sender */ | |
1513 | err = len; | |
1514 | goto out_free; | |
1515 | } | |
1516 | ||
1c92b4e5 | 1517 | unix_state_lock(other); |
1da177e4 LT |
1518 | err = -EPERM; |
1519 | if (!unix_may_send(sk, other)) | |
1520 | goto out_unlock; | |
1521 | ||
1522 | if (sock_flag(other, SOCK_DEAD)) { | |
1523 | /* | |
1524 | * Check with 1003.1g - what should | |
1525 | * datagram error | |
1526 | */ | |
1c92b4e5 | 1527 | unix_state_unlock(other); |
1da177e4 LT |
1528 | sock_put(other); |
1529 | ||
1530 | err = 0; | |
1c92b4e5 | 1531 | unix_state_lock(sk); |
1da177e4 | 1532 | if (unix_peer(sk) == other) { |
e27dfcea | 1533 | unix_peer(sk) = NULL; |
1c92b4e5 | 1534 | unix_state_unlock(sk); |
1da177e4 LT |
1535 | |
1536 | unix_dgram_disconnected(sk, other); | |
1537 | sock_put(other); | |
1538 | err = -ECONNREFUSED; | |
1539 | } else { | |
1c92b4e5 | 1540 | unix_state_unlock(sk); |
1da177e4 LT |
1541 | } |
1542 | ||
1543 | other = NULL; | |
1544 | if (err) | |
1545 | goto out_free; | |
1546 | goto restart; | |
1547 | } | |
1548 | ||
1549 | err = -EPIPE; | |
1550 | if (other->sk_shutdown & RCV_SHUTDOWN) | |
1551 | goto out_unlock; | |
1552 | ||
1553 | if (sk->sk_type != SOCK_SEQPACKET) { | |
1554 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1555 | if (err) | |
1556 | goto out_unlock; | |
1557 | } | |
1558 | ||
3c73419c | 1559 | if (unix_peer(other) != sk && unix_recvq_full(other)) { |
1da177e4 LT |
1560 | if (!timeo) { |
1561 | err = -EAGAIN; | |
1562 | goto out_unlock; | |
1563 | } | |
1564 | ||
1565 | timeo = unix_wait_for_peer(other, timeo); | |
1566 | ||
1567 | err = sock_intr_errno(timeo); | |
1568 | if (signal_pending(current)) | |
1569 | goto out_free; | |
1570 | ||
1571 | goto restart; | |
1572 | } | |
1573 | ||
3f66116e AC |
1574 | if (sock_flag(other, SOCK_RCVTSTAMP)) |
1575 | __net_timestamp(skb); | |
16e57262 | 1576 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1577 | skb_queue_tail(&other->sk_receive_queue, skb); |
25888e30 ED |
1578 | if (max_level > unix_sk(other)->recursion_level) |
1579 | unix_sk(other)->recursion_level = max_level; | |
1c92b4e5 | 1580 | unix_state_unlock(other); |
1da177e4 LT |
1581 | other->sk_data_ready(other, len); |
1582 | sock_put(other); | |
f78a5fda | 1583 | scm_destroy(siocb->scm); |
1da177e4 LT |
1584 | return len; |
1585 | ||
1586 | out_unlock: | |
1c92b4e5 | 1587 | unix_state_unlock(other); |
1da177e4 LT |
1588 | out_free: |
1589 | kfree_skb(skb); | |
1590 | out: | |
1591 | if (other) | |
1592 | sock_put(other); | |
f78a5fda | 1593 | scm_destroy(siocb->scm); |
1da177e4 LT |
1594 | return err; |
1595 | } | |
1596 | ||
ac7bfa62 | 1597 | |
1da177e4 LT |
1598 | static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock, |
1599 | struct msghdr *msg, size_t len) | |
1600 | { | |
1601 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
1602 | struct sock *sk = sock->sk; | |
1603 | struct sock *other = NULL; | |
6eba6a37 | 1604 | int err, size; |
f78a5fda | 1605 | struct sk_buff *skb; |
e27dfcea | 1606 | int sent = 0; |
1da177e4 | 1607 | struct scm_cookie tmp_scm; |
8ba69ba6 | 1608 | bool fds_sent = false; |
25888e30 | 1609 | int max_level; |
1da177e4 LT |
1610 | |
1611 | if (NULL == siocb->scm) | |
1612 | siocb->scm = &tmp_scm; | |
5f23b734 | 1613 | wait_for_unix_gc(); |
1da177e4 LT |
1614 | err = scm_send(sock, msg, siocb->scm); |
1615 | if (err < 0) | |
1616 | return err; | |
1617 | ||
1618 | err = -EOPNOTSUPP; | |
1619 | if (msg->msg_flags&MSG_OOB) | |
1620 | goto out_err; | |
1621 | ||
1622 | if (msg->msg_namelen) { | |
1623 | err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; | |
1624 | goto out_err; | |
1625 | } else { | |
1da177e4 | 1626 | err = -ENOTCONN; |
830a1e5c | 1627 | other = unix_peer(sk); |
1da177e4 LT |
1628 | if (!other) |
1629 | goto out_err; | |
1630 | } | |
1631 | ||
1632 | if (sk->sk_shutdown & SEND_SHUTDOWN) | |
1633 | goto pipe_err; | |
1634 | ||
6eba6a37 | 1635 | while (sent < len) { |
1da177e4 | 1636 | /* |
e9df7d7f BL |
1637 | * Optimisation for the fact that under 0.01% of X |
1638 | * messages typically need breaking up. | |
1da177e4 LT |
1639 | */ |
1640 | ||
e9df7d7f | 1641 | size = len-sent; |
1da177e4 LT |
1642 | |
1643 | /* Keep two messages in the pipe so it schedules better */ | |
e9df7d7f BL |
1644 | if (size > ((sk->sk_sndbuf >> 1) - 64)) |
1645 | size = (sk->sk_sndbuf >> 1) - 64; | |
1da177e4 LT |
1646 | |
1647 | if (size > SKB_MAX_ALLOC) | |
1648 | size = SKB_MAX_ALLOC; | |
ac7bfa62 | 1649 | |
1da177e4 LT |
1650 | /* |
1651 | * Grab a buffer | |
1652 | */ | |
ac7bfa62 | 1653 | |
6eba6a37 ED |
1654 | skb = sock_alloc_send_skb(sk, size, msg->msg_flags&MSG_DONTWAIT, |
1655 | &err); | |
1da177e4 | 1656 | |
e27dfcea | 1657 | if (skb == NULL) |
1da177e4 LT |
1658 | goto out_err; |
1659 | ||
1660 | /* | |
1661 | * If you pass two values to the sock_alloc_send_skb | |
1662 | * it tries to grab the large buffer with GFP_NOFS | |
1663 | * (which can fail easily), and if it fails grab the | |
1664 | * fallback size buffer which is under a page and will | |
1665 | * succeed. [Alan] | |
1666 | */ | |
1667 | size = min_t(int, size, skb_tailroom(skb)); | |
1668 | ||
7361c36c | 1669 | |
f78a5fda DM |
1670 | /* Only send the fds in the first buffer */ |
1671 | err = unix_scm_to_skb(siocb->scm, skb, !fds_sent); | |
25888e30 | 1672 | if (err < 0) { |
7361c36c | 1673 | kfree_skb(skb); |
f78a5fda | 1674 | goto out_err; |
6209344f | 1675 | } |
25888e30 | 1676 | max_level = err + 1; |
7361c36c | 1677 | fds_sent = true; |
1da177e4 | 1678 | |
6eba6a37 ED |
1679 | err = memcpy_fromiovec(skb_put(skb, size), msg->msg_iov, size); |
1680 | if (err) { | |
1da177e4 | 1681 | kfree_skb(skb); |
f78a5fda | 1682 | goto out_err; |
1da177e4 LT |
1683 | } |
1684 | ||
1c92b4e5 | 1685 | unix_state_lock(other); |
1da177e4 LT |
1686 | |
1687 | if (sock_flag(other, SOCK_DEAD) || | |
1688 | (other->sk_shutdown & RCV_SHUTDOWN)) | |
1689 | goto pipe_err_free; | |
1690 | ||
16e57262 | 1691 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1692 | skb_queue_tail(&other->sk_receive_queue, skb); |
25888e30 ED |
1693 | if (max_level > unix_sk(other)->recursion_level) |
1694 | unix_sk(other)->recursion_level = max_level; | |
1c92b4e5 | 1695 | unix_state_unlock(other); |
1da177e4 | 1696 | other->sk_data_ready(other, size); |
e27dfcea | 1697 | sent += size; |
1da177e4 | 1698 | } |
1da177e4 | 1699 | |
f78a5fda | 1700 | scm_destroy(siocb->scm); |
1da177e4 LT |
1701 | siocb->scm = NULL; |
1702 | ||
1703 | return sent; | |
1704 | ||
1705 | pipe_err_free: | |
1c92b4e5 | 1706 | unix_state_unlock(other); |
1da177e4 LT |
1707 | kfree_skb(skb); |
1708 | pipe_err: | |
6eba6a37 ED |
1709 | if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL)) |
1710 | send_sig(SIGPIPE, current, 0); | |
1da177e4 LT |
1711 | err = -EPIPE; |
1712 | out_err: | |
f78a5fda | 1713 | scm_destroy(siocb->scm); |
1da177e4 LT |
1714 | siocb->scm = NULL; |
1715 | return sent ? : err; | |
1716 | } | |
1717 | ||
1718 | static int unix_seqpacket_sendmsg(struct kiocb *kiocb, struct socket *sock, | |
1719 | struct msghdr *msg, size_t len) | |
1720 | { | |
1721 | int err; | |
1722 | struct sock *sk = sock->sk; | |
ac7bfa62 | 1723 | |
1da177e4 LT |
1724 | err = sock_error(sk); |
1725 | if (err) | |
1726 | return err; | |
1727 | ||
1728 | if (sk->sk_state != TCP_ESTABLISHED) | |
1729 | return -ENOTCONN; | |
1730 | ||
1731 | if (msg->msg_namelen) | |
1732 | msg->msg_namelen = 0; | |
1733 | ||
1734 | return unix_dgram_sendmsg(kiocb, sock, msg, len); | |
1735 | } | |
ac7bfa62 | 1736 | |
a05d2ad1 EB |
1737 | static int unix_seqpacket_recvmsg(struct kiocb *iocb, struct socket *sock, |
1738 | struct msghdr *msg, size_t size, | |
1739 | int flags) | |
1740 | { | |
1741 | struct sock *sk = sock->sk; | |
1742 | ||
1743 | if (sk->sk_state != TCP_ESTABLISHED) | |
1744 | return -ENOTCONN; | |
1745 | ||
1746 | return unix_dgram_recvmsg(iocb, sock, msg, size, flags); | |
1747 | } | |
1748 | ||
1da177e4 LT |
1749 | static void unix_copy_addr(struct msghdr *msg, struct sock *sk) |
1750 | { | |
1751 | struct unix_sock *u = unix_sk(sk); | |
1752 | ||
1753 | msg->msg_namelen = 0; | |
1754 | if (u->addr) { | |
1755 | msg->msg_namelen = u->addr->len; | |
1756 | memcpy(msg->msg_name, u->addr->name, u->addr->len); | |
1757 | } | |
1758 | } | |
1759 | ||
1760 | static int unix_dgram_recvmsg(struct kiocb *iocb, struct socket *sock, | |
1761 | struct msghdr *msg, size_t size, | |
1762 | int flags) | |
1763 | { | |
1764 | struct sock_iocb *siocb = kiocb_to_siocb(iocb); | |
1765 | struct scm_cookie tmp_scm; | |
1766 | struct sock *sk = sock->sk; | |
1767 | struct unix_sock *u = unix_sk(sk); | |
1768 | int noblock = flags & MSG_DONTWAIT; | |
1769 | struct sk_buff *skb; | |
1770 | int err; | |
f55bb7f9 | 1771 | int peeked, skip; |
1da177e4 LT |
1772 | |
1773 | err = -EOPNOTSUPP; | |
1774 | if (flags&MSG_OOB) | |
1775 | goto out; | |
1776 | ||
1777 | msg->msg_namelen = 0; | |
1778 | ||
b3ca9b02 RW |
1779 | err = mutex_lock_interruptible(&u->readlock); |
1780 | if (err) { | |
1781 | err = sock_intr_errno(sock_rcvtimeo(sk, noblock)); | |
1782 | goto out; | |
1783 | } | |
1da177e4 | 1784 | |
f55bb7f9 PE |
1785 | skip = sk_peek_offset(sk, flags); |
1786 | ||
1787 | skb = __skb_recv_datagram(sk, flags, &peeked, &skip, &err); | |
0a112258 FZ |
1788 | if (!skb) { |
1789 | unix_state_lock(sk); | |
1790 | /* Signal EOF on disconnected non-blocking SEQPACKET socket. */ | |
1791 | if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN && | |
1792 | (sk->sk_shutdown & RCV_SHUTDOWN)) | |
1793 | err = 0; | |
1794 | unix_state_unlock(sk); | |
1da177e4 | 1795 | goto out_unlock; |
0a112258 | 1796 | } |
1da177e4 | 1797 | |
67426b75 ED |
1798 | wake_up_interruptible_sync_poll(&u->peer_wait, |
1799 | POLLOUT | POLLWRNORM | POLLWRBAND); | |
1da177e4 LT |
1800 | |
1801 | if (msg->msg_name) | |
1802 | unix_copy_addr(msg, skb->sk); | |
1803 | ||
f55bb7f9 PE |
1804 | if (size > skb->len - skip) |
1805 | size = skb->len - skip; | |
1806 | else if (size < skb->len - skip) | |
1da177e4 LT |
1807 | msg->msg_flags |= MSG_TRUNC; |
1808 | ||
f55bb7f9 | 1809 | err = skb_copy_datagram_iovec(skb, skip, msg->msg_iov, size); |
1da177e4 LT |
1810 | if (err) |
1811 | goto out_free; | |
1812 | ||
3f66116e AC |
1813 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
1814 | __sock_recv_timestamp(msg, sk, skb); | |
1815 | ||
1da177e4 LT |
1816 | if (!siocb->scm) { |
1817 | siocb->scm = &tmp_scm; | |
1818 | memset(&tmp_scm, 0, sizeof(tmp_scm)); | |
1819 | } | |
f78a5fda | 1820 | scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).cred); |
877ce7c1 | 1821 | unix_set_secdata(siocb->scm, skb); |
1da177e4 | 1822 | |
6eba6a37 | 1823 | if (!(flags & MSG_PEEK)) { |
1da177e4 LT |
1824 | if (UNIXCB(skb).fp) |
1825 | unix_detach_fds(siocb->scm, skb); | |
f55bb7f9 PE |
1826 | |
1827 | sk_peek_offset_bwd(sk, skb->len); | |
6eba6a37 | 1828 | } else { |
1da177e4 LT |
1829 | /* It is questionable: on PEEK we could: |
1830 | - do not return fds - good, but too simple 8) | |
1831 | - return fds, and do not return them on read (old strategy, | |
1832 | apparently wrong) | |
1833 | - clone fds (I chose it for now, it is the most universal | |
1834 | solution) | |
ac7bfa62 YH |
1835 | |
1836 | POSIX 1003.1g does not actually define this clearly | |
1837 | at all. POSIX 1003.1g doesn't define a lot of things | |
1838 | clearly however! | |
1839 | ||
1da177e4 | 1840 | */ |
f55bb7f9 PE |
1841 | |
1842 | sk_peek_offset_fwd(sk, size); | |
1843 | ||
1da177e4 LT |
1844 | if (UNIXCB(skb).fp) |
1845 | siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp); | |
1846 | } | |
1847 | err = size; | |
1848 | ||
1849 | scm_recv(sock, msg, siocb->scm, flags); | |
1850 | ||
1851 | out_free: | |
6eba6a37 | 1852 | skb_free_datagram(sk, skb); |
1da177e4 | 1853 | out_unlock: |
57b47a53 | 1854 | mutex_unlock(&u->readlock); |
1da177e4 LT |
1855 | out: |
1856 | return err; | |
1857 | } | |
1858 | ||
1859 | /* | |
1860 | * Sleep until data has arrive. But check for races.. | |
1861 | */ | |
ac7bfa62 | 1862 | |
6eba6a37 | 1863 | static long unix_stream_data_wait(struct sock *sk, long timeo) |
1da177e4 LT |
1864 | { |
1865 | DEFINE_WAIT(wait); | |
1866 | ||
1c92b4e5 | 1867 | unix_state_lock(sk); |
1da177e4 LT |
1868 | |
1869 | for (;;) { | |
aa395145 | 1870 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 1871 | |
b03efcfb | 1872 | if (!skb_queue_empty(&sk->sk_receive_queue) || |
1da177e4 LT |
1873 | sk->sk_err || |
1874 | (sk->sk_shutdown & RCV_SHUTDOWN) || | |
1875 | signal_pending(current) || | |
1876 | !timeo) | |
1877 | break; | |
1878 | ||
1879 | set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); | |
1c92b4e5 | 1880 | unix_state_unlock(sk); |
1da177e4 | 1881 | timeo = schedule_timeout(timeo); |
1c92b4e5 | 1882 | unix_state_lock(sk); |
1da177e4 LT |
1883 | clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); |
1884 | } | |
1885 | ||
aa395145 | 1886 | finish_wait(sk_sleep(sk), &wait); |
1c92b4e5 | 1887 | unix_state_unlock(sk); |
1da177e4 LT |
1888 | return timeo; |
1889 | } | |
1890 | ||
1891 | ||
1892 | ||
1893 | static int unix_stream_recvmsg(struct kiocb *iocb, struct socket *sock, | |
1894 | struct msghdr *msg, size_t size, | |
1895 | int flags) | |
1896 | { | |
1897 | struct sock_iocb *siocb = kiocb_to_siocb(iocb); | |
1898 | struct scm_cookie tmp_scm; | |
1899 | struct sock *sk = sock->sk; | |
1900 | struct unix_sock *u = unix_sk(sk); | |
e27dfcea | 1901 | struct sockaddr_un *sunaddr = msg->msg_name; |
1da177e4 LT |
1902 | int copied = 0; |
1903 | int check_creds = 0; | |
1904 | int target; | |
1905 | int err = 0; | |
1906 | long timeo; | |
1907 | ||
1908 | err = -EINVAL; | |
1909 | if (sk->sk_state != TCP_ESTABLISHED) | |
1910 | goto out; | |
1911 | ||
1912 | err = -EOPNOTSUPP; | |
1913 | if (flags&MSG_OOB) | |
1914 | goto out; | |
1915 | ||
1916 | target = sock_rcvlowat(sk, flags&MSG_WAITALL, size); | |
1917 | timeo = sock_rcvtimeo(sk, flags&MSG_DONTWAIT); | |
1918 | ||
1919 | msg->msg_namelen = 0; | |
1920 | ||
1921 | /* Lock the socket to prevent queue disordering | |
1922 | * while sleeps in memcpy_tomsg | |
1923 | */ | |
1924 | ||
1925 | if (!siocb->scm) { | |
1926 | siocb->scm = &tmp_scm; | |
1927 | memset(&tmp_scm, 0, sizeof(tmp_scm)); | |
1928 | } | |
1929 | ||
b3ca9b02 RW |
1930 | err = mutex_lock_interruptible(&u->readlock); |
1931 | if (err) { | |
1932 | err = sock_intr_errno(timeo); | |
1933 | goto out; | |
1934 | } | |
1da177e4 | 1935 | |
6eba6a37 | 1936 | do { |
1da177e4 LT |
1937 | int chunk; |
1938 | struct sk_buff *skb; | |
1939 | ||
3c0d2f37 | 1940 | unix_state_lock(sk); |
6f01fd6e | 1941 | skb = skb_peek(&sk->sk_receive_queue); |
6eba6a37 | 1942 | if (skb == NULL) { |
25888e30 | 1943 | unix_sk(sk)->recursion_level = 0; |
1da177e4 | 1944 | if (copied >= target) |
3c0d2f37 | 1945 | goto unlock; |
1da177e4 LT |
1946 | |
1947 | /* | |
1948 | * POSIX 1003.1g mandates this order. | |
1949 | */ | |
ac7bfa62 | 1950 | |
6eba6a37 ED |
1951 | err = sock_error(sk); |
1952 | if (err) | |
3c0d2f37 | 1953 | goto unlock; |
1da177e4 | 1954 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
3c0d2f37 MS |
1955 | goto unlock; |
1956 | ||
1957 | unix_state_unlock(sk); | |
1da177e4 LT |
1958 | err = -EAGAIN; |
1959 | if (!timeo) | |
1960 | break; | |
57b47a53 | 1961 | mutex_unlock(&u->readlock); |
1da177e4 LT |
1962 | |
1963 | timeo = unix_stream_data_wait(sk, timeo); | |
1964 | ||
b3ca9b02 RW |
1965 | if (signal_pending(current) |
1966 | || mutex_lock_interruptible(&u->readlock)) { | |
1da177e4 LT |
1967 | err = sock_intr_errno(timeo); |
1968 | goto out; | |
1969 | } | |
b3ca9b02 | 1970 | |
1da177e4 | 1971 | continue; |
3c0d2f37 MS |
1972 | unlock: |
1973 | unix_state_unlock(sk); | |
1974 | break; | |
1da177e4 | 1975 | } |
3c0d2f37 | 1976 | unix_state_unlock(sk); |
1da177e4 LT |
1977 | |
1978 | if (check_creds) { | |
1979 | /* Never glue messages from different writers */ | |
7361c36c | 1980 | if ((UNIXCB(skb).pid != siocb->scm->pid) || |
6f01fd6e | 1981 | (UNIXCB(skb).cred != siocb->scm->cred)) |
1da177e4 | 1982 | break; |
1da177e4 LT |
1983 | } else { |
1984 | /* Copy credentials */ | |
f78a5fda | 1985 | scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).cred); |
1da177e4 LT |
1986 | check_creds = 1; |
1987 | } | |
1988 | ||
1989 | /* Copy address just once */ | |
6eba6a37 | 1990 | if (sunaddr) { |
1da177e4 LT |
1991 | unix_copy_addr(msg, skb->sk); |
1992 | sunaddr = NULL; | |
1993 | } | |
1994 | ||
1995 | chunk = min_t(unsigned int, skb->len, size); | |
1996 | if (memcpy_toiovec(msg->msg_iov, skb->data, chunk)) { | |
1da177e4 LT |
1997 | if (copied == 0) |
1998 | copied = -EFAULT; | |
1999 | break; | |
2000 | } | |
2001 | copied += chunk; | |
2002 | size -= chunk; | |
2003 | ||
2004 | /* Mark read part of skb as used */ | |
6eba6a37 | 2005 | if (!(flags & MSG_PEEK)) { |
1da177e4 LT |
2006 | skb_pull(skb, chunk); |
2007 | ||
2008 | if (UNIXCB(skb).fp) | |
2009 | unix_detach_fds(siocb->scm, skb); | |
2010 | ||
6f01fd6e | 2011 | if (skb->len) |
1da177e4 | 2012 | break; |
1da177e4 | 2013 | |
6f01fd6e | 2014 | skb_unlink(skb, &sk->sk_receive_queue); |
70d4bf6d | 2015 | consume_skb(skb); |
1da177e4 LT |
2016 | |
2017 | if (siocb->scm->fp) | |
2018 | break; | |
6eba6a37 | 2019 | } else { |
1da177e4 LT |
2020 | /* It is questionable, see note in unix_dgram_recvmsg. |
2021 | */ | |
2022 | if (UNIXCB(skb).fp) | |
2023 | siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp); | |
2024 | ||
1da177e4 LT |
2025 | break; |
2026 | } | |
2027 | } while (size); | |
2028 | ||
57b47a53 | 2029 | mutex_unlock(&u->readlock); |
1da177e4 LT |
2030 | scm_recv(sock, msg, siocb->scm, flags); |
2031 | out: | |
2032 | return copied ? : err; | |
2033 | } | |
2034 | ||
2035 | static int unix_shutdown(struct socket *sock, int mode) | |
2036 | { | |
2037 | struct sock *sk = sock->sk; | |
2038 | struct sock *other; | |
2039 | ||
2040 | mode = (mode+1)&(RCV_SHUTDOWN|SEND_SHUTDOWN); | |
2041 | ||
7180a031 AC |
2042 | if (!mode) |
2043 | return 0; | |
2044 | ||
2045 | unix_state_lock(sk); | |
2046 | sk->sk_shutdown |= mode; | |
2047 | other = unix_peer(sk); | |
2048 | if (other) | |
2049 | sock_hold(other); | |
2050 | unix_state_unlock(sk); | |
2051 | sk->sk_state_change(sk); | |
2052 | ||
2053 | if (other && | |
2054 | (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) { | |
2055 | ||
2056 | int peer_mode = 0; | |
2057 | ||
2058 | if (mode&RCV_SHUTDOWN) | |
2059 | peer_mode |= SEND_SHUTDOWN; | |
2060 | if (mode&SEND_SHUTDOWN) | |
2061 | peer_mode |= RCV_SHUTDOWN; | |
2062 | unix_state_lock(other); | |
2063 | other->sk_shutdown |= peer_mode; | |
2064 | unix_state_unlock(other); | |
2065 | other->sk_state_change(other); | |
2066 | if (peer_mode == SHUTDOWN_MASK) | |
2067 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP); | |
2068 | else if (peer_mode & RCV_SHUTDOWN) | |
2069 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN); | |
1da177e4 | 2070 | } |
7180a031 AC |
2071 | if (other) |
2072 | sock_put(other); | |
2073 | ||
1da177e4 LT |
2074 | return 0; |
2075 | } | |
2076 | ||
885ee74d PE |
2077 | long unix_inq_len(struct sock *sk) |
2078 | { | |
2079 | struct sk_buff *skb; | |
2080 | long amount = 0; | |
2081 | ||
2082 | if (sk->sk_state == TCP_LISTEN) | |
2083 | return -EINVAL; | |
2084 | ||
2085 | spin_lock(&sk->sk_receive_queue.lock); | |
2086 | if (sk->sk_type == SOCK_STREAM || | |
2087 | sk->sk_type == SOCK_SEQPACKET) { | |
2088 | skb_queue_walk(&sk->sk_receive_queue, skb) | |
2089 | amount += skb->len; | |
2090 | } else { | |
2091 | skb = skb_peek(&sk->sk_receive_queue); | |
2092 | if (skb) | |
2093 | amount = skb->len; | |
2094 | } | |
2095 | spin_unlock(&sk->sk_receive_queue.lock); | |
2096 | ||
2097 | return amount; | |
2098 | } | |
2099 | EXPORT_SYMBOL_GPL(unix_inq_len); | |
2100 | ||
2101 | long unix_outq_len(struct sock *sk) | |
2102 | { | |
2103 | return sk_wmem_alloc_get(sk); | |
2104 | } | |
2105 | EXPORT_SYMBOL_GPL(unix_outq_len); | |
2106 | ||
1da177e4 LT |
2107 | static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
2108 | { | |
2109 | struct sock *sk = sock->sk; | |
e27dfcea | 2110 | long amount = 0; |
1da177e4 LT |
2111 | int err; |
2112 | ||
6eba6a37 ED |
2113 | switch (cmd) { |
2114 | case SIOCOUTQ: | |
885ee74d | 2115 | amount = unix_outq_len(sk); |
6eba6a37 ED |
2116 | err = put_user(amount, (int __user *)arg); |
2117 | break; | |
2118 | case SIOCINQ: | |
885ee74d PE |
2119 | amount = unix_inq_len(sk); |
2120 | if (amount < 0) | |
2121 | err = amount; | |
2122 | else | |
1da177e4 | 2123 | err = put_user(amount, (int __user *)arg); |
885ee74d | 2124 | break; |
6eba6a37 ED |
2125 | default: |
2126 | err = -ENOIOCTLCMD; | |
2127 | break; | |
1da177e4 LT |
2128 | } |
2129 | return err; | |
2130 | } | |
2131 | ||
6eba6a37 | 2132 | static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait) |
1da177e4 LT |
2133 | { |
2134 | struct sock *sk = sock->sk; | |
2135 | unsigned int mask; | |
2136 | ||
aa395145 | 2137 | sock_poll_wait(file, sk_sleep(sk), wait); |
1da177e4 LT |
2138 | mask = 0; |
2139 | ||
2140 | /* exceptional events? */ | |
2141 | if (sk->sk_err) | |
2142 | mask |= POLLERR; | |
2143 | if (sk->sk_shutdown == SHUTDOWN_MASK) | |
2144 | mask |= POLLHUP; | |
f348d70a | 2145 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
db40980f | 2146 | mask |= POLLRDHUP | POLLIN | POLLRDNORM; |
1da177e4 LT |
2147 | |
2148 | /* readable? */ | |
db40980f | 2149 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
2150 | mask |= POLLIN | POLLRDNORM; |
2151 | ||
2152 | /* Connection-based need to check for termination and startup */ | |
6eba6a37 ED |
2153 | if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) && |
2154 | sk->sk_state == TCP_CLOSE) | |
1da177e4 LT |
2155 | mask |= POLLHUP; |
2156 | ||
2157 | /* | |
2158 | * we set writable also when the other side has shut down the | |
2159 | * connection. This prevents stuck sockets. | |
2160 | */ | |
2161 | if (unix_writable(sk)) | |
2162 | mask |= POLLOUT | POLLWRNORM | POLLWRBAND; | |
2163 | ||
2164 | return mask; | |
2165 | } | |
2166 | ||
ec0d215f RW |
2167 | static unsigned int unix_dgram_poll(struct file *file, struct socket *sock, |
2168 | poll_table *wait) | |
3c73419c | 2169 | { |
ec0d215f RW |
2170 | struct sock *sk = sock->sk, *other; |
2171 | unsigned int mask, writable; | |
3c73419c | 2172 | |
aa395145 | 2173 | sock_poll_wait(file, sk_sleep(sk), wait); |
3c73419c RW |
2174 | mask = 0; |
2175 | ||
2176 | /* exceptional events? */ | |
2177 | if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue)) | |
2178 | mask |= POLLERR; | |
2179 | if (sk->sk_shutdown & RCV_SHUTDOWN) | |
5456f09a | 2180 | mask |= POLLRDHUP | POLLIN | POLLRDNORM; |
3c73419c RW |
2181 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
2182 | mask |= POLLHUP; | |
2183 | ||
2184 | /* readable? */ | |
5456f09a | 2185 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
3c73419c RW |
2186 | mask |= POLLIN | POLLRDNORM; |
2187 | ||
2188 | /* Connection-based need to check for termination and startup */ | |
2189 | if (sk->sk_type == SOCK_SEQPACKET) { | |
2190 | if (sk->sk_state == TCP_CLOSE) | |
2191 | mask |= POLLHUP; | |
2192 | /* connection hasn't started yet? */ | |
2193 | if (sk->sk_state == TCP_SYN_SENT) | |
2194 | return mask; | |
2195 | } | |
2196 | ||
973a34aa ED |
2197 | /* No write status requested, avoid expensive OUT tests. */ |
2198 | if (wait && !(wait->key & (POLLWRBAND | POLLWRNORM | POLLOUT))) | |
2199 | return mask; | |
2200 | ||
ec0d215f | 2201 | writable = unix_writable(sk); |
5456f09a ED |
2202 | other = unix_peer_get(sk); |
2203 | if (other) { | |
2204 | if (unix_peer(other) != sk) { | |
2205 | sock_poll_wait(file, &unix_sk(other)->peer_wait, wait); | |
2206 | if (unix_recvq_full(other)) | |
2207 | writable = 0; | |
ec0d215f | 2208 | } |
5456f09a | 2209 | sock_put(other); |
ec0d215f RW |
2210 | } |
2211 | ||
2212 | if (writable) | |
3c73419c RW |
2213 | mask |= POLLOUT | POLLWRNORM | POLLWRBAND; |
2214 | else | |
2215 | set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); | |
2216 | ||
3c73419c RW |
2217 | return mask; |
2218 | } | |
1da177e4 LT |
2219 | |
2220 | #ifdef CONFIG_PROC_FS | |
a53eb3fe PE |
2221 | static struct sock *first_unix_socket(int *i) |
2222 | { | |
2223 | for (*i = 0; *i <= UNIX_HASH_SIZE; (*i)++) { | |
2224 | if (!hlist_empty(&unix_socket_table[*i])) | |
2225 | return __sk_head(&unix_socket_table[*i]); | |
2226 | } | |
2227 | return NULL; | |
2228 | } | |
2229 | ||
2230 | static struct sock *next_unix_socket(int *i, struct sock *s) | |
2231 | { | |
2232 | struct sock *next = sk_next(s); | |
2233 | /* More in this chain? */ | |
2234 | if (next) | |
2235 | return next; | |
2236 | /* Look for next non-empty chain. */ | |
2237 | for ((*i)++; *i <= UNIX_HASH_SIZE; (*i)++) { | |
2238 | if (!hlist_empty(&unix_socket_table[*i])) | |
2239 | return __sk_head(&unix_socket_table[*i]); | |
2240 | } | |
2241 | return NULL; | |
2242 | } | |
2243 | ||
097e66c5 | 2244 | struct unix_iter_state { |
e372c414 | 2245 | struct seq_net_private p; |
097e66c5 DL |
2246 | int i; |
2247 | }; | |
e27dfcea | 2248 | |
1218854a | 2249 | static struct sock *unix_seq_idx(struct seq_file *seq, loff_t pos) |
1da177e4 | 2250 | { |
1218854a | 2251 | struct unix_iter_state *iter = seq->private; |
1da177e4 LT |
2252 | loff_t off = 0; |
2253 | struct sock *s; | |
2254 | ||
097e66c5 | 2255 | for (s = first_unix_socket(&iter->i); s; s = next_unix_socket(&iter->i, s)) { |
1218854a | 2256 | if (sock_net(s) != seq_file_net(seq)) |
097e66c5 | 2257 | continue; |
ac7bfa62 | 2258 | if (off == pos) |
1da177e4 LT |
2259 | return s; |
2260 | ++off; | |
2261 | } | |
2262 | return NULL; | |
2263 | } | |
2264 | ||
1da177e4 | 2265 | static void *unix_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 2266 | __acquires(unix_table_lock) |
1da177e4 | 2267 | { |
fbe9cc4a | 2268 | spin_lock(&unix_table_lock); |
b9f3124f | 2269 | return *pos ? unix_seq_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
1da177e4 LT |
2270 | } |
2271 | ||
2272 | static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2273 | { | |
097e66c5 DL |
2274 | struct unix_iter_state *iter = seq->private; |
2275 | struct sock *sk = v; | |
1da177e4 LT |
2276 | ++*pos; |
2277 | ||
b9f3124f | 2278 | if (v == SEQ_START_TOKEN) |
097e66c5 DL |
2279 | sk = first_unix_socket(&iter->i); |
2280 | else | |
2281 | sk = next_unix_socket(&iter->i, sk); | |
1218854a | 2282 | while (sk && (sock_net(sk) != seq_file_net(seq))) |
097e66c5 DL |
2283 | sk = next_unix_socket(&iter->i, sk); |
2284 | return sk; | |
1da177e4 LT |
2285 | } |
2286 | ||
2287 | static void unix_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2288 | __releases(unix_table_lock) |
1da177e4 | 2289 | { |
fbe9cc4a | 2290 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
2291 | } |
2292 | ||
2293 | static int unix_seq_show(struct seq_file *seq, void *v) | |
2294 | { | |
ac7bfa62 | 2295 | |
b9f3124f | 2296 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2297 | seq_puts(seq, "Num RefCount Protocol Flags Type St " |
2298 | "Inode Path\n"); | |
2299 | else { | |
2300 | struct sock *s = v; | |
2301 | struct unix_sock *u = unix_sk(s); | |
1c92b4e5 | 2302 | unix_state_lock(s); |
1da177e4 | 2303 | |
71338aa7 | 2304 | seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu", |
1da177e4 LT |
2305 | s, |
2306 | atomic_read(&s->sk_refcnt), | |
2307 | 0, | |
2308 | s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0, | |
2309 | s->sk_type, | |
2310 | s->sk_socket ? | |
2311 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) : | |
2312 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING), | |
2313 | sock_i_ino(s)); | |
2314 | ||
2315 | if (u->addr) { | |
2316 | int i, len; | |
2317 | seq_putc(seq, ' '); | |
2318 | ||
2319 | i = 0; | |
2320 | len = u->addr->len - sizeof(short); | |
2321 | if (!UNIX_ABSTRACT(s)) | |
2322 | len--; | |
2323 | else { | |
2324 | seq_putc(seq, '@'); | |
2325 | i++; | |
2326 | } | |
2327 | for ( ; i < len; i++) | |
2328 | seq_putc(seq, u->addr->name->sun_path[i]); | |
2329 | } | |
1c92b4e5 | 2330 | unix_state_unlock(s); |
1da177e4 LT |
2331 | seq_putc(seq, '\n'); |
2332 | } | |
2333 | ||
2334 | return 0; | |
2335 | } | |
2336 | ||
56b3d975 | 2337 | static const struct seq_operations unix_seq_ops = { |
1da177e4 LT |
2338 | .start = unix_seq_start, |
2339 | .next = unix_seq_next, | |
2340 | .stop = unix_seq_stop, | |
2341 | .show = unix_seq_show, | |
2342 | }; | |
2343 | ||
1da177e4 LT |
2344 | static int unix_seq_open(struct inode *inode, struct file *file) |
2345 | { | |
e372c414 DL |
2346 | return seq_open_net(inode, file, &unix_seq_ops, |
2347 | sizeof(struct unix_iter_state)); | |
1da177e4 LT |
2348 | } |
2349 | ||
da7071d7 | 2350 | static const struct file_operations unix_seq_fops = { |
1da177e4 LT |
2351 | .owner = THIS_MODULE, |
2352 | .open = unix_seq_open, | |
2353 | .read = seq_read, | |
2354 | .llseek = seq_lseek, | |
e372c414 | 2355 | .release = seq_release_net, |
1da177e4 LT |
2356 | }; |
2357 | ||
2358 | #endif | |
2359 | ||
ec1b4cf7 | 2360 | static const struct net_proto_family unix_family_ops = { |
1da177e4 LT |
2361 | .family = PF_UNIX, |
2362 | .create = unix_create, | |
2363 | .owner = THIS_MODULE, | |
2364 | }; | |
2365 | ||
097e66c5 | 2366 | |
2c8c1e72 | 2367 | static int __net_init unix_net_init(struct net *net) |
097e66c5 DL |
2368 | { |
2369 | int error = -ENOMEM; | |
2370 | ||
a0a53c8b | 2371 | net->unx.sysctl_max_dgram_qlen = 10; |
1597fbc0 PE |
2372 | if (unix_sysctl_register(net)) |
2373 | goto out; | |
d392e497 | 2374 | |
097e66c5 | 2375 | #ifdef CONFIG_PROC_FS |
1597fbc0 PE |
2376 | if (!proc_net_fops_create(net, "unix", 0, &unix_seq_fops)) { |
2377 | unix_sysctl_unregister(net); | |
097e66c5 | 2378 | goto out; |
1597fbc0 | 2379 | } |
097e66c5 DL |
2380 | #endif |
2381 | error = 0; | |
2382 | out: | |
48dcc33e | 2383 | return error; |
097e66c5 DL |
2384 | } |
2385 | ||
2c8c1e72 | 2386 | static void __net_exit unix_net_exit(struct net *net) |
097e66c5 | 2387 | { |
1597fbc0 | 2388 | unix_sysctl_unregister(net); |
097e66c5 DL |
2389 | proc_net_remove(net, "unix"); |
2390 | } | |
2391 | ||
2392 | static struct pernet_operations unix_net_ops = { | |
2393 | .init = unix_net_init, | |
2394 | .exit = unix_net_exit, | |
2395 | }; | |
2396 | ||
1da177e4 LT |
2397 | static int __init af_unix_init(void) |
2398 | { | |
2399 | int rc = -1; | |
2400 | struct sk_buff *dummy_skb; | |
2401 | ||
ef047f5e | 2402 | BUILD_BUG_ON(sizeof(struct unix_skb_parms) > sizeof(dummy_skb->cb)); |
1da177e4 LT |
2403 | |
2404 | rc = proto_register(&unix_proto, 1); | |
ac7bfa62 YH |
2405 | if (rc != 0) { |
2406 | printk(KERN_CRIT "%s: Cannot create unix_sock SLAB cache!\n", | |
0dc47877 | 2407 | __func__); |
1da177e4 LT |
2408 | goto out; |
2409 | } | |
2410 | ||
2411 | sock_register(&unix_family_ops); | |
097e66c5 | 2412 | register_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2413 | out: |
2414 | return rc; | |
2415 | } | |
2416 | ||
2417 | static void __exit af_unix_exit(void) | |
2418 | { | |
2419 | sock_unregister(PF_UNIX); | |
1da177e4 | 2420 | proto_unregister(&unix_proto); |
097e66c5 | 2421 | unregister_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2422 | } |
2423 | ||
3d366960 DW |
2424 | /* Earlier than device_initcall() so that other drivers invoking |
2425 | request_module() don't end up in a loop when modprobe tries | |
2426 | to use a UNIX socket. But later than subsys_initcall() because | |
2427 | we depend on stuff initialised there */ | |
2428 | fs_initcall(af_unix_init); | |
1da177e4 LT |
2429 | module_exit(af_unix_exit); |
2430 | ||
2431 | MODULE_LICENSE("GPL"); | |
2432 | MODULE_ALIAS_NETPROTO(PF_UNIX); |