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 | ||
7123aaa3 | 118 | struct hlist_head unix_socket_table[2 * UNIX_HASH_SIZE]; |
fa7ff56f PE |
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 | 123 | |
1da177e4 | 124 | |
7123aaa3 ED |
125 | static struct hlist_head *unix_sockets_unbound(void *addr) |
126 | { | |
127 | unsigned long hash = (unsigned long)addr; | |
128 | ||
129 | hash ^= hash >> 16; | |
130 | hash ^= hash >> 8; | |
131 | hash %= UNIX_HASH_SIZE; | |
132 | return &unix_socket_table[UNIX_HASH_SIZE + hash]; | |
133 | } | |
134 | ||
135 | #define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash < UNIX_HASH_SIZE) | |
1da177e4 | 136 | |
877ce7c1 | 137 | #ifdef CONFIG_SECURITY_NETWORK |
dc49c1f9 | 138 | static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 | 139 | { |
dc49c1f9 | 140 | memcpy(UNIXSID(skb), &scm->secid, sizeof(u32)); |
877ce7c1 CZ |
141 | } |
142 | ||
143 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
144 | { | |
dc49c1f9 | 145 | scm->secid = *UNIXSID(skb); |
877ce7c1 CZ |
146 | } |
147 | #else | |
dc49c1f9 | 148 | static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 CZ |
149 | { } |
150 | ||
151 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
152 | { } | |
153 | #endif /* CONFIG_SECURITY_NETWORK */ | |
154 | ||
1da177e4 LT |
155 | /* |
156 | * SMP locking strategy: | |
fbe9cc4a | 157 | * hash table is protected with spinlock unix_table_lock |
663717f6 | 158 | * each socket state is protected by separate spin lock. |
1da177e4 LT |
159 | */ |
160 | ||
95c96174 | 161 | static inline unsigned int unix_hash_fold(__wsum n) |
1da177e4 | 162 | { |
83bb80f4 | 163 | unsigned int hash = (__force unsigned int)csum_fold(n); |
95c96174 | 164 | |
1da177e4 LT |
165 | hash ^= hash>>8; |
166 | return hash&(UNIX_HASH_SIZE-1); | |
167 | } | |
168 | ||
169 | #define unix_peer(sk) (unix_sk(sk)->peer) | |
170 | ||
171 | static inline int unix_our_peer(struct sock *sk, struct sock *osk) | |
172 | { | |
173 | return unix_peer(osk) == sk; | |
174 | } | |
175 | ||
176 | static inline int unix_may_send(struct sock *sk, struct sock *osk) | |
177 | { | |
6eba6a37 | 178 | return unix_peer(osk) == NULL || unix_our_peer(sk, osk); |
1da177e4 LT |
179 | } |
180 | ||
3c73419c RW |
181 | static inline int unix_recvq_full(struct sock const *sk) |
182 | { | |
183 | return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog; | |
184 | } | |
185 | ||
fa7ff56f | 186 | struct sock *unix_peer_get(struct sock *s) |
1da177e4 LT |
187 | { |
188 | struct sock *peer; | |
189 | ||
1c92b4e5 | 190 | unix_state_lock(s); |
1da177e4 LT |
191 | peer = unix_peer(s); |
192 | if (peer) | |
193 | sock_hold(peer); | |
1c92b4e5 | 194 | unix_state_unlock(s); |
1da177e4 LT |
195 | return peer; |
196 | } | |
fa7ff56f | 197 | EXPORT_SYMBOL_GPL(unix_peer_get); |
1da177e4 LT |
198 | |
199 | static inline void unix_release_addr(struct unix_address *addr) | |
200 | { | |
201 | if (atomic_dec_and_test(&addr->refcnt)) | |
202 | kfree(addr); | |
203 | } | |
204 | ||
205 | /* | |
206 | * Check unix socket name: | |
207 | * - should be not zero length. | |
208 | * - if started by not zero, should be NULL terminated (FS object) | |
209 | * - if started by zero, it is abstract name. | |
210 | */ | |
ac7bfa62 | 211 | |
95c96174 | 212 | static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp) |
1da177e4 LT |
213 | { |
214 | if (len <= sizeof(short) || len > sizeof(*sunaddr)) | |
215 | return -EINVAL; | |
216 | if (!sunaddr || sunaddr->sun_family != AF_UNIX) | |
217 | return -EINVAL; | |
218 | if (sunaddr->sun_path[0]) { | |
219 | /* | |
220 | * This may look like an off by one error but it is a bit more | |
221 | * subtle. 108 is the longest valid AF_UNIX path for a binding. | |
25985edc | 222 | * sun_path[108] doesn't as such exist. However in kernel space |
1da177e4 LT |
223 | * we are guaranteed that it is a valid memory location in our |
224 | * kernel address buffer. | |
225 | */ | |
e27dfcea | 226 | ((char *)sunaddr)[len] = 0; |
1da177e4 LT |
227 | len = strlen(sunaddr->sun_path)+1+sizeof(short); |
228 | return len; | |
229 | } | |
230 | ||
07f0757a | 231 | *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0)); |
1da177e4 LT |
232 | return len; |
233 | } | |
234 | ||
235 | static void __unix_remove_socket(struct sock *sk) | |
236 | { | |
237 | sk_del_node_init(sk); | |
238 | } | |
239 | ||
240 | static void __unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
241 | { | |
547b792c | 242 | WARN_ON(!sk_unhashed(sk)); |
1da177e4 LT |
243 | sk_add_node(sk, list); |
244 | } | |
245 | ||
246 | static inline void unix_remove_socket(struct sock *sk) | |
247 | { | |
fbe9cc4a | 248 | spin_lock(&unix_table_lock); |
1da177e4 | 249 | __unix_remove_socket(sk); |
fbe9cc4a | 250 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
251 | } |
252 | ||
253 | static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
254 | { | |
fbe9cc4a | 255 | spin_lock(&unix_table_lock); |
1da177e4 | 256 | __unix_insert_socket(list, sk); |
fbe9cc4a | 257 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
258 | } |
259 | ||
097e66c5 DL |
260 | static struct sock *__unix_find_socket_byname(struct net *net, |
261 | struct sockaddr_un *sunname, | |
95c96174 | 262 | int len, int type, unsigned int hash) |
1da177e4 LT |
263 | { |
264 | struct sock *s; | |
1da177e4 | 265 | |
b67bfe0d | 266 | sk_for_each(s, &unix_socket_table[hash ^ type]) { |
1da177e4 LT |
267 | struct unix_sock *u = unix_sk(s); |
268 | ||
878628fb | 269 | if (!net_eq(sock_net(s), net)) |
097e66c5 DL |
270 | continue; |
271 | ||
1da177e4 LT |
272 | if (u->addr->len == len && |
273 | !memcmp(u->addr->name, sunname, len)) | |
274 | goto found; | |
275 | } | |
276 | s = NULL; | |
277 | found: | |
278 | return s; | |
279 | } | |
280 | ||
097e66c5 DL |
281 | static inline struct sock *unix_find_socket_byname(struct net *net, |
282 | struct sockaddr_un *sunname, | |
1da177e4 | 283 | int len, int type, |
95c96174 | 284 | unsigned int hash) |
1da177e4 LT |
285 | { |
286 | struct sock *s; | |
287 | ||
fbe9cc4a | 288 | spin_lock(&unix_table_lock); |
097e66c5 | 289 | s = __unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
290 | if (s) |
291 | sock_hold(s); | |
fbe9cc4a | 292 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
293 | return s; |
294 | } | |
295 | ||
6616f788 | 296 | static struct sock *unix_find_socket_byinode(struct inode *i) |
1da177e4 LT |
297 | { |
298 | struct sock *s; | |
1da177e4 | 299 | |
fbe9cc4a | 300 | spin_lock(&unix_table_lock); |
b67bfe0d | 301 | sk_for_each(s, |
1da177e4 | 302 | &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) { |
40ffe67d | 303 | struct dentry *dentry = unix_sk(s)->path.dentry; |
1da177e4 | 304 | |
6eba6a37 | 305 | if (dentry && dentry->d_inode == i) { |
1da177e4 LT |
306 | sock_hold(s); |
307 | goto found; | |
308 | } | |
309 | } | |
310 | s = NULL; | |
311 | found: | |
fbe9cc4a | 312 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
313 | return s; |
314 | } | |
315 | ||
316 | static inline int unix_writable(struct sock *sk) | |
317 | { | |
318 | return (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf; | |
319 | } | |
320 | ||
321 | static void unix_write_space(struct sock *sk) | |
322 | { | |
43815482 ED |
323 | struct socket_wq *wq; |
324 | ||
325 | rcu_read_lock(); | |
1da177e4 | 326 | if (unix_writable(sk)) { |
43815482 ED |
327 | wq = rcu_dereference(sk->sk_wq); |
328 | if (wq_has_sleeper(wq)) | |
67426b75 ED |
329 | wake_up_interruptible_sync_poll(&wq->wait, |
330 | POLLOUT | POLLWRNORM | POLLWRBAND); | |
8d8ad9d7 | 331 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
1da177e4 | 332 | } |
43815482 | 333 | rcu_read_unlock(); |
1da177e4 LT |
334 | } |
335 | ||
336 | /* When dgram socket disconnects (or changes its peer), we clear its receive | |
337 | * queue of packets arrived from previous peer. First, it allows to do | |
338 | * flow control based only on wmem_alloc; second, sk connected to peer | |
339 | * may receive messages only from that peer. */ | |
340 | static void unix_dgram_disconnected(struct sock *sk, struct sock *other) | |
341 | { | |
b03efcfb | 342 | if (!skb_queue_empty(&sk->sk_receive_queue)) { |
1da177e4 LT |
343 | skb_queue_purge(&sk->sk_receive_queue); |
344 | wake_up_interruptible_all(&unix_sk(sk)->peer_wait); | |
345 | ||
346 | /* If one link of bidirectional dgram pipe is disconnected, | |
347 | * we signal error. Messages are lost. Do not make this, | |
348 | * when peer was not connected to us. | |
349 | */ | |
350 | if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) { | |
351 | other->sk_err = ECONNRESET; | |
352 | other->sk_error_report(other); | |
353 | } | |
354 | } | |
355 | } | |
356 | ||
357 | static void unix_sock_destructor(struct sock *sk) | |
358 | { | |
359 | struct unix_sock *u = unix_sk(sk); | |
360 | ||
361 | skb_queue_purge(&sk->sk_receive_queue); | |
362 | ||
547b792c IJ |
363 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); |
364 | WARN_ON(!sk_unhashed(sk)); | |
365 | WARN_ON(sk->sk_socket); | |
1da177e4 | 366 | if (!sock_flag(sk, SOCK_DEAD)) { |
6b41e7dd | 367 | printk(KERN_INFO "Attempt to release alive unix socket: %p\n", sk); |
1da177e4 LT |
368 | return; |
369 | } | |
370 | ||
371 | if (u->addr) | |
372 | unix_release_addr(u->addr); | |
373 | ||
518de9b3 | 374 | atomic_long_dec(&unix_nr_socks); |
6f756a8c | 375 | local_bh_disable(); |
a8076d8d | 376 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
6f756a8c | 377 | local_bh_enable(); |
1da177e4 | 378 | #ifdef UNIX_REFCNT_DEBUG |
518de9b3 ED |
379 | printk(KERN_DEBUG "UNIX %p is destroyed, %ld are still alive.\n", sk, |
380 | atomic_long_read(&unix_nr_socks)); | |
1da177e4 LT |
381 | #endif |
382 | } | |
383 | ||
ded34e0f | 384 | static void unix_release_sock(struct sock *sk, int embrion) |
1da177e4 LT |
385 | { |
386 | struct unix_sock *u = unix_sk(sk); | |
40ffe67d | 387 | struct path path; |
1da177e4 LT |
388 | struct sock *skpair; |
389 | struct sk_buff *skb; | |
390 | int state; | |
391 | ||
392 | unix_remove_socket(sk); | |
393 | ||
394 | /* Clear state */ | |
1c92b4e5 | 395 | unix_state_lock(sk); |
1da177e4 LT |
396 | sock_orphan(sk); |
397 | sk->sk_shutdown = SHUTDOWN_MASK; | |
40ffe67d AV |
398 | path = u->path; |
399 | u->path.dentry = NULL; | |
400 | u->path.mnt = NULL; | |
1da177e4 LT |
401 | state = sk->sk_state; |
402 | sk->sk_state = TCP_CLOSE; | |
1c92b4e5 | 403 | unix_state_unlock(sk); |
1da177e4 LT |
404 | |
405 | wake_up_interruptible_all(&u->peer_wait); | |
406 | ||
e27dfcea | 407 | skpair = unix_peer(sk); |
1da177e4 | 408 | |
e27dfcea | 409 | if (skpair != NULL) { |
1da177e4 | 410 | if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) { |
1c92b4e5 | 411 | unix_state_lock(skpair); |
1da177e4 LT |
412 | /* No more writes */ |
413 | skpair->sk_shutdown = SHUTDOWN_MASK; | |
414 | if (!skb_queue_empty(&sk->sk_receive_queue) || embrion) | |
415 | skpair->sk_err = ECONNRESET; | |
1c92b4e5 | 416 | unix_state_unlock(skpair); |
1da177e4 | 417 | skpair->sk_state_change(skpair); |
8d8ad9d7 | 418 | sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP); |
1da177e4 LT |
419 | } |
420 | sock_put(skpair); /* It may now die */ | |
421 | unix_peer(sk) = NULL; | |
422 | } | |
423 | ||
424 | /* Try to flush out this socket. Throw out buffers at least */ | |
425 | ||
426 | while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) { | |
e27dfcea | 427 | if (state == TCP_LISTEN) |
1da177e4 LT |
428 | unix_release_sock(skb->sk, 1); |
429 | /* passed fds are erased in the kfree_skb hook */ | |
430 | kfree_skb(skb); | |
431 | } | |
432 | ||
40ffe67d AV |
433 | if (path.dentry) |
434 | path_put(&path); | |
1da177e4 LT |
435 | |
436 | sock_put(sk); | |
437 | ||
438 | /* ---- Socket is dead now and most probably destroyed ---- */ | |
439 | ||
440 | /* | |
e04dae84 | 441 | * Fixme: BSD difference: In BSD all sockets connected to us get |
1da177e4 LT |
442 | * ECONNRESET and we die on the spot. In Linux we behave |
443 | * like files and pipes do and wait for the last | |
444 | * dereference. | |
445 | * | |
446 | * Can't we simply set sock->err? | |
447 | * | |
448 | * What the above comment does talk about? --ANK(980817) | |
449 | */ | |
450 | ||
9305cfa4 | 451 | if (unix_tot_inflight) |
ac7bfa62 | 452 | unix_gc(); /* Garbage collect fds */ |
1da177e4 LT |
453 | } |
454 | ||
109f6e39 EB |
455 | static void init_peercred(struct sock *sk) |
456 | { | |
457 | put_pid(sk->sk_peer_pid); | |
458 | if (sk->sk_peer_cred) | |
459 | put_cred(sk->sk_peer_cred); | |
460 | sk->sk_peer_pid = get_pid(task_tgid(current)); | |
461 | sk->sk_peer_cred = get_current_cred(); | |
462 | } | |
463 | ||
464 | static void copy_peercred(struct sock *sk, struct sock *peersk) | |
465 | { | |
466 | put_pid(sk->sk_peer_pid); | |
467 | if (sk->sk_peer_cred) | |
468 | put_cred(sk->sk_peer_cred); | |
469 | sk->sk_peer_pid = get_pid(peersk->sk_peer_pid); | |
470 | sk->sk_peer_cred = get_cred(peersk->sk_peer_cred); | |
471 | } | |
472 | ||
1da177e4 LT |
473 | static int unix_listen(struct socket *sock, int backlog) |
474 | { | |
475 | int err; | |
476 | struct sock *sk = sock->sk; | |
477 | struct unix_sock *u = unix_sk(sk); | |
109f6e39 | 478 | struct pid *old_pid = 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 | 499 | put_pid(old_pid); |
1da177e4 LT |
500 | out: |
501 | return err; | |
502 | } | |
503 | ||
504 | static int unix_release(struct socket *); | |
505 | static int unix_bind(struct socket *, struct sockaddr *, int); | |
506 | static int unix_stream_connect(struct socket *, struct sockaddr *, | |
507 | int addr_len, int flags); | |
508 | static int unix_socketpair(struct socket *, struct socket *); | |
509 | static int unix_accept(struct socket *, struct socket *, int); | |
510 | static int unix_getname(struct socket *, struct sockaddr *, int *, int); | |
511 | static unsigned int unix_poll(struct file *, struct socket *, poll_table *); | |
ec0d215f RW |
512 | static unsigned int unix_dgram_poll(struct file *, struct socket *, |
513 | poll_table *); | |
1da177e4 LT |
514 | static int unix_ioctl(struct socket *, unsigned int, unsigned long); |
515 | static int unix_shutdown(struct socket *, int); | |
516 | static int unix_stream_sendmsg(struct kiocb *, struct socket *, | |
517 | struct msghdr *, size_t); | |
518 | static int unix_stream_recvmsg(struct kiocb *, struct socket *, | |
519 | struct msghdr *, size_t, int); | |
520 | static int unix_dgram_sendmsg(struct kiocb *, struct socket *, | |
521 | struct msghdr *, size_t); | |
522 | static int unix_dgram_recvmsg(struct kiocb *, struct socket *, | |
523 | struct msghdr *, size_t, int); | |
524 | static int unix_dgram_connect(struct socket *, struct sockaddr *, | |
525 | int, int); | |
526 | static int unix_seqpacket_sendmsg(struct kiocb *, struct socket *, | |
527 | struct msghdr *, size_t); | |
a05d2ad1 EB |
528 | static int unix_seqpacket_recvmsg(struct kiocb *, struct socket *, |
529 | struct msghdr *, size_t, int); | |
1da177e4 | 530 | |
d90d9ff6 | 531 | static int unix_set_peek_off(struct sock *sk, int val) |
f55bb7f9 PE |
532 | { |
533 | struct unix_sock *u = unix_sk(sk); | |
534 | ||
d90d9ff6 SL |
535 | if (mutex_lock_interruptible(&u->readlock)) |
536 | return -EINTR; | |
537 | ||
f55bb7f9 PE |
538 | sk->sk_peek_off = val; |
539 | mutex_unlock(&u->readlock); | |
d90d9ff6 SL |
540 | |
541 | return 0; | |
f55bb7f9 PE |
542 | } |
543 | ||
544 | ||
90ddc4f0 | 545 | static const struct proto_ops unix_stream_ops = { |
1da177e4 LT |
546 | .family = PF_UNIX, |
547 | .owner = THIS_MODULE, | |
548 | .release = unix_release, | |
549 | .bind = unix_bind, | |
550 | .connect = unix_stream_connect, | |
551 | .socketpair = unix_socketpair, | |
552 | .accept = unix_accept, | |
553 | .getname = unix_getname, | |
554 | .poll = unix_poll, | |
555 | .ioctl = unix_ioctl, | |
556 | .listen = unix_listen, | |
557 | .shutdown = unix_shutdown, | |
558 | .setsockopt = sock_no_setsockopt, | |
559 | .getsockopt = sock_no_getsockopt, | |
560 | .sendmsg = unix_stream_sendmsg, | |
561 | .recvmsg = unix_stream_recvmsg, | |
562 | .mmap = sock_no_mmap, | |
563 | .sendpage = sock_no_sendpage, | |
fc0d7536 | 564 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
565 | }; |
566 | ||
90ddc4f0 | 567 | static const struct proto_ops unix_dgram_ops = { |
1da177e4 LT |
568 | .family = PF_UNIX, |
569 | .owner = THIS_MODULE, | |
570 | .release = unix_release, | |
571 | .bind = unix_bind, | |
572 | .connect = unix_dgram_connect, | |
573 | .socketpair = unix_socketpair, | |
574 | .accept = sock_no_accept, | |
575 | .getname = unix_getname, | |
ec0d215f | 576 | .poll = unix_dgram_poll, |
1da177e4 LT |
577 | .ioctl = unix_ioctl, |
578 | .listen = sock_no_listen, | |
579 | .shutdown = unix_shutdown, | |
580 | .setsockopt = sock_no_setsockopt, | |
581 | .getsockopt = sock_no_getsockopt, | |
582 | .sendmsg = unix_dgram_sendmsg, | |
583 | .recvmsg = unix_dgram_recvmsg, | |
584 | .mmap = sock_no_mmap, | |
585 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 586 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
587 | }; |
588 | ||
90ddc4f0 | 589 | static const struct proto_ops unix_seqpacket_ops = { |
1da177e4 LT |
590 | .family = PF_UNIX, |
591 | .owner = THIS_MODULE, | |
592 | .release = unix_release, | |
593 | .bind = unix_bind, | |
594 | .connect = unix_stream_connect, | |
595 | .socketpair = unix_socketpair, | |
596 | .accept = unix_accept, | |
597 | .getname = unix_getname, | |
ec0d215f | 598 | .poll = unix_dgram_poll, |
1da177e4 LT |
599 | .ioctl = unix_ioctl, |
600 | .listen = unix_listen, | |
601 | .shutdown = unix_shutdown, | |
602 | .setsockopt = sock_no_setsockopt, | |
603 | .getsockopt = sock_no_getsockopt, | |
604 | .sendmsg = unix_seqpacket_sendmsg, | |
a05d2ad1 | 605 | .recvmsg = unix_seqpacket_recvmsg, |
1da177e4 LT |
606 | .mmap = sock_no_mmap, |
607 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 608 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
609 | }; |
610 | ||
611 | static struct proto unix_proto = { | |
248969ae ED |
612 | .name = "UNIX", |
613 | .owner = THIS_MODULE, | |
248969ae | 614 | .obj_size = sizeof(struct unix_sock), |
1da177e4 LT |
615 | }; |
616 | ||
a09785a2 IM |
617 | /* |
618 | * AF_UNIX sockets do not interact with hardware, hence they | |
619 | * dont trigger interrupts - so it's safe for them to have | |
620 | * bh-unsafe locking for their sk_receive_queue.lock. Split off | |
621 | * this special lock-class by reinitializing the spinlock key: | |
622 | */ | |
623 | static struct lock_class_key af_unix_sk_receive_queue_lock_key; | |
624 | ||
6eba6a37 | 625 | static struct sock *unix_create1(struct net *net, struct socket *sock) |
1da177e4 LT |
626 | { |
627 | struct sock *sk = NULL; | |
628 | struct unix_sock *u; | |
629 | ||
518de9b3 ED |
630 | atomic_long_inc(&unix_nr_socks); |
631 | if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files()) | |
1da177e4 LT |
632 | goto out; |
633 | ||
6257ff21 | 634 | sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto); |
1da177e4 LT |
635 | if (!sk) |
636 | goto out; | |
637 | ||
6eba6a37 | 638 | sock_init_data(sock, sk); |
a09785a2 IM |
639 | lockdep_set_class(&sk->sk_receive_queue.lock, |
640 | &af_unix_sk_receive_queue_lock_key); | |
1da177e4 LT |
641 | |
642 | sk->sk_write_space = unix_write_space; | |
a0a53c8b | 643 | sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen; |
1da177e4 LT |
644 | sk->sk_destruct = unix_sock_destructor; |
645 | u = unix_sk(sk); | |
40ffe67d AV |
646 | u->path.dentry = NULL; |
647 | u->path.mnt = NULL; | |
fd19f329 | 648 | spin_lock_init(&u->lock); |
516e0cc5 | 649 | atomic_long_set(&u->inflight, 0); |
1fd05ba5 | 650 | INIT_LIST_HEAD(&u->link); |
57b47a53 | 651 | mutex_init(&u->readlock); /* single task reading lock */ |
1da177e4 | 652 | init_waitqueue_head(&u->peer_wait); |
7123aaa3 | 653 | unix_insert_socket(unix_sockets_unbound(sk), sk); |
1da177e4 | 654 | out: |
284b327b | 655 | if (sk == NULL) |
518de9b3 | 656 | atomic_long_dec(&unix_nr_socks); |
920de804 ED |
657 | else { |
658 | local_bh_disable(); | |
a8076d8d | 659 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
920de804 ED |
660 | local_bh_enable(); |
661 | } | |
1da177e4 LT |
662 | return sk; |
663 | } | |
664 | ||
3f378b68 EP |
665 | static int unix_create(struct net *net, struct socket *sock, int protocol, |
666 | int kern) | |
1da177e4 LT |
667 | { |
668 | if (protocol && protocol != PF_UNIX) | |
669 | return -EPROTONOSUPPORT; | |
670 | ||
671 | sock->state = SS_UNCONNECTED; | |
672 | ||
673 | switch (sock->type) { | |
674 | case SOCK_STREAM: | |
675 | sock->ops = &unix_stream_ops; | |
676 | break; | |
677 | /* | |
678 | * Believe it or not BSD has AF_UNIX, SOCK_RAW though | |
679 | * nothing uses it. | |
680 | */ | |
681 | case SOCK_RAW: | |
e27dfcea | 682 | sock->type = SOCK_DGRAM; |
1da177e4 LT |
683 | case SOCK_DGRAM: |
684 | sock->ops = &unix_dgram_ops; | |
685 | break; | |
686 | case SOCK_SEQPACKET: | |
687 | sock->ops = &unix_seqpacket_ops; | |
688 | break; | |
689 | default: | |
690 | return -ESOCKTNOSUPPORT; | |
691 | } | |
692 | ||
1b8d7ae4 | 693 | return unix_create1(net, sock) ? 0 : -ENOMEM; |
1da177e4 LT |
694 | } |
695 | ||
696 | static int unix_release(struct socket *sock) | |
697 | { | |
698 | struct sock *sk = sock->sk; | |
699 | ||
700 | if (!sk) | |
701 | return 0; | |
702 | ||
ded34e0f | 703 | unix_release_sock(sk, 0); |
1da177e4 LT |
704 | sock->sk = NULL; |
705 | ||
ded34e0f | 706 | return 0; |
1da177e4 LT |
707 | } |
708 | ||
709 | static int unix_autobind(struct socket *sock) | |
710 | { | |
711 | struct sock *sk = sock->sk; | |
3b1e0a65 | 712 | struct net *net = sock_net(sk); |
1da177e4 LT |
713 | struct unix_sock *u = unix_sk(sk); |
714 | static u32 ordernum = 1; | |
6eba6a37 | 715 | struct unix_address *addr; |
1da177e4 | 716 | int err; |
8df73ff9 | 717 | unsigned int retries = 0; |
1da177e4 | 718 | |
57bc52eb SL |
719 | err = mutex_lock_interruptible(&u->readlock); |
720 | if (err) | |
721 | return err; | |
1da177e4 LT |
722 | |
723 | err = 0; | |
724 | if (u->addr) | |
725 | goto out; | |
726 | ||
727 | err = -ENOMEM; | |
0da974f4 | 728 | addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL); |
1da177e4 LT |
729 | if (!addr) |
730 | goto out; | |
731 | ||
1da177e4 LT |
732 | addr->name->sun_family = AF_UNIX; |
733 | atomic_set(&addr->refcnt, 1); | |
734 | ||
735 | retry: | |
736 | addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short); | |
07f0757a | 737 | addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0)); |
1da177e4 | 738 | |
fbe9cc4a | 739 | spin_lock(&unix_table_lock); |
1da177e4 LT |
740 | ordernum = (ordernum+1)&0xFFFFF; |
741 | ||
097e66c5 | 742 | if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type, |
1da177e4 | 743 | addr->hash)) { |
fbe9cc4a | 744 | spin_unlock(&unix_table_lock); |
8df73ff9 TH |
745 | /* |
746 | * __unix_find_socket_byname() may take long time if many names | |
747 | * are already in use. | |
748 | */ | |
749 | cond_resched(); | |
750 | /* Give up if all names seems to be in use. */ | |
751 | if (retries++ == 0xFFFFF) { | |
752 | err = -ENOSPC; | |
753 | kfree(addr); | |
754 | goto out; | |
755 | } | |
1da177e4 LT |
756 | goto retry; |
757 | } | |
758 | addr->hash ^= sk->sk_type; | |
759 | ||
760 | __unix_remove_socket(sk); | |
761 | u->addr = addr; | |
762 | __unix_insert_socket(&unix_socket_table[addr->hash], sk); | |
fbe9cc4a | 763 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
764 | err = 0; |
765 | ||
57b47a53 | 766 | out: mutex_unlock(&u->readlock); |
1da177e4 LT |
767 | return err; |
768 | } | |
769 | ||
097e66c5 DL |
770 | static struct sock *unix_find_other(struct net *net, |
771 | struct sockaddr_un *sunname, int len, | |
95c96174 | 772 | int type, unsigned int hash, int *error) |
1da177e4 LT |
773 | { |
774 | struct sock *u; | |
421748ec | 775 | struct path path; |
1da177e4 | 776 | int err = 0; |
ac7bfa62 | 777 | |
1da177e4 | 778 | if (sunname->sun_path[0]) { |
421748ec AV |
779 | struct inode *inode; |
780 | err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path); | |
1da177e4 LT |
781 | if (err) |
782 | goto fail; | |
421748ec AV |
783 | inode = path.dentry->d_inode; |
784 | err = inode_permission(inode, MAY_WRITE); | |
1da177e4 LT |
785 | if (err) |
786 | goto put_fail; | |
787 | ||
788 | err = -ECONNREFUSED; | |
421748ec | 789 | if (!S_ISSOCK(inode->i_mode)) |
1da177e4 | 790 | goto put_fail; |
6616f788 | 791 | u = unix_find_socket_byinode(inode); |
1da177e4 LT |
792 | if (!u) |
793 | goto put_fail; | |
794 | ||
795 | if (u->sk_type == type) | |
68ac1234 | 796 | touch_atime(&path); |
1da177e4 | 797 | |
421748ec | 798 | path_put(&path); |
1da177e4 | 799 | |
e27dfcea | 800 | err = -EPROTOTYPE; |
1da177e4 LT |
801 | if (u->sk_type != type) { |
802 | sock_put(u); | |
803 | goto fail; | |
804 | } | |
805 | } else { | |
806 | err = -ECONNREFUSED; | |
e27dfcea | 807 | u = unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
808 | if (u) { |
809 | struct dentry *dentry; | |
40ffe67d | 810 | dentry = unix_sk(u)->path.dentry; |
1da177e4 | 811 | if (dentry) |
68ac1234 | 812 | touch_atime(&unix_sk(u)->path); |
1da177e4 LT |
813 | } else |
814 | goto fail; | |
815 | } | |
816 | return u; | |
817 | ||
818 | put_fail: | |
421748ec | 819 | path_put(&path); |
1da177e4 | 820 | fail: |
e27dfcea | 821 | *error = err; |
1da177e4 LT |
822 | return NULL; |
823 | } | |
824 | ||
faf02010 AV |
825 | static int unix_mknod(const char *sun_path, umode_t mode, struct path *res) |
826 | { | |
827 | struct dentry *dentry; | |
828 | struct path path; | |
829 | int err = 0; | |
830 | /* | |
831 | * Get the parent directory, calculate the hash for last | |
832 | * component. | |
833 | */ | |
834 | dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0); | |
835 | err = PTR_ERR(dentry); | |
836 | if (IS_ERR(dentry)) | |
837 | return err; | |
838 | ||
839 | /* | |
840 | * All right, let's create it. | |
841 | */ | |
842 | err = security_path_mknod(&path, dentry, mode, 0); | |
843 | if (!err) { | |
844 | err = vfs_mknod(path.dentry->d_inode, dentry, mode, 0); | |
845 | if (!err) { | |
846 | res->mnt = mntget(path.mnt); | |
847 | res->dentry = dget(dentry); | |
848 | } | |
849 | } | |
850 | done_path_create(&path, dentry); | |
851 | return err; | |
852 | } | |
1da177e4 LT |
853 | |
854 | static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
855 | { | |
856 | struct sock *sk = sock->sk; | |
3b1e0a65 | 857 | struct net *net = sock_net(sk); |
1da177e4 | 858 | struct unix_sock *u = unix_sk(sk); |
e27dfcea | 859 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
dae6ad8f | 860 | char *sun_path = sunaddr->sun_path; |
1da177e4 | 861 | int err; |
95c96174 | 862 | unsigned int hash; |
1da177e4 LT |
863 | struct unix_address *addr; |
864 | struct hlist_head *list; | |
865 | ||
866 | err = -EINVAL; | |
867 | if (sunaddr->sun_family != AF_UNIX) | |
868 | goto out; | |
869 | ||
e27dfcea | 870 | if (addr_len == sizeof(short)) { |
1da177e4 LT |
871 | err = unix_autobind(sock); |
872 | goto out; | |
873 | } | |
874 | ||
875 | err = unix_mkname(sunaddr, addr_len, &hash); | |
876 | if (err < 0) | |
877 | goto out; | |
878 | addr_len = err; | |
879 | ||
57bc52eb SL |
880 | err = mutex_lock_interruptible(&u->readlock); |
881 | if (err) | |
882 | goto out; | |
1da177e4 LT |
883 | |
884 | err = -EINVAL; | |
885 | if (u->addr) | |
886 | goto out_up; | |
887 | ||
888 | err = -ENOMEM; | |
889 | addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL); | |
890 | if (!addr) | |
891 | goto out_up; | |
892 | ||
893 | memcpy(addr->name, sunaddr, addr_len); | |
894 | addr->len = addr_len; | |
895 | addr->hash = hash ^ sk->sk_type; | |
896 | atomic_set(&addr->refcnt, 1); | |
897 | ||
dae6ad8f | 898 | if (sun_path[0]) { |
faf02010 AV |
899 | struct path path; |
900 | umode_t mode = S_IFSOCK | | |
ce3b0f8d | 901 | (SOCK_INODE(sock)->i_mode & ~current_umask()); |
faf02010 AV |
902 | err = unix_mknod(sun_path, mode, &path); |
903 | if (err) { | |
904 | if (err == -EEXIST) | |
905 | err = -EADDRINUSE; | |
906 | unix_release_addr(addr); | |
907 | goto out_up; | |
908 | } | |
1da177e4 | 909 | addr->hash = UNIX_HASH_SIZE; |
faf02010 AV |
910 | hash = path.dentry->d_inode->i_ino & (UNIX_HASH_SIZE-1); |
911 | spin_lock(&unix_table_lock); | |
912 | u->path = path; | |
913 | list = &unix_socket_table[hash]; | |
914 | } else { | |
915 | spin_lock(&unix_table_lock); | |
1da177e4 | 916 | err = -EADDRINUSE; |
097e66c5 | 917 | if (__unix_find_socket_byname(net, sunaddr, addr_len, |
1da177e4 LT |
918 | sk->sk_type, hash)) { |
919 | unix_release_addr(addr); | |
920 | goto out_unlock; | |
921 | } | |
922 | ||
923 | list = &unix_socket_table[addr->hash]; | |
1da177e4 LT |
924 | } |
925 | ||
926 | err = 0; | |
927 | __unix_remove_socket(sk); | |
928 | u->addr = addr; | |
929 | __unix_insert_socket(list, sk); | |
930 | ||
931 | out_unlock: | |
fbe9cc4a | 932 | spin_unlock(&unix_table_lock); |
1da177e4 | 933 | out_up: |
57b47a53 | 934 | mutex_unlock(&u->readlock); |
1da177e4 LT |
935 | out: |
936 | return err; | |
1da177e4 LT |
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 | 970 | struct sock *other; |
95c96174 | 971 | unsigned int hash; |
1da177e4 LT |
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; | |
95c96174 | 1069 | unsigned int hash; |
1da177e4 LT |
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 | } | |
40ffe67d AV |
1195 | if (otheru->path.dentry) { |
1196 | path_get(&otheru->path); | |
1197 | newu->path = otheru->path; | |
1da177e4 LT |
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 | ||
a769ad65 DB |
1255 | static void unix_sock_inherit_flags(const struct socket *old, |
1256 | struct socket *new) | |
1257 | { | |
1258 | if (test_bit(SOCK_PASSCRED, &old->flags)) | |
1259 | set_bit(SOCK_PASSCRED, &new->flags); | |
1260 | if (test_bit(SOCK_PASSSEC, &old->flags)) | |
1261 | set_bit(SOCK_PASSSEC, &new->flags); | |
1262 | } | |
1263 | ||
1da177e4 LT |
1264 | static int unix_accept(struct socket *sock, struct socket *newsock, int flags) |
1265 | { | |
1266 | struct sock *sk = sock->sk; | |
1267 | struct sock *tsk; | |
1268 | struct sk_buff *skb; | |
1269 | int err; | |
1270 | ||
1271 | err = -EOPNOTSUPP; | |
6eba6a37 | 1272 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
1da177e4 LT |
1273 | goto out; |
1274 | ||
1275 | err = -EINVAL; | |
1276 | if (sk->sk_state != TCP_LISTEN) | |
1277 | goto out; | |
1278 | ||
1279 | /* If socket state is TCP_LISTEN it cannot change (for now...), | |
1280 | * so that no locks are necessary. | |
1281 | */ | |
1282 | ||
1283 | skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err); | |
1284 | if (!skb) { | |
1285 | /* This means receive shutdown. */ | |
1286 | if (err == 0) | |
1287 | err = -EINVAL; | |
1288 | goto out; | |
1289 | } | |
1290 | ||
1291 | tsk = skb->sk; | |
1292 | skb_free_datagram(sk, skb); | |
1293 | wake_up_interruptible(&unix_sk(sk)->peer_wait); | |
1294 | ||
1295 | /* attach accepted sock to socket */ | |
1c92b4e5 | 1296 | unix_state_lock(tsk); |
1da177e4 | 1297 | newsock->state = SS_CONNECTED; |
a769ad65 | 1298 | unix_sock_inherit_flags(sock, newsock); |
1da177e4 | 1299 | sock_graft(tsk, newsock); |
1c92b4e5 | 1300 | unix_state_unlock(tsk); |
1da177e4 LT |
1301 | return 0; |
1302 | ||
1303 | out: | |
1304 | return err; | |
1305 | } | |
1306 | ||
1307 | ||
1308 | static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer) | |
1309 | { | |
1310 | struct sock *sk = sock->sk; | |
1311 | struct unix_sock *u; | |
13cfa97b | 1312 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr); |
1da177e4 LT |
1313 | int err = 0; |
1314 | ||
1315 | if (peer) { | |
1316 | sk = unix_peer_get(sk); | |
1317 | ||
1318 | err = -ENOTCONN; | |
1319 | if (!sk) | |
1320 | goto out; | |
1321 | err = 0; | |
1322 | } else { | |
1323 | sock_hold(sk); | |
1324 | } | |
1325 | ||
1326 | u = unix_sk(sk); | |
1c92b4e5 | 1327 | unix_state_lock(sk); |
1da177e4 LT |
1328 | if (!u->addr) { |
1329 | sunaddr->sun_family = AF_UNIX; | |
1330 | sunaddr->sun_path[0] = 0; | |
1331 | *uaddr_len = sizeof(short); | |
1332 | } else { | |
1333 | struct unix_address *addr = u->addr; | |
1334 | ||
1335 | *uaddr_len = addr->len; | |
1336 | memcpy(sunaddr, addr->name, *uaddr_len); | |
1337 | } | |
1c92b4e5 | 1338 | unix_state_unlock(sk); |
1da177e4 LT |
1339 | sock_put(sk); |
1340 | out: | |
1341 | return err; | |
1342 | } | |
1343 | ||
1344 | static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb) | |
1345 | { | |
1346 | int i; | |
1347 | ||
1348 | scm->fp = UNIXCB(skb).fp; | |
1da177e4 LT |
1349 | UNIXCB(skb).fp = NULL; |
1350 | ||
6eba6a37 | 1351 | for (i = scm->fp->count-1; i >= 0; i--) |
1da177e4 LT |
1352 | unix_notinflight(scm->fp->fp[i]); |
1353 | } | |
1354 | ||
7361c36c | 1355 | static void unix_destruct_scm(struct sk_buff *skb) |
1da177e4 LT |
1356 | { |
1357 | struct scm_cookie scm; | |
1358 | memset(&scm, 0, sizeof(scm)); | |
7361c36c | 1359 | scm.pid = UNIXCB(skb).pid; |
7361c36c EB |
1360 | if (UNIXCB(skb).fp) |
1361 | unix_detach_fds(&scm, skb); | |
1da177e4 LT |
1362 | |
1363 | /* Alas, it calls VFS */ | |
1364 | /* So fscking what? fput() had been SMP-safe since the last Summer */ | |
1365 | scm_destroy(&scm); | |
1366 | sock_wfree(skb); | |
1367 | } | |
1368 | ||
25888e30 ED |
1369 | #define MAX_RECURSION_LEVEL 4 |
1370 | ||
6209344f | 1371 | static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb) |
1da177e4 LT |
1372 | { |
1373 | int i; | |
25888e30 ED |
1374 | unsigned char max_level = 0; |
1375 | int unix_sock_count = 0; | |
1376 | ||
1377 | for (i = scm->fp->count - 1; i >= 0; i--) { | |
1378 | struct sock *sk = unix_get_socket(scm->fp->fp[i]); | |
1379 | ||
1380 | if (sk) { | |
1381 | unix_sock_count++; | |
1382 | max_level = max(max_level, | |
1383 | unix_sk(sk)->recursion_level); | |
1384 | } | |
1385 | } | |
1386 | if (unlikely(max_level > MAX_RECURSION_LEVEL)) | |
1387 | return -ETOOMANYREFS; | |
6209344f MS |
1388 | |
1389 | /* | |
1390 | * Need to duplicate file references for the sake of garbage | |
1391 | * collection. Otherwise a socket in the fps might become a | |
1392 | * candidate for GC while the skb is not yet queued. | |
1393 | */ | |
1394 | UNIXCB(skb).fp = scm_fp_dup(scm->fp); | |
1395 | if (!UNIXCB(skb).fp) | |
1396 | return -ENOMEM; | |
1397 | ||
25888e30 ED |
1398 | if (unix_sock_count) { |
1399 | for (i = scm->fp->count - 1; i >= 0; i--) | |
1400 | unix_inflight(scm->fp->fp[i]); | |
1401 | } | |
1402 | return max_level; | |
1da177e4 LT |
1403 | } |
1404 | ||
f78a5fda | 1405 | static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds) |
7361c36c EB |
1406 | { |
1407 | int err = 0; | |
16e57262 | 1408 | |
f78a5fda | 1409 | UNIXCB(skb).pid = get_pid(scm->pid); |
6b0ee8c0 EB |
1410 | UNIXCB(skb).uid = scm->creds.uid; |
1411 | UNIXCB(skb).gid = scm->creds.gid; | |
7361c36c EB |
1412 | UNIXCB(skb).fp = NULL; |
1413 | if (scm->fp && send_fds) | |
1414 | err = unix_attach_fds(scm, skb); | |
1415 | ||
1416 | skb->destructor = unix_destruct_scm; | |
1417 | return err; | |
1418 | } | |
1419 | ||
16e57262 ED |
1420 | /* |
1421 | * Some apps rely on write() giving SCM_CREDENTIALS | |
1422 | * We include credentials if source or destination socket | |
1423 | * asserted SOCK_PASSCRED. | |
1424 | */ | |
1425 | static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock, | |
1426 | const struct sock *other) | |
1427 | { | |
6b0ee8c0 | 1428 | if (UNIXCB(skb).pid) |
16e57262 ED |
1429 | return; |
1430 | if (test_bit(SOCK_PASSCRED, &sock->flags) || | |
25da0e3e EB |
1431 | !other->sk_socket || |
1432 | test_bit(SOCK_PASSCRED, &other->sk_socket->flags)) { | |
16e57262 | 1433 | UNIXCB(skb).pid = get_pid(task_tgid(current)); |
6e0895c2 | 1434 | current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid); |
16e57262 ED |
1435 | } |
1436 | } | |
1437 | ||
1da177e4 LT |
1438 | /* |
1439 | * Send AF_UNIX data. | |
1440 | */ | |
1441 | ||
1442 | static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock, | |
1443 | struct msghdr *msg, size_t len) | |
1444 | { | |
1445 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
1446 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1447 | struct net *net = sock_net(sk); |
1da177e4 | 1448 | struct unix_sock *u = unix_sk(sk); |
e27dfcea | 1449 | struct sockaddr_un *sunaddr = msg->msg_name; |
1da177e4 LT |
1450 | struct sock *other = NULL; |
1451 | int namelen = 0; /* fake GCC */ | |
1452 | int err; | |
95c96174 | 1453 | unsigned int hash; |
f78a5fda | 1454 | struct sk_buff *skb; |
1da177e4 LT |
1455 | long timeo; |
1456 | struct scm_cookie tmp_scm; | |
25888e30 | 1457 | int max_level; |
eb6a2481 | 1458 | int data_len = 0; |
1da177e4 LT |
1459 | |
1460 | if (NULL == siocb->scm) | |
1461 | siocb->scm = &tmp_scm; | |
5f23b734 | 1462 | wait_for_unix_gc(); |
e0e3cea4 | 1463 | err = scm_send(sock, msg, siocb->scm, false); |
1da177e4 LT |
1464 | if (err < 0) |
1465 | return err; | |
1466 | ||
1467 | err = -EOPNOTSUPP; | |
1468 | if (msg->msg_flags&MSG_OOB) | |
1469 | goto out; | |
1470 | ||
1471 | if (msg->msg_namelen) { | |
1472 | err = unix_mkname(sunaddr, msg->msg_namelen, &hash); | |
1473 | if (err < 0) | |
1474 | goto out; | |
1475 | namelen = err; | |
1476 | } else { | |
1477 | sunaddr = NULL; | |
1478 | err = -ENOTCONN; | |
1479 | other = unix_peer_get(sk); | |
1480 | if (!other) | |
1481 | goto out; | |
1482 | } | |
1483 | ||
f64f9e71 JP |
1484 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr |
1485 | && (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1486 | goto out; |
1487 | ||
1488 | err = -EMSGSIZE; | |
1489 | if (len > sk->sk_sndbuf - 32) | |
1490 | goto out; | |
1491 | ||
eb6a2481 ED |
1492 | if (len > SKB_MAX_ALLOC) |
1493 | data_len = min_t(size_t, | |
1494 | len - SKB_MAX_ALLOC, | |
1495 | MAX_SKB_FRAGS * PAGE_SIZE); | |
1496 | ||
1497 | skb = sock_alloc_send_pskb(sk, len - data_len, data_len, | |
1498 | msg->msg_flags & MSG_DONTWAIT, &err); | |
e27dfcea | 1499 | if (skb == NULL) |
1da177e4 LT |
1500 | goto out; |
1501 | ||
f78a5fda | 1502 | err = unix_scm_to_skb(siocb->scm, skb, true); |
25888e30 | 1503 | if (err < 0) |
7361c36c | 1504 | goto out_free; |
25888e30 | 1505 | max_level = err + 1; |
dc49c1f9 | 1506 | unix_get_secdata(siocb->scm, skb); |
877ce7c1 | 1507 | |
eb6a2481 ED |
1508 | skb_put(skb, len - data_len); |
1509 | skb->data_len = data_len; | |
1510 | skb->len = len; | |
1511 | err = skb_copy_datagram_from_iovec(skb, 0, msg->msg_iov, 0, len); | |
1da177e4 LT |
1512 | if (err) |
1513 | goto out_free; | |
1514 | ||
1515 | timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); | |
1516 | ||
1517 | restart: | |
1518 | if (!other) { | |
1519 | err = -ECONNRESET; | |
1520 | if (sunaddr == NULL) | |
1521 | goto out_free; | |
1522 | ||
097e66c5 | 1523 | other = unix_find_other(net, sunaddr, namelen, sk->sk_type, |
1da177e4 | 1524 | hash, &err); |
e27dfcea | 1525 | if (other == NULL) |
1da177e4 LT |
1526 | goto out_free; |
1527 | } | |
1528 | ||
d6ae3bae AC |
1529 | if (sk_filter(other, skb) < 0) { |
1530 | /* Toss the packet but do not return any error to the sender */ | |
1531 | err = len; | |
1532 | goto out_free; | |
1533 | } | |
1534 | ||
1c92b4e5 | 1535 | unix_state_lock(other); |
1da177e4 LT |
1536 | err = -EPERM; |
1537 | if (!unix_may_send(sk, other)) | |
1538 | goto out_unlock; | |
1539 | ||
1540 | if (sock_flag(other, SOCK_DEAD)) { | |
1541 | /* | |
1542 | * Check with 1003.1g - what should | |
1543 | * datagram error | |
1544 | */ | |
1c92b4e5 | 1545 | unix_state_unlock(other); |
1da177e4 LT |
1546 | sock_put(other); |
1547 | ||
1548 | err = 0; | |
1c92b4e5 | 1549 | unix_state_lock(sk); |
1da177e4 | 1550 | if (unix_peer(sk) == other) { |
e27dfcea | 1551 | unix_peer(sk) = NULL; |
1c92b4e5 | 1552 | unix_state_unlock(sk); |
1da177e4 LT |
1553 | |
1554 | unix_dgram_disconnected(sk, other); | |
1555 | sock_put(other); | |
1556 | err = -ECONNREFUSED; | |
1557 | } else { | |
1c92b4e5 | 1558 | unix_state_unlock(sk); |
1da177e4 LT |
1559 | } |
1560 | ||
1561 | other = NULL; | |
1562 | if (err) | |
1563 | goto out_free; | |
1564 | goto restart; | |
1565 | } | |
1566 | ||
1567 | err = -EPIPE; | |
1568 | if (other->sk_shutdown & RCV_SHUTDOWN) | |
1569 | goto out_unlock; | |
1570 | ||
1571 | if (sk->sk_type != SOCK_SEQPACKET) { | |
1572 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1573 | if (err) | |
1574 | goto out_unlock; | |
1575 | } | |
1576 | ||
3c73419c | 1577 | if (unix_peer(other) != sk && unix_recvq_full(other)) { |
1da177e4 LT |
1578 | if (!timeo) { |
1579 | err = -EAGAIN; | |
1580 | goto out_unlock; | |
1581 | } | |
1582 | ||
1583 | timeo = unix_wait_for_peer(other, timeo); | |
1584 | ||
1585 | err = sock_intr_errno(timeo); | |
1586 | if (signal_pending(current)) | |
1587 | goto out_free; | |
1588 | ||
1589 | goto restart; | |
1590 | } | |
1591 | ||
3f66116e AC |
1592 | if (sock_flag(other, SOCK_RCVTSTAMP)) |
1593 | __net_timestamp(skb); | |
16e57262 | 1594 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1595 | skb_queue_tail(&other->sk_receive_queue, skb); |
25888e30 ED |
1596 | if (max_level > unix_sk(other)->recursion_level) |
1597 | unix_sk(other)->recursion_level = max_level; | |
1c92b4e5 | 1598 | unix_state_unlock(other); |
1da177e4 LT |
1599 | other->sk_data_ready(other, len); |
1600 | sock_put(other); | |
f78a5fda | 1601 | scm_destroy(siocb->scm); |
1da177e4 LT |
1602 | return len; |
1603 | ||
1604 | out_unlock: | |
1c92b4e5 | 1605 | unix_state_unlock(other); |
1da177e4 LT |
1606 | out_free: |
1607 | kfree_skb(skb); | |
1608 | out: | |
1609 | if (other) | |
1610 | sock_put(other); | |
f78a5fda | 1611 | scm_destroy(siocb->scm); |
1da177e4 LT |
1612 | return err; |
1613 | } | |
1614 | ||
ac7bfa62 | 1615 | |
1da177e4 LT |
1616 | static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock, |
1617 | struct msghdr *msg, size_t len) | |
1618 | { | |
1619 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
1620 | struct sock *sk = sock->sk; | |
1621 | struct sock *other = NULL; | |
6eba6a37 | 1622 | int err, size; |
f78a5fda | 1623 | struct sk_buff *skb; |
e27dfcea | 1624 | int sent = 0; |
1da177e4 | 1625 | struct scm_cookie tmp_scm; |
8ba69ba6 | 1626 | bool fds_sent = false; |
25888e30 | 1627 | int max_level; |
1da177e4 LT |
1628 | |
1629 | if (NULL == siocb->scm) | |
1630 | siocb->scm = &tmp_scm; | |
5f23b734 | 1631 | wait_for_unix_gc(); |
e0e3cea4 | 1632 | err = scm_send(sock, msg, siocb->scm, false); |
1da177e4 LT |
1633 | if (err < 0) |
1634 | return err; | |
1635 | ||
1636 | err = -EOPNOTSUPP; | |
1637 | if (msg->msg_flags&MSG_OOB) | |
1638 | goto out_err; | |
1639 | ||
1640 | if (msg->msg_namelen) { | |
1641 | err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; | |
1642 | goto out_err; | |
1643 | } else { | |
1da177e4 | 1644 | err = -ENOTCONN; |
830a1e5c | 1645 | other = unix_peer(sk); |
1da177e4 LT |
1646 | if (!other) |
1647 | goto out_err; | |
1648 | } | |
1649 | ||
1650 | if (sk->sk_shutdown & SEND_SHUTDOWN) | |
1651 | goto pipe_err; | |
1652 | ||
6eba6a37 | 1653 | while (sent < len) { |
1da177e4 | 1654 | /* |
e9df7d7f BL |
1655 | * Optimisation for the fact that under 0.01% of X |
1656 | * messages typically need breaking up. | |
1da177e4 LT |
1657 | */ |
1658 | ||
e9df7d7f | 1659 | size = len-sent; |
1da177e4 LT |
1660 | |
1661 | /* Keep two messages in the pipe so it schedules better */ | |
e9df7d7f BL |
1662 | if (size > ((sk->sk_sndbuf >> 1) - 64)) |
1663 | size = (sk->sk_sndbuf >> 1) - 64; | |
1da177e4 LT |
1664 | |
1665 | if (size > SKB_MAX_ALLOC) | |
1666 | size = SKB_MAX_ALLOC; | |
ac7bfa62 | 1667 | |
1da177e4 LT |
1668 | /* |
1669 | * Grab a buffer | |
1670 | */ | |
ac7bfa62 | 1671 | |
6eba6a37 ED |
1672 | skb = sock_alloc_send_skb(sk, size, msg->msg_flags&MSG_DONTWAIT, |
1673 | &err); | |
1da177e4 | 1674 | |
e27dfcea | 1675 | if (skb == NULL) |
1da177e4 LT |
1676 | goto out_err; |
1677 | ||
1678 | /* | |
1679 | * If you pass two values to the sock_alloc_send_skb | |
1680 | * it tries to grab the large buffer with GFP_NOFS | |
1681 | * (which can fail easily), and if it fails grab the | |
1682 | * fallback size buffer which is under a page and will | |
1683 | * succeed. [Alan] | |
1684 | */ | |
1685 | size = min_t(int, size, skb_tailroom(skb)); | |
1686 | ||
7361c36c | 1687 | |
f78a5fda DM |
1688 | /* Only send the fds in the first buffer */ |
1689 | err = unix_scm_to_skb(siocb->scm, skb, !fds_sent); | |
25888e30 | 1690 | if (err < 0) { |
7361c36c | 1691 | kfree_skb(skb); |
f78a5fda | 1692 | goto out_err; |
6209344f | 1693 | } |
25888e30 | 1694 | max_level = err + 1; |
7361c36c | 1695 | fds_sent = true; |
1da177e4 | 1696 | |
6eba6a37 ED |
1697 | err = memcpy_fromiovec(skb_put(skb, size), msg->msg_iov, size); |
1698 | if (err) { | |
1da177e4 | 1699 | kfree_skb(skb); |
f78a5fda | 1700 | goto out_err; |
1da177e4 LT |
1701 | } |
1702 | ||
1c92b4e5 | 1703 | unix_state_lock(other); |
1da177e4 LT |
1704 | |
1705 | if (sock_flag(other, SOCK_DEAD) || | |
1706 | (other->sk_shutdown & RCV_SHUTDOWN)) | |
1707 | goto pipe_err_free; | |
1708 | ||
16e57262 | 1709 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1710 | skb_queue_tail(&other->sk_receive_queue, skb); |
25888e30 ED |
1711 | if (max_level > unix_sk(other)->recursion_level) |
1712 | unix_sk(other)->recursion_level = max_level; | |
1c92b4e5 | 1713 | unix_state_unlock(other); |
1da177e4 | 1714 | other->sk_data_ready(other, size); |
e27dfcea | 1715 | sent += size; |
1da177e4 | 1716 | } |
1da177e4 | 1717 | |
f78a5fda | 1718 | scm_destroy(siocb->scm); |
1da177e4 LT |
1719 | siocb->scm = NULL; |
1720 | ||
1721 | return sent; | |
1722 | ||
1723 | pipe_err_free: | |
1c92b4e5 | 1724 | unix_state_unlock(other); |
1da177e4 LT |
1725 | kfree_skb(skb); |
1726 | pipe_err: | |
6eba6a37 ED |
1727 | if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL)) |
1728 | send_sig(SIGPIPE, current, 0); | |
1da177e4 LT |
1729 | err = -EPIPE; |
1730 | out_err: | |
f78a5fda | 1731 | scm_destroy(siocb->scm); |
1da177e4 LT |
1732 | siocb->scm = NULL; |
1733 | return sent ? : err; | |
1734 | } | |
1735 | ||
1736 | static int unix_seqpacket_sendmsg(struct kiocb *kiocb, struct socket *sock, | |
1737 | struct msghdr *msg, size_t len) | |
1738 | { | |
1739 | int err; | |
1740 | struct sock *sk = sock->sk; | |
ac7bfa62 | 1741 | |
1da177e4 LT |
1742 | err = sock_error(sk); |
1743 | if (err) | |
1744 | return err; | |
1745 | ||
1746 | if (sk->sk_state != TCP_ESTABLISHED) | |
1747 | return -ENOTCONN; | |
1748 | ||
1749 | if (msg->msg_namelen) | |
1750 | msg->msg_namelen = 0; | |
1751 | ||
1752 | return unix_dgram_sendmsg(kiocb, sock, msg, len); | |
1753 | } | |
ac7bfa62 | 1754 | |
a05d2ad1 EB |
1755 | static int unix_seqpacket_recvmsg(struct kiocb *iocb, struct socket *sock, |
1756 | struct msghdr *msg, size_t size, | |
1757 | int flags) | |
1758 | { | |
1759 | struct sock *sk = sock->sk; | |
1760 | ||
1761 | if (sk->sk_state != TCP_ESTABLISHED) | |
1762 | return -ENOTCONN; | |
1763 | ||
1764 | return unix_dgram_recvmsg(iocb, sock, msg, size, flags); | |
1765 | } | |
1766 | ||
1da177e4 LT |
1767 | static void unix_copy_addr(struct msghdr *msg, struct sock *sk) |
1768 | { | |
1769 | struct unix_sock *u = unix_sk(sk); | |
1770 | ||
1da177e4 LT |
1771 | if (u->addr) { |
1772 | msg->msg_namelen = u->addr->len; | |
1773 | memcpy(msg->msg_name, u->addr->name, u->addr->len); | |
1774 | } | |
1775 | } | |
1776 | ||
1777 | static int unix_dgram_recvmsg(struct kiocb *iocb, struct socket *sock, | |
1778 | struct msghdr *msg, size_t size, | |
1779 | int flags) | |
1780 | { | |
1781 | struct sock_iocb *siocb = kiocb_to_siocb(iocb); | |
1782 | struct scm_cookie tmp_scm; | |
1783 | struct sock *sk = sock->sk; | |
1784 | struct unix_sock *u = unix_sk(sk); | |
1785 | int noblock = flags & MSG_DONTWAIT; | |
1786 | struct sk_buff *skb; | |
1787 | int err; | |
f55bb7f9 | 1788 | int peeked, skip; |
1da177e4 LT |
1789 | |
1790 | err = -EOPNOTSUPP; | |
1791 | if (flags&MSG_OOB) | |
1792 | goto out; | |
1793 | ||
b3ca9b02 | 1794 | err = mutex_lock_interruptible(&u->readlock); |
f423fefe ED |
1795 | if (unlikely(err)) { |
1796 | /* recvmsg() in non blocking mode is supposed to return -EAGAIN | |
1797 | * sk_rcvtimeo is not honored by mutex_lock_interruptible() | |
1798 | */ | |
1799 | err = noblock ? -EAGAIN : -ERESTARTSYS; | |
b3ca9b02 RW |
1800 | goto out; |
1801 | } | |
1da177e4 | 1802 | |
f55bb7f9 PE |
1803 | skip = sk_peek_offset(sk, flags); |
1804 | ||
1805 | skb = __skb_recv_datagram(sk, flags, &peeked, &skip, &err); | |
0a112258 FZ |
1806 | if (!skb) { |
1807 | unix_state_lock(sk); | |
1808 | /* Signal EOF on disconnected non-blocking SEQPACKET socket. */ | |
1809 | if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN && | |
1810 | (sk->sk_shutdown & RCV_SHUTDOWN)) | |
1811 | err = 0; | |
1812 | unix_state_unlock(sk); | |
1da177e4 | 1813 | goto out_unlock; |
0a112258 | 1814 | } |
1da177e4 | 1815 | |
67426b75 ED |
1816 | wake_up_interruptible_sync_poll(&u->peer_wait, |
1817 | POLLOUT | POLLWRNORM | POLLWRBAND); | |
1da177e4 LT |
1818 | |
1819 | if (msg->msg_name) | |
1820 | unix_copy_addr(msg, skb->sk); | |
1821 | ||
f55bb7f9 PE |
1822 | if (size > skb->len - skip) |
1823 | size = skb->len - skip; | |
1824 | else if (size < skb->len - skip) | |
1da177e4 LT |
1825 | msg->msg_flags |= MSG_TRUNC; |
1826 | ||
f55bb7f9 | 1827 | err = skb_copy_datagram_iovec(skb, skip, msg->msg_iov, size); |
1da177e4 LT |
1828 | if (err) |
1829 | goto out_free; | |
1830 | ||
3f66116e AC |
1831 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
1832 | __sock_recv_timestamp(msg, sk, skb); | |
1833 | ||
1da177e4 LT |
1834 | if (!siocb->scm) { |
1835 | siocb->scm = &tmp_scm; | |
1836 | memset(&tmp_scm, 0, sizeof(tmp_scm)); | |
1837 | } | |
6b0ee8c0 | 1838 | scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); |
877ce7c1 | 1839 | unix_set_secdata(siocb->scm, skb); |
1da177e4 | 1840 | |
6eba6a37 | 1841 | if (!(flags & MSG_PEEK)) { |
1da177e4 LT |
1842 | if (UNIXCB(skb).fp) |
1843 | unix_detach_fds(siocb->scm, skb); | |
f55bb7f9 PE |
1844 | |
1845 | sk_peek_offset_bwd(sk, skb->len); | |
6eba6a37 | 1846 | } else { |
1da177e4 LT |
1847 | /* It is questionable: on PEEK we could: |
1848 | - do not return fds - good, but too simple 8) | |
1849 | - return fds, and do not return them on read (old strategy, | |
1850 | apparently wrong) | |
1851 | - clone fds (I chose it for now, it is the most universal | |
1852 | solution) | |
ac7bfa62 YH |
1853 | |
1854 | POSIX 1003.1g does not actually define this clearly | |
1855 | at all. POSIX 1003.1g doesn't define a lot of things | |
1856 | clearly however! | |
1857 | ||
1da177e4 | 1858 | */ |
f55bb7f9 PE |
1859 | |
1860 | sk_peek_offset_fwd(sk, size); | |
1861 | ||
1da177e4 LT |
1862 | if (UNIXCB(skb).fp) |
1863 | siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp); | |
1864 | } | |
9f6f9af7 | 1865 | err = (flags & MSG_TRUNC) ? skb->len - skip : size; |
1da177e4 LT |
1866 | |
1867 | scm_recv(sock, msg, siocb->scm, flags); | |
1868 | ||
1869 | out_free: | |
6eba6a37 | 1870 | skb_free_datagram(sk, skb); |
1da177e4 | 1871 | out_unlock: |
57b47a53 | 1872 | mutex_unlock(&u->readlock); |
1da177e4 LT |
1873 | out: |
1874 | return err; | |
1875 | } | |
1876 | ||
1877 | /* | |
79f632c7 | 1878 | * Sleep until more data has arrived. But check for races.. |
1da177e4 | 1879 | */ |
79f632c7 BP |
1880 | static long unix_stream_data_wait(struct sock *sk, long timeo, |
1881 | struct sk_buff *last) | |
1da177e4 LT |
1882 | { |
1883 | DEFINE_WAIT(wait); | |
1884 | ||
1c92b4e5 | 1885 | unix_state_lock(sk); |
1da177e4 LT |
1886 | |
1887 | for (;;) { | |
aa395145 | 1888 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 1889 | |
79f632c7 | 1890 | if (skb_peek_tail(&sk->sk_receive_queue) != last || |
1da177e4 LT |
1891 | sk->sk_err || |
1892 | (sk->sk_shutdown & RCV_SHUTDOWN) || | |
1893 | signal_pending(current) || | |
1894 | !timeo) | |
1895 | break; | |
1896 | ||
1897 | set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); | |
1c92b4e5 | 1898 | unix_state_unlock(sk); |
1da177e4 | 1899 | timeo = schedule_timeout(timeo); |
1c92b4e5 | 1900 | unix_state_lock(sk); |
7659c934 MS |
1901 | |
1902 | if (sock_flag(sk, SOCK_DEAD)) | |
1903 | break; | |
1904 | ||
1da177e4 LT |
1905 | clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); |
1906 | } | |
1907 | ||
aa395145 | 1908 | finish_wait(sk_sleep(sk), &wait); |
1c92b4e5 | 1909 | unix_state_unlock(sk); |
1da177e4 LT |
1910 | return timeo; |
1911 | } | |
1912 | ||
1da177e4 LT |
1913 | static int unix_stream_recvmsg(struct kiocb *iocb, struct socket *sock, |
1914 | struct msghdr *msg, size_t size, | |
1915 | int flags) | |
1916 | { | |
1917 | struct sock_iocb *siocb = kiocb_to_siocb(iocb); | |
1918 | struct scm_cookie tmp_scm; | |
1919 | struct sock *sk = sock->sk; | |
1920 | struct unix_sock *u = unix_sk(sk); | |
e27dfcea | 1921 | struct sockaddr_un *sunaddr = msg->msg_name; |
1da177e4 | 1922 | int copied = 0; |
f423fefe | 1923 | int noblock = flags & MSG_DONTWAIT; |
1da177e4 LT |
1924 | int check_creds = 0; |
1925 | int target; | |
1926 | int err = 0; | |
1927 | long timeo; | |
fc0d7536 | 1928 | int skip; |
1da177e4 LT |
1929 | |
1930 | err = -EINVAL; | |
1931 | if (sk->sk_state != TCP_ESTABLISHED) | |
1932 | goto out; | |
1933 | ||
1934 | err = -EOPNOTSUPP; | |
1935 | if (flags&MSG_OOB) | |
1936 | goto out; | |
1937 | ||
1938 | target = sock_rcvlowat(sk, flags&MSG_WAITALL, size); | |
f423fefe | 1939 | timeo = sock_rcvtimeo(sk, noblock); |
1da177e4 | 1940 | |
1da177e4 LT |
1941 | /* Lock the socket to prevent queue disordering |
1942 | * while sleeps in memcpy_tomsg | |
1943 | */ | |
1944 | ||
1945 | if (!siocb->scm) { | |
1946 | siocb->scm = &tmp_scm; | |
1947 | memset(&tmp_scm, 0, sizeof(tmp_scm)); | |
1948 | } | |
1949 | ||
b3ca9b02 | 1950 | err = mutex_lock_interruptible(&u->readlock); |
f423fefe ED |
1951 | if (unlikely(err)) { |
1952 | /* recvmsg() in non blocking mode is supposed to return -EAGAIN | |
1953 | * sk_rcvtimeo is not honored by mutex_lock_interruptible() | |
1954 | */ | |
1955 | err = noblock ? -EAGAIN : -ERESTARTSYS; | |
b3ca9b02 RW |
1956 | goto out; |
1957 | } | |
1da177e4 | 1958 | |
6eba6a37 | 1959 | do { |
1da177e4 | 1960 | int chunk; |
79f632c7 | 1961 | struct sk_buff *skb, *last; |
1da177e4 | 1962 | |
3c0d2f37 | 1963 | unix_state_lock(sk); |
7659c934 MS |
1964 | if (sock_flag(sk, SOCK_DEAD)) { |
1965 | err = -ECONNRESET; | |
1966 | goto unlock; | |
1967 | } | |
79f632c7 | 1968 | last = skb = skb_peek(&sk->sk_receive_queue); |
fc0d7536 | 1969 | again: |
6eba6a37 | 1970 | if (skb == NULL) { |
25888e30 | 1971 | unix_sk(sk)->recursion_level = 0; |
1da177e4 | 1972 | if (copied >= target) |
3c0d2f37 | 1973 | goto unlock; |
1da177e4 LT |
1974 | |
1975 | /* | |
1976 | * POSIX 1003.1g mandates this order. | |
1977 | */ | |
ac7bfa62 | 1978 | |
6eba6a37 ED |
1979 | err = sock_error(sk); |
1980 | if (err) | |
3c0d2f37 | 1981 | goto unlock; |
1da177e4 | 1982 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
3c0d2f37 MS |
1983 | goto unlock; |
1984 | ||
1985 | unix_state_unlock(sk); | |
1da177e4 LT |
1986 | err = -EAGAIN; |
1987 | if (!timeo) | |
1988 | break; | |
57b47a53 | 1989 | mutex_unlock(&u->readlock); |
1da177e4 | 1990 | |
79f632c7 | 1991 | timeo = unix_stream_data_wait(sk, timeo, last); |
1da177e4 | 1992 | |
b3ca9b02 RW |
1993 | if (signal_pending(current) |
1994 | || mutex_lock_interruptible(&u->readlock)) { | |
1da177e4 LT |
1995 | err = sock_intr_errno(timeo); |
1996 | goto out; | |
1997 | } | |
b3ca9b02 | 1998 | |
1da177e4 | 1999 | continue; |
3c0d2f37 MS |
2000 | unlock: |
2001 | unix_state_unlock(sk); | |
2002 | break; | |
1da177e4 | 2003 | } |
fc0d7536 | 2004 | |
79f632c7 BP |
2005 | skip = sk_peek_offset(sk, flags); |
2006 | while (skip >= skb->len) { | |
fc0d7536 | 2007 | skip -= skb->len; |
79f632c7 | 2008 | last = skb; |
fc0d7536 | 2009 | skb = skb_peek_next(skb, &sk->sk_receive_queue); |
79f632c7 BP |
2010 | if (!skb) |
2011 | goto again; | |
fc0d7536 PE |
2012 | } |
2013 | ||
3c0d2f37 | 2014 | unix_state_unlock(sk); |
1da177e4 LT |
2015 | |
2016 | if (check_creds) { | |
2017 | /* Never glue messages from different writers */ | |
7361c36c | 2018 | if ((UNIXCB(skb).pid != siocb->scm->pid) || |
6b0ee8c0 EB |
2019 | !uid_eq(UNIXCB(skb).uid, siocb->scm->creds.uid) || |
2020 | !gid_eq(UNIXCB(skb).gid, siocb->scm->creds.gid)) | |
1da177e4 | 2021 | break; |
0e82e7f6 | 2022 | } else if (test_bit(SOCK_PASSCRED, &sock->flags)) { |
1da177e4 | 2023 | /* Copy credentials */ |
6b0ee8c0 | 2024 | scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); |
1da177e4 LT |
2025 | check_creds = 1; |
2026 | } | |
2027 | ||
2028 | /* Copy address just once */ | |
6eba6a37 | 2029 | if (sunaddr) { |
1da177e4 LT |
2030 | unix_copy_addr(msg, skb->sk); |
2031 | sunaddr = NULL; | |
2032 | } | |
2033 | ||
fc0d7536 PE |
2034 | chunk = min_t(unsigned int, skb->len - skip, size); |
2035 | if (memcpy_toiovec(msg->msg_iov, skb->data + skip, chunk)) { | |
1da177e4 LT |
2036 | if (copied == 0) |
2037 | copied = -EFAULT; | |
2038 | break; | |
2039 | } | |
2040 | copied += chunk; | |
2041 | size -= chunk; | |
2042 | ||
2043 | /* Mark read part of skb as used */ | |
6eba6a37 | 2044 | if (!(flags & MSG_PEEK)) { |
1da177e4 LT |
2045 | skb_pull(skb, chunk); |
2046 | ||
fc0d7536 PE |
2047 | sk_peek_offset_bwd(sk, chunk); |
2048 | ||
1da177e4 LT |
2049 | if (UNIXCB(skb).fp) |
2050 | unix_detach_fds(siocb->scm, skb); | |
2051 | ||
6f01fd6e | 2052 | if (skb->len) |
1da177e4 | 2053 | break; |
1da177e4 | 2054 | |
6f01fd6e | 2055 | skb_unlink(skb, &sk->sk_receive_queue); |
70d4bf6d | 2056 | consume_skb(skb); |
1da177e4 LT |
2057 | |
2058 | if (siocb->scm->fp) | |
2059 | break; | |
6eba6a37 | 2060 | } else { |
1da177e4 LT |
2061 | /* It is questionable, see note in unix_dgram_recvmsg. |
2062 | */ | |
2063 | if (UNIXCB(skb).fp) | |
2064 | siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp); | |
2065 | ||
1f21dc67 AC |
2066 | if (skip) { |
2067 | sk_peek_offset_fwd(sk, chunk); | |
2068 | skip -= chunk; | |
2069 | } | |
2070 | ||
2071 | if (UNIXCB(skb).fp) | |
2072 | break; | |
fc0d7536 | 2073 | |
1f21dc67 AC |
2074 | last = skb; |
2075 | unix_state_lock(sk); | |
2076 | skb = skb_peek_next(skb, &sk->sk_receive_queue); | |
2077 | if (skb) | |
2078 | goto again; | |
2079 | unix_state_unlock(sk); | |
1da177e4 LT |
2080 | break; |
2081 | } | |
2082 | } while (size); | |
2083 | ||
57b47a53 | 2084 | mutex_unlock(&u->readlock); |
1da177e4 LT |
2085 | scm_recv(sock, msg, siocb->scm, flags); |
2086 | out: | |
2087 | return copied ? : err; | |
2088 | } | |
2089 | ||
2090 | static int unix_shutdown(struct socket *sock, int mode) | |
2091 | { | |
2092 | struct sock *sk = sock->sk; | |
2093 | struct sock *other; | |
2094 | ||
fc61b928 XW |
2095 | if (mode < SHUT_RD || mode > SHUT_RDWR) |
2096 | return -EINVAL; | |
2097 | /* This maps: | |
2098 | * SHUT_RD (0) -> RCV_SHUTDOWN (1) | |
2099 | * SHUT_WR (1) -> SEND_SHUTDOWN (2) | |
2100 | * SHUT_RDWR (2) -> SHUTDOWN_MASK (3) | |
2101 | */ | |
2102 | ++mode; | |
7180a031 AC |
2103 | |
2104 | unix_state_lock(sk); | |
2105 | sk->sk_shutdown |= mode; | |
2106 | other = unix_peer(sk); | |
2107 | if (other) | |
2108 | sock_hold(other); | |
2109 | unix_state_unlock(sk); | |
2110 | sk->sk_state_change(sk); | |
2111 | ||
2112 | if (other && | |
2113 | (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) { | |
2114 | ||
2115 | int peer_mode = 0; | |
2116 | ||
2117 | if (mode&RCV_SHUTDOWN) | |
2118 | peer_mode |= SEND_SHUTDOWN; | |
2119 | if (mode&SEND_SHUTDOWN) | |
2120 | peer_mode |= RCV_SHUTDOWN; | |
2121 | unix_state_lock(other); | |
2122 | other->sk_shutdown |= peer_mode; | |
2123 | unix_state_unlock(other); | |
2124 | other->sk_state_change(other); | |
2125 | if (peer_mode == SHUTDOWN_MASK) | |
2126 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP); | |
2127 | else if (peer_mode & RCV_SHUTDOWN) | |
2128 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN); | |
1da177e4 | 2129 | } |
7180a031 AC |
2130 | if (other) |
2131 | sock_put(other); | |
2132 | ||
1da177e4 LT |
2133 | return 0; |
2134 | } | |
2135 | ||
885ee74d PE |
2136 | long unix_inq_len(struct sock *sk) |
2137 | { | |
2138 | struct sk_buff *skb; | |
2139 | long amount = 0; | |
2140 | ||
2141 | if (sk->sk_state == TCP_LISTEN) | |
2142 | return -EINVAL; | |
2143 | ||
2144 | spin_lock(&sk->sk_receive_queue.lock); | |
2145 | if (sk->sk_type == SOCK_STREAM || | |
2146 | sk->sk_type == SOCK_SEQPACKET) { | |
2147 | skb_queue_walk(&sk->sk_receive_queue, skb) | |
2148 | amount += skb->len; | |
2149 | } else { | |
2150 | skb = skb_peek(&sk->sk_receive_queue); | |
2151 | if (skb) | |
2152 | amount = skb->len; | |
2153 | } | |
2154 | spin_unlock(&sk->sk_receive_queue.lock); | |
2155 | ||
2156 | return amount; | |
2157 | } | |
2158 | EXPORT_SYMBOL_GPL(unix_inq_len); | |
2159 | ||
2160 | long unix_outq_len(struct sock *sk) | |
2161 | { | |
2162 | return sk_wmem_alloc_get(sk); | |
2163 | } | |
2164 | EXPORT_SYMBOL_GPL(unix_outq_len); | |
2165 | ||
1da177e4 LT |
2166 | static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
2167 | { | |
2168 | struct sock *sk = sock->sk; | |
e27dfcea | 2169 | long amount = 0; |
1da177e4 LT |
2170 | int err; |
2171 | ||
6eba6a37 ED |
2172 | switch (cmd) { |
2173 | case SIOCOUTQ: | |
885ee74d | 2174 | amount = unix_outq_len(sk); |
6eba6a37 ED |
2175 | err = put_user(amount, (int __user *)arg); |
2176 | break; | |
2177 | case SIOCINQ: | |
885ee74d PE |
2178 | amount = unix_inq_len(sk); |
2179 | if (amount < 0) | |
2180 | err = amount; | |
2181 | else | |
1da177e4 | 2182 | err = put_user(amount, (int __user *)arg); |
885ee74d | 2183 | break; |
6eba6a37 ED |
2184 | default: |
2185 | err = -ENOIOCTLCMD; | |
2186 | break; | |
1da177e4 LT |
2187 | } |
2188 | return err; | |
2189 | } | |
2190 | ||
6eba6a37 | 2191 | static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait) |
1da177e4 LT |
2192 | { |
2193 | struct sock *sk = sock->sk; | |
2194 | unsigned int mask; | |
2195 | ||
aa395145 | 2196 | sock_poll_wait(file, sk_sleep(sk), wait); |
1da177e4 LT |
2197 | mask = 0; |
2198 | ||
2199 | /* exceptional events? */ | |
2200 | if (sk->sk_err) | |
2201 | mask |= POLLERR; | |
2202 | if (sk->sk_shutdown == SHUTDOWN_MASK) | |
2203 | mask |= POLLHUP; | |
f348d70a | 2204 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
db40980f | 2205 | mask |= POLLRDHUP | POLLIN | POLLRDNORM; |
1da177e4 LT |
2206 | |
2207 | /* readable? */ | |
db40980f | 2208 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
2209 | mask |= POLLIN | POLLRDNORM; |
2210 | ||
2211 | /* Connection-based need to check for termination and startup */ | |
6eba6a37 ED |
2212 | if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) && |
2213 | sk->sk_state == TCP_CLOSE) | |
1da177e4 LT |
2214 | mask |= POLLHUP; |
2215 | ||
2216 | /* | |
2217 | * we set writable also when the other side has shut down the | |
2218 | * connection. This prevents stuck sockets. | |
2219 | */ | |
2220 | if (unix_writable(sk)) | |
2221 | mask |= POLLOUT | POLLWRNORM | POLLWRBAND; | |
2222 | ||
2223 | return mask; | |
2224 | } | |
2225 | ||
ec0d215f RW |
2226 | static unsigned int unix_dgram_poll(struct file *file, struct socket *sock, |
2227 | poll_table *wait) | |
3c73419c | 2228 | { |
ec0d215f RW |
2229 | struct sock *sk = sock->sk, *other; |
2230 | unsigned int mask, writable; | |
3c73419c | 2231 | |
aa395145 | 2232 | sock_poll_wait(file, sk_sleep(sk), wait); |
3c73419c RW |
2233 | mask = 0; |
2234 | ||
2235 | /* exceptional events? */ | |
2236 | if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue)) | |
7d4c04fc | 2237 | mask |= POLLERR | |
8facd5fb | 2238 | (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0); |
7d4c04fc | 2239 | |
3c73419c | 2240 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
5456f09a | 2241 | mask |= POLLRDHUP | POLLIN | POLLRDNORM; |
3c73419c RW |
2242 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
2243 | mask |= POLLHUP; | |
2244 | ||
2245 | /* readable? */ | |
5456f09a | 2246 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
3c73419c RW |
2247 | mask |= POLLIN | POLLRDNORM; |
2248 | ||
2249 | /* Connection-based need to check for termination and startup */ | |
2250 | if (sk->sk_type == SOCK_SEQPACKET) { | |
2251 | if (sk->sk_state == TCP_CLOSE) | |
2252 | mask |= POLLHUP; | |
2253 | /* connection hasn't started yet? */ | |
2254 | if (sk->sk_state == TCP_SYN_SENT) | |
2255 | return mask; | |
2256 | } | |
2257 | ||
973a34aa | 2258 | /* No write status requested, avoid expensive OUT tests. */ |
626cf236 | 2259 | if (!(poll_requested_events(wait) & (POLLWRBAND|POLLWRNORM|POLLOUT))) |
973a34aa ED |
2260 | return mask; |
2261 | ||
ec0d215f | 2262 | writable = unix_writable(sk); |
5456f09a ED |
2263 | other = unix_peer_get(sk); |
2264 | if (other) { | |
2265 | if (unix_peer(other) != sk) { | |
2266 | sock_poll_wait(file, &unix_sk(other)->peer_wait, wait); | |
2267 | if (unix_recvq_full(other)) | |
2268 | writable = 0; | |
ec0d215f | 2269 | } |
5456f09a | 2270 | sock_put(other); |
ec0d215f RW |
2271 | } |
2272 | ||
2273 | if (writable) | |
3c73419c RW |
2274 | mask |= POLLOUT | POLLWRNORM | POLLWRBAND; |
2275 | else | |
2276 | set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); | |
2277 | ||
3c73419c RW |
2278 | return mask; |
2279 | } | |
1da177e4 LT |
2280 | |
2281 | #ifdef CONFIG_PROC_FS | |
a53eb3fe | 2282 | |
7123aaa3 ED |
2283 | #define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1) |
2284 | ||
2285 | #define get_bucket(x) ((x) >> BUCKET_SPACE) | |
2286 | #define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1)) | |
2287 | #define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o)) | |
a53eb3fe | 2288 | |
7123aaa3 | 2289 | static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos) |
1da177e4 | 2290 | { |
7123aaa3 ED |
2291 | unsigned long offset = get_offset(*pos); |
2292 | unsigned long bucket = get_bucket(*pos); | |
2293 | struct sock *sk; | |
2294 | unsigned long count = 0; | |
1da177e4 | 2295 | |
7123aaa3 ED |
2296 | for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) { |
2297 | if (sock_net(sk) != seq_file_net(seq)) | |
097e66c5 | 2298 | continue; |
7123aaa3 ED |
2299 | if (++count == offset) |
2300 | break; | |
2301 | } | |
2302 | ||
2303 | return sk; | |
2304 | } | |
2305 | ||
2306 | static struct sock *unix_next_socket(struct seq_file *seq, | |
2307 | struct sock *sk, | |
2308 | loff_t *pos) | |
2309 | { | |
2310 | unsigned long bucket; | |
2311 | ||
2312 | while (sk > (struct sock *)SEQ_START_TOKEN) { | |
2313 | sk = sk_next(sk); | |
2314 | if (!sk) | |
2315 | goto next_bucket; | |
2316 | if (sock_net(sk) == seq_file_net(seq)) | |
2317 | return sk; | |
1da177e4 | 2318 | } |
7123aaa3 ED |
2319 | |
2320 | do { | |
2321 | sk = unix_from_bucket(seq, pos); | |
2322 | if (sk) | |
2323 | return sk; | |
2324 | ||
2325 | next_bucket: | |
2326 | bucket = get_bucket(*pos) + 1; | |
2327 | *pos = set_bucket_offset(bucket, 1); | |
2328 | } while (bucket < ARRAY_SIZE(unix_socket_table)); | |
2329 | ||
1da177e4 LT |
2330 | return NULL; |
2331 | } | |
2332 | ||
1da177e4 | 2333 | static void *unix_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 2334 | __acquires(unix_table_lock) |
1da177e4 | 2335 | { |
fbe9cc4a | 2336 | spin_lock(&unix_table_lock); |
7123aaa3 ED |
2337 | |
2338 | if (!*pos) | |
2339 | return SEQ_START_TOKEN; | |
2340 | ||
2341 | if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table)) | |
2342 | return NULL; | |
2343 | ||
2344 | return unix_next_socket(seq, NULL, pos); | |
1da177e4 LT |
2345 | } |
2346 | ||
2347 | static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2348 | { | |
2349 | ++*pos; | |
7123aaa3 | 2350 | return unix_next_socket(seq, v, pos); |
1da177e4 LT |
2351 | } |
2352 | ||
2353 | static void unix_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2354 | __releases(unix_table_lock) |
1da177e4 | 2355 | { |
fbe9cc4a | 2356 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
2357 | } |
2358 | ||
2359 | static int unix_seq_show(struct seq_file *seq, void *v) | |
2360 | { | |
ac7bfa62 | 2361 | |
b9f3124f | 2362 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2363 | seq_puts(seq, "Num RefCount Protocol Flags Type St " |
2364 | "Inode Path\n"); | |
2365 | else { | |
2366 | struct sock *s = v; | |
2367 | struct unix_sock *u = unix_sk(s); | |
1c92b4e5 | 2368 | unix_state_lock(s); |
1da177e4 | 2369 | |
71338aa7 | 2370 | seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu", |
1da177e4 LT |
2371 | s, |
2372 | atomic_read(&s->sk_refcnt), | |
2373 | 0, | |
2374 | s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0, | |
2375 | s->sk_type, | |
2376 | s->sk_socket ? | |
2377 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) : | |
2378 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING), | |
2379 | sock_i_ino(s)); | |
2380 | ||
2381 | if (u->addr) { | |
2382 | int i, len; | |
2383 | seq_putc(seq, ' '); | |
2384 | ||
2385 | i = 0; | |
2386 | len = u->addr->len - sizeof(short); | |
2387 | if (!UNIX_ABSTRACT(s)) | |
2388 | len--; | |
2389 | else { | |
2390 | seq_putc(seq, '@'); | |
2391 | i++; | |
2392 | } | |
2393 | for ( ; i < len; i++) | |
2394 | seq_putc(seq, u->addr->name->sun_path[i]); | |
2395 | } | |
1c92b4e5 | 2396 | unix_state_unlock(s); |
1da177e4 LT |
2397 | seq_putc(seq, '\n'); |
2398 | } | |
2399 | ||
2400 | return 0; | |
2401 | } | |
2402 | ||
56b3d975 | 2403 | static const struct seq_operations unix_seq_ops = { |
1da177e4 LT |
2404 | .start = unix_seq_start, |
2405 | .next = unix_seq_next, | |
2406 | .stop = unix_seq_stop, | |
2407 | .show = unix_seq_show, | |
2408 | }; | |
2409 | ||
1da177e4 LT |
2410 | static int unix_seq_open(struct inode *inode, struct file *file) |
2411 | { | |
e372c414 | 2412 | return seq_open_net(inode, file, &unix_seq_ops, |
8b51b064 | 2413 | sizeof(struct seq_net_private)); |
1da177e4 LT |
2414 | } |
2415 | ||
da7071d7 | 2416 | static const struct file_operations unix_seq_fops = { |
1da177e4 LT |
2417 | .owner = THIS_MODULE, |
2418 | .open = unix_seq_open, | |
2419 | .read = seq_read, | |
2420 | .llseek = seq_lseek, | |
e372c414 | 2421 | .release = seq_release_net, |
1da177e4 LT |
2422 | }; |
2423 | ||
2424 | #endif | |
2425 | ||
ec1b4cf7 | 2426 | static const struct net_proto_family unix_family_ops = { |
1da177e4 LT |
2427 | .family = PF_UNIX, |
2428 | .create = unix_create, | |
2429 | .owner = THIS_MODULE, | |
2430 | }; | |
2431 | ||
097e66c5 | 2432 | |
2c8c1e72 | 2433 | static int __net_init unix_net_init(struct net *net) |
097e66c5 DL |
2434 | { |
2435 | int error = -ENOMEM; | |
2436 | ||
a0a53c8b | 2437 | net->unx.sysctl_max_dgram_qlen = 10; |
1597fbc0 PE |
2438 | if (unix_sysctl_register(net)) |
2439 | goto out; | |
d392e497 | 2440 | |
097e66c5 | 2441 | #ifdef CONFIG_PROC_FS |
d4beaa66 | 2442 | if (!proc_create("unix", 0, net->proc_net, &unix_seq_fops)) { |
1597fbc0 | 2443 | unix_sysctl_unregister(net); |
097e66c5 | 2444 | goto out; |
1597fbc0 | 2445 | } |
097e66c5 DL |
2446 | #endif |
2447 | error = 0; | |
2448 | out: | |
48dcc33e | 2449 | return error; |
097e66c5 DL |
2450 | } |
2451 | ||
2c8c1e72 | 2452 | static void __net_exit unix_net_exit(struct net *net) |
097e66c5 | 2453 | { |
1597fbc0 | 2454 | unix_sysctl_unregister(net); |
ece31ffd | 2455 | remove_proc_entry("unix", net->proc_net); |
097e66c5 DL |
2456 | } |
2457 | ||
2458 | static struct pernet_operations unix_net_ops = { | |
2459 | .init = unix_net_init, | |
2460 | .exit = unix_net_exit, | |
2461 | }; | |
2462 | ||
1da177e4 LT |
2463 | static int __init af_unix_init(void) |
2464 | { | |
2465 | int rc = -1; | |
1da177e4 | 2466 | |
b4fff5f8 | 2467 | BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 LT |
2468 | |
2469 | rc = proto_register(&unix_proto, 1); | |
ac7bfa62 YH |
2470 | if (rc != 0) { |
2471 | printk(KERN_CRIT "%s: Cannot create unix_sock SLAB cache!\n", | |
0dc47877 | 2472 | __func__); |
1da177e4 LT |
2473 | goto out; |
2474 | } | |
2475 | ||
2476 | sock_register(&unix_family_ops); | |
097e66c5 | 2477 | register_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2478 | out: |
2479 | return rc; | |
2480 | } | |
2481 | ||
2482 | static void __exit af_unix_exit(void) | |
2483 | { | |
2484 | sock_unregister(PF_UNIX); | |
1da177e4 | 2485 | proto_unregister(&unix_proto); |
097e66c5 | 2486 | unregister_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2487 | } |
2488 | ||
3d366960 DW |
2489 | /* Earlier than device_initcall() so that other drivers invoking |
2490 | request_module() don't end up in a loop when modprobe tries | |
2491 | to use a UNIX socket. But later than subsys_initcall() because | |
2492 | we depend on stuff initialised there */ | |
2493 | fs_initcall(af_unix_init); | |
1da177e4 LT |
2494 | module_exit(af_unix_exit); |
2495 | ||
2496 | MODULE_LICENSE("GPL"); | |
2497 | MODULE_ALIAS_NETPROTO(PF_UNIX); |