]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NET4: Implementation of BSD Unix domain sockets. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
5 | * |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License | |
8 | * as published by the Free Software Foundation; either version | |
9 | * 2 of the License, or (at your option) any later version. | |
10 | * | |
1da177e4 LT |
11 | * Fixes: |
12 | * Linus Torvalds : Assorted bug cures. | |
13 | * Niibe Yutaka : async I/O support. | |
14 | * Carsten Paeth : PF_UNIX check, address fixes. | |
15 | * Alan Cox : Limit size of allocated blocks. | |
16 | * Alan Cox : Fixed the stupid socketpair bug. | |
17 | * Alan Cox : BSD compatibility fine tuning. | |
18 | * Alan Cox : Fixed a bug in connect when interrupted. | |
19 | * Alan Cox : Sorted out a proper draft version of | |
20 | * file descriptor passing hacked up from | |
21 | * Mike Shaver's work. | |
22 | * Marty Leisner : Fixes to fd passing | |
23 | * Nick Nevin : recvmsg bugfix. | |
24 | * Alan Cox : Started proper garbage collector | |
25 | * Heiko EiBfeldt : Missing verify_area check | |
26 | * Alan Cox : Started POSIXisms | |
27 | * Andreas Schwab : Replace inode by dentry for proper | |
28 | * reference counting | |
29 | * Kirk Petersen : Made this a module | |
30 | * Christoph Rohland : Elegant non-blocking accept/connect algorithm. | |
31 | * Lots of bug fixes. | |
32 | * Alexey Kuznetosv : Repaired (I hope) bugs introduces | |
33 | * by above two patches. | |
34 | * Andrea Arcangeli : If possible we block in connect(2) | |
35 | * if the max backlog of the listen socket | |
36 | * is been reached. This won't break | |
37 | * old apps and it will avoid huge amount | |
38 | * of socks hashed (this for unix_gc() | |
39 | * performances reasons). | |
40 | * Security fix that limits the max | |
41 | * number of socks to 2*max_files and | |
42 | * the number of skb queueable in the | |
43 | * dgram receiver. | |
44 | * Artur Skawina : Hash function optimizations | |
45 | * Alexey Kuznetsov : Full scale SMP. Lot of bugs are introduced 8) | |
46 | * Malcolm Beattie : Set peercred for socketpair | |
47 | * Michal Ostrowski : Module initialization cleanup. | |
48 | * Arnaldo C. Melo : Remove MOD_{INC,DEC}_USE_COUNT, | |
49 | * the core infrastructure is doing that | |
50 | * for all net proto families now (2.5.69+) | |
51 | * | |
52 | * | |
53 | * Known differences from reference BSD that was tested: | |
54 | * | |
55 | * [TO FIX] | |
56 | * ECONNREFUSED is not returned from one end of a connected() socket to the | |
57 | * other the moment one end closes. | |
58 | * fstat() doesn't return st_dev=0, and give the blksize as high water mark | |
59 | * and a fake inode identifier (nor the BSD first socket fstat twice bug). | |
60 | * [NOT TO FIX] | |
61 | * accept() returns a path name even if the connecting socket has closed | |
62 | * in the meantime (BSD loses the path and gives up). | |
63 | * accept() returns 0 length path for an unbound connector. BSD returns 16 | |
64 | * and a null first byte in the path (but not for gethost/peername - BSD bug ??) | |
65 | * socketpair(...SOCK_RAW..) doesn't panic the kernel. | |
66 | * BSD af_unix apparently has connect forgetting to block properly. | |
67 | * (need to check this with the POSIX spec in detail) | |
68 | * | |
69 | * Differences from 2.0.0-11-... (ANK) | |
70 | * Bug fixes and improvements. | |
71 | * - client shutdown killed server socket. | |
72 | * - removed all useless cli/sti pairs. | |
73 | * | |
74 | * Semantic changes/extensions. | |
75 | * - generic control message passing. | |
76 | * - SCM_CREDENTIALS control message. | |
77 | * - "Abstract" (not FS based) socket bindings. | |
78 | * Abstract names are sequences of bytes (not zero terminated) | |
79 | * started by 0, so that this name space does not intersect | |
80 | * with BSD names. | |
81 | */ | |
82 | ||
5cc208be | 83 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
84 | ||
1da177e4 | 85 | #include <linux/module.h> |
1da177e4 | 86 | #include <linux/kernel.h> |
1da177e4 | 87 | #include <linux/signal.h> |
3f07c014 | 88 | #include <linux/sched/signal.h> |
1da177e4 LT |
89 | #include <linux/errno.h> |
90 | #include <linux/string.h> | |
91 | #include <linux/stat.h> | |
92 | #include <linux/dcache.h> | |
93 | #include <linux/namei.h> | |
94 | #include <linux/socket.h> | |
95 | #include <linux/un.h> | |
96 | #include <linux/fcntl.h> | |
97 | #include <linux/termios.h> | |
98 | #include <linux/sockios.h> | |
99 | #include <linux/net.h> | |
100 | #include <linux/in.h> | |
101 | #include <linux/fs.h> | |
102 | #include <linux/slab.h> | |
7c0f6ba6 | 103 | #include <linux/uaccess.h> |
1da177e4 LT |
104 | #include <linux/skbuff.h> |
105 | #include <linux/netdevice.h> | |
457c4cbc | 106 | #include <net/net_namespace.h> |
1da177e4 | 107 | #include <net/sock.h> |
c752f073 | 108 | #include <net/tcp_states.h> |
1da177e4 LT |
109 | #include <net/af_unix.h> |
110 | #include <linux/proc_fs.h> | |
111 | #include <linux/seq_file.h> | |
112 | #include <net/scm.h> | |
113 | #include <linux/init.h> | |
114 | #include <linux/poll.h> | |
1da177e4 LT |
115 | #include <linux/rtnetlink.h> |
116 | #include <linux/mount.h> | |
117 | #include <net/checksum.h> | |
118 | #include <linux/security.h> | |
2b15af6f | 119 | #include <linux/freezer.h> |
ba94f308 | 120 | #include <linux/file.h> |
1da177e4 | 121 | |
f4e65870 JA |
122 | #include "scm.h" |
123 | ||
7123aaa3 | 124 | struct hlist_head unix_socket_table[2 * UNIX_HASH_SIZE]; |
fa7ff56f PE |
125 | EXPORT_SYMBOL_GPL(unix_socket_table); |
126 | DEFINE_SPINLOCK(unix_table_lock); | |
127 | EXPORT_SYMBOL_GPL(unix_table_lock); | |
518de9b3 | 128 | static atomic_long_t unix_nr_socks; |
1da177e4 | 129 | |
1da177e4 | 130 | |
7123aaa3 ED |
131 | static struct hlist_head *unix_sockets_unbound(void *addr) |
132 | { | |
133 | unsigned long hash = (unsigned long)addr; | |
134 | ||
135 | hash ^= hash >> 16; | |
136 | hash ^= hash >> 8; | |
137 | hash %= UNIX_HASH_SIZE; | |
138 | return &unix_socket_table[UNIX_HASH_SIZE + hash]; | |
139 | } | |
140 | ||
141 | #define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash < UNIX_HASH_SIZE) | |
1da177e4 | 142 | |
877ce7c1 | 143 | #ifdef CONFIG_SECURITY_NETWORK |
dc49c1f9 | 144 | static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 | 145 | { |
37a9a8df | 146 | UNIXCB(skb).secid = scm->secid; |
877ce7c1 CZ |
147 | } |
148 | ||
149 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
150 | { | |
37a9a8df SS |
151 | scm->secid = UNIXCB(skb).secid; |
152 | } | |
153 | ||
154 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
155 | { | |
156 | return (scm->secid == UNIXCB(skb).secid); | |
877ce7c1 CZ |
157 | } |
158 | #else | |
dc49c1f9 | 159 | static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 CZ |
160 | { } |
161 | ||
162 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
163 | { } | |
37a9a8df SS |
164 | |
165 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
166 | { | |
167 | return true; | |
168 | } | |
877ce7c1 CZ |
169 | #endif /* CONFIG_SECURITY_NETWORK */ |
170 | ||
1da177e4 LT |
171 | /* |
172 | * SMP locking strategy: | |
fbe9cc4a | 173 | * hash table is protected with spinlock unix_table_lock |
663717f6 | 174 | * each socket state is protected by separate spin lock. |
1da177e4 LT |
175 | */ |
176 | ||
95c96174 | 177 | static inline unsigned int unix_hash_fold(__wsum n) |
1da177e4 | 178 | { |
0a13404d | 179 | unsigned int hash = (__force unsigned int)csum_fold(n); |
95c96174 | 180 | |
1da177e4 LT |
181 | hash ^= hash>>8; |
182 | return hash&(UNIX_HASH_SIZE-1); | |
183 | } | |
184 | ||
185 | #define unix_peer(sk) (unix_sk(sk)->peer) | |
186 | ||
187 | static inline int unix_our_peer(struct sock *sk, struct sock *osk) | |
188 | { | |
189 | return unix_peer(osk) == sk; | |
190 | } | |
191 | ||
192 | static inline int unix_may_send(struct sock *sk, struct sock *osk) | |
193 | { | |
6eba6a37 | 194 | return unix_peer(osk) == NULL || unix_our_peer(sk, osk); |
1da177e4 LT |
195 | } |
196 | ||
3c73419c RW |
197 | static inline int unix_recvq_full(struct sock const *sk) |
198 | { | |
199 | return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog; | |
200 | } | |
201 | ||
fa7ff56f | 202 | struct sock *unix_peer_get(struct sock *s) |
1da177e4 LT |
203 | { |
204 | struct sock *peer; | |
205 | ||
1c92b4e5 | 206 | unix_state_lock(s); |
1da177e4 LT |
207 | peer = unix_peer(s); |
208 | if (peer) | |
209 | sock_hold(peer); | |
1c92b4e5 | 210 | unix_state_unlock(s); |
1da177e4 LT |
211 | return peer; |
212 | } | |
fa7ff56f | 213 | EXPORT_SYMBOL_GPL(unix_peer_get); |
1da177e4 LT |
214 | |
215 | static inline void unix_release_addr(struct unix_address *addr) | |
216 | { | |
8c9814b9 | 217 | if (refcount_dec_and_test(&addr->refcnt)) |
1da177e4 LT |
218 | kfree(addr); |
219 | } | |
220 | ||
221 | /* | |
222 | * Check unix socket name: | |
223 | * - should be not zero length. | |
224 | * - if started by not zero, should be NULL terminated (FS object) | |
225 | * - if started by zero, it is abstract name. | |
226 | */ | |
ac7bfa62 | 227 | |
95c96174 | 228 | static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp) |
1da177e4 | 229 | { |
33c4368e KK |
230 | *hashp = 0; |
231 | ||
1da177e4 LT |
232 | if (len <= sizeof(short) || len > sizeof(*sunaddr)) |
233 | return -EINVAL; | |
234 | if (!sunaddr || sunaddr->sun_family != AF_UNIX) | |
235 | return -EINVAL; | |
236 | if (sunaddr->sun_path[0]) { | |
237 | /* | |
238 | * This may look like an off by one error but it is a bit more | |
239 | * subtle. 108 is the longest valid AF_UNIX path for a binding. | |
25985edc | 240 | * sun_path[108] doesn't as such exist. However in kernel space |
1da177e4 LT |
241 | * we are guaranteed that it is a valid memory location in our |
242 | * kernel address buffer. | |
243 | */ | |
e27dfcea | 244 | ((char *)sunaddr)[len] = 0; |
1da177e4 LT |
245 | len = strlen(sunaddr->sun_path)+1+sizeof(short); |
246 | return len; | |
247 | } | |
248 | ||
07f0757a | 249 | *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0)); |
1da177e4 LT |
250 | return len; |
251 | } | |
252 | ||
253 | static void __unix_remove_socket(struct sock *sk) | |
254 | { | |
255 | sk_del_node_init(sk); | |
256 | } | |
257 | ||
258 | static void __unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
259 | { | |
547b792c | 260 | WARN_ON(!sk_unhashed(sk)); |
1da177e4 LT |
261 | sk_add_node(sk, list); |
262 | } | |
263 | ||
264 | static inline void unix_remove_socket(struct sock *sk) | |
265 | { | |
fbe9cc4a | 266 | spin_lock(&unix_table_lock); |
1da177e4 | 267 | __unix_remove_socket(sk); |
fbe9cc4a | 268 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
269 | } |
270 | ||
271 | static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
272 | { | |
fbe9cc4a | 273 | spin_lock(&unix_table_lock); |
1da177e4 | 274 | __unix_insert_socket(list, sk); |
fbe9cc4a | 275 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
276 | } |
277 | ||
097e66c5 DL |
278 | static struct sock *__unix_find_socket_byname(struct net *net, |
279 | struct sockaddr_un *sunname, | |
95c96174 | 280 | int len, int type, unsigned int hash) |
1da177e4 LT |
281 | { |
282 | struct sock *s; | |
1da177e4 | 283 | |
b67bfe0d | 284 | sk_for_each(s, &unix_socket_table[hash ^ type]) { |
1da177e4 LT |
285 | struct unix_sock *u = unix_sk(s); |
286 | ||
878628fb | 287 | if (!net_eq(sock_net(s), net)) |
097e66c5 DL |
288 | continue; |
289 | ||
1da177e4 LT |
290 | if (u->addr->len == len && |
291 | !memcmp(u->addr->name, sunname, len)) | |
292 | goto found; | |
293 | } | |
294 | s = NULL; | |
295 | found: | |
296 | return s; | |
297 | } | |
298 | ||
097e66c5 DL |
299 | static inline struct sock *unix_find_socket_byname(struct net *net, |
300 | struct sockaddr_un *sunname, | |
1da177e4 | 301 | int len, int type, |
95c96174 | 302 | unsigned int hash) |
1da177e4 LT |
303 | { |
304 | struct sock *s; | |
305 | ||
fbe9cc4a | 306 | spin_lock(&unix_table_lock); |
097e66c5 | 307 | s = __unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
308 | if (s) |
309 | sock_hold(s); | |
fbe9cc4a | 310 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
311 | return s; |
312 | } | |
313 | ||
6616f788 | 314 | static struct sock *unix_find_socket_byinode(struct inode *i) |
1da177e4 LT |
315 | { |
316 | struct sock *s; | |
1da177e4 | 317 | |
fbe9cc4a | 318 | spin_lock(&unix_table_lock); |
b67bfe0d | 319 | sk_for_each(s, |
1da177e4 | 320 | &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) { |
40ffe67d | 321 | struct dentry *dentry = unix_sk(s)->path.dentry; |
1da177e4 | 322 | |
beef5121 | 323 | if (dentry && d_backing_inode(dentry) == i) { |
1da177e4 LT |
324 | sock_hold(s); |
325 | goto found; | |
326 | } | |
327 | } | |
328 | s = NULL; | |
329 | found: | |
fbe9cc4a | 330 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
331 | return s; |
332 | } | |
333 | ||
7d267278 RW |
334 | /* Support code for asymmetrically connected dgram sockets |
335 | * | |
336 | * If a datagram socket is connected to a socket not itself connected | |
337 | * to the first socket (eg, /dev/log), clients may only enqueue more | |
338 | * messages if the present receive queue of the server socket is not | |
339 | * "too large". This means there's a second writeability condition | |
340 | * poll and sendmsg need to test. The dgram recv code will do a wake | |
341 | * up on the peer_wait wait queue of a socket upon reception of a | |
342 | * datagram which needs to be propagated to sleeping would-be writers | |
343 | * since these might not have sent anything so far. This can't be | |
344 | * accomplished via poll_wait because the lifetime of the server | |
345 | * socket might be less than that of its clients if these break their | |
346 | * association with it or if the server socket is closed while clients | |
347 | * are still connected to it and there's no way to inform "a polling | |
348 | * implementation" that it should let go of a certain wait queue | |
349 | * | |
ac6424b9 | 350 | * In order to propagate a wake up, a wait_queue_entry_t of the client |
7d267278 RW |
351 | * socket is enqueued on the peer_wait queue of the server socket |
352 | * whose wake function does a wake_up on the ordinary client socket | |
353 | * wait queue. This connection is established whenever a write (or | |
354 | * poll for write) hit the flow control condition and broken when the | |
355 | * association to the server socket is dissolved or after a wake up | |
356 | * was relayed. | |
357 | */ | |
358 | ||
ac6424b9 | 359 | static int unix_dgram_peer_wake_relay(wait_queue_entry_t *q, unsigned mode, int flags, |
7d267278 RW |
360 | void *key) |
361 | { | |
362 | struct unix_sock *u; | |
363 | wait_queue_head_t *u_sleep; | |
364 | ||
365 | u = container_of(q, struct unix_sock, peer_wake); | |
366 | ||
367 | __remove_wait_queue(&unix_sk(u->peer_wake.private)->peer_wait, | |
368 | q); | |
369 | u->peer_wake.private = NULL; | |
370 | ||
371 | /* relaying can only happen while the wq still exists */ | |
372 | u_sleep = sk_sleep(&u->sk); | |
373 | if (u_sleep) | |
3ad6f93e | 374 | wake_up_interruptible_poll(u_sleep, key_to_poll(key)); |
7d267278 RW |
375 | |
376 | return 0; | |
377 | } | |
378 | ||
379 | static int unix_dgram_peer_wake_connect(struct sock *sk, struct sock *other) | |
380 | { | |
381 | struct unix_sock *u, *u_other; | |
382 | int rc; | |
383 | ||
384 | u = unix_sk(sk); | |
385 | u_other = unix_sk(other); | |
386 | rc = 0; | |
387 | spin_lock(&u_other->peer_wait.lock); | |
388 | ||
389 | if (!u->peer_wake.private) { | |
390 | u->peer_wake.private = other; | |
391 | __add_wait_queue(&u_other->peer_wait, &u->peer_wake); | |
392 | ||
393 | rc = 1; | |
394 | } | |
395 | ||
396 | spin_unlock(&u_other->peer_wait.lock); | |
397 | return rc; | |
398 | } | |
399 | ||
400 | static void unix_dgram_peer_wake_disconnect(struct sock *sk, | |
401 | struct sock *other) | |
402 | { | |
403 | struct unix_sock *u, *u_other; | |
404 | ||
405 | u = unix_sk(sk); | |
406 | u_other = unix_sk(other); | |
407 | spin_lock(&u_other->peer_wait.lock); | |
408 | ||
409 | if (u->peer_wake.private == other) { | |
410 | __remove_wait_queue(&u_other->peer_wait, &u->peer_wake); | |
411 | u->peer_wake.private = NULL; | |
412 | } | |
413 | ||
414 | spin_unlock(&u_other->peer_wait.lock); | |
415 | } | |
416 | ||
417 | static void unix_dgram_peer_wake_disconnect_wakeup(struct sock *sk, | |
418 | struct sock *other) | |
419 | { | |
420 | unix_dgram_peer_wake_disconnect(sk, other); | |
421 | wake_up_interruptible_poll(sk_sleep(sk), | |
a9a08845 LT |
422 | EPOLLOUT | |
423 | EPOLLWRNORM | | |
424 | EPOLLWRBAND); | |
7d267278 RW |
425 | } |
426 | ||
427 | /* preconditions: | |
428 | * - unix_peer(sk) == other | |
429 | * - association is stable | |
430 | */ | |
431 | static int unix_dgram_peer_wake_me(struct sock *sk, struct sock *other) | |
432 | { | |
433 | int connected; | |
434 | ||
435 | connected = unix_dgram_peer_wake_connect(sk, other); | |
436 | ||
51f7e951 JB |
437 | /* If other is SOCK_DEAD, we want to make sure we signal |
438 | * POLLOUT, such that a subsequent write() can get a | |
439 | * -ECONNREFUSED. Otherwise, if we haven't queued any skbs | |
440 | * to other and its full, we will hang waiting for POLLOUT. | |
441 | */ | |
442 | if (unix_recvq_full(other) && !sock_flag(other, SOCK_DEAD)) | |
7d267278 RW |
443 | return 1; |
444 | ||
445 | if (connected) | |
446 | unix_dgram_peer_wake_disconnect(sk, other); | |
447 | ||
448 | return 0; | |
449 | } | |
450 | ||
1586a587 | 451 | static int unix_writable(const struct sock *sk) |
1da177e4 | 452 | { |
1586a587 | 453 | return sk->sk_state != TCP_LISTEN && |
14afee4b | 454 | (refcount_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf; |
1da177e4 LT |
455 | } |
456 | ||
457 | static void unix_write_space(struct sock *sk) | |
458 | { | |
43815482 ED |
459 | struct socket_wq *wq; |
460 | ||
461 | rcu_read_lock(); | |
1da177e4 | 462 | if (unix_writable(sk)) { |
43815482 | 463 | wq = rcu_dereference(sk->sk_wq); |
1ce0bf50 | 464 | if (skwq_has_sleeper(wq)) |
67426b75 | 465 | wake_up_interruptible_sync_poll(&wq->wait, |
a9a08845 | 466 | EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND); |
8d8ad9d7 | 467 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
1da177e4 | 468 | } |
43815482 | 469 | rcu_read_unlock(); |
1da177e4 LT |
470 | } |
471 | ||
472 | /* When dgram socket disconnects (or changes its peer), we clear its receive | |
473 | * queue of packets arrived from previous peer. First, it allows to do | |
474 | * flow control based only on wmem_alloc; second, sk connected to peer | |
475 | * may receive messages only from that peer. */ | |
476 | static void unix_dgram_disconnected(struct sock *sk, struct sock *other) | |
477 | { | |
b03efcfb | 478 | if (!skb_queue_empty(&sk->sk_receive_queue)) { |
1da177e4 LT |
479 | skb_queue_purge(&sk->sk_receive_queue); |
480 | wake_up_interruptible_all(&unix_sk(sk)->peer_wait); | |
481 | ||
482 | /* If one link of bidirectional dgram pipe is disconnected, | |
483 | * we signal error. Messages are lost. Do not make this, | |
484 | * when peer was not connected to us. | |
485 | */ | |
486 | if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) { | |
487 | other->sk_err = ECONNRESET; | |
488 | other->sk_error_report(other); | |
489 | } | |
490 | } | |
491 | } | |
492 | ||
493 | static void unix_sock_destructor(struct sock *sk) | |
494 | { | |
495 | struct unix_sock *u = unix_sk(sk); | |
496 | ||
497 | skb_queue_purge(&sk->sk_receive_queue); | |
498 | ||
14afee4b | 499 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
547b792c IJ |
500 | WARN_ON(!sk_unhashed(sk)); |
501 | WARN_ON(sk->sk_socket); | |
1da177e4 | 502 | if (!sock_flag(sk, SOCK_DEAD)) { |
5cc208be | 503 | pr_info("Attempt to release alive unix socket: %p\n", sk); |
1da177e4 LT |
504 | return; |
505 | } | |
506 | ||
507 | if (u->addr) | |
508 | unix_release_addr(u->addr); | |
509 | ||
518de9b3 | 510 | atomic_long_dec(&unix_nr_socks); |
6f756a8c | 511 | local_bh_disable(); |
a8076d8d | 512 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
6f756a8c | 513 | local_bh_enable(); |
1da177e4 | 514 | #ifdef UNIX_REFCNT_DEBUG |
5cc208be | 515 | pr_debug("UNIX %p is destroyed, %ld are still alive.\n", sk, |
518de9b3 | 516 | atomic_long_read(&unix_nr_socks)); |
1da177e4 LT |
517 | #endif |
518 | } | |
519 | ||
ded34e0f | 520 | static void unix_release_sock(struct sock *sk, int embrion) |
1da177e4 LT |
521 | { |
522 | struct unix_sock *u = unix_sk(sk); | |
40ffe67d | 523 | struct path path; |
1da177e4 LT |
524 | struct sock *skpair; |
525 | struct sk_buff *skb; | |
526 | int state; | |
527 | ||
528 | unix_remove_socket(sk); | |
529 | ||
530 | /* Clear state */ | |
1c92b4e5 | 531 | unix_state_lock(sk); |
1da177e4 LT |
532 | sock_orphan(sk); |
533 | sk->sk_shutdown = SHUTDOWN_MASK; | |
40ffe67d AV |
534 | path = u->path; |
535 | u->path.dentry = NULL; | |
536 | u->path.mnt = NULL; | |
1da177e4 LT |
537 | state = sk->sk_state; |
538 | sk->sk_state = TCP_CLOSE; | |
1c92b4e5 | 539 | unix_state_unlock(sk); |
1da177e4 LT |
540 | |
541 | wake_up_interruptible_all(&u->peer_wait); | |
542 | ||
e27dfcea | 543 | skpair = unix_peer(sk); |
1da177e4 | 544 | |
e27dfcea | 545 | if (skpair != NULL) { |
1da177e4 | 546 | if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) { |
1c92b4e5 | 547 | unix_state_lock(skpair); |
1da177e4 LT |
548 | /* No more writes */ |
549 | skpair->sk_shutdown = SHUTDOWN_MASK; | |
550 | if (!skb_queue_empty(&sk->sk_receive_queue) || embrion) | |
551 | skpair->sk_err = ECONNRESET; | |
1c92b4e5 | 552 | unix_state_unlock(skpair); |
1da177e4 | 553 | skpair->sk_state_change(skpair); |
8d8ad9d7 | 554 | sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP); |
1da177e4 | 555 | } |
7d267278 RW |
556 | |
557 | unix_dgram_peer_wake_disconnect(sk, skpair); | |
1da177e4 LT |
558 | sock_put(skpair); /* It may now die */ |
559 | unix_peer(sk) = NULL; | |
560 | } | |
561 | ||
562 | /* Try to flush out this socket. Throw out buffers at least */ | |
563 | ||
564 | while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) { | |
e27dfcea | 565 | if (state == TCP_LISTEN) |
1da177e4 LT |
566 | unix_release_sock(skb->sk, 1); |
567 | /* passed fds are erased in the kfree_skb hook */ | |
73ed5d25 | 568 | UNIXCB(skb).consumed = skb->len; |
1da177e4 LT |
569 | kfree_skb(skb); |
570 | } | |
571 | ||
40ffe67d AV |
572 | if (path.dentry) |
573 | path_put(&path); | |
1da177e4 LT |
574 | |
575 | sock_put(sk); | |
576 | ||
577 | /* ---- Socket is dead now and most probably destroyed ---- */ | |
578 | ||
579 | /* | |
e04dae84 | 580 | * Fixme: BSD difference: In BSD all sockets connected to us get |
1da177e4 LT |
581 | * ECONNRESET and we die on the spot. In Linux we behave |
582 | * like files and pipes do and wait for the last | |
583 | * dereference. | |
584 | * | |
585 | * Can't we simply set sock->err? | |
586 | * | |
587 | * What the above comment does talk about? --ANK(980817) | |
588 | */ | |
589 | ||
9305cfa4 | 590 | if (unix_tot_inflight) |
ac7bfa62 | 591 | unix_gc(); /* Garbage collect fds */ |
1da177e4 LT |
592 | } |
593 | ||
109f6e39 EB |
594 | static void init_peercred(struct sock *sk) |
595 | { | |
596 | put_pid(sk->sk_peer_pid); | |
597 | if (sk->sk_peer_cred) | |
598 | put_cred(sk->sk_peer_cred); | |
599 | sk->sk_peer_pid = get_pid(task_tgid(current)); | |
600 | sk->sk_peer_cred = get_current_cred(); | |
601 | } | |
602 | ||
603 | static void copy_peercred(struct sock *sk, struct sock *peersk) | |
604 | { | |
605 | put_pid(sk->sk_peer_pid); | |
606 | if (sk->sk_peer_cred) | |
607 | put_cred(sk->sk_peer_cred); | |
608 | sk->sk_peer_pid = get_pid(peersk->sk_peer_pid); | |
609 | sk->sk_peer_cred = get_cred(peersk->sk_peer_cred); | |
610 | } | |
611 | ||
1da177e4 LT |
612 | static int unix_listen(struct socket *sock, int backlog) |
613 | { | |
614 | int err; | |
615 | struct sock *sk = sock->sk; | |
616 | struct unix_sock *u = unix_sk(sk); | |
109f6e39 | 617 | struct pid *old_pid = NULL; |
1da177e4 LT |
618 | |
619 | err = -EOPNOTSUPP; | |
6eba6a37 ED |
620 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
621 | goto out; /* Only stream/seqpacket sockets accept */ | |
1da177e4 LT |
622 | err = -EINVAL; |
623 | if (!u->addr) | |
6eba6a37 | 624 | goto out; /* No listens on an unbound socket */ |
1c92b4e5 | 625 | unix_state_lock(sk); |
1da177e4 LT |
626 | if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN) |
627 | goto out_unlock; | |
628 | if (backlog > sk->sk_max_ack_backlog) | |
629 | wake_up_interruptible_all(&u->peer_wait); | |
630 | sk->sk_max_ack_backlog = backlog; | |
631 | sk->sk_state = TCP_LISTEN; | |
632 | /* set credentials so connect can copy them */ | |
109f6e39 | 633 | init_peercred(sk); |
1da177e4 LT |
634 | err = 0; |
635 | ||
636 | out_unlock: | |
1c92b4e5 | 637 | unix_state_unlock(sk); |
109f6e39 | 638 | put_pid(old_pid); |
1da177e4 LT |
639 | out: |
640 | return err; | |
641 | } | |
642 | ||
643 | static int unix_release(struct socket *); | |
644 | static int unix_bind(struct socket *, struct sockaddr *, int); | |
645 | static int unix_stream_connect(struct socket *, struct sockaddr *, | |
646 | int addr_len, int flags); | |
647 | static int unix_socketpair(struct socket *, struct socket *); | |
cdfbabfb | 648 | static int unix_accept(struct socket *, struct socket *, int, bool); |
9b2c45d4 | 649 | static int unix_getname(struct socket *, struct sockaddr *, int); |
a11e1d43 LT |
650 | static __poll_t unix_poll(struct file *, struct socket *, poll_table *); |
651 | static __poll_t unix_dgram_poll(struct file *, struct socket *, | |
652 | poll_table *); | |
1da177e4 LT |
653 | static int unix_ioctl(struct socket *, unsigned int, unsigned long); |
654 | static int unix_shutdown(struct socket *, int); | |
1b784140 YX |
655 | static int unix_stream_sendmsg(struct socket *, struct msghdr *, size_t); |
656 | static int unix_stream_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
869e7c62 HFS |
657 | static ssize_t unix_stream_sendpage(struct socket *, struct page *, int offset, |
658 | size_t size, int flags); | |
2b514574 HFS |
659 | static ssize_t unix_stream_splice_read(struct socket *, loff_t *ppos, |
660 | struct pipe_inode_info *, size_t size, | |
661 | unsigned int flags); | |
1b784140 YX |
662 | static int unix_dgram_sendmsg(struct socket *, struct msghdr *, size_t); |
663 | static int unix_dgram_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
1da177e4 LT |
664 | static int unix_dgram_connect(struct socket *, struct sockaddr *, |
665 | int, int); | |
1b784140 YX |
666 | static int unix_seqpacket_sendmsg(struct socket *, struct msghdr *, size_t); |
667 | static int unix_seqpacket_recvmsg(struct socket *, struct msghdr *, size_t, | |
668 | int); | |
1da177e4 | 669 | |
12663bfc | 670 | static int unix_set_peek_off(struct sock *sk, int val) |
f55bb7f9 PE |
671 | { |
672 | struct unix_sock *u = unix_sk(sk); | |
673 | ||
6e1ce3c3 | 674 | if (mutex_lock_interruptible(&u->iolock)) |
12663bfc SL |
675 | return -EINTR; |
676 | ||
f55bb7f9 | 677 | sk->sk_peek_off = val; |
6e1ce3c3 | 678 | mutex_unlock(&u->iolock); |
12663bfc SL |
679 | |
680 | return 0; | |
f55bb7f9 PE |
681 | } |
682 | ||
683 | ||
90ddc4f0 | 684 | static const struct proto_ops unix_stream_ops = { |
1da177e4 LT |
685 | .family = PF_UNIX, |
686 | .owner = THIS_MODULE, | |
687 | .release = unix_release, | |
688 | .bind = unix_bind, | |
689 | .connect = unix_stream_connect, | |
690 | .socketpair = unix_socketpair, | |
691 | .accept = unix_accept, | |
692 | .getname = unix_getname, | |
a11e1d43 | 693 | .poll = unix_poll, |
1da177e4 LT |
694 | .ioctl = unix_ioctl, |
695 | .listen = unix_listen, | |
696 | .shutdown = unix_shutdown, | |
697 | .setsockopt = sock_no_setsockopt, | |
698 | .getsockopt = sock_no_getsockopt, | |
699 | .sendmsg = unix_stream_sendmsg, | |
700 | .recvmsg = unix_stream_recvmsg, | |
701 | .mmap = sock_no_mmap, | |
869e7c62 | 702 | .sendpage = unix_stream_sendpage, |
2b514574 | 703 | .splice_read = unix_stream_splice_read, |
fc0d7536 | 704 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
705 | }; |
706 | ||
90ddc4f0 | 707 | static const struct proto_ops unix_dgram_ops = { |
1da177e4 LT |
708 | .family = PF_UNIX, |
709 | .owner = THIS_MODULE, | |
710 | .release = unix_release, | |
711 | .bind = unix_bind, | |
712 | .connect = unix_dgram_connect, | |
713 | .socketpair = unix_socketpair, | |
714 | .accept = sock_no_accept, | |
715 | .getname = unix_getname, | |
a11e1d43 | 716 | .poll = unix_dgram_poll, |
1da177e4 LT |
717 | .ioctl = unix_ioctl, |
718 | .listen = sock_no_listen, | |
719 | .shutdown = unix_shutdown, | |
720 | .setsockopt = sock_no_setsockopt, | |
721 | .getsockopt = sock_no_getsockopt, | |
722 | .sendmsg = unix_dgram_sendmsg, | |
723 | .recvmsg = unix_dgram_recvmsg, | |
724 | .mmap = sock_no_mmap, | |
725 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 726 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
727 | }; |
728 | ||
90ddc4f0 | 729 | static const struct proto_ops unix_seqpacket_ops = { |
1da177e4 LT |
730 | .family = PF_UNIX, |
731 | .owner = THIS_MODULE, | |
732 | .release = unix_release, | |
733 | .bind = unix_bind, | |
734 | .connect = unix_stream_connect, | |
735 | .socketpair = unix_socketpair, | |
736 | .accept = unix_accept, | |
737 | .getname = unix_getname, | |
a11e1d43 | 738 | .poll = unix_dgram_poll, |
1da177e4 LT |
739 | .ioctl = unix_ioctl, |
740 | .listen = unix_listen, | |
741 | .shutdown = unix_shutdown, | |
742 | .setsockopt = sock_no_setsockopt, | |
743 | .getsockopt = sock_no_getsockopt, | |
744 | .sendmsg = unix_seqpacket_sendmsg, | |
a05d2ad1 | 745 | .recvmsg = unix_seqpacket_recvmsg, |
1da177e4 LT |
746 | .mmap = sock_no_mmap, |
747 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 748 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
749 | }; |
750 | ||
751 | static struct proto unix_proto = { | |
248969ae ED |
752 | .name = "UNIX", |
753 | .owner = THIS_MODULE, | |
248969ae | 754 | .obj_size = sizeof(struct unix_sock), |
1da177e4 LT |
755 | }; |
756 | ||
11aa9c28 | 757 | static struct sock *unix_create1(struct net *net, struct socket *sock, int kern) |
1da177e4 LT |
758 | { |
759 | struct sock *sk = NULL; | |
760 | struct unix_sock *u; | |
761 | ||
518de9b3 ED |
762 | atomic_long_inc(&unix_nr_socks); |
763 | if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files()) | |
1da177e4 LT |
764 | goto out; |
765 | ||
11aa9c28 | 766 | sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto, kern); |
1da177e4 LT |
767 | if (!sk) |
768 | goto out; | |
769 | ||
6eba6a37 | 770 | sock_init_data(sock, sk); |
1da177e4 | 771 | |
3aa9799e | 772 | sk->sk_allocation = GFP_KERNEL_ACCOUNT; |
1da177e4 | 773 | sk->sk_write_space = unix_write_space; |
a0a53c8b | 774 | sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen; |
1da177e4 LT |
775 | sk->sk_destruct = unix_sock_destructor; |
776 | u = unix_sk(sk); | |
40ffe67d AV |
777 | u->path.dentry = NULL; |
778 | u->path.mnt = NULL; | |
fd19f329 | 779 | spin_lock_init(&u->lock); |
516e0cc5 | 780 | atomic_long_set(&u->inflight, 0); |
1fd05ba5 | 781 | INIT_LIST_HEAD(&u->link); |
6e1ce3c3 LT |
782 | mutex_init(&u->iolock); /* single task reading lock */ |
783 | mutex_init(&u->bindlock); /* single task binding lock */ | |
1da177e4 | 784 | init_waitqueue_head(&u->peer_wait); |
7d267278 | 785 | init_waitqueue_func_entry(&u->peer_wake, unix_dgram_peer_wake_relay); |
7123aaa3 | 786 | unix_insert_socket(unix_sockets_unbound(sk), sk); |
1da177e4 | 787 | out: |
284b327b | 788 | if (sk == NULL) |
518de9b3 | 789 | atomic_long_dec(&unix_nr_socks); |
920de804 ED |
790 | else { |
791 | local_bh_disable(); | |
a8076d8d | 792 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
920de804 ED |
793 | local_bh_enable(); |
794 | } | |
1da177e4 LT |
795 | return sk; |
796 | } | |
797 | ||
3f378b68 EP |
798 | static int unix_create(struct net *net, struct socket *sock, int protocol, |
799 | int kern) | |
1da177e4 LT |
800 | { |
801 | if (protocol && protocol != PF_UNIX) | |
802 | return -EPROTONOSUPPORT; | |
803 | ||
804 | sock->state = SS_UNCONNECTED; | |
805 | ||
806 | switch (sock->type) { | |
807 | case SOCK_STREAM: | |
808 | sock->ops = &unix_stream_ops; | |
809 | break; | |
810 | /* | |
811 | * Believe it or not BSD has AF_UNIX, SOCK_RAW though | |
812 | * nothing uses it. | |
813 | */ | |
814 | case SOCK_RAW: | |
e27dfcea | 815 | sock->type = SOCK_DGRAM; |
110af3ac | 816 | /* fall through */ |
1da177e4 LT |
817 | case SOCK_DGRAM: |
818 | sock->ops = &unix_dgram_ops; | |
819 | break; | |
820 | case SOCK_SEQPACKET: | |
821 | sock->ops = &unix_seqpacket_ops; | |
822 | break; | |
823 | default: | |
824 | return -ESOCKTNOSUPPORT; | |
825 | } | |
826 | ||
11aa9c28 | 827 | return unix_create1(net, sock, kern) ? 0 : -ENOMEM; |
1da177e4 LT |
828 | } |
829 | ||
830 | static int unix_release(struct socket *sock) | |
831 | { | |
832 | struct sock *sk = sock->sk; | |
833 | ||
834 | if (!sk) | |
835 | return 0; | |
836 | ||
ded34e0f | 837 | unix_release_sock(sk, 0); |
1da177e4 LT |
838 | sock->sk = NULL; |
839 | ||
ded34e0f | 840 | return 0; |
1da177e4 LT |
841 | } |
842 | ||
843 | static int unix_autobind(struct socket *sock) | |
844 | { | |
845 | struct sock *sk = sock->sk; | |
3b1e0a65 | 846 | struct net *net = sock_net(sk); |
1da177e4 LT |
847 | struct unix_sock *u = unix_sk(sk); |
848 | static u32 ordernum = 1; | |
6eba6a37 | 849 | struct unix_address *addr; |
1da177e4 | 850 | int err; |
8df73ff9 | 851 | unsigned int retries = 0; |
1da177e4 | 852 | |
6e1ce3c3 | 853 | err = mutex_lock_interruptible(&u->bindlock); |
37ab4fa7 SL |
854 | if (err) |
855 | return err; | |
1da177e4 LT |
856 | |
857 | err = 0; | |
858 | if (u->addr) | |
859 | goto out; | |
860 | ||
861 | err = -ENOMEM; | |
0da974f4 | 862 | addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL); |
1da177e4 LT |
863 | if (!addr) |
864 | goto out; | |
865 | ||
1da177e4 | 866 | addr->name->sun_family = AF_UNIX; |
8c9814b9 | 867 | refcount_set(&addr->refcnt, 1); |
1da177e4 LT |
868 | |
869 | retry: | |
870 | addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short); | |
07f0757a | 871 | addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0)); |
1da177e4 | 872 | |
fbe9cc4a | 873 | spin_lock(&unix_table_lock); |
1da177e4 LT |
874 | ordernum = (ordernum+1)&0xFFFFF; |
875 | ||
097e66c5 | 876 | if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type, |
1da177e4 | 877 | addr->hash)) { |
fbe9cc4a | 878 | spin_unlock(&unix_table_lock); |
8df73ff9 TH |
879 | /* |
880 | * __unix_find_socket_byname() may take long time if many names | |
881 | * are already in use. | |
882 | */ | |
883 | cond_resched(); | |
884 | /* Give up if all names seems to be in use. */ | |
885 | if (retries++ == 0xFFFFF) { | |
886 | err = -ENOSPC; | |
887 | kfree(addr); | |
888 | goto out; | |
889 | } | |
1da177e4 LT |
890 | goto retry; |
891 | } | |
892 | addr->hash ^= sk->sk_type; | |
893 | ||
894 | __unix_remove_socket(sk); | |
ae3b5641 | 895 | smp_store_release(&u->addr, addr); |
1da177e4 | 896 | __unix_insert_socket(&unix_socket_table[addr->hash], sk); |
fbe9cc4a | 897 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
898 | err = 0; |
899 | ||
6e1ce3c3 | 900 | out: mutex_unlock(&u->bindlock); |
1da177e4 LT |
901 | return err; |
902 | } | |
903 | ||
097e66c5 DL |
904 | static struct sock *unix_find_other(struct net *net, |
905 | struct sockaddr_un *sunname, int len, | |
95c96174 | 906 | int type, unsigned int hash, int *error) |
1da177e4 LT |
907 | { |
908 | struct sock *u; | |
421748ec | 909 | struct path path; |
1da177e4 | 910 | int err = 0; |
ac7bfa62 | 911 | |
1da177e4 | 912 | if (sunname->sun_path[0]) { |
421748ec AV |
913 | struct inode *inode; |
914 | err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path); | |
1da177e4 LT |
915 | if (err) |
916 | goto fail; | |
beef5121 | 917 | inode = d_backing_inode(path.dentry); |
421748ec | 918 | err = inode_permission(inode, MAY_WRITE); |
1da177e4 LT |
919 | if (err) |
920 | goto put_fail; | |
921 | ||
922 | err = -ECONNREFUSED; | |
421748ec | 923 | if (!S_ISSOCK(inode->i_mode)) |
1da177e4 | 924 | goto put_fail; |
6616f788 | 925 | u = unix_find_socket_byinode(inode); |
1da177e4 LT |
926 | if (!u) |
927 | goto put_fail; | |
928 | ||
929 | if (u->sk_type == type) | |
68ac1234 | 930 | touch_atime(&path); |
1da177e4 | 931 | |
421748ec | 932 | path_put(&path); |
1da177e4 | 933 | |
e27dfcea | 934 | err = -EPROTOTYPE; |
1da177e4 LT |
935 | if (u->sk_type != type) { |
936 | sock_put(u); | |
937 | goto fail; | |
938 | } | |
939 | } else { | |
940 | err = -ECONNREFUSED; | |
e27dfcea | 941 | u = unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
942 | if (u) { |
943 | struct dentry *dentry; | |
40ffe67d | 944 | dentry = unix_sk(u)->path.dentry; |
1da177e4 | 945 | if (dentry) |
68ac1234 | 946 | touch_atime(&unix_sk(u)->path); |
1da177e4 LT |
947 | } else |
948 | goto fail; | |
949 | } | |
950 | return u; | |
951 | ||
952 | put_fail: | |
421748ec | 953 | path_put(&path); |
1da177e4 | 954 | fail: |
e27dfcea | 955 | *error = err; |
1da177e4 LT |
956 | return NULL; |
957 | } | |
958 | ||
38f7bd94 | 959 | static int unix_mknod(const char *sun_path, umode_t mode, struct path *res) |
faf02010 | 960 | { |
38f7bd94 LT |
961 | struct dentry *dentry; |
962 | struct path path; | |
963 | int err = 0; | |
964 | /* | |
965 | * Get the parent directory, calculate the hash for last | |
966 | * component. | |
967 | */ | |
968 | dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0); | |
969 | err = PTR_ERR(dentry); | |
970 | if (IS_ERR(dentry)) | |
971 | return err; | |
faf02010 | 972 | |
38f7bd94 LT |
973 | /* |
974 | * All right, let's create it. | |
975 | */ | |
976 | err = security_path_mknod(&path, dentry, mode, 0); | |
faf02010 | 977 | if (!err) { |
38f7bd94 | 978 | err = vfs_mknod(d_inode(path.dentry), dentry, mode, 0); |
faf02010 | 979 | if (!err) { |
38f7bd94 | 980 | res->mnt = mntget(path.mnt); |
faf02010 AV |
981 | res->dentry = dget(dentry); |
982 | } | |
983 | } | |
38f7bd94 | 984 | done_path_create(&path, dentry); |
faf02010 AV |
985 | return err; |
986 | } | |
1da177e4 LT |
987 | |
988 | static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
989 | { | |
990 | struct sock *sk = sock->sk; | |
3b1e0a65 | 991 | struct net *net = sock_net(sk); |
1da177e4 | 992 | struct unix_sock *u = unix_sk(sk); |
e27dfcea | 993 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
dae6ad8f | 994 | char *sun_path = sunaddr->sun_path; |
38f7bd94 | 995 | int err; |
95c96174 | 996 | unsigned int hash; |
1da177e4 LT |
997 | struct unix_address *addr; |
998 | struct hlist_head *list; | |
82fe0d2b | 999 | struct path path = { }; |
1da177e4 LT |
1000 | |
1001 | err = -EINVAL; | |
defbcf2d MJ |
1002 | if (addr_len < offsetofend(struct sockaddr_un, sun_family) || |
1003 | sunaddr->sun_family != AF_UNIX) | |
1da177e4 LT |
1004 | goto out; |
1005 | ||
e27dfcea | 1006 | if (addr_len == sizeof(short)) { |
1da177e4 LT |
1007 | err = unix_autobind(sock); |
1008 | goto out; | |
1009 | } | |
1010 | ||
1011 | err = unix_mkname(sunaddr, addr_len, &hash); | |
1012 | if (err < 0) | |
1013 | goto out; | |
1014 | addr_len = err; | |
1015 | ||
0fb44559 WC |
1016 | if (sun_path[0]) { |
1017 | umode_t mode = S_IFSOCK | | |
1018 | (SOCK_INODE(sock)->i_mode & ~current_umask()); | |
1019 | err = unix_mknod(sun_path, mode, &path); | |
1020 | if (err) { | |
1021 | if (err == -EEXIST) | |
1022 | err = -EADDRINUSE; | |
1023 | goto out; | |
1024 | } | |
1025 | } | |
1026 | ||
6e1ce3c3 | 1027 | err = mutex_lock_interruptible(&u->bindlock); |
37ab4fa7 | 1028 | if (err) |
0fb44559 | 1029 | goto out_put; |
1da177e4 LT |
1030 | |
1031 | err = -EINVAL; | |
1032 | if (u->addr) | |
1033 | goto out_up; | |
1034 | ||
1035 | err = -ENOMEM; | |
1036 | addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL); | |
1037 | if (!addr) | |
1038 | goto out_up; | |
1039 | ||
1040 | memcpy(addr->name, sunaddr, addr_len); | |
1041 | addr->len = addr_len; | |
1042 | addr->hash = hash ^ sk->sk_type; | |
8c9814b9 | 1043 | refcount_set(&addr->refcnt, 1); |
1da177e4 | 1044 | |
38f7bd94 | 1045 | if (sun_path[0]) { |
1da177e4 | 1046 | addr->hash = UNIX_HASH_SIZE; |
beef5121 | 1047 | hash = d_backing_inode(path.dentry)->i_ino & (UNIX_HASH_SIZE - 1); |
faf02010 | 1048 | spin_lock(&unix_table_lock); |
38f7bd94 | 1049 | u->path = path; |
faf02010 AV |
1050 | list = &unix_socket_table[hash]; |
1051 | } else { | |
1052 | spin_lock(&unix_table_lock); | |
1da177e4 | 1053 | err = -EADDRINUSE; |
097e66c5 | 1054 | if (__unix_find_socket_byname(net, sunaddr, addr_len, |
1da177e4 LT |
1055 | sk->sk_type, hash)) { |
1056 | unix_release_addr(addr); | |
1057 | goto out_unlock; | |
1058 | } | |
1059 | ||
1060 | list = &unix_socket_table[addr->hash]; | |
1da177e4 LT |
1061 | } |
1062 | ||
1063 | err = 0; | |
1064 | __unix_remove_socket(sk); | |
ae3b5641 | 1065 | smp_store_release(&u->addr, addr); |
1da177e4 LT |
1066 | __unix_insert_socket(list, sk); |
1067 | ||
1068 | out_unlock: | |
fbe9cc4a | 1069 | spin_unlock(&unix_table_lock); |
1da177e4 | 1070 | out_up: |
6e1ce3c3 | 1071 | mutex_unlock(&u->bindlock); |
0fb44559 WC |
1072 | out_put: |
1073 | if (err) | |
1074 | path_put(&path); | |
1da177e4 LT |
1075 | out: |
1076 | return err; | |
1da177e4 LT |
1077 | } |
1078 | ||
278a3de5 DM |
1079 | static void unix_state_double_lock(struct sock *sk1, struct sock *sk2) |
1080 | { | |
1081 | if (unlikely(sk1 == sk2) || !sk2) { | |
1082 | unix_state_lock(sk1); | |
1083 | return; | |
1084 | } | |
1085 | if (sk1 < sk2) { | |
1086 | unix_state_lock(sk1); | |
1087 | unix_state_lock_nested(sk2); | |
1088 | } else { | |
1089 | unix_state_lock(sk2); | |
1090 | unix_state_lock_nested(sk1); | |
1091 | } | |
1092 | } | |
1093 | ||
1094 | static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2) | |
1095 | { | |
1096 | if (unlikely(sk1 == sk2) || !sk2) { | |
1097 | unix_state_unlock(sk1); | |
1098 | return; | |
1099 | } | |
1100 | unix_state_unlock(sk1); | |
1101 | unix_state_unlock(sk2); | |
1102 | } | |
1103 | ||
1da177e4 LT |
1104 | static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr, |
1105 | int alen, int flags) | |
1106 | { | |
1107 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1108 | struct net *net = sock_net(sk); |
e27dfcea | 1109 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr; |
1da177e4 | 1110 | struct sock *other; |
95c96174 | 1111 | unsigned int hash; |
1da177e4 LT |
1112 | int err; |
1113 | ||
defbcf2d MJ |
1114 | err = -EINVAL; |
1115 | if (alen < offsetofend(struct sockaddr, sa_family)) | |
1116 | goto out; | |
1117 | ||
1da177e4 LT |
1118 | if (addr->sa_family != AF_UNSPEC) { |
1119 | err = unix_mkname(sunaddr, alen, &hash); | |
1120 | if (err < 0) | |
1121 | goto out; | |
1122 | alen = err; | |
1123 | ||
1124 | if (test_bit(SOCK_PASSCRED, &sock->flags) && | |
1125 | !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0) | |
1126 | goto out; | |
1127 | ||
278a3de5 | 1128 | restart: |
e27dfcea | 1129 | other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err); |
1da177e4 LT |
1130 | if (!other) |
1131 | goto out; | |
1132 | ||
278a3de5 DM |
1133 | unix_state_double_lock(sk, other); |
1134 | ||
1135 | /* Apparently VFS overslept socket death. Retry. */ | |
1136 | if (sock_flag(other, SOCK_DEAD)) { | |
1137 | unix_state_double_unlock(sk, other); | |
1138 | sock_put(other); | |
1139 | goto restart; | |
1140 | } | |
1da177e4 LT |
1141 | |
1142 | err = -EPERM; | |
1143 | if (!unix_may_send(sk, other)) | |
1144 | goto out_unlock; | |
1145 | ||
1146 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1147 | if (err) | |
1148 | goto out_unlock; | |
1149 | ||
1150 | } else { | |
1151 | /* | |
1152 | * 1003.1g breaking connected state with AF_UNSPEC | |
1153 | */ | |
1154 | other = NULL; | |
278a3de5 | 1155 | unix_state_double_lock(sk, other); |
1da177e4 LT |
1156 | } |
1157 | ||
1158 | /* | |
1159 | * If it was connected, reconnect. | |
1160 | */ | |
1161 | if (unix_peer(sk)) { | |
1162 | struct sock *old_peer = unix_peer(sk); | |
e27dfcea | 1163 | unix_peer(sk) = other; |
7d267278 RW |
1164 | unix_dgram_peer_wake_disconnect_wakeup(sk, old_peer); |
1165 | ||
278a3de5 | 1166 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1167 | |
1168 | if (other != old_peer) | |
1169 | unix_dgram_disconnected(sk, old_peer); | |
1170 | sock_put(old_peer); | |
1171 | } else { | |
e27dfcea | 1172 | unix_peer(sk) = other; |
278a3de5 | 1173 | unix_state_double_unlock(sk, other); |
1da177e4 | 1174 | } |
ac7bfa62 | 1175 | return 0; |
1da177e4 LT |
1176 | |
1177 | out_unlock: | |
278a3de5 | 1178 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1179 | sock_put(other); |
1180 | out: | |
1181 | return err; | |
1182 | } | |
1183 | ||
1184 | static long unix_wait_for_peer(struct sock *other, long timeo) | |
1185 | { | |
1186 | struct unix_sock *u = unix_sk(other); | |
1187 | int sched; | |
1188 | DEFINE_WAIT(wait); | |
1189 | ||
1190 | prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE); | |
1191 | ||
1192 | sched = !sock_flag(other, SOCK_DEAD) && | |
1193 | !(other->sk_shutdown & RCV_SHUTDOWN) && | |
3c73419c | 1194 | unix_recvq_full(other); |
1da177e4 | 1195 | |
1c92b4e5 | 1196 | unix_state_unlock(other); |
1da177e4 LT |
1197 | |
1198 | if (sched) | |
1199 | timeo = schedule_timeout(timeo); | |
1200 | ||
1201 | finish_wait(&u->peer_wait, &wait); | |
1202 | return timeo; | |
1203 | } | |
1204 | ||
1205 | static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
1206 | int addr_len, int flags) | |
1207 | { | |
e27dfcea | 1208 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
1da177e4 | 1209 | struct sock *sk = sock->sk; |
3b1e0a65 | 1210 | struct net *net = sock_net(sk); |
1da177e4 LT |
1211 | struct unix_sock *u = unix_sk(sk), *newu, *otheru; |
1212 | struct sock *newsk = NULL; | |
1213 | struct sock *other = NULL; | |
1214 | struct sk_buff *skb = NULL; | |
95c96174 | 1215 | unsigned int hash; |
1da177e4 LT |
1216 | int st; |
1217 | int err; | |
1218 | long timeo; | |
1219 | ||
1220 | err = unix_mkname(sunaddr, addr_len, &hash); | |
1221 | if (err < 0) | |
1222 | goto out; | |
1223 | addr_len = err; | |
1224 | ||
f64f9e71 JP |
1225 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr && |
1226 | (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1227 | goto out; |
1228 | ||
1229 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
1230 | ||
1231 | /* First of all allocate resources. | |
1232 | If we will make it after state is locked, | |
1233 | we will have to recheck all again in any case. | |
1234 | */ | |
1235 | ||
1236 | err = -ENOMEM; | |
1237 | ||
1238 | /* create new sock for complete connection */ | |
11aa9c28 | 1239 | newsk = unix_create1(sock_net(sk), NULL, 0); |
1da177e4 LT |
1240 | if (newsk == NULL) |
1241 | goto out; | |
1242 | ||
1243 | /* Allocate skb for sending to listening sock */ | |
1244 | skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL); | |
1245 | if (skb == NULL) | |
1246 | goto out; | |
1247 | ||
1248 | restart: | |
1249 | /* Find listening sock. */ | |
097e66c5 | 1250 | other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err); |
1da177e4 LT |
1251 | if (!other) |
1252 | goto out; | |
1253 | ||
1254 | /* Latch state of peer */ | |
1c92b4e5 | 1255 | unix_state_lock(other); |
1da177e4 LT |
1256 | |
1257 | /* Apparently VFS overslept socket death. Retry. */ | |
1258 | if (sock_flag(other, SOCK_DEAD)) { | |
1c92b4e5 | 1259 | unix_state_unlock(other); |
1da177e4 LT |
1260 | sock_put(other); |
1261 | goto restart; | |
1262 | } | |
1263 | ||
1264 | err = -ECONNREFUSED; | |
1265 | if (other->sk_state != TCP_LISTEN) | |
1266 | goto out_unlock; | |
77238f2b TS |
1267 | if (other->sk_shutdown & RCV_SHUTDOWN) |
1268 | goto out_unlock; | |
1da177e4 | 1269 | |
3c73419c | 1270 | if (unix_recvq_full(other)) { |
1da177e4 LT |
1271 | err = -EAGAIN; |
1272 | if (!timeo) | |
1273 | goto out_unlock; | |
1274 | ||
1275 | timeo = unix_wait_for_peer(other, timeo); | |
1276 | ||
1277 | err = sock_intr_errno(timeo); | |
1278 | if (signal_pending(current)) | |
1279 | goto out; | |
1280 | sock_put(other); | |
1281 | goto restart; | |
ac7bfa62 | 1282 | } |
1da177e4 LT |
1283 | |
1284 | /* Latch our state. | |
1285 | ||
e5537bfc | 1286 | It is tricky place. We need to grab our state lock and cannot |
1da177e4 LT |
1287 | drop lock on peer. It is dangerous because deadlock is |
1288 | possible. Connect to self case and simultaneous | |
1289 | attempt to connect are eliminated by checking socket | |
1290 | state. other is TCP_LISTEN, if sk is TCP_LISTEN we | |
1291 | check this before attempt to grab lock. | |
1292 | ||
1293 | Well, and we have to recheck the state after socket locked. | |
1294 | */ | |
1295 | st = sk->sk_state; | |
1296 | ||
1297 | switch (st) { | |
1298 | case TCP_CLOSE: | |
1299 | /* This is ok... continue with connect */ | |
1300 | break; | |
1301 | case TCP_ESTABLISHED: | |
1302 | /* Socket is already connected */ | |
1303 | err = -EISCONN; | |
1304 | goto out_unlock; | |
1305 | default: | |
1306 | err = -EINVAL; | |
1307 | goto out_unlock; | |
1308 | } | |
1309 | ||
1c92b4e5 | 1310 | unix_state_lock_nested(sk); |
1da177e4 LT |
1311 | |
1312 | if (sk->sk_state != st) { | |
1c92b4e5 DM |
1313 | unix_state_unlock(sk); |
1314 | unix_state_unlock(other); | |
1da177e4 LT |
1315 | sock_put(other); |
1316 | goto restart; | |
1317 | } | |
1318 | ||
3610cda5 | 1319 | err = security_unix_stream_connect(sk, other, newsk); |
1da177e4 | 1320 | if (err) { |
1c92b4e5 | 1321 | unix_state_unlock(sk); |
1da177e4 LT |
1322 | goto out_unlock; |
1323 | } | |
1324 | ||
1325 | /* The way is open! Fastly set all the necessary fields... */ | |
1326 | ||
1327 | sock_hold(sk); | |
1328 | unix_peer(newsk) = sk; | |
1329 | newsk->sk_state = TCP_ESTABLISHED; | |
1330 | newsk->sk_type = sk->sk_type; | |
109f6e39 | 1331 | init_peercred(newsk); |
1da177e4 | 1332 | newu = unix_sk(newsk); |
eaefd110 | 1333 | RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq); |
1da177e4 LT |
1334 | otheru = unix_sk(other); |
1335 | ||
ae3b5641 AV |
1336 | /* copy address information from listening to new sock |
1337 | * | |
1338 | * The contents of *(otheru->addr) and otheru->path | |
1339 | * are seen fully set up here, since we have found | |
1340 | * otheru in hash under unix_table_lock. Insertion | |
1341 | * into the hash chain we'd found it in had been done | |
1342 | * in an earlier critical area protected by unix_table_lock, | |
1343 | * the same one where we'd set *(otheru->addr) contents, | |
1344 | * as well as otheru->path and otheru->addr itself. | |
1345 | * | |
1346 | * Using smp_store_release() here to set newu->addr | |
1347 | * is enough to make those stores, as well as stores | |
1348 | * to newu->path visible to anyone who gets newu->addr | |
1349 | * by smp_load_acquire(). IOW, the same warranties | |
1350 | * as for unix_sock instances bound in unix_bind() or | |
1351 | * in unix_autobind(). | |
1352 | */ | |
40ffe67d AV |
1353 | if (otheru->path.dentry) { |
1354 | path_get(&otheru->path); | |
1355 | newu->path = otheru->path; | |
1da177e4 | 1356 | } |
ae3b5641 AV |
1357 | refcount_inc(&otheru->addr->refcnt); |
1358 | smp_store_release(&newu->addr, otheru->addr); | |
1da177e4 LT |
1359 | |
1360 | /* Set credentials */ | |
109f6e39 | 1361 | copy_peercred(sk, other); |
1da177e4 | 1362 | |
1da177e4 LT |
1363 | sock->state = SS_CONNECTED; |
1364 | sk->sk_state = TCP_ESTABLISHED; | |
830a1e5c BL |
1365 | sock_hold(newsk); |
1366 | ||
4e857c58 | 1367 | smp_mb__after_atomic(); /* sock_hold() does an atomic_inc() */ |
830a1e5c | 1368 | unix_peer(sk) = newsk; |
1da177e4 | 1369 | |
1c92b4e5 | 1370 | unix_state_unlock(sk); |
1da177e4 LT |
1371 | |
1372 | /* take ten and and send info to listening sock */ | |
1373 | spin_lock(&other->sk_receive_queue.lock); | |
1374 | __skb_queue_tail(&other->sk_receive_queue, skb); | |
1da177e4 | 1375 | spin_unlock(&other->sk_receive_queue.lock); |
1c92b4e5 | 1376 | unix_state_unlock(other); |
676d2369 | 1377 | other->sk_data_ready(other); |
1da177e4 LT |
1378 | sock_put(other); |
1379 | return 0; | |
1380 | ||
1381 | out_unlock: | |
1382 | if (other) | |
1c92b4e5 | 1383 | unix_state_unlock(other); |
1da177e4 LT |
1384 | |
1385 | out: | |
40d44446 | 1386 | kfree_skb(skb); |
1da177e4 LT |
1387 | if (newsk) |
1388 | unix_release_sock(newsk, 0); | |
1389 | if (other) | |
1390 | sock_put(other); | |
1391 | return err; | |
1392 | } | |
1393 | ||
1394 | static int unix_socketpair(struct socket *socka, struct socket *sockb) | |
1395 | { | |
e27dfcea | 1396 | struct sock *ska = socka->sk, *skb = sockb->sk; |
1da177e4 LT |
1397 | |
1398 | /* Join our sockets back to back */ | |
1399 | sock_hold(ska); | |
1400 | sock_hold(skb); | |
e27dfcea JK |
1401 | unix_peer(ska) = skb; |
1402 | unix_peer(skb) = ska; | |
109f6e39 EB |
1403 | init_peercred(ska); |
1404 | init_peercred(skb); | |
1da177e4 LT |
1405 | |
1406 | if (ska->sk_type != SOCK_DGRAM) { | |
1407 | ska->sk_state = TCP_ESTABLISHED; | |
1408 | skb->sk_state = TCP_ESTABLISHED; | |
1409 | socka->state = SS_CONNECTED; | |
1410 | sockb->state = SS_CONNECTED; | |
1411 | } | |
1412 | return 0; | |
1413 | } | |
1414 | ||
90c6bd34 DB |
1415 | static void unix_sock_inherit_flags(const struct socket *old, |
1416 | struct socket *new) | |
1417 | { | |
1418 | if (test_bit(SOCK_PASSCRED, &old->flags)) | |
1419 | set_bit(SOCK_PASSCRED, &new->flags); | |
1420 | if (test_bit(SOCK_PASSSEC, &old->flags)) | |
1421 | set_bit(SOCK_PASSSEC, &new->flags); | |
1422 | } | |
1423 | ||
cdfbabfb DH |
1424 | static int unix_accept(struct socket *sock, struct socket *newsock, int flags, |
1425 | bool kern) | |
1da177e4 LT |
1426 | { |
1427 | struct sock *sk = sock->sk; | |
1428 | struct sock *tsk; | |
1429 | struct sk_buff *skb; | |
1430 | int err; | |
1431 | ||
1432 | err = -EOPNOTSUPP; | |
6eba6a37 | 1433 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
1da177e4 LT |
1434 | goto out; |
1435 | ||
1436 | err = -EINVAL; | |
1437 | if (sk->sk_state != TCP_LISTEN) | |
1438 | goto out; | |
1439 | ||
1440 | /* If socket state is TCP_LISTEN it cannot change (for now...), | |
1441 | * so that no locks are necessary. | |
1442 | */ | |
1443 | ||
1444 | skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err); | |
1445 | if (!skb) { | |
1446 | /* This means receive shutdown. */ | |
1447 | if (err == 0) | |
1448 | err = -EINVAL; | |
1449 | goto out; | |
1450 | } | |
1451 | ||
1452 | tsk = skb->sk; | |
1453 | skb_free_datagram(sk, skb); | |
1454 | wake_up_interruptible(&unix_sk(sk)->peer_wait); | |
1455 | ||
1456 | /* attach accepted sock to socket */ | |
1c92b4e5 | 1457 | unix_state_lock(tsk); |
1da177e4 | 1458 | newsock->state = SS_CONNECTED; |
90c6bd34 | 1459 | unix_sock_inherit_flags(sock, newsock); |
1da177e4 | 1460 | sock_graft(tsk, newsock); |
1c92b4e5 | 1461 | unix_state_unlock(tsk); |
1da177e4 LT |
1462 | return 0; |
1463 | ||
1464 | out: | |
1465 | return err; | |
1466 | } | |
1467 | ||
1468 | ||
9b2c45d4 | 1469 | static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int peer) |
1da177e4 LT |
1470 | { |
1471 | struct sock *sk = sock->sk; | |
ae3b5641 | 1472 | struct unix_address *addr; |
13cfa97b | 1473 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr); |
1da177e4 LT |
1474 | int err = 0; |
1475 | ||
1476 | if (peer) { | |
1477 | sk = unix_peer_get(sk); | |
1478 | ||
1479 | err = -ENOTCONN; | |
1480 | if (!sk) | |
1481 | goto out; | |
1482 | err = 0; | |
1483 | } else { | |
1484 | sock_hold(sk); | |
1485 | } | |
1486 | ||
ae3b5641 AV |
1487 | addr = smp_load_acquire(&unix_sk(sk)->addr); |
1488 | if (!addr) { | |
1da177e4 LT |
1489 | sunaddr->sun_family = AF_UNIX; |
1490 | sunaddr->sun_path[0] = 0; | |
9b2c45d4 | 1491 | err = sizeof(short); |
1da177e4 | 1492 | } else { |
9b2c45d4 DV |
1493 | err = addr->len; |
1494 | memcpy(sunaddr, addr->name, addr->len); | |
1da177e4 | 1495 | } |
1da177e4 LT |
1496 | sock_put(sk); |
1497 | out: | |
1498 | return err; | |
1499 | } | |
1500 | ||
f78a5fda | 1501 | static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds) |
7361c36c EB |
1502 | { |
1503 | int err = 0; | |
16e57262 | 1504 | |
f78a5fda | 1505 | UNIXCB(skb).pid = get_pid(scm->pid); |
6b0ee8c0 EB |
1506 | UNIXCB(skb).uid = scm->creds.uid; |
1507 | UNIXCB(skb).gid = scm->creds.gid; | |
7361c36c | 1508 | UNIXCB(skb).fp = NULL; |
37a9a8df | 1509 | unix_get_secdata(scm, skb); |
7361c36c EB |
1510 | if (scm->fp && send_fds) |
1511 | err = unix_attach_fds(scm, skb); | |
1512 | ||
1513 | skb->destructor = unix_destruct_scm; | |
1514 | return err; | |
1515 | } | |
1516 | ||
9490f886 HFS |
1517 | static bool unix_passcred_enabled(const struct socket *sock, |
1518 | const struct sock *other) | |
1519 | { | |
1520 | return test_bit(SOCK_PASSCRED, &sock->flags) || | |
1521 | !other->sk_socket || | |
1522 | test_bit(SOCK_PASSCRED, &other->sk_socket->flags); | |
1523 | } | |
1524 | ||
16e57262 ED |
1525 | /* |
1526 | * Some apps rely on write() giving SCM_CREDENTIALS | |
1527 | * We include credentials if source or destination socket | |
1528 | * asserted SOCK_PASSCRED. | |
1529 | */ | |
1530 | static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock, | |
1531 | const struct sock *other) | |
1532 | { | |
6b0ee8c0 | 1533 | if (UNIXCB(skb).pid) |
16e57262 | 1534 | return; |
9490f886 | 1535 | if (unix_passcred_enabled(sock, other)) { |
16e57262 | 1536 | UNIXCB(skb).pid = get_pid(task_tgid(current)); |
6e0895c2 | 1537 | current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid); |
16e57262 ED |
1538 | } |
1539 | } | |
1540 | ||
9490f886 HFS |
1541 | static int maybe_init_creds(struct scm_cookie *scm, |
1542 | struct socket *socket, | |
1543 | const struct sock *other) | |
1544 | { | |
1545 | int err; | |
1546 | struct msghdr msg = { .msg_controllen = 0 }; | |
1547 | ||
1548 | err = scm_send(socket, &msg, scm, false); | |
1549 | if (err) | |
1550 | return err; | |
1551 | ||
1552 | if (unix_passcred_enabled(socket, other)) { | |
1553 | scm->pid = get_pid(task_tgid(current)); | |
1554 | current_uid_gid(&scm->creds.uid, &scm->creds.gid); | |
1555 | } | |
1556 | return err; | |
1557 | } | |
1558 | ||
1559 | static bool unix_skb_scm_eq(struct sk_buff *skb, | |
1560 | struct scm_cookie *scm) | |
1561 | { | |
1562 | const struct unix_skb_parms *u = &UNIXCB(skb); | |
1563 | ||
1564 | return u->pid == scm->pid && | |
1565 | uid_eq(u->uid, scm->creds.uid) && | |
1566 | gid_eq(u->gid, scm->creds.gid) && | |
1567 | unix_secdata_eq(scm, skb); | |
1568 | } | |
1569 | ||
1da177e4 LT |
1570 | /* |
1571 | * Send AF_UNIX data. | |
1572 | */ | |
1573 | ||
1b784140 YX |
1574 | static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg, |
1575 | size_t len) | |
1da177e4 | 1576 | { |
1da177e4 | 1577 | struct sock *sk = sock->sk; |
3b1e0a65 | 1578 | struct net *net = sock_net(sk); |
1da177e4 | 1579 | struct unix_sock *u = unix_sk(sk); |
342dfc30 | 1580 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name); |
1da177e4 LT |
1581 | struct sock *other = NULL; |
1582 | int namelen = 0; /* fake GCC */ | |
1583 | int err; | |
95c96174 | 1584 | unsigned int hash; |
f78a5fda | 1585 | struct sk_buff *skb; |
1da177e4 | 1586 | long timeo; |
7cc05662 | 1587 | struct scm_cookie scm; |
eb6a2481 | 1588 | int data_len = 0; |
7d267278 | 1589 | int sk_locked; |
1da177e4 | 1590 | |
5f23b734 | 1591 | wait_for_unix_gc(); |
7cc05662 | 1592 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1593 | if (err < 0) |
1594 | return err; | |
1595 | ||
1596 | err = -EOPNOTSUPP; | |
1597 | if (msg->msg_flags&MSG_OOB) | |
1598 | goto out; | |
1599 | ||
1600 | if (msg->msg_namelen) { | |
1601 | err = unix_mkname(sunaddr, msg->msg_namelen, &hash); | |
1602 | if (err < 0) | |
1603 | goto out; | |
1604 | namelen = err; | |
1605 | } else { | |
1606 | sunaddr = NULL; | |
1607 | err = -ENOTCONN; | |
1608 | other = unix_peer_get(sk); | |
1609 | if (!other) | |
1610 | goto out; | |
1611 | } | |
1612 | ||
f64f9e71 JP |
1613 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr |
1614 | && (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1615 | goto out; |
1616 | ||
1617 | err = -EMSGSIZE; | |
1618 | if (len > sk->sk_sndbuf - 32) | |
1619 | goto out; | |
1620 | ||
31ff6aa5 | 1621 | if (len > SKB_MAX_ALLOC) { |
eb6a2481 ED |
1622 | data_len = min_t(size_t, |
1623 | len - SKB_MAX_ALLOC, | |
1624 | MAX_SKB_FRAGS * PAGE_SIZE); | |
31ff6aa5 KT |
1625 | data_len = PAGE_ALIGN(data_len); |
1626 | ||
1627 | BUILD_BUG_ON(SKB_MAX_ALLOC < PAGE_SIZE); | |
1628 | } | |
eb6a2481 ED |
1629 | |
1630 | skb = sock_alloc_send_pskb(sk, len - data_len, data_len, | |
28d64271 ED |
1631 | msg->msg_flags & MSG_DONTWAIT, &err, |
1632 | PAGE_ALLOC_COSTLY_ORDER); | |
e27dfcea | 1633 | if (skb == NULL) |
1da177e4 LT |
1634 | goto out; |
1635 | ||
7cc05662 | 1636 | err = unix_scm_to_skb(&scm, skb, true); |
25888e30 | 1637 | if (err < 0) |
7361c36c | 1638 | goto out_free; |
877ce7c1 | 1639 | |
eb6a2481 ED |
1640 | skb_put(skb, len - data_len); |
1641 | skb->data_len = data_len; | |
1642 | skb->len = len; | |
c0371da6 | 1643 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len); |
1da177e4 LT |
1644 | if (err) |
1645 | goto out_free; | |
1646 | ||
1647 | timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); | |
1648 | ||
1649 | restart: | |
1650 | if (!other) { | |
1651 | err = -ECONNRESET; | |
1652 | if (sunaddr == NULL) | |
1653 | goto out_free; | |
1654 | ||
097e66c5 | 1655 | other = unix_find_other(net, sunaddr, namelen, sk->sk_type, |
1da177e4 | 1656 | hash, &err); |
e27dfcea | 1657 | if (other == NULL) |
1da177e4 LT |
1658 | goto out_free; |
1659 | } | |
1660 | ||
d6ae3bae AC |
1661 | if (sk_filter(other, skb) < 0) { |
1662 | /* Toss the packet but do not return any error to the sender */ | |
1663 | err = len; | |
1664 | goto out_free; | |
1665 | } | |
1666 | ||
7d267278 | 1667 | sk_locked = 0; |
1c92b4e5 | 1668 | unix_state_lock(other); |
7d267278 | 1669 | restart_locked: |
1da177e4 LT |
1670 | err = -EPERM; |
1671 | if (!unix_may_send(sk, other)) | |
1672 | goto out_unlock; | |
1673 | ||
7d267278 | 1674 | if (unlikely(sock_flag(other, SOCK_DEAD))) { |
1da177e4 LT |
1675 | /* |
1676 | * Check with 1003.1g - what should | |
1677 | * datagram error | |
1678 | */ | |
1c92b4e5 | 1679 | unix_state_unlock(other); |
1da177e4 LT |
1680 | sock_put(other); |
1681 | ||
7d267278 RW |
1682 | if (!sk_locked) |
1683 | unix_state_lock(sk); | |
1684 | ||
1da177e4 | 1685 | err = 0; |
1da177e4 | 1686 | if (unix_peer(sk) == other) { |
e27dfcea | 1687 | unix_peer(sk) = NULL; |
7d267278 RW |
1688 | unix_dgram_peer_wake_disconnect_wakeup(sk, other); |
1689 | ||
1c92b4e5 | 1690 | unix_state_unlock(sk); |
1da177e4 LT |
1691 | |
1692 | unix_dgram_disconnected(sk, other); | |
1693 | sock_put(other); | |
1694 | err = -ECONNREFUSED; | |
1695 | } else { | |
1c92b4e5 | 1696 | unix_state_unlock(sk); |
1da177e4 LT |
1697 | } |
1698 | ||
1699 | other = NULL; | |
1700 | if (err) | |
1701 | goto out_free; | |
1702 | goto restart; | |
1703 | } | |
1704 | ||
1705 | err = -EPIPE; | |
1706 | if (other->sk_shutdown & RCV_SHUTDOWN) | |
1707 | goto out_unlock; | |
1708 | ||
1709 | if (sk->sk_type != SOCK_SEQPACKET) { | |
1710 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1711 | if (err) | |
1712 | goto out_unlock; | |
1713 | } | |
1714 | ||
a5527dda RW |
1715 | /* other == sk && unix_peer(other) != sk if |
1716 | * - unix_peer(sk) == NULL, destination address bound to sk | |
1717 | * - unix_peer(sk) == sk by time of get but disconnected before lock | |
1718 | */ | |
1719 | if (other != sk && | |
1720 | unlikely(unix_peer(other) != sk && unix_recvq_full(other))) { | |
7d267278 RW |
1721 | if (timeo) { |
1722 | timeo = unix_wait_for_peer(other, timeo); | |
1723 | ||
1724 | err = sock_intr_errno(timeo); | |
1725 | if (signal_pending(current)) | |
1726 | goto out_free; | |
1727 | ||
1728 | goto restart; | |
1da177e4 LT |
1729 | } |
1730 | ||
7d267278 RW |
1731 | if (!sk_locked) { |
1732 | unix_state_unlock(other); | |
1733 | unix_state_double_lock(sk, other); | |
1734 | } | |
1da177e4 | 1735 | |
7d267278 RW |
1736 | if (unix_peer(sk) != other || |
1737 | unix_dgram_peer_wake_me(sk, other)) { | |
1738 | err = -EAGAIN; | |
1739 | sk_locked = 1; | |
1740 | goto out_unlock; | |
1741 | } | |
1da177e4 | 1742 | |
7d267278 RW |
1743 | if (!sk_locked) { |
1744 | sk_locked = 1; | |
1745 | goto restart_locked; | |
1746 | } | |
1da177e4 LT |
1747 | } |
1748 | ||
7d267278 RW |
1749 | if (unlikely(sk_locked)) |
1750 | unix_state_unlock(sk); | |
1751 | ||
3f66116e AC |
1752 | if (sock_flag(other, SOCK_RCVTSTAMP)) |
1753 | __net_timestamp(skb); | |
16e57262 | 1754 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1755 | skb_queue_tail(&other->sk_receive_queue, skb); |
1c92b4e5 | 1756 | unix_state_unlock(other); |
676d2369 | 1757 | other->sk_data_ready(other); |
1da177e4 | 1758 | sock_put(other); |
7cc05662 | 1759 | scm_destroy(&scm); |
1da177e4 LT |
1760 | return len; |
1761 | ||
1762 | out_unlock: | |
7d267278 RW |
1763 | if (sk_locked) |
1764 | unix_state_unlock(sk); | |
1c92b4e5 | 1765 | unix_state_unlock(other); |
1da177e4 LT |
1766 | out_free: |
1767 | kfree_skb(skb); | |
1768 | out: | |
1769 | if (other) | |
1770 | sock_put(other); | |
7cc05662 | 1771 | scm_destroy(&scm); |
1da177e4 LT |
1772 | return err; |
1773 | } | |
1774 | ||
e370a723 | 1775 | /* We use paged skbs for stream sockets, and limit occupancy to 32768 |
d4e9a408 | 1776 | * bytes, and a minimum of a full page. |
e370a723 ED |
1777 | */ |
1778 | #define UNIX_SKB_FRAGS_SZ (PAGE_SIZE << get_order(32768)) | |
ac7bfa62 | 1779 | |
1b784140 YX |
1780 | static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg, |
1781 | size_t len) | |
1da177e4 | 1782 | { |
1da177e4 LT |
1783 | struct sock *sk = sock->sk; |
1784 | struct sock *other = NULL; | |
6eba6a37 | 1785 | int err, size; |
f78a5fda | 1786 | struct sk_buff *skb; |
e27dfcea | 1787 | int sent = 0; |
7cc05662 | 1788 | struct scm_cookie scm; |
8ba69ba6 | 1789 | bool fds_sent = false; |
e370a723 | 1790 | int data_len; |
1da177e4 | 1791 | |
5f23b734 | 1792 | wait_for_unix_gc(); |
7cc05662 | 1793 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1794 | if (err < 0) |
1795 | return err; | |
1796 | ||
1797 | err = -EOPNOTSUPP; | |
1798 | if (msg->msg_flags&MSG_OOB) | |
1799 | goto out_err; | |
1800 | ||
1801 | if (msg->msg_namelen) { | |
1802 | err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; | |
1803 | goto out_err; | |
1804 | } else { | |
1da177e4 | 1805 | err = -ENOTCONN; |
830a1e5c | 1806 | other = unix_peer(sk); |
1da177e4 LT |
1807 | if (!other) |
1808 | goto out_err; | |
1809 | } | |
1810 | ||
1811 | if (sk->sk_shutdown & SEND_SHUTDOWN) | |
1812 | goto pipe_err; | |
1813 | ||
6eba6a37 | 1814 | while (sent < len) { |
e370a723 | 1815 | size = len - sent; |
1da177e4 LT |
1816 | |
1817 | /* Keep two messages in the pipe so it schedules better */ | |
e370a723 | 1818 | size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64); |
1da177e4 | 1819 | |
e370a723 ED |
1820 | /* allow fallback to order-0 allocations */ |
1821 | size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ); | |
ac7bfa62 | 1822 | |
e370a723 | 1823 | data_len = max_t(int, 0, size - SKB_MAX_HEAD(0)); |
1da177e4 | 1824 | |
31ff6aa5 KT |
1825 | data_len = min_t(size_t, size, PAGE_ALIGN(data_len)); |
1826 | ||
e370a723 | 1827 | skb = sock_alloc_send_pskb(sk, size - data_len, data_len, |
28d64271 ED |
1828 | msg->msg_flags & MSG_DONTWAIT, &err, |
1829 | get_order(UNIX_SKB_FRAGS_SZ)); | |
e370a723 | 1830 | if (!skb) |
1da177e4 LT |
1831 | goto out_err; |
1832 | ||
f78a5fda | 1833 | /* Only send the fds in the first buffer */ |
7cc05662 | 1834 | err = unix_scm_to_skb(&scm, skb, !fds_sent); |
25888e30 | 1835 | if (err < 0) { |
7361c36c | 1836 | kfree_skb(skb); |
f78a5fda | 1837 | goto out_err; |
6209344f | 1838 | } |
7361c36c | 1839 | fds_sent = true; |
1da177e4 | 1840 | |
e370a723 ED |
1841 | skb_put(skb, size - data_len); |
1842 | skb->data_len = data_len; | |
1843 | skb->len = size; | |
c0371da6 | 1844 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size); |
6eba6a37 | 1845 | if (err) { |
1da177e4 | 1846 | kfree_skb(skb); |
f78a5fda | 1847 | goto out_err; |
1da177e4 LT |
1848 | } |
1849 | ||
1c92b4e5 | 1850 | unix_state_lock(other); |
1da177e4 LT |
1851 | |
1852 | if (sock_flag(other, SOCK_DEAD) || | |
1853 | (other->sk_shutdown & RCV_SHUTDOWN)) | |
1854 | goto pipe_err_free; | |
1855 | ||
16e57262 | 1856 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1857 | skb_queue_tail(&other->sk_receive_queue, skb); |
1c92b4e5 | 1858 | unix_state_unlock(other); |
676d2369 | 1859 | other->sk_data_ready(other); |
e27dfcea | 1860 | sent += size; |
1da177e4 | 1861 | } |
1da177e4 | 1862 | |
7cc05662 | 1863 | scm_destroy(&scm); |
1da177e4 LT |
1864 | |
1865 | return sent; | |
1866 | ||
1867 | pipe_err_free: | |
1c92b4e5 | 1868 | unix_state_unlock(other); |
1da177e4 LT |
1869 | kfree_skb(skb); |
1870 | pipe_err: | |
6eba6a37 ED |
1871 | if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL)) |
1872 | send_sig(SIGPIPE, current, 0); | |
1da177e4 LT |
1873 | err = -EPIPE; |
1874 | out_err: | |
7cc05662 | 1875 | scm_destroy(&scm); |
1da177e4 LT |
1876 | return sent ? : err; |
1877 | } | |
1878 | ||
869e7c62 HFS |
1879 | static ssize_t unix_stream_sendpage(struct socket *socket, struct page *page, |
1880 | int offset, size_t size, int flags) | |
1881 | { | |
9490f886 HFS |
1882 | int err; |
1883 | bool send_sigpipe = false; | |
1884 | bool init_scm = true; | |
1885 | struct scm_cookie scm; | |
869e7c62 HFS |
1886 | struct sock *other, *sk = socket->sk; |
1887 | struct sk_buff *skb, *newskb = NULL, *tail = NULL; | |
1888 | ||
1889 | if (flags & MSG_OOB) | |
1890 | return -EOPNOTSUPP; | |
1891 | ||
1892 | other = unix_peer(sk); | |
1893 | if (!other || sk->sk_state != TCP_ESTABLISHED) | |
1894 | return -ENOTCONN; | |
1895 | ||
1896 | if (false) { | |
1897 | alloc_skb: | |
1898 | unix_state_unlock(other); | |
6e1ce3c3 | 1899 | mutex_unlock(&unix_sk(other)->iolock); |
869e7c62 HFS |
1900 | newskb = sock_alloc_send_pskb(sk, 0, 0, flags & MSG_DONTWAIT, |
1901 | &err, 0); | |
1902 | if (!newskb) | |
9490f886 | 1903 | goto err; |
869e7c62 HFS |
1904 | } |
1905 | ||
6e1ce3c3 | 1906 | /* we must acquire iolock as we modify already present |
869e7c62 HFS |
1907 | * skbs in the sk_receive_queue and mess with skb->len |
1908 | */ | |
6e1ce3c3 | 1909 | err = mutex_lock_interruptible(&unix_sk(other)->iolock); |
869e7c62 HFS |
1910 | if (err) { |
1911 | err = flags & MSG_DONTWAIT ? -EAGAIN : -ERESTARTSYS; | |
869e7c62 HFS |
1912 | goto err; |
1913 | } | |
1914 | ||
1915 | if (sk->sk_shutdown & SEND_SHUTDOWN) { | |
1916 | err = -EPIPE; | |
9490f886 | 1917 | send_sigpipe = true; |
869e7c62 HFS |
1918 | goto err_unlock; |
1919 | } | |
1920 | ||
1921 | unix_state_lock(other); | |
1922 | ||
1923 | if (sock_flag(other, SOCK_DEAD) || | |
1924 | other->sk_shutdown & RCV_SHUTDOWN) { | |
1925 | err = -EPIPE; | |
9490f886 | 1926 | send_sigpipe = true; |
869e7c62 HFS |
1927 | goto err_state_unlock; |
1928 | } | |
1929 | ||
9490f886 HFS |
1930 | if (init_scm) { |
1931 | err = maybe_init_creds(&scm, socket, other); | |
1932 | if (err) | |
1933 | goto err_state_unlock; | |
1934 | init_scm = false; | |
1935 | } | |
1936 | ||
869e7c62 HFS |
1937 | skb = skb_peek_tail(&other->sk_receive_queue); |
1938 | if (tail && tail == skb) { | |
1939 | skb = newskb; | |
9490f886 HFS |
1940 | } else if (!skb || !unix_skb_scm_eq(skb, &scm)) { |
1941 | if (newskb) { | |
869e7c62 | 1942 | skb = newskb; |
9490f886 HFS |
1943 | } else { |
1944 | tail = skb; | |
869e7c62 | 1945 | goto alloc_skb; |
9490f886 | 1946 | } |
869e7c62 HFS |
1947 | } else if (newskb) { |
1948 | /* this is fast path, we don't necessarily need to | |
1949 | * call to kfree_skb even though with newskb == NULL | |
1950 | * this - does no harm | |
1951 | */ | |
1952 | consume_skb(newskb); | |
8844f972 | 1953 | newskb = NULL; |
869e7c62 HFS |
1954 | } |
1955 | ||
1956 | if (skb_append_pagefrags(skb, page, offset, size)) { | |
1957 | tail = skb; | |
1958 | goto alloc_skb; | |
1959 | } | |
1960 | ||
1961 | skb->len += size; | |
1962 | skb->data_len += size; | |
1963 | skb->truesize += size; | |
14afee4b | 1964 | refcount_add(size, &sk->sk_wmem_alloc); |
869e7c62 | 1965 | |
a3a116e0 | 1966 | if (newskb) { |
9490f886 HFS |
1967 | err = unix_scm_to_skb(&scm, skb, false); |
1968 | if (err) | |
1969 | goto err_state_unlock; | |
a3a116e0 | 1970 | spin_lock(&other->sk_receive_queue.lock); |
869e7c62 | 1971 | __skb_queue_tail(&other->sk_receive_queue, newskb); |
a3a116e0 HFS |
1972 | spin_unlock(&other->sk_receive_queue.lock); |
1973 | } | |
869e7c62 HFS |
1974 | |
1975 | unix_state_unlock(other); | |
6e1ce3c3 | 1976 | mutex_unlock(&unix_sk(other)->iolock); |
869e7c62 HFS |
1977 | |
1978 | other->sk_data_ready(other); | |
9490f886 | 1979 | scm_destroy(&scm); |
869e7c62 HFS |
1980 | return size; |
1981 | ||
1982 | err_state_unlock: | |
1983 | unix_state_unlock(other); | |
1984 | err_unlock: | |
6e1ce3c3 | 1985 | mutex_unlock(&unix_sk(other)->iolock); |
869e7c62 HFS |
1986 | err: |
1987 | kfree_skb(newskb); | |
1988 | if (send_sigpipe && !(flags & MSG_NOSIGNAL)) | |
1989 | send_sig(SIGPIPE, current, 0); | |
9490f886 HFS |
1990 | if (!init_scm) |
1991 | scm_destroy(&scm); | |
869e7c62 HFS |
1992 | return err; |
1993 | } | |
1994 | ||
1b784140 YX |
1995 | static int unix_seqpacket_sendmsg(struct socket *sock, struct msghdr *msg, |
1996 | size_t len) | |
1da177e4 LT |
1997 | { |
1998 | int err; | |
1999 | struct sock *sk = sock->sk; | |
ac7bfa62 | 2000 | |
1da177e4 LT |
2001 | err = sock_error(sk); |
2002 | if (err) | |
2003 | return err; | |
2004 | ||
2005 | if (sk->sk_state != TCP_ESTABLISHED) | |
2006 | return -ENOTCONN; | |
2007 | ||
2008 | if (msg->msg_namelen) | |
2009 | msg->msg_namelen = 0; | |
2010 | ||
1b784140 | 2011 | return unix_dgram_sendmsg(sock, msg, len); |
1da177e4 | 2012 | } |
ac7bfa62 | 2013 | |
1b784140 YX |
2014 | static int unix_seqpacket_recvmsg(struct socket *sock, struct msghdr *msg, |
2015 | size_t size, int flags) | |
a05d2ad1 EB |
2016 | { |
2017 | struct sock *sk = sock->sk; | |
2018 | ||
2019 | if (sk->sk_state != TCP_ESTABLISHED) | |
2020 | return -ENOTCONN; | |
2021 | ||
1b784140 | 2022 | return unix_dgram_recvmsg(sock, msg, size, flags); |
a05d2ad1 EB |
2023 | } |
2024 | ||
1da177e4 LT |
2025 | static void unix_copy_addr(struct msghdr *msg, struct sock *sk) |
2026 | { | |
ae3b5641 | 2027 | struct unix_address *addr = smp_load_acquire(&unix_sk(sk)->addr); |
1da177e4 | 2028 | |
ae3b5641 AV |
2029 | if (addr) { |
2030 | msg->msg_namelen = addr->len; | |
2031 | memcpy(msg->msg_name, addr->name, addr->len); | |
1da177e4 LT |
2032 | } |
2033 | } | |
2034 | ||
1b784140 YX |
2035 | static int unix_dgram_recvmsg(struct socket *sock, struct msghdr *msg, |
2036 | size_t size, int flags) | |
1da177e4 | 2037 | { |
7cc05662 | 2038 | struct scm_cookie scm; |
1da177e4 LT |
2039 | struct sock *sk = sock->sk; |
2040 | struct unix_sock *u = unix_sk(sk); | |
64874280 RW |
2041 | struct sk_buff *skb, *last; |
2042 | long timeo; | |
1da177e4 | 2043 | int err; |
f55bb7f9 | 2044 | int peeked, skip; |
1da177e4 LT |
2045 | |
2046 | err = -EOPNOTSUPP; | |
2047 | if (flags&MSG_OOB) | |
2048 | goto out; | |
2049 | ||
64874280 | 2050 | timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT); |
1da177e4 | 2051 | |
64874280 | 2052 | do { |
6e1ce3c3 | 2053 | mutex_lock(&u->iolock); |
f55bb7f9 | 2054 | |
64874280 | 2055 | skip = sk_peek_offset(sk, flags); |
7c13f97f PA |
2056 | skb = __skb_try_recv_datagram(sk, flags, NULL, &peeked, &skip, |
2057 | &err, &last); | |
64874280 RW |
2058 | if (skb) |
2059 | break; | |
2060 | ||
6e1ce3c3 | 2061 | mutex_unlock(&u->iolock); |
64874280 RW |
2062 | |
2063 | if (err != -EAGAIN) | |
2064 | break; | |
2065 | } while (timeo && | |
2066 | !__skb_wait_for_more_packets(sk, &err, &timeo, last)); | |
2067 | ||
6e1ce3c3 | 2068 | if (!skb) { /* implies iolock unlocked */ |
0a112258 FZ |
2069 | unix_state_lock(sk); |
2070 | /* Signal EOF on disconnected non-blocking SEQPACKET socket. */ | |
2071 | if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN && | |
2072 | (sk->sk_shutdown & RCV_SHUTDOWN)) | |
2073 | err = 0; | |
2074 | unix_state_unlock(sk); | |
64874280 | 2075 | goto out; |
0a112258 | 2076 | } |
1da177e4 | 2077 | |
77b75f4d RW |
2078 | if (wq_has_sleeper(&u->peer_wait)) |
2079 | wake_up_interruptible_sync_poll(&u->peer_wait, | |
a9a08845 LT |
2080 | EPOLLOUT | EPOLLWRNORM | |
2081 | EPOLLWRBAND); | |
1da177e4 LT |
2082 | |
2083 | if (msg->msg_name) | |
2084 | unix_copy_addr(msg, skb->sk); | |
2085 | ||
f55bb7f9 PE |
2086 | if (size > skb->len - skip) |
2087 | size = skb->len - skip; | |
2088 | else if (size < skb->len - skip) | |
1da177e4 LT |
2089 | msg->msg_flags |= MSG_TRUNC; |
2090 | ||
51f3d02b | 2091 | err = skb_copy_datagram_msg(skb, skip, msg, size); |
1da177e4 LT |
2092 | if (err) |
2093 | goto out_free; | |
2094 | ||
3f66116e AC |
2095 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
2096 | __sock_recv_timestamp(msg, sk, skb); | |
2097 | ||
7cc05662 CH |
2098 | memset(&scm, 0, sizeof(scm)); |
2099 | ||
2100 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); | |
2101 | unix_set_secdata(&scm, skb); | |
1da177e4 | 2102 | |
6eba6a37 | 2103 | if (!(flags & MSG_PEEK)) { |
1da177e4 | 2104 | if (UNIXCB(skb).fp) |
7cc05662 | 2105 | unix_detach_fds(&scm, skb); |
f55bb7f9 PE |
2106 | |
2107 | sk_peek_offset_bwd(sk, skb->len); | |
6eba6a37 | 2108 | } else { |
1da177e4 LT |
2109 | /* It is questionable: on PEEK we could: |
2110 | - do not return fds - good, but too simple 8) | |
2111 | - return fds, and do not return them on read (old strategy, | |
2112 | apparently wrong) | |
2113 | - clone fds (I chose it for now, it is the most universal | |
2114 | solution) | |
ac7bfa62 YH |
2115 | |
2116 | POSIX 1003.1g does not actually define this clearly | |
2117 | at all. POSIX 1003.1g doesn't define a lot of things | |
2118 | clearly however! | |
2119 | ||
1da177e4 | 2120 | */ |
f55bb7f9 PE |
2121 | |
2122 | sk_peek_offset_fwd(sk, size); | |
2123 | ||
1da177e4 | 2124 | if (UNIXCB(skb).fp) |
7cc05662 | 2125 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 2126 | } |
9f6f9af7 | 2127 | err = (flags & MSG_TRUNC) ? skb->len - skip : size; |
1da177e4 | 2128 | |
7cc05662 | 2129 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
2130 | |
2131 | out_free: | |
6eba6a37 | 2132 | skb_free_datagram(sk, skb); |
6e1ce3c3 | 2133 | mutex_unlock(&u->iolock); |
1da177e4 LT |
2134 | out: |
2135 | return err; | |
2136 | } | |
2137 | ||
2138 | /* | |
79f632c7 | 2139 | * Sleep until more data has arrived. But check for races.. |
1da177e4 | 2140 | */ |
79f632c7 | 2141 | static long unix_stream_data_wait(struct sock *sk, long timeo, |
06a77b07 WC |
2142 | struct sk_buff *last, unsigned int last_len, |
2143 | bool freezable) | |
1da177e4 | 2144 | { |
2b514574 | 2145 | struct sk_buff *tail; |
1da177e4 LT |
2146 | DEFINE_WAIT(wait); |
2147 | ||
1c92b4e5 | 2148 | unix_state_lock(sk); |
1da177e4 LT |
2149 | |
2150 | for (;;) { | |
aa395145 | 2151 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 2152 | |
2b514574 HFS |
2153 | tail = skb_peek_tail(&sk->sk_receive_queue); |
2154 | if (tail != last || | |
2155 | (tail && tail->len != last_len) || | |
1da177e4 LT |
2156 | sk->sk_err || |
2157 | (sk->sk_shutdown & RCV_SHUTDOWN) || | |
2158 | signal_pending(current) || | |
2159 | !timeo) | |
2160 | break; | |
2161 | ||
9cd3e072 | 2162 | sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk); |
1c92b4e5 | 2163 | unix_state_unlock(sk); |
06a77b07 WC |
2164 | if (freezable) |
2165 | timeo = freezable_schedule_timeout(timeo); | |
2166 | else | |
2167 | timeo = schedule_timeout(timeo); | |
1c92b4e5 | 2168 | unix_state_lock(sk); |
b48732e4 MS |
2169 | |
2170 | if (sock_flag(sk, SOCK_DEAD)) | |
2171 | break; | |
2172 | ||
9cd3e072 | 2173 | sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk); |
1da177e4 LT |
2174 | } |
2175 | ||
aa395145 | 2176 | finish_wait(sk_sleep(sk), &wait); |
1c92b4e5 | 2177 | unix_state_unlock(sk); |
1da177e4 LT |
2178 | return timeo; |
2179 | } | |
2180 | ||
e370a723 ED |
2181 | static unsigned int unix_skb_len(const struct sk_buff *skb) |
2182 | { | |
2183 | return skb->len - UNIXCB(skb).consumed; | |
2184 | } | |
2185 | ||
2b514574 HFS |
2186 | struct unix_stream_read_state { |
2187 | int (*recv_actor)(struct sk_buff *, int, int, | |
2188 | struct unix_stream_read_state *); | |
2189 | struct socket *socket; | |
2190 | struct msghdr *msg; | |
2191 | struct pipe_inode_info *pipe; | |
2192 | size_t size; | |
2193 | int flags; | |
2194 | unsigned int splice_flags; | |
2195 | }; | |
2196 | ||
06a77b07 WC |
2197 | static int unix_stream_read_generic(struct unix_stream_read_state *state, |
2198 | bool freezable) | |
1da177e4 | 2199 | { |
7cc05662 | 2200 | struct scm_cookie scm; |
2b514574 | 2201 | struct socket *sock = state->socket; |
1da177e4 LT |
2202 | struct sock *sk = sock->sk; |
2203 | struct unix_sock *u = unix_sk(sk); | |
1da177e4 | 2204 | int copied = 0; |
2b514574 | 2205 | int flags = state->flags; |
de144391 | 2206 | int noblock = flags & MSG_DONTWAIT; |
2b514574 | 2207 | bool check_creds = false; |
1da177e4 LT |
2208 | int target; |
2209 | int err = 0; | |
2210 | long timeo; | |
fc0d7536 | 2211 | int skip; |
2b514574 HFS |
2212 | size_t size = state->size; |
2213 | unsigned int last_len; | |
1da177e4 | 2214 | |
1b92ee3d RW |
2215 | if (unlikely(sk->sk_state != TCP_ESTABLISHED)) { |
2216 | err = -EINVAL; | |
1da177e4 | 2217 | goto out; |
1b92ee3d | 2218 | } |
1da177e4 | 2219 | |
1b92ee3d RW |
2220 | if (unlikely(flags & MSG_OOB)) { |
2221 | err = -EOPNOTSUPP; | |
1da177e4 | 2222 | goto out; |
1b92ee3d | 2223 | } |
1da177e4 | 2224 | |
2b514574 | 2225 | target = sock_rcvlowat(sk, flags & MSG_WAITALL, size); |
de144391 | 2226 | timeo = sock_rcvtimeo(sk, noblock); |
1da177e4 | 2227 | |
2b514574 HFS |
2228 | memset(&scm, 0, sizeof(scm)); |
2229 | ||
1da177e4 LT |
2230 | /* Lock the socket to prevent queue disordering |
2231 | * while sleeps in memcpy_tomsg | |
2232 | */ | |
6e1ce3c3 | 2233 | mutex_lock(&u->iolock); |
1da177e4 | 2234 | |
a0917e0b | 2235 | skip = max(sk_peek_offset(sk, flags), 0); |
e9193d60 | 2236 | |
6eba6a37 | 2237 | do { |
1da177e4 | 2238 | int chunk; |
73ed5d25 | 2239 | bool drop_skb; |
79f632c7 | 2240 | struct sk_buff *skb, *last; |
1da177e4 | 2241 | |
18eceb81 | 2242 | redo: |
3c0d2f37 | 2243 | unix_state_lock(sk); |
b48732e4 MS |
2244 | if (sock_flag(sk, SOCK_DEAD)) { |
2245 | err = -ECONNRESET; | |
2246 | goto unlock; | |
2247 | } | |
79f632c7 | 2248 | last = skb = skb_peek(&sk->sk_receive_queue); |
2b514574 | 2249 | last_len = last ? last->len : 0; |
fc0d7536 | 2250 | again: |
6eba6a37 | 2251 | if (skb == NULL) { |
1da177e4 | 2252 | if (copied >= target) |
3c0d2f37 | 2253 | goto unlock; |
1da177e4 LT |
2254 | |
2255 | /* | |
2256 | * POSIX 1003.1g mandates this order. | |
2257 | */ | |
ac7bfa62 | 2258 | |
6eba6a37 ED |
2259 | err = sock_error(sk); |
2260 | if (err) | |
3c0d2f37 | 2261 | goto unlock; |
1da177e4 | 2262 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
3c0d2f37 MS |
2263 | goto unlock; |
2264 | ||
2265 | unix_state_unlock(sk); | |
1b92ee3d RW |
2266 | if (!timeo) { |
2267 | err = -EAGAIN; | |
1da177e4 | 2268 | break; |
1b92ee3d RW |
2269 | } |
2270 | ||
6e1ce3c3 | 2271 | mutex_unlock(&u->iolock); |
1da177e4 | 2272 | |
2b514574 | 2273 | timeo = unix_stream_data_wait(sk, timeo, last, |
06a77b07 | 2274 | last_len, freezable); |
1da177e4 | 2275 | |
3822b5c2 | 2276 | if (signal_pending(current)) { |
1da177e4 | 2277 | err = sock_intr_errno(timeo); |
fa0dc04d | 2278 | scm_destroy(&scm); |
1da177e4 LT |
2279 | goto out; |
2280 | } | |
b3ca9b02 | 2281 | |
6e1ce3c3 | 2282 | mutex_lock(&u->iolock); |
18eceb81 | 2283 | goto redo; |
2b514574 | 2284 | unlock: |
3c0d2f37 MS |
2285 | unix_state_unlock(sk); |
2286 | break; | |
1da177e4 | 2287 | } |
fc0d7536 | 2288 | |
e370a723 ED |
2289 | while (skip >= unix_skb_len(skb)) { |
2290 | skip -= unix_skb_len(skb); | |
79f632c7 | 2291 | last = skb; |
2b514574 | 2292 | last_len = skb->len; |
fc0d7536 | 2293 | skb = skb_peek_next(skb, &sk->sk_receive_queue); |
79f632c7 BP |
2294 | if (!skb) |
2295 | goto again; | |
fc0d7536 PE |
2296 | } |
2297 | ||
3c0d2f37 | 2298 | unix_state_unlock(sk); |
1da177e4 LT |
2299 | |
2300 | if (check_creds) { | |
2301 | /* Never glue messages from different writers */ | |
9490f886 | 2302 | if (!unix_skb_scm_eq(skb, &scm)) |
1da177e4 | 2303 | break; |
0e82e7f6 | 2304 | } else if (test_bit(SOCK_PASSCRED, &sock->flags)) { |
1da177e4 | 2305 | /* Copy credentials */ |
7cc05662 | 2306 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); |
37a9a8df | 2307 | unix_set_secdata(&scm, skb); |
2b514574 | 2308 | check_creds = true; |
1da177e4 LT |
2309 | } |
2310 | ||
2311 | /* Copy address just once */ | |
2b514574 HFS |
2312 | if (state->msg && state->msg->msg_name) { |
2313 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, | |
2314 | state->msg->msg_name); | |
2315 | unix_copy_addr(state->msg, skb->sk); | |
1da177e4 LT |
2316 | sunaddr = NULL; |
2317 | } | |
2318 | ||
e370a723 | 2319 | chunk = min_t(unsigned int, unix_skb_len(skb) - skip, size); |
73ed5d25 | 2320 | skb_get(skb); |
2b514574 | 2321 | chunk = state->recv_actor(skb, skip, chunk, state); |
73ed5d25 HFS |
2322 | drop_skb = !unix_skb_len(skb); |
2323 | /* skb is only safe to use if !drop_skb */ | |
2324 | consume_skb(skb); | |
2b514574 | 2325 | if (chunk < 0) { |
1da177e4 LT |
2326 | if (copied == 0) |
2327 | copied = -EFAULT; | |
2328 | break; | |
2329 | } | |
2330 | copied += chunk; | |
2331 | size -= chunk; | |
2332 | ||
73ed5d25 HFS |
2333 | if (drop_skb) { |
2334 | /* the skb was touched by a concurrent reader; | |
2335 | * we should not expect anything from this skb | |
2336 | * anymore and assume it invalid - we can be | |
2337 | * sure it was dropped from the socket queue | |
2338 | * | |
2339 | * let's report a short read | |
2340 | */ | |
2341 | err = 0; | |
2342 | break; | |
2343 | } | |
2344 | ||
1da177e4 | 2345 | /* Mark read part of skb as used */ |
6eba6a37 | 2346 | if (!(flags & MSG_PEEK)) { |
e370a723 | 2347 | UNIXCB(skb).consumed += chunk; |
1da177e4 | 2348 | |
fc0d7536 PE |
2349 | sk_peek_offset_bwd(sk, chunk); |
2350 | ||
1da177e4 | 2351 | if (UNIXCB(skb).fp) |
7cc05662 | 2352 | unix_detach_fds(&scm, skb); |
1da177e4 | 2353 | |
e370a723 | 2354 | if (unix_skb_len(skb)) |
1da177e4 | 2355 | break; |
1da177e4 | 2356 | |
6f01fd6e | 2357 | skb_unlink(skb, &sk->sk_receive_queue); |
70d4bf6d | 2358 | consume_skb(skb); |
1da177e4 | 2359 | |
7cc05662 | 2360 | if (scm.fp) |
1da177e4 | 2361 | break; |
6eba6a37 | 2362 | } else { |
1da177e4 LT |
2363 | /* It is questionable, see note in unix_dgram_recvmsg. |
2364 | */ | |
2365 | if (UNIXCB(skb).fp) | |
7cc05662 | 2366 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 2367 | |
e9193d60 | 2368 | sk_peek_offset_fwd(sk, chunk); |
fc0d7536 | 2369 | |
9f389e35 AC |
2370 | if (UNIXCB(skb).fp) |
2371 | break; | |
2372 | ||
e9193d60 | 2373 | skip = 0; |
9f389e35 AC |
2374 | last = skb; |
2375 | last_len = skb->len; | |
2376 | unix_state_lock(sk); | |
2377 | skb = skb_peek_next(skb, &sk->sk_receive_queue); | |
2378 | if (skb) | |
2379 | goto again; | |
2380 | unix_state_unlock(sk); | |
1da177e4 LT |
2381 | break; |
2382 | } | |
2383 | } while (size); | |
2384 | ||
6e1ce3c3 | 2385 | mutex_unlock(&u->iolock); |
2b514574 HFS |
2386 | if (state->msg) |
2387 | scm_recv(sock, state->msg, &scm, flags); | |
2388 | else | |
2389 | scm_destroy(&scm); | |
1da177e4 LT |
2390 | out: |
2391 | return copied ? : err; | |
2392 | } | |
2393 | ||
2b514574 HFS |
2394 | static int unix_stream_read_actor(struct sk_buff *skb, |
2395 | int skip, int chunk, | |
2396 | struct unix_stream_read_state *state) | |
2397 | { | |
2398 | int ret; | |
2399 | ||
2400 | ret = skb_copy_datagram_msg(skb, UNIXCB(skb).consumed + skip, | |
2401 | state->msg, chunk); | |
2402 | return ret ?: chunk; | |
2403 | } | |
2404 | ||
2405 | static int unix_stream_recvmsg(struct socket *sock, struct msghdr *msg, | |
2406 | size_t size, int flags) | |
2407 | { | |
2408 | struct unix_stream_read_state state = { | |
2409 | .recv_actor = unix_stream_read_actor, | |
2410 | .socket = sock, | |
2411 | .msg = msg, | |
2412 | .size = size, | |
2413 | .flags = flags | |
2414 | }; | |
2415 | ||
06a77b07 | 2416 | return unix_stream_read_generic(&state, true); |
2b514574 HFS |
2417 | } |
2418 | ||
2b514574 HFS |
2419 | static int unix_stream_splice_actor(struct sk_buff *skb, |
2420 | int skip, int chunk, | |
2421 | struct unix_stream_read_state *state) | |
2422 | { | |
2423 | return skb_splice_bits(skb, state->socket->sk, | |
2424 | UNIXCB(skb).consumed + skip, | |
25869262 | 2425 | state->pipe, chunk, state->splice_flags); |
2b514574 HFS |
2426 | } |
2427 | ||
2428 | static ssize_t unix_stream_splice_read(struct socket *sock, loff_t *ppos, | |
2429 | struct pipe_inode_info *pipe, | |
2430 | size_t size, unsigned int flags) | |
2431 | { | |
2432 | struct unix_stream_read_state state = { | |
2433 | .recv_actor = unix_stream_splice_actor, | |
2434 | .socket = sock, | |
2435 | .pipe = pipe, | |
2436 | .size = size, | |
2437 | .splice_flags = flags, | |
2438 | }; | |
2439 | ||
2440 | if (unlikely(*ppos)) | |
2441 | return -ESPIPE; | |
2442 | ||
2443 | if (sock->file->f_flags & O_NONBLOCK || | |
2444 | flags & SPLICE_F_NONBLOCK) | |
2445 | state.flags = MSG_DONTWAIT; | |
2446 | ||
06a77b07 | 2447 | return unix_stream_read_generic(&state, false); |
2b514574 HFS |
2448 | } |
2449 | ||
1da177e4 LT |
2450 | static int unix_shutdown(struct socket *sock, int mode) |
2451 | { | |
2452 | struct sock *sk = sock->sk; | |
2453 | struct sock *other; | |
2454 | ||
fc61b928 XW |
2455 | if (mode < SHUT_RD || mode > SHUT_RDWR) |
2456 | return -EINVAL; | |
2457 | /* This maps: | |
2458 | * SHUT_RD (0) -> RCV_SHUTDOWN (1) | |
2459 | * SHUT_WR (1) -> SEND_SHUTDOWN (2) | |
2460 | * SHUT_RDWR (2) -> SHUTDOWN_MASK (3) | |
2461 | */ | |
2462 | ++mode; | |
7180a031 AC |
2463 | |
2464 | unix_state_lock(sk); | |
2465 | sk->sk_shutdown |= mode; | |
2466 | other = unix_peer(sk); | |
2467 | if (other) | |
2468 | sock_hold(other); | |
2469 | unix_state_unlock(sk); | |
2470 | sk->sk_state_change(sk); | |
2471 | ||
2472 | if (other && | |
2473 | (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) { | |
2474 | ||
2475 | int peer_mode = 0; | |
2476 | ||
2477 | if (mode&RCV_SHUTDOWN) | |
2478 | peer_mode |= SEND_SHUTDOWN; | |
2479 | if (mode&SEND_SHUTDOWN) | |
2480 | peer_mode |= RCV_SHUTDOWN; | |
2481 | unix_state_lock(other); | |
2482 | other->sk_shutdown |= peer_mode; | |
2483 | unix_state_unlock(other); | |
2484 | other->sk_state_change(other); | |
2485 | if (peer_mode == SHUTDOWN_MASK) | |
2486 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP); | |
2487 | else if (peer_mode & RCV_SHUTDOWN) | |
2488 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN); | |
1da177e4 | 2489 | } |
7180a031 AC |
2490 | if (other) |
2491 | sock_put(other); | |
2492 | ||
1da177e4 LT |
2493 | return 0; |
2494 | } | |
2495 | ||
885ee74d PE |
2496 | long unix_inq_len(struct sock *sk) |
2497 | { | |
2498 | struct sk_buff *skb; | |
2499 | long amount = 0; | |
2500 | ||
2501 | if (sk->sk_state == TCP_LISTEN) | |
2502 | return -EINVAL; | |
2503 | ||
2504 | spin_lock(&sk->sk_receive_queue.lock); | |
2505 | if (sk->sk_type == SOCK_STREAM || | |
2506 | sk->sk_type == SOCK_SEQPACKET) { | |
2507 | skb_queue_walk(&sk->sk_receive_queue, skb) | |
e370a723 | 2508 | amount += unix_skb_len(skb); |
885ee74d PE |
2509 | } else { |
2510 | skb = skb_peek(&sk->sk_receive_queue); | |
2511 | if (skb) | |
2512 | amount = skb->len; | |
2513 | } | |
2514 | spin_unlock(&sk->sk_receive_queue.lock); | |
2515 | ||
2516 | return amount; | |
2517 | } | |
2518 | EXPORT_SYMBOL_GPL(unix_inq_len); | |
2519 | ||
2520 | long unix_outq_len(struct sock *sk) | |
2521 | { | |
2522 | return sk_wmem_alloc_get(sk); | |
2523 | } | |
2524 | EXPORT_SYMBOL_GPL(unix_outq_len); | |
2525 | ||
ba94f308 AV |
2526 | static int unix_open_file(struct sock *sk) |
2527 | { | |
2528 | struct path path; | |
2529 | struct file *f; | |
2530 | int fd; | |
2531 | ||
2532 | if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) | |
2533 | return -EPERM; | |
2534 | ||
ae3b5641 AV |
2535 | if (!smp_load_acquire(&unix_sk(sk)->addr)) |
2536 | return -ENOENT; | |
2537 | ||
ba94f308 | 2538 | path = unix_sk(sk)->path; |
ae3b5641 | 2539 | if (!path.dentry) |
ba94f308 | 2540 | return -ENOENT; |
ba94f308 AV |
2541 | |
2542 | path_get(&path); | |
ba94f308 AV |
2543 | |
2544 | fd = get_unused_fd_flags(O_CLOEXEC); | |
2545 | if (fd < 0) | |
2546 | goto out; | |
2547 | ||
2548 | f = dentry_open(&path, O_PATH, current_cred()); | |
2549 | if (IS_ERR(f)) { | |
2550 | put_unused_fd(fd); | |
2551 | fd = PTR_ERR(f); | |
2552 | goto out; | |
2553 | } | |
2554 | ||
2555 | fd_install(fd, f); | |
2556 | out: | |
2557 | path_put(&path); | |
2558 | ||
2559 | return fd; | |
2560 | } | |
2561 | ||
1da177e4 LT |
2562 | static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
2563 | { | |
2564 | struct sock *sk = sock->sk; | |
e27dfcea | 2565 | long amount = 0; |
1da177e4 LT |
2566 | int err; |
2567 | ||
6eba6a37 ED |
2568 | switch (cmd) { |
2569 | case SIOCOUTQ: | |
885ee74d | 2570 | amount = unix_outq_len(sk); |
6eba6a37 ED |
2571 | err = put_user(amount, (int __user *)arg); |
2572 | break; | |
2573 | case SIOCINQ: | |
885ee74d PE |
2574 | amount = unix_inq_len(sk); |
2575 | if (amount < 0) | |
2576 | err = amount; | |
2577 | else | |
1da177e4 | 2578 | err = put_user(amount, (int __user *)arg); |
885ee74d | 2579 | break; |
ba94f308 AV |
2580 | case SIOCUNIXFILE: |
2581 | err = unix_open_file(sk); | |
2582 | break; | |
6eba6a37 ED |
2583 | default: |
2584 | err = -ENOIOCTLCMD; | |
2585 | break; | |
1da177e4 LT |
2586 | } |
2587 | return err; | |
2588 | } | |
2589 | ||
a11e1d43 | 2590 | static __poll_t unix_poll(struct file *file, struct socket *sock, poll_table *wait) |
1da177e4 LT |
2591 | { |
2592 | struct sock *sk = sock->sk; | |
a11e1d43 LT |
2593 | __poll_t mask; |
2594 | ||
89ab066d | 2595 | sock_poll_wait(file, sock, wait); |
a11e1d43 | 2596 | mask = 0; |
1da177e4 LT |
2597 | |
2598 | /* exceptional events? */ | |
2599 | if (sk->sk_err) | |
a9a08845 | 2600 | mask |= EPOLLERR; |
1da177e4 | 2601 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
a9a08845 | 2602 | mask |= EPOLLHUP; |
f348d70a | 2603 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
a9a08845 | 2604 | mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; |
1da177e4 LT |
2605 | |
2606 | /* readable? */ | |
db40980f | 2607 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
a9a08845 | 2608 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 LT |
2609 | |
2610 | /* Connection-based need to check for termination and startup */ | |
6eba6a37 ED |
2611 | if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) && |
2612 | sk->sk_state == TCP_CLOSE) | |
a9a08845 | 2613 | mask |= EPOLLHUP; |
1da177e4 LT |
2614 | |
2615 | /* | |
2616 | * we set writable also when the other side has shut down the | |
2617 | * connection. This prevents stuck sockets. | |
2618 | */ | |
2619 | if (unix_writable(sk)) | |
a9a08845 | 2620 | mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; |
1da177e4 LT |
2621 | |
2622 | return mask; | |
2623 | } | |
2624 | ||
a11e1d43 LT |
2625 | static __poll_t unix_dgram_poll(struct file *file, struct socket *sock, |
2626 | poll_table *wait) | |
3c73419c | 2627 | { |
ec0d215f | 2628 | struct sock *sk = sock->sk, *other; |
a11e1d43 LT |
2629 | unsigned int writable; |
2630 | __poll_t mask; | |
2631 | ||
89ab066d | 2632 | sock_poll_wait(file, sock, wait); |
a11e1d43 | 2633 | mask = 0; |
3c73419c RW |
2634 | |
2635 | /* exceptional events? */ | |
2636 | if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue)) | |
a9a08845 LT |
2637 | mask |= EPOLLERR | |
2638 | (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0); | |
7d4c04fc | 2639 | |
3c73419c | 2640 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
a9a08845 | 2641 | mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM; |
3c73419c | 2642 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
a9a08845 | 2643 | mask |= EPOLLHUP; |
3c73419c RW |
2644 | |
2645 | /* readable? */ | |
5456f09a | 2646 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
a9a08845 | 2647 | mask |= EPOLLIN | EPOLLRDNORM; |
3c73419c RW |
2648 | |
2649 | /* Connection-based need to check for termination and startup */ | |
2650 | if (sk->sk_type == SOCK_SEQPACKET) { | |
2651 | if (sk->sk_state == TCP_CLOSE) | |
a9a08845 | 2652 | mask |= EPOLLHUP; |
3c73419c RW |
2653 | /* connection hasn't started yet? */ |
2654 | if (sk->sk_state == TCP_SYN_SENT) | |
2655 | return mask; | |
2656 | } | |
2657 | ||
973a34aa | 2658 | /* No write status requested, avoid expensive OUT tests. */ |
a11e1d43 | 2659 | if (!(poll_requested_events(wait) & (EPOLLWRBAND|EPOLLWRNORM|EPOLLOUT))) |
973a34aa ED |
2660 | return mask; |
2661 | ||
ec0d215f | 2662 | writable = unix_writable(sk); |
7d267278 RW |
2663 | if (writable) { |
2664 | unix_state_lock(sk); | |
2665 | ||
2666 | other = unix_peer(sk); | |
2667 | if (other && unix_peer(other) != sk && | |
2668 | unix_recvq_full(other) && | |
2669 | unix_dgram_peer_wake_me(sk, other)) | |
2670 | writable = 0; | |
2671 | ||
2672 | unix_state_unlock(sk); | |
ec0d215f RW |
2673 | } |
2674 | ||
2675 | if (writable) | |
a9a08845 | 2676 | mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND; |
3c73419c | 2677 | else |
9cd3e072 | 2678 | sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk); |
3c73419c | 2679 | |
3c73419c RW |
2680 | return mask; |
2681 | } | |
1da177e4 LT |
2682 | |
2683 | #ifdef CONFIG_PROC_FS | |
a53eb3fe | 2684 | |
7123aaa3 ED |
2685 | #define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1) |
2686 | ||
2687 | #define get_bucket(x) ((x) >> BUCKET_SPACE) | |
2688 | #define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1)) | |
2689 | #define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o)) | |
a53eb3fe | 2690 | |
7123aaa3 | 2691 | static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos) |
1da177e4 | 2692 | { |
7123aaa3 ED |
2693 | unsigned long offset = get_offset(*pos); |
2694 | unsigned long bucket = get_bucket(*pos); | |
2695 | struct sock *sk; | |
2696 | unsigned long count = 0; | |
1da177e4 | 2697 | |
7123aaa3 ED |
2698 | for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) { |
2699 | if (sock_net(sk) != seq_file_net(seq)) | |
097e66c5 | 2700 | continue; |
7123aaa3 ED |
2701 | if (++count == offset) |
2702 | break; | |
2703 | } | |
2704 | ||
2705 | return sk; | |
2706 | } | |
2707 | ||
2708 | static struct sock *unix_next_socket(struct seq_file *seq, | |
2709 | struct sock *sk, | |
2710 | loff_t *pos) | |
2711 | { | |
2712 | unsigned long bucket; | |
2713 | ||
2714 | while (sk > (struct sock *)SEQ_START_TOKEN) { | |
2715 | sk = sk_next(sk); | |
2716 | if (!sk) | |
2717 | goto next_bucket; | |
2718 | if (sock_net(sk) == seq_file_net(seq)) | |
2719 | return sk; | |
1da177e4 | 2720 | } |
7123aaa3 ED |
2721 | |
2722 | do { | |
2723 | sk = unix_from_bucket(seq, pos); | |
2724 | if (sk) | |
2725 | return sk; | |
2726 | ||
2727 | next_bucket: | |
2728 | bucket = get_bucket(*pos) + 1; | |
2729 | *pos = set_bucket_offset(bucket, 1); | |
2730 | } while (bucket < ARRAY_SIZE(unix_socket_table)); | |
2731 | ||
1da177e4 LT |
2732 | return NULL; |
2733 | } | |
2734 | ||
1da177e4 | 2735 | static void *unix_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 2736 | __acquires(unix_table_lock) |
1da177e4 | 2737 | { |
fbe9cc4a | 2738 | spin_lock(&unix_table_lock); |
7123aaa3 ED |
2739 | |
2740 | if (!*pos) | |
2741 | return SEQ_START_TOKEN; | |
2742 | ||
2743 | if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table)) | |
2744 | return NULL; | |
2745 | ||
2746 | return unix_next_socket(seq, NULL, pos); | |
1da177e4 LT |
2747 | } |
2748 | ||
2749 | static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2750 | { | |
2751 | ++*pos; | |
7123aaa3 | 2752 | return unix_next_socket(seq, v, pos); |
1da177e4 LT |
2753 | } |
2754 | ||
2755 | static void unix_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2756 | __releases(unix_table_lock) |
1da177e4 | 2757 | { |
fbe9cc4a | 2758 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
2759 | } |
2760 | ||
2761 | static int unix_seq_show(struct seq_file *seq, void *v) | |
2762 | { | |
ac7bfa62 | 2763 | |
b9f3124f | 2764 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2765 | seq_puts(seq, "Num RefCount Protocol Flags Type St " |
2766 | "Inode Path\n"); | |
2767 | else { | |
2768 | struct sock *s = v; | |
2769 | struct unix_sock *u = unix_sk(s); | |
1c92b4e5 | 2770 | unix_state_lock(s); |
1da177e4 | 2771 | |
71338aa7 | 2772 | seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu", |
1da177e4 | 2773 | s, |
41c6d650 | 2774 | refcount_read(&s->sk_refcnt), |
1da177e4 LT |
2775 | 0, |
2776 | s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0, | |
2777 | s->sk_type, | |
2778 | s->sk_socket ? | |
2779 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) : | |
2780 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING), | |
2781 | sock_i_ino(s)); | |
2782 | ||
ae3b5641 | 2783 | if (u->addr) { // under unix_table_lock here |
1da177e4 LT |
2784 | int i, len; |
2785 | seq_putc(seq, ' '); | |
2786 | ||
2787 | i = 0; | |
2788 | len = u->addr->len - sizeof(short); | |
2789 | if (!UNIX_ABSTRACT(s)) | |
2790 | len--; | |
2791 | else { | |
2792 | seq_putc(seq, '@'); | |
2793 | i++; | |
2794 | } | |
2795 | for ( ; i < len; i++) | |
e7947ea7 IB |
2796 | seq_putc(seq, u->addr->name->sun_path[i] ?: |
2797 | '@'); | |
1da177e4 | 2798 | } |
1c92b4e5 | 2799 | unix_state_unlock(s); |
1da177e4 LT |
2800 | seq_putc(seq, '\n'); |
2801 | } | |
2802 | ||
2803 | return 0; | |
2804 | } | |
2805 | ||
56b3d975 | 2806 | static const struct seq_operations unix_seq_ops = { |
1da177e4 LT |
2807 | .start = unix_seq_start, |
2808 | .next = unix_seq_next, | |
2809 | .stop = unix_seq_stop, | |
2810 | .show = unix_seq_show, | |
2811 | }; | |
1da177e4 LT |
2812 | #endif |
2813 | ||
ec1b4cf7 | 2814 | static const struct net_proto_family unix_family_ops = { |
1da177e4 LT |
2815 | .family = PF_UNIX, |
2816 | .create = unix_create, | |
2817 | .owner = THIS_MODULE, | |
2818 | }; | |
2819 | ||
097e66c5 | 2820 | |
2c8c1e72 | 2821 | static int __net_init unix_net_init(struct net *net) |
097e66c5 DL |
2822 | { |
2823 | int error = -ENOMEM; | |
2824 | ||
a0a53c8b | 2825 | net->unx.sysctl_max_dgram_qlen = 10; |
1597fbc0 PE |
2826 | if (unix_sysctl_register(net)) |
2827 | goto out; | |
d392e497 | 2828 | |
097e66c5 | 2829 | #ifdef CONFIG_PROC_FS |
c3506372 CH |
2830 | if (!proc_create_net("unix", 0, net->proc_net, &unix_seq_ops, |
2831 | sizeof(struct seq_net_private))) { | |
1597fbc0 | 2832 | unix_sysctl_unregister(net); |
097e66c5 | 2833 | goto out; |
1597fbc0 | 2834 | } |
097e66c5 DL |
2835 | #endif |
2836 | error = 0; | |
2837 | out: | |
48dcc33e | 2838 | return error; |
097e66c5 DL |
2839 | } |
2840 | ||
2c8c1e72 | 2841 | static void __net_exit unix_net_exit(struct net *net) |
097e66c5 | 2842 | { |
1597fbc0 | 2843 | unix_sysctl_unregister(net); |
ece31ffd | 2844 | remove_proc_entry("unix", net->proc_net); |
097e66c5 DL |
2845 | } |
2846 | ||
2847 | static struct pernet_operations unix_net_ops = { | |
2848 | .init = unix_net_init, | |
2849 | .exit = unix_net_exit, | |
2850 | }; | |
2851 | ||
1da177e4 LT |
2852 | static int __init af_unix_init(void) |
2853 | { | |
2854 | int rc = -1; | |
1da177e4 | 2855 | |
b4fff5f8 | 2856 | BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 LT |
2857 | |
2858 | rc = proto_register(&unix_proto, 1); | |
ac7bfa62 | 2859 | if (rc != 0) { |
5cc208be | 2860 | pr_crit("%s: Cannot create unix_sock SLAB cache!\n", __func__); |
1da177e4 LT |
2861 | goto out; |
2862 | } | |
2863 | ||
2864 | sock_register(&unix_family_ops); | |
097e66c5 | 2865 | register_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2866 | out: |
2867 | return rc; | |
2868 | } | |
2869 | ||
2870 | static void __exit af_unix_exit(void) | |
2871 | { | |
2872 | sock_unregister(PF_UNIX); | |
1da177e4 | 2873 | proto_unregister(&unix_proto); |
097e66c5 | 2874 | unregister_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2875 | } |
2876 | ||
3d366960 DW |
2877 | /* Earlier than device_initcall() so that other drivers invoking |
2878 | request_module() don't end up in a loop when modprobe tries | |
2879 | to use a UNIX socket. But later than subsys_initcall() because | |
2880 | we depend on stuff initialised there */ | |
2881 | fs_initcall(af_unix_init); | |
1da177e4 LT |
2882 | module_exit(af_unix_exit); |
2883 | ||
2884 | MODULE_LICENSE("GPL"); | |
2885 | MODULE_ALIAS_NETPROTO(PF_UNIX); |