]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NET4: Implementation of BSD Unix domain sockets. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
5 | * |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License | |
8 | * as published by the Free Software Foundation; either version | |
9 | * 2 of the License, or (at your option) any later version. | |
10 | * | |
1da177e4 LT |
11 | * Fixes: |
12 | * Linus Torvalds : Assorted bug cures. | |
13 | * Niibe Yutaka : async I/O support. | |
14 | * Carsten Paeth : PF_UNIX check, address fixes. | |
15 | * Alan Cox : Limit size of allocated blocks. | |
16 | * Alan Cox : Fixed the stupid socketpair bug. | |
17 | * Alan Cox : BSD compatibility fine tuning. | |
18 | * Alan Cox : Fixed a bug in connect when interrupted. | |
19 | * Alan Cox : Sorted out a proper draft version of | |
20 | * file descriptor passing hacked up from | |
21 | * Mike Shaver's work. | |
22 | * Marty Leisner : Fixes to fd passing | |
23 | * Nick Nevin : recvmsg bugfix. | |
24 | * Alan Cox : Started proper garbage collector | |
25 | * Heiko EiBfeldt : Missing verify_area check | |
26 | * Alan Cox : Started POSIXisms | |
27 | * Andreas Schwab : Replace inode by dentry for proper | |
28 | * reference counting | |
29 | * Kirk Petersen : Made this a module | |
30 | * Christoph Rohland : Elegant non-blocking accept/connect algorithm. | |
31 | * Lots of bug fixes. | |
32 | * Alexey Kuznetosv : Repaired (I hope) bugs introduces | |
33 | * by above two patches. | |
34 | * Andrea Arcangeli : If possible we block in connect(2) | |
35 | * if the max backlog of the listen socket | |
36 | * is been reached. This won't break | |
37 | * old apps and it will avoid huge amount | |
38 | * of socks hashed (this for unix_gc() | |
39 | * performances reasons). | |
40 | * Security fix that limits the max | |
41 | * number of socks to 2*max_files and | |
42 | * the number of skb queueable in the | |
43 | * dgram receiver. | |
44 | * Artur Skawina : Hash function optimizations | |
45 | * Alexey Kuznetsov : Full scale SMP. Lot of bugs are introduced 8) | |
46 | * Malcolm Beattie : Set peercred for socketpair | |
47 | * Michal Ostrowski : Module initialization cleanup. | |
48 | * Arnaldo C. Melo : Remove MOD_{INC,DEC}_USE_COUNT, | |
49 | * the core infrastructure is doing that | |
50 | * for all net proto families now (2.5.69+) | |
51 | * | |
52 | * | |
53 | * Known differences from reference BSD that was tested: | |
54 | * | |
55 | * [TO FIX] | |
56 | * ECONNREFUSED is not returned from one end of a connected() socket to the | |
57 | * other the moment one end closes. | |
58 | * fstat() doesn't return st_dev=0, and give the blksize as high water mark | |
59 | * and a fake inode identifier (nor the BSD first socket fstat twice bug). | |
60 | * [NOT TO FIX] | |
61 | * accept() returns a path name even if the connecting socket has closed | |
62 | * in the meantime (BSD loses the path and gives up). | |
63 | * accept() returns 0 length path for an unbound connector. BSD returns 16 | |
64 | * and a null first byte in the path (but not for gethost/peername - BSD bug ??) | |
65 | * socketpair(...SOCK_RAW..) doesn't panic the kernel. | |
66 | * BSD af_unix apparently has connect forgetting to block properly. | |
67 | * (need to check this with the POSIX spec in detail) | |
68 | * | |
69 | * Differences from 2.0.0-11-... (ANK) | |
70 | * Bug fixes and improvements. | |
71 | * - client shutdown killed server socket. | |
72 | * - removed all useless cli/sti pairs. | |
73 | * | |
74 | * Semantic changes/extensions. | |
75 | * - generic control message passing. | |
76 | * - SCM_CREDENTIALS control message. | |
77 | * - "Abstract" (not FS based) socket bindings. | |
78 | * Abstract names are sequences of bytes (not zero terminated) | |
79 | * started by 0, so that this name space does not intersect | |
80 | * with BSD names. | |
81 | */ | |
82 | ||
5cc208be | 83 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
84 | ||
1da177e4 | 85 | #include <linux/module.h> |
1da177e4 | 86 | #include <linux/kernel.h> |
1da177e4 LT |
87 | #include <linux/signal.h> |
88 | #include <linux/sched.h> | |
89 | #include <linux/errno.h> | |
90 | #include <linux/string.h> | |
91 | #include <linux/stat.h> | |
92 | #include <linux/dcache.h> | |
93 | #include <linux/namei.h> | |
94 | #include <linux/socket.h> | |
95 | #include <linux/un.h> | |
96 | #include <linux/fcntl.h> | |
97 | #include <linux/termios.h> | |
98 | #include <linux/sockios.h> | |
99 | #include <linux/net.h> | |
100 | #include <linux/in.h> | |
101 | #include <linux/fs.h> | |
102 | #include <linux/slab.h> | |
103 | #include <asm/uaccess.h> | |
104 | #include <linux/skbuff.h> | |
105 | #include <linux/netdevice.h> | |
457c4cbc | 106 | #include <net/net_namespace.h> |
1da177e4 | 107 | #include <net/sock.h> |
c752f073 | 108 | #include <net/tcp_states.h> |
1da177e4 LT |
109 | #include <net/af_unix.h> |
110 | #include <linux/proc_fs.h> | |
111 | #include <linux/seq_file.h> | |
112 | #include <net/scm.h> | |
113 | #include <linux/init.h> | |
114 | #include <linux/poll.h> | |
1da177e4 LT |
115 | #include <linux/rtnetlink.h> |
116 | #include <linux/mount.h> | |
117 | #include <net/checksum.h> | |
118 | #include <linux/security.h> | |
2b15af6f | 119 | #include <linux/freezer.h> |
1da177e4 | 120 | |
7123aaa3 | 121 | struct hlist_head unix_socket_table[2 * UNIX_HASH_SIZE]; |
fa7ff56f PE |
122 | EXPORT_SYMBOL_GPL(unix_socket_table); |
123 | DEFINE_SPINLOCK(unix_table_lock); | |
124 | EXPORT_SYMBOL_GPL(unix_table_lock); | |
518de9b3 | 125 | static atomic_long_t unix_nr_socks; |
1da177e4 | 126 | |
1da177e4 | 127 | |
7123aaa3 ED |
128 | static struct hlist_head *unix_sockets_unbound(void *addr) |
129 | { | |
130 | unsigned long hash = (unsigned long)addr; | |
131 | ||
132 | hash ^= hash >> 16; | |
133 | hash ^= hash >> 8; | |
134 | hash %= UNIX_HASH_SIZE; | |
135 | return &unix_socket_table[UNIX_HASH_SIZE + hash]; | |
136 | } | |
137 | ||
138 | #define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash < UNIX_HASH_SIZE) | |
1da177e4 | 139 | |
877ce7c1 | 140 | #ifdef CONFIG_SECURITY_NETWORK |
dc49c1f9 | 141 | static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 | 142 | { |
37a9a8df | 143 | UNIXCB(skb).secid = scm->secid; |
877ce7c1 CZ |
144 | } |
145 | ||
146 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
147 | { | |
37a9a8df SS |
148 | scm->secid = UNIXCB(skb).secid; |
149 | } | |
150 | ||
151 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
152 | { | |
153 | return (scm->secid == UNIXCB(skb).secid); | |
877ce7c1 CZ |
154 | } |
155 | #else | |
dc49c1f9 | 156 | static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb) |
877ce7c1 CZ |
157 | { } |
158 | ||
159 | static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb) | |
160 | { } | |
37a9a8df SS |
161 | |
162 | static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb) | |
163 | { | |
164 | return true; | |
165 | } | |
877ce7c1 CZ |
166 | #endif /* CONFIG_SECURITY_NETWORK */ |
167 | ||
1da177e4 LT |
168 | /* |
169 | * SMP locking strategy: | |
fbe9cc4a | 170 | * hash table is protected with spinlock unix_table_lock |
663717f6 | 171 | * each socket state is protected by separate spin lock. |
1da177e4 LT |
172 | */ |
173 | ||
95c96174 | 174 | static inline unsigned int unix_hash_fold(__wsum n) |
1da177e4 | 175 | { |
0a13404d | 176 | unsigned int hash = (__force unsigned int)csum_fold(n); |
95c96174 | 177 | |
1da177e4 LT |
178 | hash ^= hash>>8; |
179 | return hash&(UNIX_HASH_SIZE-1); | |
180 | } | |
181 | ||
182 | #define unix_peer(sk) (unix_sk(sk)->peer) | |
183 | ||
184 | static inline int unix_our_peer(struct sock *sk, struct sock *osk) | |
185 | { | |
186 | return unix_peer(osk) == sk; | |
187 | } | |
188 | ||
189 | static inline int unix_may_send(struct sock *sk, struct sock *osk) | |
190 | { | |
6eba6a37 | 191 | return unix_peer(osk) == NULL || unix_our_peer(sk, osk); |
1da177e4 LT |
192 | } |
193 | ||
3c73419c RW |
194 | static inline int unix_recvq_full(struct sock const *sk) |
195 | { | |
196 | return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog; | |
197 | } | |
198 | ||
fa7ff56f | 199 | struct sock *unix_peer_get(struct sock *s) |
1da177e4 LT |
200 | { |
201 | struct sock *peer; | |
202 | ||
1c92b4e5 | 203 | unix_state_lock(s); |
1da177e4 LT |
204 | peer = unix_peer(s); |
205 | if (peer) | |
206 | sock_hold(peer); | |
1c92b4e5 | 207 | unix_state_unlock(s); |
1da177e4 LT |
208 | return peer; |
209 | } | |
fa7ff56f | 210 | EXPORT_SYMBOL_GPL(unix_peer_get); |
1da177e4 LT |
211 | |
212 | static inline void unix_release_addr(struct unix_address *addr) | |
213 | { | |
214 | if (atomic_dec_and_test(&addr->refcnt)) | |
215 | kfree(addr); | |
216 | } | |
217 | ||
218 | /* | |
219 | * Check unix socket name: | |
220 | * - should be not zero length. | |
221 | * - if started by not zero, should be NULL terminated (FS object) | |
222 | * - if started by zero, it is abstract name. | |
223 | */ | |
ac7bfa62 | 224 | |
95c96174 | 225 | static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp) |
1da177e4 LT |
226 | { |
227 | if (len <= sizeof(short) || len > sizeof(*sunaddr)) | |
228 | return -EINVAL; | |
229 | if (!sunaddr || sunaddr->sun_family != AF_UNIX) | |
230 | return -EINVAL; | |
231 | if (sunaddr->sun_path[0]) { | |
232 | /* | |
233 | * This may look like an off by one error but it is a bit more | |
234 | * subtle. 108 is the longest valid AF_UNIX path for a binding. | |
25985edc | 235 | * sun_path[108] doesn't as such exist. However in kernel space |
1da177e4 LT |
236 | * we are guaranteed that it is a valid memory location in our |
237 | * kernel address buffer. | |
238 | */ | |
e27dfcea | 239 | ((char *)sunaddr)[len] = 0; |
1da177e4 LT |
240 | len = strlen(sunaddr->sun_path)+1+sizeof(short); |
241 | return len; | |
242 | } | |
243 | ||
07f0757a | 244 | *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0)); |
1da177e4 LT |
245 | return len; |
246 | } | |
247 | ||
248 | static void __unix_remove_socket(struct sock *sk) | |
249 | { | |
250 | sk_del_node_init(sk); | |
251 | } | |
252 | ||
253 | static void __unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
254 | { | |
547b792c | 255 | WARN_ON(!sk_unhashed(sk)); |
1da177e4 LT |
256 | sk_add_node(sk, list); |
257 | } | |
258 | ||
259 | static inline void unix_remove_socket(struct sock *sk) | |
260 | { | |
fbe9cc4a | 261 | spin_lock(&unix_table_lock); |
1da177e4 | 262 | __unix_remove_socket(sk); |
fbe9cc4a | 263 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
264 | } |
265 | ||
266 | static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk) | |
267 | { | |
fbe9cc4a | 268 | spin_lock(&unix_table_lock); |
1da177e4 | 269 | __unix_insert_socket(list, sk); |
fbe9cc4a | 270 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
271 | } |
272 | ||
097e66c5 DL |
273 | static struct sock *__unix_find_socket_byname(struct net *net, |
274 | struct sockaddr_un *sunname, | |
95c96174 | 275 | int len, int type, unsigned int hash) |
1da177e4 LT |
276 | { |
277 | struct sock *s; | |
1da177e4 | 278 | |
b67bfe0d | 279 | sk_for_each(s, &unix_socket_table[hash ^ type]) { |
1da177e4 LT |
280 | struct unix_sock *u = unix_sk(s); |
281 | ||
878628fb | 282 | if (!net_eq(sock_net(s), net)) |
097e66c5 DL |
283 | continue; |
284 | ||
1da177e4 LT |
285 | if (u->addr->len == len && |
286 | !memcmp(u->addr->name, sunname, len)) | |
287 | goto found; | |
288 | } | |
289 | s = NULL; | |
290 | found: | |
291 | return s; | |
292 | } | |
293 | ||
097e66c5 DL |
294 | static inline struct sock *unix_find_socket_byname(struct net *net, |
295 | struct sockaddr_un *sunname, | |
1da177e4 | 296 | int len, int type, |
95c96174 | 297 | unsigned int hash) |
1da177e4 LT |
298 | { |
299 | struct sock *s; | |
300 | ||
fbe9cc4a | 301 | spin_lock(&unix_table_lock); |
097e66c5 | 302 | s = __unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
303 | if (s) |
304 | sock_hold(s); | |
fbe9cc4a | 305 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
306 | return s; |
307 | } | |
308 | ||
6616f788 | 309 | static struct sock *unix_find_socket_byinode(struct inode *i) |
1da177e4 LT |
310 | { |
311 | struct sock *s; | |
1da177e4 | 312 | |
fbe9cc4a | 313 | spin_lock(&unix_table_lock); |
b67bfe0d | 314 | sk_for_each(s, |
1da177e4 | 315 | &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) { |
40ffe67d | 316 | struct dentry *dentry = unix_sk(s)->path.dentry; |
1da177e4 | 317 | |
a25b376b | 318 | if (dentry && d_backing_inode(dentry) == i) { |
1da177e4 LT |
319 | sock_hold(s); |
320 | goto found; | |
321 | } | |
322 | } | |
323 | s = NULL; | |
324 | found: | |
fbe9cc4a | 325 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
326 | return s; |
327 | } | |
328 | ||
1586a587 | 329 | static int unix_writable(const struct sock *sk) |
1da177e4 | 330 | { |
1586a587 ED |
331 | return sk->sk_state != TCP_LISTEN && |
332 | (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf; | |
1da177e4 LT |
333 | } |
334 | ||
335 | static void unix_write_space(struct sock *sk) | |
336 | { | |
43815482 ED |
337 | struct socket_wq *wq; |
338 | ||
339 | rcu_read_lock(); | |
1da177e4 | 340 | if (unix_writable(sk)) { |
43815482 ED |
341 | wq = rcu_dereference(sk->sk_wq); |
342 | if (wq_has_sleeper(wq)) | |
67426b75 ED |
343 | wake_up_interruptible_sync_poll(&wq->wait, |
344 | POLLOUT | POLLWRNORM | POLLWRBAND); | |
8d8ad9d7 | 345 | sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT); |
1da177e4 | 346 | } |
43815482 | 347 | rcu_read_unlock(); |
1da177e4 LT |
348 | } |
349 | ||
350 | /* When dgram socket disconnects (or changes its peer), we clear its receive | |
351 | * queue of packets arrived from previous peer. First, it allows to do | |
352 | * flow control based only on wmem_alloc; second, sk connected to peer | |
353 | * may receive messages only from that peer. */ | |
354 | static void unix_dgram_disconnected(struct sock *sk, struct sock *other) | |
355 | { | |
b03efcfb | 356 | if (!skb_queue_empty(&sk->sk_receive_queue)) { |
1da177e4 LT |
357 | skb_queue_purge(&sk->sk_receive_queue); |
358 | wake_up_interruptible_all(&unix_sk(sk)->peer_wait); | |
359 | ||
360 | /* If one link of bidirectional dgram pipe is disconnected, | |
361 | * we signal error. Messages are lost. Do not make this, | |
362 | * when peer was not connected to us. | |
363 | */ | |
364 | if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) { | |
365 | other->sk_err = ECONNRESET; | |
366 | other->sk_error_report(other); | |
367 | } | |
368 | } | |
369 | } | |
370 | ||
371 | static void unix_sock_destructor(struct sock *sk) | |
372 | { | |
373 | struct unix_sock *u = unix_sk(sk); | |
374 | ||
375 | skb_queue_purge(&sk->sk_receive_queue); | |
376 | ||
547b792c IJ |
377 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); |
378 | WARN_ON(!sk_unhashed(sk)); | |
379 | WARN_ON(sk->sk_socket); | |
1da177e4 | 380 | if (!sock_flag(sk, SOCK_DEAD)) { |
5cc208be | 381 | pr_info("Attempt to release alive unix socket: %p\n", sk); |
1da177e4 LT |
382 | return; |
383 | } | |
384 | ||
385 | if (u->addr) | |
386 | unix_release_addr(u->addr); | |
387 | ||
518de9b3 | 388 | atomic_long_dec(&unix_nr_socks); |
6f756a8c | 389 | local_bh_disable(); |
a8076d8d | 390 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
6f756a8c | 391 | local_bh_enable(); |
1da177e4 | 392 | #ifdef UNIX_REFCNT_DEBUG |
5cc208be | 393 | pr_debug("UNIX %p is destroyed, %ld are still alive.\n", sk, |
518de9b3 | 394 | atomic_long_read(&unix_nr_socks)); |
1da177e4 LT |
395 | #endif |
396 | } | |
397 | ||
ded34e0f | 398 | static void unix_release_sock(struct sock *sk, int embrion) |
1da177e4 LT |
399 | { |
400 | struct unix_sock *u = unix_sk(sk); | |
40ffe67d | 401 | struct path path; |
1da177e4 LT |
402 | struct sock *skpair; |
403 | struct sk_buff *skb; | |
404 | int state; | |
405 | ||
406 | unix_remove_socket(sk); | |
407 | ||
408 | /* Clear state */ | |
1c92b4e5 | 409 | unix_state_lock(sk); |
1da177e4 LT |
410 | sock_orphan(sk); |
411 | sk->sk_shutdown = SHUTDOWN_MASK; | |
40ffe67d AV |
412 | path = u->path; |
413 | u->path.dentry = NULL; | |
414 | u->path.mnt = NULL; | |
1da177e4 LT |
415 | state = sk->sk_state; |
416 | sk->sk_state = TCP_CLOSE; | |
1c92b4e5 | 417 | unix_state_unlock(sk); |
1da177e4 LT |
418 | |
419 | wake_up_interruptible_all(&u->peer_wait); | |
420 | ||
e27dfcea | 421 | skpair = unix_peer(sk); |
1da177e4 | 422 | |
e27dfcea | 423 | if (skpair != NULL) { |
1da177e4 | 424 | if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) { |
1c92b4e5 | 425 | unix_state_lock(skpair); |
1da177e4 LT |
426 | /* No more writes */ |
427 | skpair->sk_shutdown = SHUTDOWN_MASK; | |
428 | if (!skb_queue_empty(&sk->sk_receive_queue) || embrion) | |
429 | skpair->sk_err = ECONNRESET; | |
1c92b4e5 | 430 | unix_state_unlock(skpair); |
1da177e4 | 431 | skpair->sk_state_change(skpair); |
8d8ad9d7 | 432 | sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP); |
1da177e4 LT |
433 | } |
434 | sock_put(skpair); /* It may now die */ | |
435 | unix_peer(sk) = NULL; | |
436 | } | |
437 | ||
438 | /* Try to flush out this socket. Throw out buffers at least */ | |
439 | ||
440 | while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) { | |
e27dfcea | 441 | if (state == TCP_LISTEN) |
1da177e4 LT |
442 | unix_release_sock(skb->sk, 1); |
443 | /* passed fds are erased in the kfree_skb hook */ | |
73ed5d25 | 444 | UNIXCB(skb).consumed = skb->len; |
1da177e4 LT |
445 | kfree_skb(skb); |
446 | } | |
447 | ||
40ffe67d AV |
448 | if (path.dentry) |
449 | path_put(&path); | |
1da177e4 LT |
450 | |
451 | sock_put(sk); | |
452 | ||
453 | /* ---- Socket is dead now and most probably destroyed ---- */ | |
454 | ||
455 | /* | |
e04dae84 | 456 | * Fixme: BSD difference: In BSD all sockets connected to us get |
1da177e4 LT |
457 | * ECONNRESET and we die on the spot. In Linux we behave |
458 | * like files and pipes do and wait for the last | |
459 | * dereference. | |
460 | * | |
461 | * Can't we simply set sock->err? | |
462 | * | |
463 | * What the above comment does talk about? --ANK(980817) | |
464 | */ | |
465 | ||
9305cfa4 | 466 | if (unix_tot_inflight) |
ac7bfa62 | 467 | unix_gc(); /* Garbage collect fds */ |
1da177e4 LT |
468 | } |
469 | ||
109f6e39 EB |
470 | static void init_peercred(struct sock *sk) |
471 | { | |
472 | put_pid(sk->sk_peer_pid); | |
473 | if (sk->sk_peer_cred) | |
474 | put_cred(sk->sk_peer_cred); | |
475 | sk->sk_peer_pid = get_pid(task_tgid(current)); | |
476 | sk->sk_peer_cred = get_current_cred(); | |
477 | } | |
478 | ||
479 | static void copy_peercred(struct sock *sk, struct sock *peersk) | |
480 | { | |
481 | put_pid(sk->sk_peer_pid); | |
482 | if (sk->sk_peer_cred) | |
483 | put_cred(sk->sk_peer_cred); | |
484 | sk->sk_peer_pid = get_pid(peersk->sk_peer_pid); | |
485 | sk->sk_peer_cred = get_cred(peersk->sk_peer_cred); | |
486 | } | |
487 | ||
1da177e4 LT |
488 | static int unix_listen(struct socket *sock, int backlog) |
489 | { | |
490 | int err; | |
491 | struct sock *sk = sock->sk; | |
492 | struct unix_sock *u = unix_sk(sk); | |
109f6e39 | 493 | struct pid *old_pid = NULL; |
1da177e4 LT |
494 | |
495 | err = -EOPNOTSUPP; | |
6eba6a37 ED |
496 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
497 | goto out; /* Only stream/seqpacket sockets accept */ | |
1da177e4 LT |
498 | err = -EINVAL; |
499 | if (!u->addr) | |
6eba6a37 | 500 | goto out; /* No listens on an unbound socket */ |
1c92b4e5 | 501 | unix_state_lock(sk); |
1da177e4 LT |
502 | if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN) |
503 | goto out_unlock; | |
504 | if (backlog > sk->sk_max_ack_backlog) | |
505 | wake_up_interruptible_all(&u->peer_wait); | |
506 | sk->sk_max_ack_backlog = backlog; | |
507 | sk->sk_state = TCP_LISTEN; | |
508 | /* set credentials so connect can copy them */ | |
109f6e39 | 509 | init_peercred(sk); |
1da177e4 LT |
510 | err = 0; |
511 | ||
512 | out_unlock: | |
1c92b4e5 | 513 | unix_state_unlock(sk); |
109f6e39 | 514 | put_pid(old_pid); |
1da177e4 LT |
515 | out: |
516 | return err; | |
517 | } | |
518 | ||
519 | static int unix_release(struct socket *); | |
520 | static int unix_bind(struct socket *, struct sockaddr *, int); | |
521 | static int unix_stream_connect(struct socket *, struct sockaddr *, | |
522 | int addr_len, int flags); | |
523 | static int unix_socketpair(struct socket *, struct socket *); | |
524 | static int unix_accept(struct socket *, struct socket *, int); | |
525 | static int unix_getname(struct socket *, struct sockaddr *, int *, int); | |
526 | static unsigned int unix_poll(struct file *, struct socket *, poll_table *); | |
ec0d215f RW |
527 | static unsigned int unix_dgram_poll(struct file *, struct socket *, |
528 | poll_table *); | |
1da177e4 LT |
529 | static int unix_ioctl(struct socket *, unsigned int, unsigned long); |
530 | static int unix_shutdown(struct socket *, int); | |
1b784140 YX |
531 | static int unix_stream_sendmsg(struct socket *, struct msghdr *, size_t); |
532 | static int unix_stream_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
869e7c62 HFS |
533 | static ssize_t unix_stream_sendpage(struct socket *, struct page *, int offset, |
534 | size_t size, int flags); | |
2b514574 HFS |
535 | static ssize_t unix_stream_splice_read(struct socket *, loff_t *ppos, |
536 | struct pipe_inode_info *, size_t size, | |
537 | unsigned int flags); | |
1b784140 YX |
538 | static int unix_dgram_sendmsg(struct socket *, struct msghdr *, size_t); |
539 | static int unix_dgram_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
1da177e4 LT |
540 | static int unix_dgram_connect(struct socket *, struct sockaddr *, |
541 | int, int); | |
1b784140 YX |
542 | static int unix_seqpacket_sendmsg(struct socket *, struct msghdr *, size_t); |
543 | static int unix_seqpacket_recvmsg(struct socket *, struct msghdr *, size_t, | |
544 | int); | |
1da177e4 | 545 | |
12663bfc | 546 | static int unix_set_peek_off(struct sock *sk, int val) |
f55bb7f9 PE |
547 | { |
548 | struct unix_sock *u = unix_sk(sk); | |
549 | ||
12663bfc SL |
550 | if (mutex_lock_interruptible(&u->readlock)) |
551 | return -EINTR; | |
552 | ||
f55bb7f9 PE |
553 | sk->sk_peek_off = val; |
554 | mutex_unlock(&u->readlock); | |
12663bfc SL |
555 | |
556 | return 0; | |
f55bb7f9 PE |
557 | } |
558 | ||
559 | ||
90ddc4f0 | 560 | static const struct proto_ops unix_stream_ops = { |
1da177e4 LT |
561 | .family = PF_UNIX, |
562 | .owner = THIS_MODULE, | |
563 | .release = unix_release, | |
564 | .bind = unix_bind, | |
565 | .connect = unix_stream_connect, | |
566 | .socketpair = unix_socketpair, | |
567 | .accept = unix_accept, | |
568 | .getname = unix_getname, | |
569 | .poll = unix_poll, | |
570 | .ioctl = unix_ioctl, | |
571 | .listen = unix_listen, | |
572 | .shutdown = unix_shutdown, | |
573 | .setsockopt = sock_no_setsockopt, | |
574 | .getsockopt = sock_no_getsockopt, | |
575 | .sendmsg = unix_stream_sendmsg, | |
576 | .recvmsg = unix_stream_recvmsg, | |
577 | .mmap = sock_no_mmap, | |
869e7c62 | 578 | .sendpage = unix_stream_sendpage, |
2b514574 | 579 | .splice_read = unix_stream_splice_read, |
fc0d7536 | 580 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
581 | }; |
582 | ||
90ddc4f0 | 583 | static const struct proto_ops unix_dgram_ops = { |
1da177e4 LT |
584 | .family = PF_UNIX, |
585 | .owner = THIS_MODULE, | |
586 | .release = unix_release, | |
587 | .bind = unix_bind, | |
588 | .connect = unix_dgram_connect, | |
589 | .socketpair = unix_socketpair, | |
590 | .accept = sock_no_accept, | |
591 | .getname = unix_getname, | |
ec0d215f | 592 | .poll = unix_dgram_poll, |
1da177e4 LT |
593 | .ioctl = unix_ioctl, |
594 | .listen = sock_no_listen, | |
595 | .shutdown = unix_shutdown, | |
596 | .setsockopt = sock_no_setsockopt, | |
597 | .getsockopt = sock_no_getsockopt, | |
598 | .sendmsg = unix_dgram_sendmsg, | |
599 | .recvmsg = unix_dgram_recvmsg, | |
600 | .mmap = sock_no_mmap, | |
601 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 602 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
603 | }; |
604 | ||
90ddc4f0 | 605 | static const struct proto_ops unix_seqpacket_ops = { |
1da177e4 LT |
606 | .family = PF_UNIX, |
607 | .owner = THIS_MODULE, | |
608 | .release = unix_release, | |
609 | .bind = unix_bind, | |
610 | .connect = unix_stream_connect, | |
611 | .socketpair = unix_socketpair, | |
612 | .accept = unix_accept, | |
613 | .getname = unix_getname, | |
ec0d215f | 614 | .poll = unix_dgram_poll, |
1da177e4 LT |
615 | .ioctl = unix_ioctl, |
616 | .listen = unix_listen, | |
617 | .shutdown = unix_shutdown, | |
618 | .setsockopt = sock_no_setsockopt, | |
619 | .getsockopt = sock_no_getsockopt, | |
620 | .sendmsg = unix_seqpacket_sendmsg, | |
a05d2ad1 | 621 | .recvmsg = unix_seqpacket_recvmsg, |
1da177e4 LT |
622 | .mmap = sock_no_mmap, |
623 | .sendpage = sock_no_sendpage, | |
f55bb7f9 | 624 | .set_peek_off = unix_set_peek_off, |
1da177e4 LT |
625 | }; |
626 | ||
627 | static struct proto unix_proto = { | |
248969ae ED |
628 | .name = "UNIX", |
629 | .owner = THIS_MODULE, | |
248969ae | 630 | .obj_size = sizeof(struct unix_sock), |
1da177e4 LT |
631 | }; |
632 | ||
a09785a2 IM |
633 | /* |
634 | * AF_UNIX sockets do not interact with hardware, hence they | |
635 | * dont trigger interrupts - so it's safe for them to have | |
636 | * bh-unsafe locking for their sk_receive_queue.lock. Split off | |
637 | * this special lock-class by reinitializing the spinlock key: | |
638 | */ | |
639 | static struct lock_class_key af_unix_sk_receive_queue_lock_key; | |
640 | ||
11aa9c28 | 641 | static struct sock *unix_create1(struct net *net, struct socket *sock, int kern) |
1da177e4 LT |
642 | { |
643 | struct sock *sk = NULL; | |
644 | struct unix_sock *u; | |
645 | ||
518de9b3 ED |
646 | atomic_long_inc(&unix_nr_socks); |
647 | if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files()) | |
1da177e4 LT |
648 | goto out; |
649 | ||
11aa9c28 | 650 | sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto, kern); |
1da177e4 LT |
651 | if (!sk) |
652 | goto out; | |
653 | ||
6eba6a37 | 654 | sock_init_data(sock, sk); |
a09785a2 IM |
655 | lockdep_set_class(&sk->sk_receive_queue.lock, |
656 | &af_unix_sk_receive_queue_lock_key); | |
1da177e4 LT |
657 | |
658 | sk->sk_write_space = unix_write_space; | |
a0a53c8b | 659 | sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen; |
1da177e4 LT |
660 | sk->sk_destruct = unix_sock_destructor; |
661 | u = unix_sk(sk); | |
40ffe67d AV |
662 | u->path.dentry = NULL; |
663 | u->path.mnt = NULL; | |
fd19f329 | 664 | spin_lock_init(&u->lock); |
516e0cc5 | 665 | atomic_long_set(&u->inflight, 0); |
1fd05ba5 | 666 | INIT_LIST_HEAD(&u->link); |
57b47a53 | 667 | mutex_init(&u->readlock); /* single task reading lock */ |
1da177e4 | 668 | init_waitqueue_head(&u->peer_wait); |
7123aaa3 | 669 | unix_insert_socket(unix_sockets_unbound(sk), sk); |
1da177e4 | 670 | out: |
284b327b | 671 | if (sk == NULL) |
518de9b3 | 672 | atomic_long_dec(&unix_nr_socks); |
920de804 ED |
673 | else { |
674 | local_bh_disable(); | |
a8076d8d | 675 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
920de804 ED |
676 | local_bh_enable(); |
677 | } | |
1da177e4 LT |
678 | return sk; |
679 | } | |
680 | ||
3f378b68 EP |
681 | static int unix_create(struct net *net, struct socket *sock, int protocol, |
682 | int kern) | |
1da177e4 LT |
683 | { |
684 | if (protocol && protocol != PF_UNIX) | |
685 | return -EPROTONOSUPPORT; | |
686 | ||
687 | sock->state = SS_UNCONNECTED; | |
688 | ||
689 | switch (sock->type) { | |
690 | case SOCK_STREAM: | |
691 | sock->ops = &unix_stream_ops; | |
692 | break; | |
693 | /* | |
694 | * Believe it or not BSD has AF_UNIX, SOCK_RAW though | |
695 | * nothing uses it. | |
696 | */ | |
697 | case SOCK_RAW: | |
e27dfcea | 698 | sock->type = SOCK_DGRAM; |
1da177e4 LT |
699 | case SOCK_DGRAM: |
700 | sock->ops = &unix_dgram_ops; | |
701 | break; | |
702 | case SOCK_SEQPACKET: | |
703 | sock->ops = &unix_seqpacket_ops; | |
704 | break; | |
705 | default: | |
706 | return -ESOCKTNOSUPPORT; | |
707 | } | |
708 | ||
11aa9c28 | 709 | return unix_create1(net, sock, kern) ? 0 : -ENOMEM; |
1da177e4 LT |
710 | } |
711 | ||
712 | static int unix_release(struct socket *sock) | |
713 | { | |
714 | struct sock *sk = sock->sk; | |
715 | ||
716 | if (!sk) | |
717 | return 0; | |
718 | ||
ded34e0f | 719 | unix_release_sock(sk, 0); |
1da177e4 LT |
720 | sock->sk = NULL; |
721 | ||
ded34e0f | 722 | return 0; |
1da177e4 LT |
723 | } |
724 | ||
725 | static int unix_autobind(struct socket *sock) | |
726 | { | |
727 | struct sock *sk = sock->sk; | |
3b1e0a65 | 728 | struct net *net = sock_net(sk); |
1da177e4 LT |
729 | struct unix_sock *u = unix_sk(sk); |
730 | static u32 ordernum = 1; | |
6eba6a37 | 731 | struct unix_address *addr; |
1da177e4 | 732 | int err; |
8df73ff9 | 733 | unsigned int retries = 0; |
1da177e4 | 734 | |
37ab4fa7 SL |
735 | err = mutex_lock_interruptible(&u->readlock); |
736 | if (err) | |
737 | return err; | |
1da177e4 LT |
738 | |
739 | err = 0; | |
740 | if (u->addr) | |
741 | goto out; | |
742 | ||
743 | err = -ENOMEM; | |
0da974f4 | 744 | addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL); |
1da177e4 LT |
745 | if (!addr) |
746 | goto out; | |
747 | ||
1da177e4 LT |
748 | addr->name->sun_family = AF_UNIX; |
749 | atomic_set(&addr->refcnt, 1); | |
750 | ||
751 | retry: | |
752 | addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short); | |
07f0757a | 753 | addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0)); |
1da177e4 | 754 | |
fbe9cc4a | 755 | spin_lock(&unix_table_lock); |
1da177e4 LT |
756 | ordernum = (ordernum+1)&0xFFFFF; |
757 | ||
097e66c5 | 758 | if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type, |
1da177e4 | 759 | addr->hash)) { |
fbe9cc4a | 760 | spin_unlock(&unix_table_lock); |
8df73ff9 TH |
761 | /* |
762 | * __unix_find_socket_byname() may take long time if many names | |
763 | * are already in use. | |
764 | */ | |
765 | cond_resched(); | |
766 | /* Give up if all names seems to be in use. */ | |
767 | if (retries++ == 0xFFFFF) { | |
768 | err = -ENOSPC; | |
769 | kfree(addr); | |
770 | goto out; | |
771 | } | |
1da177e4 LT |
772 | goto retry; |
773 | } | |
774 | addr->hash ^= sk->sk_type; | |
775 | ||
776 | __unix_remove_socket(sk); | |
777 | u->addr = addr; | |
778 | __unix_insert_socket(&unix_socket_table[addr->hash], sk); | |
fbe9cc4a | 779 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
780 | err = 0; |
781 | ||
57b47a53 | 782 | out: mutex_unlock(&u->readlock); |
1da177e4 LT |
783 | return err; |
784 | } | |
785 | ||
097e66c5 DL |
786 | static struct sock *unix_find_other(struct net *net, |
787 | struct sockaddr_un *sunname, int len, | |
95c96174 | 788 | int type, unsigned int hash, int *error) |
1da177e4 LT |
789 | { |
790 | struct sock *u; | |
421748ec | 791 | struct path path; |
1da177e4 | 792 | int err = 0; |
ac7bfa62 | 793 | |
1da177e4 | 794 | if (sunname->sun_path[0]) { |
421748ec AV |
795 | struct inode *inode; |
796 | err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path); | |
1da177e4 LT |
797 | if (err) |
798 | goto fail; | |
a25b376b | 799 | inode = d_backing_inode(path.dentry); |
421748ec | 800 | err = inode_permission(inode, MAY_WRITE); |
1da177e4 LT |
801 | if (err) |
802 | goto put_fail; | |
803 | ||
804 | err = -ECONNREFUSED; | |
421748ec | 805 | if (!S_ISSOCK(inode->i_mode)) |
1da177e4 | 806 | goto put_fail; |
6616f788 | 807 | u = unix_find_socket_byinode(inode); |
1da177e4 LT |
808 | if (!u) |
809 | goto put_fail; | |
810 | ||
811 | if (u->sk_type == type) | |
68ac1234 | 812 | touch_atime(&path); |
1da177e4 | 813 | |
421748ec | 814 | path_put(&path); |
1da177e4 | 815 | |
e27dfcea | 816 | err = -EPROTOTYPE; |
1da177e4 LT |
817 | if (u->sk_type != type) { |
818 | sock_put(u); | |
819 | goto fail; | |
820 | } | |
821 | } else { | |
822 | err = -ECONNREFUSED; | |
e27dfcea | 823 | u = unix_find_socket_byname(net, sunname, len, type, hash); |
1da177e4 LT |
824 | if (u) { |
825 | struct dentry *dentry; | |
40ffe67d | 826 | dentry = unix_sk(u)->path.dentry; |
1da177e4 | 827 | if (dentry) |
68ac1234 | 828 | touch_atime(&unix_sk(u)->path); |
1da177e4 LT |
829 | } else |
830 | goto fail; | |
831 | } | |
832 | return u; | |
833 | ||
834 | put_fail: | |
421748ec | 835 | path_put(&path); |
1da177e4 | 836 | fail: |
e27dfcea | 837 | *error = err; |
1da177e4 LT |
838 | return NULL; |
839 | } | |
840 | ||
faf02010 AV |
841 | static int unix_mknod(const char *sun_path, umode_t mode, struct path *res) |
842 | { | |
843 | struct dentry *dentry; | |
844 | struct path path; | |
845 | int err = 0; | |
846 | /* | |
847 | * Get the parent directory, calculate the hash for last | |
848 | * component. | |
849 | */ | |
850 | dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0); | |
851 | err = PTR_ERR(dentry); | |
852 | if (IS_ERR(dentry)) | |
853 | return err; | |
854 | ||
855 | /* | |
856 | * All right, let's create it. | |
857 | */ | |
858 | err = security_path_mknod(&path, dentry, mode, 0); | |
859 | if (!err) { | |
ee8ac4d6 | 860 | err = vfs_mknod(d_inode(path.dentry), dentry, mode, 0); |
faf02010 AV |
861 | if (!err) { |
862 | res->mnt = mntget(path.mnt); | |
863 | res->dentry = dget(dentry); | |
864 | } | |
865 | } | |
866 | done_path_create(&path, dentry); | |
867 | return err; | |
868 | } | |
1da177e4 LT |
869 | |
870 | static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
871 | { | |
872 | struct sock *sk = sock->sk; | |
3b1e0a65 | 873 | struct net *net = sock_net(sk); |
1da177e4 | 874 | struct unix_sock *u = unix_sk(sk); |
e27dfcea | 875 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
dae6ad8f | 876 | char *sun_path = sunaddr->sun_path; |
1da177e4 | 877 | int err; |
95c96174 | 878 | unsigned int hash; |
1da177e4 LT |
879 | struct unix_address *addr; |
880 | struct hlist_head *list; | |
881 | ||
882 | err = -EINVAL; | |
883 | if (sunaddr->sun_family != AF_UNIX) | |
884 | goto out; | |
885 | ||
e27dfcea | 886 | if (addr_len == sizeof(short)) { |
1da177e4 LT |
887 | err = unix_autobind(sock); |
888 | goto out; | |
889 | } | |
890 | ||
891 | err = unix_mkname(sunaddr, addr_len, &hash); | |
892 | if (err < 0) | |
893 | goto out; | |
894 | addr_len = err; | |
895 | ||
37ab4fa7 SL |
896 | err = mutex_lock_interruptible(&u->readlock); |
897 | if (err) | |
898 | goto out; | |
1da177e4 LT |
899 | |
900 | err = -EINVAL; | |
901 | if (u->addr) | |
902 | goto out_up; | |
903 | ||
904 | err = -ENOMEM; | |
905 | addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL); | |
906 | if (!addr) | |
907 | goto out_up; | |
908 | ||
909 | memcpy(addr->name, sunaddr, addr_len); | |
910 | addr->len = addr_len; | |
911 | addr->hash = hash ^ sk->sk_type; | |
912 | atomic_set(&addr->refcnt, 1); | |
913 | ||
dae6ad8f | 914 | if (sun_path[0]) { |
faf02010 AV |
915 | struct path path; |
916 | umode_t mode = S_IFSOCK | | |
ce3b0f8d | 917 | (SOCK_INODE(sock)->i_mode & ~current_umask()); |
faf02010 AV |
918 | err = unix_mknod(sun_path, mode, &path); |
919 | if (err) { | |
920 | if (err == -EEXIST) | |
921 | err = -EADDRINUSE; | |
922 | unix_release_addr(addr); | |
923 | goto out_up; | |
924 | } | |
1da177e4 | 925 | addr->hash = UNIX_HASH_SIZE; |
a25b376b | 926 | hash = d_backing_inode(path.dentry)->i_ino & (UNIX_HASH_SIZE-1); |
faf02010 AV |
927 | spin_lock(&unix_table_lock); |
928 | u->path = path; | |
929 | list = &unix_socket_table[hash]; | |
930 | } else { | |
931 | spin_lock(&unix_table_lock); | |
1da177e4 | 932 | err = -EADDRINUSE; |
097e66c5 | 933 | if (__unix_find_socket_byname(net, sunaddr, addr_len, |
1da177e4 LT |
934 | sk->sk_type, hash)) { |
935 | unix_release_addr(addr); | |
936 | goto out_unlock; | |
937 | } | |
938 | ||
939 | list = &unix_socket_table[addr->hash]; | |
1da177e4 LT |
940 | } |
941 | ||
942 | err = 0; | |
943 | __unix_remove_socket(sk); | |
944 | u->addr = addr; | |
945 | __unix_insert_socket(list, sk); | |
946 | ||
947 | out_unlock: | |
fbe9cc4a | 948 | spin_unlock(&unix_table_lock); |
1da177e4 | 949 | out_up: |
57b47a53 | 950 | mutex_unlock(&u->readlock); |
1da177e4 LT |
951 | out: |
952 | return err; | |
1da177e4 LT |
953 | } |
954 | ||
278a3de5 DM |
955 | static void unix_state_double_lock(struct sock *sk1, struct sock *sk2) |
956 | { | |
957 | if (unlikely(sk1 == sk2) || !sk2) { | |
958 | unix_state_lock(sk1); | |
959 | return; | |
960 | } | |
961 | if (sk1 < sk2) { | |
962 | unix_state_lock(sk1); | |
963 | unix_state_lock_nested(sk2); | |
964 | } else { | |
965 | unix_state_lock(sk2); | |
966 | unix_state_lock_nested(sk1); | |
967 | } | |
968 | } | |
969 | ||
970 | static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2) | |
971 | { | |
972 | if (unlikely(sk1 == sk2) || !sk2) { | |
973 | unix_state_unlock(sk1); | |
974 | return; | |
975 | } | |
976 | unix_state_unlock(sk1); | |
977 | unix_state_unlock(sk2); | |
978 | } | |
979 | ||
1da177e4 LT |
980 | static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr, |
981 | int alen, int flags) | |
982 | { | |
983 | struct sock *sk = sock->sk; | |
3b1e0a65 | 984 | struct net *net = sock_net(sk); |
e27dfcea | 985 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr; |
1da177e4 | 986 | struct sock *other; |
95c96174 | 987 | unsigned int hash; |
1da177e4 LT |
988 | int err; |
989 | ||
990 | if (addr->sa_family != AF_UNSPEC) { | |
991 | err = unix_mkname(sunaddr, alen, &hash); | |
992 | if (err < 0) | |
993 | goto out; | |
994 | alen = err; | |
995 | ||
996 | if (test_bit(SOCK_PASSCRED, &sock->flags) && | |
997 | !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0) | |
998 | goto out; | |
999 | ||
278a3de5 | 1000 | restart: |
e27dfcea | 1001 | other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err); |
1da177e4 LT |
1002 | if (!other) |
1003 | goto out; | |
1004 | ||
278a3de5 DM |
1005 | unix_state_double_lock(sk, other); |
1006 | ||
1007 | /* Apparently VFS overslept socket death. Retry. */ | |
1008 | if (sock_flag(other, SOCK_DEAD)) { | |
1009 | unix_state_double_unlock(sk, other); | |
1010 | sock_put(other); | |
1011 | goto restart; | |
1012 | } | |
1da177e4 LT |
1013 | |
1014 | err = -EPERM; | |
1015 | if (!unix_may_send(sk, other)) | |
1016 | goto out_unlock; | |
1017 | ||
1018 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1019 | if (err) | |
1020 | goto out_unlock; | |
1021 | ||
1022 | } else { | |
1023 | /* | |
1024 | * 1003.1g breaking connected state with AF_UNSPEC | |
1025 | */ | |
1026 | other = NULL; | |
278a3de5 | 1027 | unix_state_double_lock(sk, other); |
1da177e4 LT |
1028 | } |
1029 | ||
1030 | /* | |
1031 | * If it was connected, reconnect. | |
1032 | */ | |
1033 | if (unix_peer(sk)) { | |
1034 | struct sock *old_peer = unix_peer(sk); | |
e27dfcea | 1035 | unix_peer(sk) = other; |
278a3de5 | 1036 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1037 | |
1038 | if (other != old_peer) | |
1039 | unix_dgram_disconnected(sk, old_peer); | |
1040 | sock_put(old_peer); | |
1041 | } else { | |
e27dfcea | 1042 | unix_peer(sk) = other; |
278a3de5 | 1043 | unix_state_double_unlock(sk, other); |
1da177e4 | 1044 | } |
ac7bfa62 | 1045 | return 0; |
1da177e4 LT |
1046 | |
1047 | out_unlock: | |
278a3de5 | 1048 | unix_state_double_unlock(sk, other); |
1da177e4 LT |
1049 | sock_put(other); |
1050 | out: | |
1051 | return err; | |
1052 | } | |
1053 | ||
1054 | static long unix_wait_for_peer(struct sock *other, long timeo) | |
1055 | { | |
1056 | struct unix_sock *u = unix_sk(other); | |
1057 | int sched; | |
1058 | DEFINE_WAIT(wait); | |
1059 | ||
1060 | prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE); | |
1061 | ||
1062 | sched = !sock_flag(other, SOCK_DEAD) && | |
1063 | !(other->sk_shutdown & RCV_SHUTDOWN) && | |
3c73419c | 1064 | unix_recvq_full(other); |
1da177e4 | 1065 | |
1c92b4e5 | 1066 | unix_state_unlock(other); |
1da177e4 LT |
1067 | |
1068 | if (sched) | |
1069 | timeo = schedule_timeout(timeo); | |
1070 | ||
1071 | finish_wait(&u->peer_wait, &wait); | |
1072 | return timeo; | |
1073 | } | |
1074 | ||
1075 | static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
1076 | int addr_len, int flags) | |
1077 | { | |
e27dfcea | 1078 | struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr; |
1da177e4 | 1079 | struct sock *sk = sock->sk; |
3b1e0a65 | 1080 | struct net *net = sock_net(sk); |
1da177e4 LT |
1081 | struct unix_sock *u = unix_sk(sk), *newu, *otheru; |
1082 | struct sock *newsk = NULL; | |
1083 | struct sock *other = NULL; | |
1084 | struct sk_buff *skb = NULL; | |
95c96174 | 1085 | unsigned int hash; |
1da177e4 LT |
1086 | int st; |
1087 | int err; | |
1088 | long timeo; | |
1089 | ||
1090 | err = unix_mkname(sunaddr, addr_len, &hash); | |
1091 | if (err < 0) | |
1092 | goto out; | |
1093 | addr_len = err; | |
1094 | ||
f64f9e71 JP |
1095 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr && |
1096 | (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1097 | goto out; |
1098 | ||
1099 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
1100 | ||
1101 | /* First of all allocate resources. | |
1102 | If we will make it after state is locked, | |
1103 | we will have to recheck all again in any case. | |
1104 | */ | |
1105 | ||
1106 | err = -ENOMEM; | |
1107 | ||
1108 | /* create new sock for complete connection */ | |
11aa9c28 | 1109 | newsk = unix_create1(sock_net(sk), NULL, 0); |
1da177e4 LT |
1110 | if (newsk == NULL) |
1111 | goto out; | |
1112 | ||
1113 | /* Allocate skb for sending to listening sock */ | |
1114 | skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL); | |
1115 | if (skb == NULL) | |
1116 | goto out; | |
1117 | ||
1118 | restart: | |
1119 | /* Find listening sock. */ | |
097e66c5 | 1120 | other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err); |
1da177e4 LT |
1121 | if (!other) |
1122 | goto out; | |
1123 | ||
1124 | /* Latch state of peer */ | |
1c92b4e5 | 1125 | unix_state_lock(other); |
1da177e4 LT |
1126 | |
1127 | /* Apparently VFS overslept socket death. Retry. */ | |
1128 | if (sock_flag(other, SOCK_DEAD)) { | |
1c92b4e5 | 1129 | unix_state_unlock(other); |
1da177e4 LT |
1130 | sock_put(other); |
1131 | goto restart; | |
1132 | } | |
1133 | ||
1134 | err = -ECONNREFUSED; | |
1135 | if (other->sk_state != TCP_LISTEN) | |
1136 | goto out_unlock; | |
77238f2b TS |
1137 | if (other->sk_shutdown & RCV_SHUTDOWN) |
1138 | goto out_unlock; | |
1da177e4 | 1139 | |
3c73419c | 1140 | if (unix_recvq_full(other)) { |
1da177e4 LT |
1141 | err = -EAGAIN; |
1142 | if (!timeo) | |
1143 | goto out_unlock; | |
1144 | ||
1145 | timeo = unix_wait_for_peer(other, timeo); | |
1146 | ||
1147 | err = sock_intr_errno(timeo); | |
1148 | if (signal_pending(current)) | |
1149 | goto out; | |
1150 | sock_put(other); | |
1151 | goto restart; | |
ac7bfa62 | 1152 | } |
1da177e4 LT |
1153 | |
1154 | /* Latch our state. | |
1155 | ||
e5537bfc | 1156 | It is tricky place. We need to grab our state lock and cannot |
1da177e4 LT |
1157 | drop lock on peer. It is dangerous because deadlock is |
1158 | possible. Connect to self case and simultaneous | |
1159 | attempt to connect are eliminated by checking socket | |
1160 | state. other is TCP_LISTEN, if sk is TCP_LISTEN we | |
1161 | check this before attempt to grab lock. | |
1162 | ||
1163 | Well, and we have to recheck the state after socket locked. | |
1164 | */ | |
1165 | st = sk->sk_state; | |
1166 | ||
1167 | switch (st) { | |
1168 | case TCP_CLOSE: | |
1169 | /* This is ok... continue with connect */ | |
1170 | break; | |
1171 | case TCP_ESTABLISHED: | |
1172 | /* Socket is already connected */ | |
1173 | err = -EISCONN; | |
1174 | goto out_unlock; | |
1175 | default: | |
1176 | err = -EINVAL; | |
1177 | goto out_unlock; | |
1178 | } | |
1179 | ||
1c92b4e5 | 1180 | unix_state_lock_nested(sk); |
1da177e4 LT |
1181 | |
1182 | if (sk->sk_state != st) { | |
1c92b4e5 DM |
1183 | unix_state_unlock(sk); |
1184 | unix_state_unlock(other); | |
1da177e4 LT |
1185 | sock_put(other); |
1186 | goto restart; | |
1187 | } | |
1188 | ||
3610cda5 | 1189 | err = security_unix_stream_connect(sk, other, newsk); |
1da177e4 | 1190 | if (err) { |
1c92b4e5 | 1191 | unix_state_unlock(sk); |
1da177e4 LT |
1192 | goto out_unlock; |
1193 | } | |
1194 | ||
1195 | /* The way is open! Fastly set all the necessary fields... */ | |
1196 | ||
1197 | sock_hold(sk); | |
1198 | unix_peer(newsk) = sk; | |
1199 | newsk->sk_state = TCP_ESTABLISHED; | |
1200 | newsk->sk_type = sk->sk_type; | |
109f6e39 | 1201 | init_peercred(newsk); |
1da177e4 | 1202 | newu = unix_sk(newsk); |
eaefd110 | 1203 | RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq); |
1da177e4 LT |
1204 | otheru = unix_sk(other); |
1205 | ||
1206 | /* copy address information from listening to new sock*/ | |
1207 | if (otheru->addr) { | |
1208 | atomic_inc(&otheru->addr->refcnt); | |
1209 | newu->addr = otheru->addr; | |
1210 | } | |
40ffe67d AV |
1211 | if (otheru->path.dentry) { |
1212 | path_get(&otheru->path); | |
1213 | newu->path = otheru->path; | |
1da177e4 LT |
1214 | } |
1215 | ||
1216 | /* Set credentials */ | |
109f6e39 | 1217 | copy_peercred(sk, other); |
1da177e4 | 1218 | |
1da177e4 LT |
1219 | sock->state = SS_CONNECTED; |
1220 | sk->sk_state = TCP_ESTABLISHED; | |
830a1e5c BL |
1221 | sock_hold(newsk); |
1222 | ||
4e857c58 | 1223 | smp_mb__after_atomic(); /* sock_hold() does an atomic_inc() */ |
830a1e5c | 1224 | unix_peer(sk) = newsk; |
1da177e4 | 1225 | |
1c92b4e5 | 1226 | unix_state_unlock(sk); |
1da177e4 LT |
1227 | |
1228 | /* take ten and and send info to listening sock */ | |
1229 | spin_lock(&other->sk_receive_queue.lock); | |
1230 | __skb_queue_tail(&other->sk_receive_queue, skb); | |
1da177e4 | 1231 | spin_unlock(&other->sk_receive_queue.lock); |
1c92b4e5 | 1232 | unix_state_unlock(other); |
676d2369 | 1233 | other->sk_data_ready(other); |
1da177e4 LT |
1234 | sock_put(other); |
1235 | return 0; | |
1236 | ||
1237 | out_unlock: | |
1238 | if (other) | |
1c92b4e5 | 1239 | unix_state_unlock(other); |
1da177e4 LT |
1240 | |
1241 | out: | |
40d44446 | 1242 | kfree_skb(skb); |
1da177e4 LT |
1243 | if (newsk) |
1244 | unix_release_sock(newsk, 0); | |
1245 | if (other) | |
1246 | sock_put(other); | |
1247 | return err; | |
1248 | } | |
1249 | ||
1250 | static int unix_socketpair(struct socket *socka, struct socket *sockb) | |
1251 | { | |
e27dfcea | 1252 | struct sock *ska = socka->sk, *skb = sockb->sk; |
1da177e4 LT |
1253 | |
1254 | /* Join our sockets back to back */ | |
1255 | sock_hold(ska); | |
1256 | sock_hold(skb); | |
e27dfcea JK |
1257 | unix_peer(ska) = skb; |
1258 | unix_peer(skb) = ska; | |
109f6e39 EB |
1259 | init_peercred(ska); |
1260 | init_peercred(skb); | |
1da177e4 LT |
1261 | |
1262 | if (ska->sk_type != SOCK_DGRAM) { | |
1263 | ska->sk_state = TCP_ESTABLISHED; | |
1264 | skb->sk_state = TCP_ESTABLISHED; | |
1265 | socka->state = SS_CONNECTED; | |
1266 | sockb->state = SS_CONNECTED; | |
1267 | } | |
1268 | return 0; | |
1269 | } | |
1270 | ||
90c6bd34 DB |
1271 | static void unix_sock_inherit_flags(const struct socket *old, |
1272 | struct socket *new) | |
1273 | { | |
1274 | if (test_bit(SOCK_PASSCRED, &old->flags)) | |
1275 | set_bit(SOCK_PASSCRED, &new->flags); | |
1276 | if (test_bit(SOCK_PASSSEC, &old->flags)) | |
1277 | set_bit(SOCK_PASSSEC, &new->flags); | |
1278 | } | |
1279 | ||
1da177e4 LT |
1280 | static int unix_accept(struct socket *sock, struct socket *newsock, int flags) |
1281 | { | |
1282 | struct sock *sk = sock->sk; | |
1283 | struct sock *tsk; | |
1284 | struct sk_buff *skb; | |
1285 | int err; | |
1286 | ||
1287 | err = -EOPNOTSUPP; | |
6eba6a37 | 1288 | if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET) |
1da177e4 LT |
1289 | goto out; |
1290 | ||
1291 | err = -EINVAL; | |
1292 | if (sk->sk_state != TCP_LISTEN) | |
1293 | goto out; | |
1294 | ||
1295 | /* If socket state is TCP_LISTEN it cannot change (for now...), | |
1296 | * so that no locks are necessary. | |
1297 | */ | |
1298 | ||
1299 | skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err); | |
1300 | if (!skb) { | |
1301 | /* This means receive shutdown. */ | |
1302 | if (err == 0) | |
1303 | err = -EINVAL; | |
1304 | goto out; | |
1305 | } | |
1306 | ||
1307 | tsk = skb->sk; | |
1308 | skb_free_datagram(sk, skb); | |
1309 | wake_up_interruptible(&unix_sk(sk)->peer_wait); | |
1310 | ||
1311 | /* attach accepted sock to socket */ | |
1c92b4e5 | 1312 | unix_state_lock(tsk); |
1da177e4 | 1313 | newsock->state = SS_CONNECTED; |
90c6bd34 | 1314 | unix_sock_inherit_flags(sock, newsock); |
1da177e4 | 1315 | sock_graft(tsk, newsock); |
1c92b4e5 | 1316 | unix_state_unlock(tsk); |
1da177e4 LT |
1317 | return 0; |
1318 | ||
1319 | out: | |
1320 | return err; | |
1321 | } | |
1322 | ||
1323 | ||
1324 | static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer) | |
1325 | { | |
1326 | struct sock *sk = sock->sk; | |
1327 | struct unix_sock *u; | |
13cfa97b | 1328 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr); |
1da177e4 LT |
1329 | int err = 0; |
1330 | ||
1331 | if (peer) { | |
1332 | sk = unix_peer_get(sk); | |
1333 | ||
1334 | err = -ENOTCONN; | |
1335 | if (!sk) | |
1336 | goto out; | |
1337 | err = 0; | |
1338 | } else { | |
1339 | sock_hold(sk); | |
1340 | } | |
1341 | ||
1342 | u = unix_sk(sk); | |
1c92b4e5 | 1343 | unix_state_lock(sk); |
1da177e4 LT |
1344 | if (!u->addr) { |
1345 | sunaddr->sun_family = AF_UNIX; | |
1346 | sunaddr->sun_path[0] = 0; | |
1347 | *uaddr_len = sizeof(short); | |
1348 | } else { | |
1349 | struct unix_address *addr = u->addr; | |
1350 | ||
1351 | *uaddr_len = addr->len; | |
1352 | memcpy(sunaddr, addr->name, *uaddr_len); | |
1353 | } | |
1c92b4e5 | 1354 | unix_state_unlock(sk); |
1da177e4 LT |
1355 | sock_put(sk); |
1356 | out: | |
1357 | return err; | |
1358 | } | |
1359 | ||
1360 | static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb) | |
1361 | { | |
1362 | int i; | |
1363 | ||
1364 | scm->fp = UNIXCB(skb).fp; | |
1da177e4 LT |
1365 | UNIXCB(skb).fp = NULL; |
1366 | ||
6eba6a37 | 1367 | for (i = scm->fp->count-1; i >= 0; i--) |
1da177e4 LT |
1368 | unix_notinflight(scm->fp->fp[i]); |
1369 | } | |
1370 | ||
7361c36c | 1371 | static void unix_destruct_scm(struct sk_buff *skb) |
1da177e4 LT |
1372 | { |
1373 | struct scm_cookie scm; | |
1374 | memset(&scm, 0, sizeof(scm)); | |
7361c36c | 1375 | scm.pid = UNIXCB(skb).pid; |
7361c36c EB |
1376 | if (UNIXCB(skb).fp) |
1377 | unix_detach_fds(&scm, skb); | |
1da177e4 LT |
1378 | |
1379 | /* Alas, it calls VFS */ | |
1380 | /* So fscking what? fput() had been SMP-safe since the last Summer */ | |
1381 | scm_destroy(&scm); | |
1382 | sock_wfree(skb); | |
1383 | } | |
1384 | ||
25888e30 ED |
1385 | #define MAX_RECURSION_LEVEL 4 |
1386 | ||
6209344f | 1387 | static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb) |
1da177e4 LT |
1388 | { |
1389 | int i; | |
25888e30 ED |
1390 | unsigned char max_level = 0; |
1391 | int unix_sock_count = 0; | |
1392 | ||
1393 | for (i = scm->fp->count - 1; i >= 0; i--) { | |
1394 | struct sock *sk = unix_get_socket(scm->fp->fp[i]); | |
1395 | ||
1396 | if (sk) { | |
1397 | unix_sock_count++; | |
1398 | max_level = max(max_level, | |
1399 | unix_sk(sk)->recursion_level); | |
1400 | } | |
1401 | } | |
1402 | if (unlikely(max_level > MAX_RECURSION_LEVEL)) | |
1403 | return -ETOOMANYREFS; | |
6209344f MS |
1404 | |
1405 | /* | |
1406 | * Need to duplicate file references for the sake of garbage | |
1407 | * collection. Otherwise a socket in the fps might become a | |
1408 | * candidate for GC while the skb is not yet queued. | |
1409 | */ | |
1410 | UNIXCB(skb).fp = scm_fp_dup(scm->fp); | |
1411 | if (!UNIXCB(skb).fp) | |
1412 | return -ENOMEM; | |
1413 | ||
25888e30 ED |
1414 | if (unix_sock_count) { |
1415 | for (i = scm->fp->count - 1; i >= 0; i--) | |
1416 | unix_inflight(scm->fp->fp[i]); | |
1417 | } | |
1418 | return max_level; | |
1da177e4 LT |
1419 | } |
1420 | ||
f78a5fda | 1421 | static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds) |
7361c36c EB |
1422 | { |
1423 | int err = 0; | |
16e57262 | 1424 | |
f78a5fda | 1425 | UNIXCB(skb).pid = get_pid(scm->pid); |
6b0ee8c0 EB |
1426 | UNIXCB(skb).uid = scm->creds.uid; |
1427 | UNIXCB(skb).gid = scm->creds.gid; | |
7361c36c | 1428 | UNIXCB(skb).fp = NULL; |
37a9a8df | 1429 | unix_get_secdata(scm, skb); |
7361c36c EB |
1430 | if (scm->fp && send_fds) |
1431 | err = unix_attach_fds(scm, skb); | |
1432 | ||
1433 | skb->destructor = unix_destruct_scm; | |
1434 | return err; | |
1435 | } | |
1436 | ||
16e57262 ED |
1437 | /* |
1438 | * Some apps rely on write() giving SCM_CREDENTIALS | |
1439 | * We include credentials if source or destination socket | |
1440 | * asserted SOCK_PASSCRED. | |
1441 | */ | |
1442 | static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock, | |
1443 | const struct sock *other) | |
1444 | { | |
6b0ee8c0 | 1445 | if (UNIXCB(skb).pid) |
16e57262 ED |
1446 | return; |
1447 | if (test_bit(SOCK_PASSCRED, &sock->flags) || | |
25da0e3e EB |
1448 | !other->sk_socket || |
1449 | test_bit(SOCK_PASSCRED, &other->sk_socket->flags)) { | |
16e57262 | 1450 | UNIXCB(skb).pid = get_pid(task_tgid(current)); |
6e0895c2 | 1451 | current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid); |
16e57262 ED |
1452 | } |
1453 | } | |
1454 | ||
1da177e4 LT |
1455 | /* |
1456 | * Send AF_UNIX data. | |
1457 | */ | |
1458 | ||
1b784140 YX |
1459 | static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg, |
1460 | size_t len) | |
1da177e4 | 1461 | { |
1da177e4 | 1462 | struct sock *sk = sock->sk; |
3b1e0a65 | 1463 | struct net *net = sock_net(sk); |
1da177e4 | 1464 | struct unix_sock *u = unix_sk(sk); |
342dfc30 | 1465 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name); |
1da177e4 LT |
1466 | struct sock *other = NULL; |
1467 | int namelen = 0; /* fake GCC */ | |
1468 | int err; | |
95c96174 | 1469 | unsigned int hash; |
f78a5fda | 1470 | struct sk_buff *skb; |
1da177e4 | 1471 | long timeo; |
7cc05662 | 1472 | struct scm_cookie scm; |
25888e30 | 1473 | int max_level; |
eb6a2481 | 1474 | int data_len = 0; |
1da177e4 | 1475 | |
5f23b734 | 1476 | wait_for_unix_gc(); |
7cc05662 | 1477 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1478 | if (err < 0) |
1479 | return err; | |
1480 | ||
1481 | err = -EOPNOTSUPP; | |
1482 | if (msg->msg_flags&MSG_OOB) | |
1483 | goto out; | |
1484 | ||
1485 | if (msg->msg_namelen) { | |
1486 | err = unix_mkname(sunaddr, msg->msg_namelen, &hash); | |
1487 | if (err < 0) | |
1488 | goto out; | |
1489 | namelen = err; | |
1490 | } else { | |
1491 | sunaddr = NULL; | |
1492 | err = -ENOTCONN; | |
1493 | other = unix_peer_get(sk); | |
1494 | if (!other) | |
1495 | goto out; | |
1496 | } | |
1497 | ||
f64f9e71 JP |
1498 | if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr |
1499 | && (err = unix_autobind(sock)) != 0) | |
1da177e4 LT |
1500 | goto out; |
1501 | ||
1502 | err = -EMSGSIZE; | |
1503 | if (len > sk->sk_sndbuf - 32) | |
1504 | goto out; | |
1505 | ||
31ff6aa5 | 1506 | if (len > SKB_MAX_ALLOC) { |
eb6a2481 ED |
1507 | data_len = min_t(size_t, |
1508 | len - SKB_MAX_ALLOC, | |
1509 | MAX_SKB_FRAGS * PAGE_SIZE); | |
31ff6aa5 KT |
1510 | data_len = PAGE_ALIGN(data_len); |
1511 | ||
1512 | BUILD_BUG_ON(SKB_MAX_ALLOC < PAGE_SIZE); | |
1513 | } | |
eb6a2481 ED |
1514 | |
1515 | skb = sock_alloc_send_pskb(sk, len - data_len, data_len, | |
28d64271 ED |
1516 | msg->msg_flags & MSG_DONTWAIT, &err, |
1517 | PAGE_ALLOC_COSTLY_ORDER); | |
e27dfcea | 1518 | if (skb == NULL) |
1da177e4 LT |
1519 | goto out; |
1520 | ||
7cc05662 | 1521 | err = unix_scm_to_skb(&scm, skb, true); |
25888e30 | 1522 | if (err < 0) |
7361c36c | 1523 | goto out_free; |
25888e30 | 1524 | max_level = err + 1; |
877ce7c1 | 1525 | |
eb6a2481 ED |
1526 | skb_put(skb, len - data_len); |
1527 | skb->data_len = data_len; | |
1528 | skb->len = len; | |
c0371da6 | 1529 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len); |
1da177e4 LT |
1530 | if (err) |
1531 | goto out_free; | |
1532 | ||
1533 | timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT); | |
1534 | ||
1535 | restart: | |
1536 | if (!other) { | |
1537 | err = -ECONNRESET; | |
1538 | if (sunaddr == NULL) | |
1539 | goto out_free; | |
1540 | ||
097e66c5 | 1541 | other = unix_find_other(net, sunaddr, namelen, sk->sk_type, |
1da177e4 | 1542 | hash, &err); |
e27dfcea | 1543 | if (other == NULL) |
1da177e4 LT |
1544 | goto out_free; |
1545 | } | |
1546 | ||
d6ae3bae AC |
1547 | if (sk_filter(other, skb) < 0) { |
1548 | /* Toss the packet but do not return any error to the sender */ | |
1549 | err = len; | |
1550 | goto out_free; | |
1551 | } | |
1552 | ||
1c92b4e5 | 1553 | unix_state_lock(other); |
1da177e4 LT |
1554 | err = -EPERM; |
1555 | if (!unix_may_send(sk, other)) | |
1556 | goto out_unlock; | |
1557 | ||
1558 | if (sock_flag(other, SOCK_DEAD)) { | |
1559 | /* | |
1560 | * Check with 1003.1g - what should | |
1561 | * datagram error | |
1562 | */ | |
1c92b4e5 | 1563 | unix_state_unlock(other); |
1da177e4 LT |
1564 | sock_put(other); |
1565 | ||
1566 | err = 0; | |
1c92b4e5 | 1567 | unix_state_lock(sk); |
1da177e4 | 1568 | if (unix_peer(sk) == other) { |
e27dfcea | 1569 | unix_peer(sk) = NULL; |
1c92b4e5 | 1570 | unix_state_unlock(sk); |
1da177e4 LT |
1571 | |
1572 | unix_dgram_disconnected(sk, other); | |
1573 | sock_put(other); | |
1574 | err = -ECONNREFUSED; | |
1575 | } else { | |
1c92b4e5 | 1576 | unix_state_unlock(sk); |
1da177e4 LT |
1577 | } |
1578 | ||
1579 | other = NULL; | |
1580 | if (err) | |
1581 | goto out_free; | |
1582 | goto restart; | |
1583 | } | |
1584 | ||
1585 | err = -EPIPE; | |
1586 | if (other->sk_shutdown & RCV_SHUTDOWN) | |
1587 | goto out_unlock; | |
1588 | ||
1589 | if (sk->sk_type != SOCK_SEQPACKET) { | |
1590 | err = security_unix_may_send(sk->sk_socket, other->sk_socket); | |
1591 | if (err) | |
1592 | goto out_unlock; | |
1593 | } | |
1594 | ||
3c73419c | 1595 | if (unix_peer(other) != sk && unix_recvq_full(other)) { |
1da177e4 LT |
1596 | if (!timeo) { |
1597 | err = -EAGAIN; | |
1598 | goto out_unlock; | |
1599 | } | |
1600 | ||
1601 | timeo = unix_wait_for_peer(other, timeo); | |
1602 | ||
1603 | err = sock_intr_errno(timeo); | |
1604 | if (signal_pending(current)) | |
1605 | goto out_free; | |
1606 | ||
1607 | goto restart; | |
1608 | } | |
1609 | ||
3f66116e AC |
1610 | if (sock_flag(other, SOCK_RCVTSTAMP)) |
1611 | __net_timestamp(skb); | |
16e57262 | 1612 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1613 | skb_queue_tail(&other->sk_receive_queue, skb); |
25888e30 ED |
1614 | if (max_level > unix_sk(other)->recursion_level) |
1615 | unix_sk(other)->recursion_level = max_level; | |
1c92b4e5 | 1616 | unix_state_unlock(other); |
676d2369 | 1617 | other->sk_data_ready(other); |
1da177e4 | 1618 | sock_put(other); |
7cc05662 | 1619 | scm_destroy(&scm); |
1da177e4 LT |
1620 | return len; |
1621 | ||
1622 | out_unlock: | |
1c92b4e5 | 1623 | unix_state_unlock(other); |
1da177e4 LT |
1624 | out_free: |
1625 | kfree_skb(skb); | |
1626 | out: | |
1627 | if (other) | |
1628 | sock_put(other); | |
7cc05662 | 1629 | scm_destroy(&scm); |
1da177e4 LT |
1630 | return err; |
1631 | } | |
1632 | ||
e370a723 ED |
1633 | /* We use paged skbs for stream sockets, and limit occupancy to 32768 |
1634 | * bytes, and a minimun of a full page. | |
1635 | */ | |
1636 | #define UNIX_SKB_FRAGS_SZ (PAGE_SIZE << get_order(32768)) | |
ac7bfa62 | 1637 | |
1b784140 YX |
1638 | static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg, |
1639 | size_t len) | |
1da177e4 | 1640 | { |
1da177e4 LT |
1641 | struct sock *sk = sock->sk; |
1642 | struct sock *other = NULL; | |
6eba6a37 | 1643 | int err, size; |
f78a5fda | 1644 | struct sk_buff *skb; |
e27dfcea | 1645 | int sent = 0; |
7cc05662 | 1646 | struct scm_cookie scm; |
8ba69ba6 | 1647 | bool fds_sent = false; |
25888e30 | 1648 | int max_level; |
e370a723 | 1649 | int data_len; |
1da177e4 | 1650 | |
5f23b734 | 1651 | wait_for_unix_gc(); |
7cc05662 | 1652 | err = scm_send(sock, msg, &scm, false); |
1da177e4 LT |
1653 | if (err < 0) |
1654 | return err; | |
1655 | ||
1656 | err = -EOPNOTSUPP; | |
1657 | if (msg->msg_flags&MSG_OOB) | |
1658 | goto out_err; | |
1659 | ||
1660 | if (msg->msg_namelen) { | |
1661 | err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP; | |
1662 | goto out_err; | |
1663 | } else { | |
1da177e4 | 1664 | err = -ENOTCONN; |
830a1e5c | 1665 | other = unix_peer(sk); |
1da177e4 LT |
1666 | if (!other) |
1667 | goto out_err; | |
1668 | } | |
1669 | ||
1670 | if (sk->sk_shutdown & SEND_SHUTDOWN) | |
1671 | goto pipe_err; | |
1672 | ||
6eba6a37 | 1673 | while (sent < len) { |
e370a723 | 1674 | size = len - sent; |
1da177e4 LT |
1675 | |
1676 | /* Keep two messages in the pipe so it schedules better */ | |
e370a723 | 1677 | size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64); |
1da177e4 | 1678 | |
e370a723 ED |
1679 | /* allow fallback to order-0 allocations */ |
1680 | size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ); | |
ac7bfa62 | 1681 | |
e370a723 | 1682 | data_len = max_t(int, 0, size - SKB_MAX_HEAD(0)); |
1da177e4 | 1683 | |
31ff6aa5 KT |
1684 | data_len = min_t(size_t, size, PAGE_ALIGN(data_len)); |
1685 | ||
e370a723 | 1686 | skb = sock_alloc_send_pskb(sk, size - data_len, data_len, |
28d64271 ED |
1687 | msg->msg_flags & MSG_DONTWAIT, &err, |
1688 | get_order(UNIX_SKB_FRAGS_SZ)); | |
e370a723 | 1689 | if (!skb) |
1da177e4 LT |
1690 | goto out_err; |
1691 | ||
f78a5fda | 1692 | /* Only send the fds in the first buffer */ |
7cc05662 | 1693 | err = unix_scm_to_skb(&scm, skb, !fds_sent); |
25888e30 | 1694 | if (err < 0) { |
7361c36c | 1695 | kfree_skb(skb); |
f78a5fda | 1696 | goto out_err; |
6209344f | 1697 | } |
25888e30 | 1698 | max_level = err + 1; |
7361c36c | 1699 | fds_sent = true; |
1da177e4 | 1700 | |
e370a723 ED |
1701 | skb_put(skb, size - data_len); |
1702 | skb->data_len = data_len; | |
1703 | skb->len = size; | |
c0371da6 | 1704 | err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size); |
6eba6a37 | 1705 | if (err) { |
1da177e4 | 1706 | kfree_skb(skb); |
f78a5fda | 1707 | goto out_err; |
1da177e4 LT |
1708 | } |
1709 | ||
1c92b4e5 | 1710 | unix_state_lock(other); |
1da177e4 LT |
1711 | |
1712 | if (sock_flag(other, SOCK_DEAD) || | |
1713 | (other->sk_shutdown & RCV_SHUTDOWN)) | |
1714 | goto pipe_err_free; | |
1715 | ||
16e57262 | 1716 | maybe_add_creds(skb, sock, other); |
1da177e4 | 1717 | skb_queue_tail(&other->sk_receive_queue, skb); |
25888e30 ED |
1718 | if (max_level > unix_sk(other)->recursion_level) |
1719 | unix_sk(other)->recursion_level = max_level; | |
1c92b4e5 | 1720 | unix_state_unlock(other); |
676d2369 | 1721 | other->sk_data_ready(other); |
e27dfcea | 1722 | sent += size; |
1da177e4 | 1723 | } |
1da177e4 | 1724 | |
7cc05662 | 1725 | scm_destroy(&scm); |
1da177e4 LT |
1726 | |
1727 | return sent; | |
1728 | ||
1729 | pipe_err_free: | |
1c92b4e5 | 1730 | unix_state_unlock(other); |
1da177e4 LT |
1731 | kfree_skb(skb); |
1732 | pipe_err: | |
6eba6a37 ED |
1733 | if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL)) |
1734 | send_sig(SIGPIPE, current, 0); | |
1da177e4 LT |
1735 | err = -EPIPE; |
1736 | out_err: | |
7cc05662 | 1737 | scm_destroy(&scm); |
1da177e4 LT |
1738 | return sent ? : err; |
1739 | } | |
1740 | ||
869e7c62 HFS |
1741 | static ssize_t unix_stream_sendpage(struct socket *socket, struct page *page, |
1742 | int offset, size_t size, int flags) | |
1743 | { | |
1744 | int err = 0; | |
1745 | bool send_sigpipe = true; | |
1746 | struct sock *other, *sk = socket->sk; | |
1747 | struct sk_buff *skb, *newskb = NULL, *tail = NULL; | |
1748 | ||
1749 | if (flags & MSG_OOB) | |
1750 | return -EOPNOTSUPP; | |
1751 | ||
1752 | other = unix_peer(sk); | |
1753 | if (!other || sk->sk_state != TCP_ESTABLISHED) | |
1754 | return -ENOTCONN; | |
1755 | ||
1756 | if (false) { | |
1757 | alloc_skb: | |
1758 | unix_state_unlock(other); | |
1759 | mutex_unlock(&unix_sk(other)->readlock); | |
1760 | newskb = sock_alloc_send_pskb(sk, 0, 0, flags & MSG_DONTWAIT, | |
1761 | &err, 0); | |
1762 | if (!newskb) | |
1763 | return err; | |
1764 | } | |
1765 | ||
1766 | /* we must acquire readlock as we modify already present | |
1767 | * skbs in the sk_receive_queue and mess with skb->len | |
1768 | */ | |
1769 | err = mutex_lock_interruptible(&unix_sk(other)->readlock); | |
1770 | if (err) { | |
1771 | err = flags & MSG_DONTWAIT ? -EAGAIN : -ERESTARTSYS; | |
1772 | send_sigpipe = false; | |
1773 | goto err; | |
1774 | } | |
1775 | ||
1776 | if (sk->sk_shutdown & SEND_SHUTDOWN) { | |
1777 | err = -EPIPE; | |
1778 | goto err_unlock; | |
1779 | } | |
1780 | ||
1781 | unix_state_lock(other); | |
1782 | ||
1783 | if (sock_flag(other, SOCK_DEAD) || | |
1784 | other->sk_shutdown & RCV_SHUTDOWN) { | |
1785 | err = -EPIPE; | |
1786 | goto err_state_unlock; | |
1787 | } | |
1788 | ||
1789 | skb = skb_peek_tail(&other->sk_receive_queue); | |
1790 | if (tail && tail == skb) { | |
1791 | skb = newskb; | |
1792 | } else if (!skb) { | |
1793 | if (newskb) | |
1794 | skb = newskb; | |
1795 | else | |
1796 | goto alloc_skb; | |
1797 | } else if (newskb) { | |
1798 | /* this is fast path, we don't necessarily need to | |
1799 | * call to kfree_skb even though with newskb == NULL | |
1800 | * this - does no harm | |
1801 | */ | |
1802 | consume_skb(newskb); | |
8844f972 | 1803 | newskb = NULL; |
869e7c62 HFS |
1804 | } |
1805 | ||
1806 | if (skb_append_pagefrags(skb, page, offset, size)) { | |
1807 | tail = skb; | |
1808 | goto alloc_skb; | |
1809 | } | |
1810 | ||
1811 | skb->len += size; | |
1812 | skb->data_len += size; | |
1813 | skb->truesize += size; | |
1814 | atomic_add(size, &sk->sk_wmem_alloc); | |
1815 | ||
1816 | if (newskb) | |
1817 | __skb_queue_tail(&other->sk_receive_queue, newskb); | |
1818 | ||
1819 | unix_state_unlock(other); | |
1820 | mutex_unlock(&unix_sk(other)->readlock); | |
1821 | ||
1822 | other->sk_data_ready(other); | |
1823 | ||
1824 | return size; | |
1825 | ||
1826 | err_state_unlock: | |
1827 | unix_state_unlock(other); | |
1828 | err_unlock: | |
1829 | mutex_unlock(&unix_sk(other)->readlock); | |
1830 | err: | |
1831 | kfree_skb(newskb); | |
1832 | if (send_sigpipe && !(flags & MSG_NOSIGNAL)) | |
1833 | send_sig(SIGPIPE, current, 0); | |
1834 | return err; | |
1835 | } | |
1836 | ||
1b784140 YX |
1837 | static int unix_seqpacket_sendmsg(struct socket *sock, struct msghdr *msg, |
1838 | size_t len) | |
1da177e4 LT |
1839 | { |
1840 | int err; | |
1841 | struct sock *sk = sock->sk; | |
ac7bfa62 | 1842 | |
1da177e4 LT |
1843 | err = sock_error(sk); |
1844 | if (err) | |
1845 | return err; | |
1846 | ||
1847 | if (sk->sk_state != TCP_ESTABLISHED) | |
1848 | return -ENOTCONN; | |
1849 | ||
1850 | if (msg->msg_namelen) | |
1851 | msg->msg_namelen = 0; | |
1852 | ||
1b784140 | 1853 | return unix_dgram_sendmsg(sock, msg, len); |
1da177e4 | 1854 | } |
ac7bfa62 | 1855 | |
1b784140 YX |
1856 | static int unix_seqpacket_recvmsg(struct socket *sock, struct msghdr *msg, |
1857 | size_t size, int flags) | |
a05d2ad1 EB |
1858 | { |
1859 | struct sock *sk = sock->sk; | |
1860 | ||
1861 | if (sk->sk_state != TCP_ESTABLISHED) | |
1862 | return -ENOTCONN; | |
1863 | ||
1b784140 | 1864 | return unix_dgram_recvmsg(sock, msg, size, flags); |
a05d2ad1 EB |
1865 | } |
1866 | ||
1da177e4 LT |
1867 | static void unix_copy_addr(struct msghdr *msg, struct sock *sk) |
1868 | { | |
1869 | struct unix_sock *u = unix_sk(sk); | |
1870 | ||
1da177e4 LT |
1871 | if (u->addr) { |
1872 | msg->msg_namelen = u->addr->len; | |
1873 | memcpy(msg->msg_name, u->addr->name, u->addr->len); | |
1874 | } | |
1875 | } | |
1876 | ||
1b784140 YX |
1877 | static int unix_dgram_recvmsg(struct socket *sock, struct msghdr *msg, |
1878 | size_t size, int flags) | |
1da177e4 | 1879 | { |
7cc05662 | 1880 | struct scm_cookie scm; |
1da177e4 LT |
1881 | struct sock *sk = sock->sk; |
1882 | struct unix_sock *u = unix_sk(sk); | |
1883 | int noblock = flags & MSG_DONTWAIT; | |
1884 | struct sk_buff *skb; | |
1885 | int err; | |
f55bb7f9 | 1886 | int peeked, skip; |
1da177e4 LT |
1887 | |
1888 | err = -EOPNOTSUPP; | |
1889 | if (flags&MSG_OOB) | |
1890 | goto out; | |
1891 | ||
b3ca9b02 | 1892 | err = mutex_lock_interruptible(&u->readlock); |
de144391 ED |
1893 | if (unlikely(err)) { |
1894 | /* recvmsg() in non blocking mode is supposed to return -EAGAIN | |
1895 | * sk_rcvtimeo is not honored by mutex_lock_interruptible() | |
1896 | */ | |
1897 | err = noblock ? -EAGAIN : -ERESTARTSYS; | |
b3ca9b02 RW |
1898 | goto out; |
1899 | } | |
1da177e4 | 1900 | |
f55bb7f9 PE |
1901 | skip = sk_peek_offset(sk, flags); |
1902 | ||
1903 | skb = __skb_recv_datagram(sk, flags, &peeked, &skip, &err); | |
0a112258 FZ |
1904 | if (!skb) { |
1905 | unix_state_lock(sk); | |
1906 | /* Signal EOF on disconnected non-blocking SEQPACKET socket. */ | |
1907 | if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN && | |
1908 | (sk->sk_shutdown & RCV_SHUTDOWN)) | |
1909 | err = 0; | |
1910 | unix_state_unlock(sk); | |
1da177e4 | 1911 | goto out_unlock; |
0a112258 | 1912 | } |
1da177e4 | 1913 | |
67426b75 ED |
1914 | wake_up_interruptible_sync_poll(&u->peer_wait, |
1915 | POLLOUT | POLLWRNORM | POLLWRBAND); | |
1da177e4 LT |
1916 | |
1917 | if (msg->msg_name) | |
1918 | unix_copy_addr(msg, skb->sk); | |
1919 | ||
f55bb7f9 PE |
1920 | if (size > skb->len - skip) |
1921 | size = skb->len - skip; | |
1922 | else if (size < skb->len - skip) | |
1da177e4 LT |
1923 | msg->msg_flags |= MSG_TRUNC; |
1924 | ||
51f3d02b | 1925 | err = skb_copy_datagram_msg(skb, skip, msg, size); |
1da177e4 LT |
1926 | if (err) |
1927 | goto out_free; | |
1928 | ||
3f66116e AC |
1929 | if (sock_flag(sk, SOCK_RCVTSTAMP)) |
1930 | __sock_recv_timestamp(msg, sk, skb); | |
1931 | ||
7cc05662 CH |
1932 | memset(&scm, 0, sizeof(scm)); |
1933 | ||
1934 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); | |
1935 | unix_set_secdata(&scm, skb); | |
1da177e4 | 1936 | |
6eba6a37 | 1937 | if (!(flags & MSG_PEEK)) { |
1da177e4 | 1938 | if (UNIXCB(skb).fp) |
7cc05662 | 1939 | unix_detach_fds(&scm, skb); |
f55bb7f9 PE |
1940 | |
1941 | sk_peek_offset_bwd(sk, skb->len); | |
6eba6a37 | 1942 | } else { |
1da177e4 LT |
1943 | /* It is questionable: on PEEK we could: |
1944 | - do not return fds - good, but too simple 8) | |
1945 | - return fds, and do not return them on read (old strategy, | |
1946 | apparently wrong) | |
1947 | - clone fds (I chose it for now, it is the most universal | |
1948 | solution) | |
ac7bfa62 YH |
1949 | |
1950 | POSIX 1003.1g does not actually define this clearly | |
1951 | at all. POSIX 1003.1g doesn't define a lot of things | |
1952 | clearly however! | |
1953 | ||
1da177e4 | 1954 | */ |
f55bb7f9 PE |
1955 | |
1956 | sk_peek_offset_fwd(sk, size); | |
1957 | ||
1da177e4 | 1958 | if (UNIXCB(skb).fp) |
7cc05662 | 1959 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 1960 | } |
9f6f9af7 | 1961 | err = (flags & MSG_TRUNC) ? skb->len - skip : size; |
1da177e4 | 1962 | |
7cc05662 | 1963 | scm_recv(sock, msg, &scm, flags); |
1da177e4 LT |
1964 | |
1965 | out_free: | |
6eba6a37 | 1966 | skb_free_datagram(sk, skb); |
1da177e4 | 1967 | out_unlock: |
57b47a53 | 1968 | mutex_unlock(&u->readlock); |
1da177e4 LT |
1969 | out: |
1970 | return err; | |
1971 | } | |
1972 | ||
1973 | /* | |
79f632c7 | 1974 | * Sleep until more data has arrived. But check for races.. |
1da177e4 | 1975 | */ |
79f632c7 | 1976 | static long unix_stream_data_wait(struct sock *sk, long timeo, |
2b514574 | 1977 | struct sk_buff *last, unsigned int last_len) |
1da177e4 | 1978 | { |
2b514574 | 1979 | struct sk_buff *tail; |
1da177e4 LT |
1980 | DEFINE_WAIT(wait); |
1981 | ||
1c92b4e5 | 1982 | unix_state_lock(sk); |
1da177e4 LT |
1983 | |
1984 | for (;;) { | |
aa395145 | 1985 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 1986 | |
2b514574 HFS |
1987 | tail = skb_peek_tail(&sk->sk_receive_queue); |
1988 | if (tail != last || | |
1989 | (tail && tail->len != last_len) || | |
1da177e4 LT |
1990 | sk->sk_err || |
1991 | (sk->sk_shutdown & RCV_SHUTDOWN) || | |
1992 | signal_pending(current) || | |
1993 | !timeo) | |
1994 | break; | |
1995 | ||
1996 | set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); | |
1c92b4e5 | 1997 | unix_state_unlock(sk); |
2b15af6f | 1998 | timeo = freezable_schedule_timeout(timeo); |
1c92b4e5 | 1999 | unix_state_lock(sk); |
b48732e4 MS |
2000 | |
2001 | if (sock_flag(sk, SOCK_DEAD)) | |
2002 | break; | |
2003 | ||
1da177e4 LT |
2004 | clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags); |
2005 | } | |
2006 | ||
aa395145 | 2007 | finish_wait(sk_sleep(sk), &wait); |
1c92b4e5 | 2008 | unix_state_unlock(sk); |
1da177e4 LT |
2009 | return timeo; |
2010 | } | |
2011 | ||
e370a723 ED |
2012 | static unsigned int unix_skb_len(const struct sk_buff *skb) |
2013 | { | |
2014 | return skb->len - UNIXCB(skb).consumed; | |
2015 | } | |
2016 | ||
2b514574 HFS |
2017 | struct unix_stream_read_state { |
2018 | int (*recv_actor)(struct sk_buff *, int, int, | |
2019 | struct unix_stream_read_state *); | |
2020 | struct socket *socket; | |
2021 | struct msghdr *msg; | |
2022 | struct pipe_inode_info *pipe; | |
2023 | size_t size; | |
2024 | int flags; | |
2025 | unsigned int splice_flags; | |
2026 | }; | |
2027 | ||
2028 | static int unix_stream_read_generic(struct unix_stream_read_state *state) | |
1da177e4 | 2029 | { |
7cc05662 | 2030 | struct scm_cookie scm; |
2b514574 | 2031 | struct socket *sock = state->socket; |
1da177e4 LT |
2032 | struct sock *sk = sock->sk; |
2033 | struct unix_sock *u = unix_sk(sk); | |
1da177e4 | 2034 | int copied = 0; |
2b514574 | 2035 | int flags = state->flags; |
de144391 | 2036 | int noblock = flags & MSG_DONTWAIT; |
2b514574 | 2037 | bool check_creds = false; |
1da177e4 LT |
2038 | int target; |
2039 | int err = 0; | |
2040 | long timeo; | |
fc0d7536 | 2041 | int skip; |
2b514574 HFS |
2042 | size_t size = state->size; |
2043 | unsigned int last_len; | |
1da177e4 LT |
2044 | |
2045 | err = -EINVAL; | |
2046 | if (sk->sk_state != TCP_ESTABLISHED) | |
2047 | goto out; | |
2048 | ||
2049 | err = -EOPNOTSUPP; | |
2b514574 | 2050 | if (flags & MSG_OOB) |
1da177e4 LT |
2051 | goto out; |
2052 | ||
2b514574 | 2053 | target = sock_rcvlowat(sk, flags & MSG_WAITALL, size); |
de144391 | 2054 | timeo = sock_rcvtimeo(sk, noblock); |
1da177e4 | 2055 | |
2b514574 HFS |
2056 | memset(&scm, 0, sizeof(scm)); |
2057 | ||
1da177e4 LT |
2058 | /* Lock the socket to prevent queue disordering |
2059 | * while sleeps in memcpy_tomsg | |
2060 | */ | |
b3ca9b02 | 2061 | err = mutex_lock_interruptible(&u->readlock); |
de144391 ED |
2062 | if (unlikely(err)) { |
2063 | /* recvmsg() in non blocking mode is supposed to return -EAGAIN | |
2064 | * sk_rcvtimeo is not honored by mutex_lock_interruptible() | |
2065 | */ | |
2066 | err = noblock ? -EAGAIN : -ERESTARTSYS; | |
b3ca9b02 RW |
2067 | goto out; |
2068 | } | |
1da177e4 | 2069 | |
e9193d60 AV |
2070 | if (flags & MSG_PEEK) |
2071 | skip = sk_peek_offset(sk, flags); | |
2072 | else | |
2073 | skip = 0; | |
2074 | ||
6eba6a37 | 2075 | do { |
1da177e4 | 2076 | int chunk; |
73ed5d25 | 2077 | bool drop_skb; |
79f632c7 | 2078 | struct sk_buff *skb, *last; |
1da177e4 | 2079 | |
3c0d2f37 | 2080 | unix_state_lock(sk); |
b48732e4 MS |
2081 | if (sock_flag(sk, SOCK_DEAD)) { |
2082 | err = -ECONNRESET; | |
2083 | goto unlock; | |
2084 | } | |
79f632c7 | 2085 | last = skb = skb_peek(&sk->sk_receive_queue); |
2b514574 | 2086 | last_len = last ? last->len : 0; |
fc0d7536 | 2087 | again: |
6eba6a37 | 2088 | if (skb == NULL) { |
25888e30 | 2089 | unix_sk(sk)->recursion_level = 0; |
1da177e4 | 2090 | if (copied >= target) |
3c0d2f37 | 2091 | goto unlock; |
1da177e4 LT |
2092 | |
2093 | /* | |
2094 | * POSIX 1003.1g mandates this order. | |
2095 | */ | |
ac7bfa62 | 2096 | |
6eba6a37 ED |
2097 | err = sock_error(sk); |
2098 | if (err) | |
3c0d2f37 | 2099 | goto unlock; |
1da177e4 | 2100 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
3c0d2f37 MS |
2101 | goto unlock; |
2102 | ||
2103 | unix_state_unlock(sk); | |
1da177e4 LT |
2104 | err = -EAGAIN; |
2105 | if (!timeo) | |
2106 | break; | |
57b47a53 | 2107 | mutex_unlock(&u->readlock); |
1da177e4 | 2108 | |
2b514574 HFS |
2109 | timeo = unix_stream_data_wait(sk, timeo, last, |
2110 | last_len); | |
1da177e4 | 2111 | |
2b514574 HFS |
2112 | if (signal_pending(current) || |
2113 | mutex_lock_interruptible(&u->readlock)) { | |
1da177e4 LT |
2114 | err = sock_intr_errno(timeo); |
2115 | goto out; | |
2116 | } | |
b3ca9b02 | 2117 | |
1da177e4 | 2118 | continue; |
2b514574 | 2119 | unlock: |
3c0d2f37 MS |
2120 | unix_state_unlock(sk); |
2121 | break; | |
1da177e4 | 2122 | } |
fc0d7536 | 2123 | |
e370a723 ED |
2124 | while (skip >= unix_skb_len(skb)) { |
2125 | skip -= unix_skb_len(skb); | |
79f632c7 | 2126 | last = skb; |
2b514574 | 2127 | last_len = skb->len; |
fc0d7536 | 2128 | skb = skb_peek_next(skb, &sk->sk_receive_queue); |
79f632c7 BP |
2129 | if (!skb) |
2130 | goto again; | |
fc0d7536 PE |
2131 | } |
2132 | ||
3c0d2f37 | 2133 | unix_state_unlock(sk); |
1da177e4 LT |
2134 | |
2135 | if (check_creds) { | |
2136 | /* Never glue messages from different writers */ | |
7cc05662 CH |
2137 | if ((UNIXCB(skb).pid != scm.pid) || |
2138 | !uid_eq(UNIXCB(skb).uid, scm.creds.uid) || | |
37a9a8df SS |
2139 | !gid_eq(UNIXCB(skb).gid, scm.creds.gid) || |
2140 | !unix_secdata_eq(&scm, skb)) | |
1da177e4 | 2141 | break; |
0e82e7f6 | 2142 | } else if (test_bit(SOCK_PASSCRED, &sock->flags)) { |
1da177e4 | 2143 | /* Copy credentials */ |
7cc05662 | 2144 | scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid); |
37a9a8df | 2145 | unix_set_secdata(&scm, skb); |
2b514574 | 2146 | check_creds = true; |
1da177e4 LT |
2147 | } |
2148 | ||
2149 | /* Copy address just once */ | |
2b514574 HFS |
2150 | if (state->msg && state->msg->msg_name) { |
2151 | DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, | |
2152 | state->msg->msg_name); | |
2153 | unix_copy_addr(state->msg, skb->sk); | |
1da177e4 LT |
2154 | sunaddr = NULL; |
2155 | } | |
2156 | ||
e370a723 | 2157 | chunk = min_t(unsigned int, unix_skb_len(skb) - skip, size); |
73ed5d25 | 2158 | skb_get(skb); |
2b514574 | 2159 | chunk = state->recv_actor(skb, skip, chunk, state); |
73ed5d25 HFS |
2160 | drop_skb = !unix_skb_len(skb); |
2161 | /* skb is only safe to use if !drop_skb */ | |
2162 | consume_skb(skb); | |
2b514574 | 2163 | if (chunk < 0) { |
1da177e4 LT |
2164 | if (copied == 0) |
2165 | copied = -EFAULT; | |
2166 | break; | |
2167 | } | |
2168 | copied += chunk; | |
2169 | size -= chunk; | |
2170 | ||
73ed5d25 HFS |
2171 | if (drop_skb) { |
2172 | /* the skb was touched by a concurrent reader; | |
2173 | * we should not expect anything from this skb | |
2174 | * anymore and assume it invalid - we can be | |
2175 | * sure it was dropped from the socket queue | |
2176 | * | |
2177 | * let's report a short read | |
2178 | */ | |
2179 | err = 0; | |
2180 | break; | |
2181 | } | |
2182 | ||
1da177e4 | 2183 | /* Mark read part of skb as used */ |
6eba6a37 | 2184 | if (!(flags & MSG_PEEK)) { |
e370a723 | 2185 | UNIXCB(skb).consumed += chunk; |
1da177e4 | 2186 | |
fc0d7536 PE |
2187 | sk_peek_offset_bwd(sk, chunk); |
2188 | ||
1da177e4 | 2189 | if (UNIXCB(skb).fp) |
7cc05662 | 2190 | unix_detach_fds(&scm, skb); |
1da177e4 | 2191 | |
e370a723 | 2192 | if (unix_skb_len(skb)) |
1da177e4 | 2193 | break; |
1da177e4 | 2194 | |
6f01fd6e | 2195 | skb_unlink(skb, &sk->sk_receive_queue); |
70d4bf6d | 2196 | consume_skb(skb); |
1da177e4 | 2197 | |
7cc05662 | 2198 | if (scm.fp) |
1da177e4 | 2199 | break; |
6eba6a37 | 2200 | } else { |
1da177e4 LT |
2201 | /* It is questionable, see note in unix_dgram_recvmsg. |
2202 | */ | |
2203 | if (UNIXCB(skb).fp) | |
7cc05662 | 2204 | scm.fp = scm_fp_dup(UNIXCB(skb).fp); |
1da177e4 | 2205 | |
e9193d60 | 2206 | sk_peek_offset_fwd(sk, chunk); |
fc0d7536 | 2207 | |
9f389e35 AC |
2208 | if (UNIXCB(skb).fp) |
2209 | break; | |
2210 | ||
e9193d60 | 2211 | skip = 0; |
9f389e35 AC |
2212 | last = skb; |
2213 | last_len = skb->len; | |
2214 | unix_state_lock(sk); | |
2215 | skb = skb_peek_next(skb, &sk->sk_receive_queue); | |
2216 | if (skb) | |
2217 | goto again; | |
2218 | unix_state_unlock(sk); | |
1da177e4 LT |
2219 | break; |
2220 | } | |
2221 | } while (size); | |
2222 | ||
57b47a53 | 2223 | mutex_unlock(&u->readlock); |
2b514574 HFS |
2224 | if (state->msg) |
2225 | scm_recv(sock, state->msg, &scm, flags); | |
2226 | else | |
2227 | scm_destroy(&scm); | |
1da177e4 LT |
2228 | out: |
2229 | return copied ? : err; | |
2230 | } | |
2231 | ||
2b514574 HFS |
2232 | static int unix_stream_read_actor(struct sk_buff *skb, |
2233 | int skip, int chunk, | |
2234 | struct unix_stream_read_state *state) | |
2235 | { | |
2236 | int ret; | |
2237 | ||
2238 | ret = skb_copy_datagram_msg(skb, UNIXCB(skb).consumed + skip, | |
2239 | state->msg, chunk); | |
2240 | return ret ?: chunk; | |
2241 | } | |
2242 | ||
2243 | static int unix_stream_recvmsg(struct socket *sock, struct msghdr *msg, | |
2244 | size_t size, int flags) | |
2245 | { | |
2246 | struct unix_stream_read_state state = { | |
2247 | .recv_actor = unix_stream_read_actor, | |
2248 | .socket = sock, | |
2249 | .msg = msg, | |
2250 | .size = size, | |
2251 | .flags = flags | |
2252 | }; | |
2253 | ||
2254 | return unix_stream_read_generic(&state); | |
2255 | } | |
2256 | ||
2257 | static ssize_t skb_unix_socket_splice(struct sock *sk, | |
2258 | struct pipe_inode_info *pipe, | |
2259 | struct splice_pipe_desc *spd) | |
2260 | { | |
2261 | int ret; | |
2262 | struct unix_sock *u = unix_sk(sk); | |
2263 | ||
2264 | mutex_unlock(&u->readlock); | |
2265 | ret = splice_to_pipe(pipe, spd); | |
2266 | mutex_lock(&u->readlock); | |
2267 | ||
2268 | return ret; | |
2269 | } | |
2270 | ||
2271 | static int unix_stream_splice_actor(struct sk_buff *skb, | |
2272 | int skip, int chunk, | |
2273 | struct unix_stream_read_state *state) | |
2274 | { | |
2275 | return skb_splice_bits(skb, state->socket->sk, | |
2276 | UNIXCB(skb).consumed + skip, | |
2277 | state->pipe, chunk, state->splice_flags, | |
2278 | skb_unix_socket_splice); | |
2279 | } | |
2280 | ||
2281 | static ssize_t unix_stream_splice_read(struct socket *sock, loff_t *ppos, | |
2282 | struct pipe_inode_info *pipe, | |
2283 | size_t size, unsigned int flags) | |
2284 | { | |
2285 | struct unix_stream_read_state state = { | |
2286 | .recv_actor = unix_stream_splice_actor, | |
2287 | .socket = sock, | |
2288 | .pipe = pipe, | |
2289 | .size = size, | |
2290 | .splice_flags = flags, | |
2291 | }; | |
2292 | ||
2293 | if (unlikely(*ppos)) | |
2294 | return -ESPIPE; | |
2295 | ||
2296 | if (sock->file->f_flags & O_NONBLOCK || | |
2297 | flags & SPLICE_F_NONBLOCK) | |
2298 | state.flags = MSG_DONTWAIT; | |
2299 | ||
2300 | return unix_stream_read_generic(&state); | |
2301 | } | |
2302 | ||
1da177e4 LT |
2303 | static int unix_shutdown(struct socket *sock, int mode) |
2304 | { | |
2305 | struct sock *sk = sock->sk; | |
2306 | struct sock *other; | |
2307 | ||
fc61b928 XW |
2308 | if (mode < SHUT_RD || mode > SHUT_RDWR) |
2309 | return -EINVAL; | |
2310 | /* This maps: | |
2311 | * SHUT_RD (0) -> RCV_SHUTDOWN (1) | |
2312 | * SHUT_WR (1) -> SEND_SHUTDOWN (2) | |
2313 | * SHUT_RDWR (2) -> SHUTDOWN_MASK (3) | |
2314 | */ | |
2315 | ++mode; | |
7180a031 AC |
2316 | |
2317 | unix_state_lock(sk); | |
2318 | sk->sk_shutdown |= mode; | |
2319 | other = unix_peer(sk); | |
2320 | if (other) | |
2321 | sock_hold(other); | |
2322 | unix_state_unlock(sk); | |
2323 | sk->sk_state_change(sk); | |
2324 | ||
2325 | if (other && | |
2326 | (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) { | |
2327 | ||
2328 | int peer_mode = 0; | |
2329 | ||
2330 | if (mode&RCV_SHUTDOWN) | |
2331 | peer_mode |= SEND_SHUTDOWN; | |
2332 | if (mode&SEND_SHUTDOWN) | |
2333 | peer_mode |= RCV_SHUTDOWN; | |
2334 | unix_state_lock(other); | |
2335 | other->sk_shutdown |= peer_mode; | |
2336 | unix_state_unlock(other); | |
2337 | other->sk_state_change(other); | |
2338 | if (peer_mode == SHUTDOWN_MASK) | |
2339 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP); | |
2340 | else if (peer_mode & RCV_SHUTDOWN) | |
2341 | sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN); | |
1da177e4 | 2342 | } |
7180a031 AC |
2343 | if (other) |
2344 | sock_put(other); | |
2345 | ||
1da177e4 LT |
2346 | return 0; |
2347 | } | |
2348 | ||
885ee74d PE |
2349 | long unix_inq_len(struct sock *sk) |
2350 | { | |
2351 | struct sk_buff *skb; | |
2352 | long amount = 0; | |
2353 | ||
2354 | if (sk->sk_state == TCP_LISTEN) | |
2355 | return -EINVAL; | |
2356 | ||
2357 | spin_lock(&sk->sk_receive_queue.lock); | |
2358 | if (sk->sk_type == SOCK_STREAM || | |
2359 | sk->sk_type == SOCK_SEQPACKET) { | |
2360 | skb_queue_walk(&sk->sk_receive_queue, skb) | |
e370a723 | 2361 | amount += unix_skb_len(skb); |
885ee74d PE |
2362 | } else { |
2363 | skb = skb_peek(&sk->sk_receive_queue); | |
2364 | if (skb) | |
2365 | amount = skb->len; | |
2366 | } | |
2367 | spin_unlock(&sk->sk_receive_queue.lock); | |
2368 | ||
2369 | return amount; | |
2370 | } | |
2371 | EXPORT_SYMBOL_GPL(unix_inq_len); | |
2372 | ||
2373 | long unix_outq_len(struct sock *sk) | |
2374 | { | |
2375 | return sk_wmem_alloc_get(sk); | |
2376 | } | |
2377 | EXPORT_SYMBOL_GPL(unix_outq_len); | |
2378 | ||
1da177e4 LT |
2379 | static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
2380 | { | |
2381 | struct sock *sk = sock->sk; | |
e27dfcea | 2382 | long amount = 0; |
1da177e4 LT |
2383 | int err; |
2384 | ||
6eba6a37 ED |
2385 | switch (cmd) { |
2386 | case SIOCOUTQ: | |
885ee74d | 2387 | amount = unix_outq_len(sk); |
6eba6a37 ED |
2388 | err = put_user(amount, (int __user *)arg); |
2389 | break; | |
2390 | case SIOCINQ: | |
885ee74d PE |
2391 | amount = unix_inq_len(sk); |
2392 | if (amount < 0) | |
2393 | err = amount; | |
2394 | else | |
1da177e4 | 2395 | err = put_user(amount, (int __user *)arg); |
885ee74d | 2396 | break; |
6eba6a37 ED |
2397 | default: |
2398 | err = -ENOIOCTLCMD; | |
2399 | break; | |
1da177e4 LT |
2400 | } |
2401 | return err; | |
2402 | } | |
2403 | ||
6eba6a37 | 2404 | static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait) |
1da177e4 LT |
2405 | { |
2406 | struct sock *sk = sock->sk; | |
2407 | unsigned int mask; | |
2408 | ||
aa395145 | 2409 | sock_poll_wait(file, sk_sleep(sk), wait); |
1da177e4 LT |
2410 | mask = 0; |
2411 | ||
2412 | /* exceptional events? */ | |
2413 | if (sk->sk_err) | |
2414 | mask |= POLLERR; | |
2415 | if (sk->sk_shutdown == SHUTDOWN_MASK) | |
2416 | mask |= POLLHUP; | |
f348d70a | 2417 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
db40980f | 2418 | mask |= POLLRDHUP | POLLIN | POLLRDNORM; |
1da177e4 LT |
2419 | |
2420 | /* readable? */ | |
db40980f | 2421 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
2422 | mask |= POLLIN | POLLRDNORM; |
2423 | ||
2424 | /* Connection-based need to check for termination and startup */ | |
6eba6a37 ED |
2425 | if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) && |
2426 | sk->sk_state == TCP_CLOSE) | |
1da177e4 LT |
2427 | mask |= POLLHUP; |
2428 | ||
2429 | /* | |
2430 | * we set writable also when the other side has shut down the | |
2431 | * connection. This prevents stuck sockets. | |
2432 | */ | |
2433 | if (unix_writable(sk)) | |
2434 | mask |= POLLOUT | POLLWRNORM | POLLWRBAND; | |
2435 | ||
2436 | return mask; | |
2437 | } | |
2438 | ||
ec0d215f RW |
2439 | static unsigned int unix_dgram_poll(struct file *file, struct socket *sock, |
2440 | poll_table *wait) | |
3c73419c | 2441 | { |
ec0d215f RW |
2442 | struct sock *sk = sock->sk, *other; |
2443 | unsigned int mask, writable; | |
3c73419c | 2444 | |
aa395145 | 2445 | sock_poll_wait(file, sk_sleep(sk), wait); |
3c73419c RW |
2446 | mask = 0; |
2447 | ||
2448 | /* exceptional events? */ | |
2449 | if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue)) | |
7d4c04fc | 2450 | mask |= POLLERR | |
8facd5fb | 2451 | (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0); |
7d4c04fc | 2452 | |
3c73419c | 2453 | if (sk->sk_shutdown & RCV_SHUTDOWN) |
5456f09a | 2454 | mask |= POLLRDHUP | POLLIN | POLLRDNORM; |
3c73419c RW |
2455 | if (sk->sk_shutdown == SHUTDOWN_MASK) |
2456 | mask |= POLLHUP; | |
2457 | ||
2458 | /* readable? */ | |
5456f09a | 2459 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
3c73419c RW |
2460 | mask |= POLLIN | POLLRDNORM; |
2461 | ||
2462 | /* Connection-based need to check for termination and startup */ | |
2463 | if (sk->sk_type == SOCK_SEQPACKET) { | |
2464 | if (sk->sk_state == TCP_CLOSE) | |
2465 | mask |= POLLHUP; | |
2466 | /* connection hasn't started yet? */ | |
2467 | if (sk->sk_state == TCP_SYN_SENT) | |
2468 | return mask; | |
2469 | } | |
2470 | ||
973a34aa | 2471 | /* No write status requested, avoid expensive OUT tests. */ |
626cf236 | 2472 | if (!(poll_requested_events(wait) & (POLLWRBAND|POLLWRNORM|POLLOUT))) |
973a34aa ED |
2473 | return mask; |
2474 | ||
ec0d215f | 2475 | writable = unix_writable(sk); |
5456f09a ED |
2476 | other = unix_peer_get(sk); |
2477 | if (other) { | |
2478 | if (unix_peer(other) != sk) { | |
2479 | sock_poll_wait(file, &unix_sk(other)->peer_wait, wait); | |
2480 | if (unix_recvq_full(other)) | |
2481 | writable = 0; | |
ec0d215f | 2482 | } |
5456f09a | 2483 | sock_put(other); |
ec0d215f RW |
2484 | } |
2485 | ||
2486 | if (writable) | |
3c73419c RW |
2487 | mask |= POLLOUT | POLLWRNORM | POLLWRBAND; |
2488 | else | |
2489 | set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); | |
2490 | ||
3c73419c RW |
2491 | return mask; |
2492 | } | |
1da177e4 LT |
2493 | |
2494 | #ifdef CONFIG_PROC_FS | |
a53eb3fe | 2495 | |
7123aaa3 ED |
2496 | #define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1) |
2497 | ||
2498 | #define get_bucket(x) ((x) >> BUCKET_SPACE) | |
2499 | #define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1)) | |
2500 | #define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o)) | |
a53eb3fe | 2501 | |
7123aaa3 | 2502 | static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos) |
1da177e4 | 2503 | { |
7123aaa3 ED |
2504 | unsigned long offset = get_offset(*pos); |
2505 | unsigned long bucket = get_bucket(*pos); | |
2506 | struct sock *sk; | |
2507 | unsigned long count = 0; | |
1da177e4 | 2508 | |
7123aaa3 ED |
2509 | for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) { |
2510 | if (sock_net(sk) != seq_file_net(seq)) | |
097e66c5 | 2511 | continue; |
7123aaa3 ED |
2512 | if (++count == offset) |
2513 | break; | |
2514 | } | |
2515 | ||
2516 | return sk; | |
2517 | } | |
2518 | ||
2519 | static struct sock *unix_next_socket(struct seq_file *seq, | |
2520 | struct sock *sk, | |
2521 | loff_t *pos) | |
2522 | { | |
2523 | unsigned long bucket; | |
2524 | ||
2525 | while (sk > (struct sock *)SEQ_START_TOKEN) { | |
2526 | sk = sk_next(sk); | |
2527 | if (!sk) | |
2528 | goto next_bucket; | |
2529 | if (sock_net(sk) == seq_file_net(seq)) | |
2530 | return sk; | |
1da177e4 | 2531 | } |
7123aaa3 ED |
2532 | |
2533 | do { | |
2534 | sk = unix_from_bucket(seq, pos); | |
2535 | if (sk) | |
2536 | return sk; | |
2537 | ||
2538 | next_bucket: | |
2539 | bucket = get_bucket(*pos) + 1; | |
2540 | *pos = set_bucket_offset(bucket, 1); | |
2541 | } while (bucket < ARRAY_SIZE(unix_socket_table)); | |
2542 | ||
1da177e4 LT |
2543 | return NULL; |
2544 | } | |
2545 | ||
1da177e4 | 2546 | static void *unix_seq_start(struct seq_file *seq, loff_t *pos) |
9a429c49 | 2547 | __acquires(unix_table_lock) |
1da177e4 | 2548 | { |
fbe9cc4a | 2549 | spin_lock(&unix_table_lock); |
7123aaa3 ED |
2550 | |
2551 | if (!*pos) | |
2552 | return SEQ_START_TOKEN; | |
2553 | ||
2554 | if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table)) | |
2555 | return NULL; | |
2556 | ||
2557 | return unix_next_socket(seq, NULL, pos); | |
1da177e4 LT |
2558 | } |
2559 | ||
2560 | static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2561 | { | |
2562 | ++*pos; | |
7123aaa3 | 2563 | return unix_next_socket(seq, v, pos); |
1da177e4 LT |
2564 | } |
2565 | ||
2566 | static void unix_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 2567 | __releases(unix_table_lock) |
1da177e4 | 2568 | { |
fbe9cc4a | 2569 | spin_unlock(&unix_table_lock); |
1da177e4 LT |
2570 | } |
2571 | ||
2572 | static int unix_seq_show(struct seq_file *seq, void *v) | |
2573 | { | |
ac7bfa62 | 2574 | |
b9f3124f | 2575 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2576 | seq_puts(seq, "Num RefCount Protocol Flags Type St " |
2577 | "Inode Path\n"); | |
2578 | else { | |
2579 | struct sock *s = v; | |
2580 | struct unix_sock *u = unix_sk(s); | |
1c92b4e5 | 2581 | unix_state_lock(s); |
1da177e4 | 2582 | |
71338aa7 | 2583 | seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu", |
1da177e4 LT |
2584 | s, |
2585 | atomic_read(&s->sk_refcnt), | |
2586 | 0, | |
2587 | s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0, | |
2588 | s->sk_type, | |
2589 | s->sk_socket ? | |
2590 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) : | |
2591 | (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING), | |
2592 | sock_i_ino(s)); | |
2593 | ||
2594 | if (u->addr) { | |
2595 | int i, len; | |
2596 | seq_putc(seq, ' '); | |
2597 | ||
2598 | i = 0; | |
2599 | len = u->addr->len - sizeof(short); | |
2600 | if (!UNIX_ABSTRACT(s)) | |
2601 | len--; | |
2602 | else { | |
2603 | seq_putc(seq, '@'); | |
2604 | i++; | |
2605 | } | |
2606 | for ( ; i < len; i++) | |
2607 | seq_putc(seq, u->addr->name->sun_path[i]); | |
2608 | } | |
1c92b4e5 | 2609 | unix_state_unlock(s); |
1da177e4 LT |
2610 | seq_putc(seq, '\n'); |
2611 | } | |
2612 | ||
2613 | return 0; | |
2614 | } | |
2615 | ||
56b3d975 | 2616 | static const struct seq_operations unix_seq_ops = { |
1da177e4 LT |
2617 | .start = unix_seq_start, |
2618 | .next = unix_seq_next, | |
2619 | .stop = unix_seq_stop, | |
2620 | .show = unix_seq_show, | |
2621 | }; | |
2622 | ||
1da177e4 LT |
2623 | static int unix_seq_open(struct inode *inode, struct file *file) |
2624 | { | |
e372c414 | 2625 | return seq_open_net(inode, file, &unix_seq_ops, |
8b51b064 | 2626 | sizeof(struct seq_net_private)); |
1da177e4 LT |
2627 | } |
2628 | ||
da7071d7 | 2629 | static const struct file_operations unix_seq_fops = { |
1da177e4 LT |
2630 | .owner = THIS_MODULE, |
2631 | .open = unix_seq_open, | |
2632 | .read = seq_read, | |
2633 | .llseek = seq_lseek, | |
e372c414 | 2634 | .release = seq_release_net, |
1da177e4 LT |
2635 | }; |
2636 | ||
2637 | #endif | |
2638 | ||
ec1b4cf7 | 2639 | static const struct net_proto_family unix_family_ops = { |
1da177e4 LT |
2640 | .family = PF_UNIX, |
2641 | .create = unix_create, | |
2642 | .owner = THIS_MODULE, | |
2643 | }; | |
2644 | ||
097e66c5 | 2645 | |
2c8c1e72 | 2646 | static int __net_init unix_net_init(struct net *net) |
097e66c5 DL |
2647 | { |
2648 | int error = -ENOMEM; | |
2649 | ||
a0a53c8b | 2650 | net->unx.sysctl_max_dgram_qlen = 10; |
1597fbc0 PE |
2651 | if (unix_sysctl_register(net)) |
2652 | goto out; | |
d392e497 | 2653 | |
097e66c5 | 2654 | #ifdef CONFIG_PROC_FS |
d4beaa66 | 2655 | if (!proc_create("unix", 0, net->proc_net, &unix_seq_fops)) { |
1597fbc0 | 2656 | unix_sysctl_unregister(net); |
097e66c5 | 2657 | goto out; |
1597fbc0 | 2658 | } |
097e66c5 DL |
2659 | #endif |
2660 | error = 0; | |
2661 | out: | |
48dcc33e | 2662 | return error; |
097e66c5 DL |
2663 | } |
2664 | ||
2c8c1e72 | 2665 | static void __net_exit unix_net_exit(struct net *net) |
097e66c5 | 2666 | { |
1597fbc0 | 2667 | unix_sysctl_unregister(net); |
ece31ffd | 2668 | remove_proc_entry("unix", net->proc_net); |
097e66c5 DL |
2669 | } |
2670 | ||
2671 | static struct pernet_operations unix_net_ops = { | |
2672 | .init = unix_net_init, | |
2673 | .exit = unix_net_exit, | |
2674 | }; | |
2675 | ||
1da177e4 LT |
2676 | static int __init af_unix_init(void) |
2677 | { | |
2678 | int rc = -1; | |
1da177e4 | 2679 | |
b4fff5f8 | 2680 | BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 LT |
2681 | |
2682 | rc = proto_register(&unix_proto, 1); | |
ac7bfa62 | 2683 | if (rc != 0) { |
5cc208be | 2684 | pr_crit("%s: Cannot create unix_sock SLAB cache!\n", __func__); |
1da177e4 LT |
2685 | goto out; |
2686 | } | |
2687 | ||
2688 | sock_register(&unix_family_ops); | |
097e66c5 | 2689 | register_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2690 | out: |
2691 | return rc; | |
2692 | } | |
2693 | ||
2694 | static void __exit af_unix_exit(void) | |
2695 | { | |
2696 | sock_unregister(PF_UNIX); | |
1da177e4 | 2697 | proto_unregister(&unix_proto); |
097e66c5 | 2698 | unregister_pernet_subsys(&unix_net_ops); |
1da177e4 LT |
2699 | } |
2700 | ||
3d366960 DW |
2701 | /* Earlier than device_initcall() so that other drivers invoking |
2702 | request_module() don't end up in a loop when modprobe tries | |
2703 | to use a UNIX socket. But later than subsys_initcall() because | |
2704 | we depend on stuff initialised there */ | |
2705 | fs_initcall(af_unix_init); | |
1da177e4 LT |
2706 | module_exit(af_unix_exit); |
2707 | ||
2708 | MODULE_LICENSE("GPL"); | |
2709 | MODULE_ALIAS_NETPROTO(PF_UNIX); |