blob: b1d75beb7e20e8e3155a84be37161871023396ac [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NET4: Implementation of BSD Unix domain sockets.
3 *
Alan Cox113aa832008-10-13 19:01:08 -07004 * Authors: Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -07005 *
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 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fixes:
12 * Linus Torvalds : Assorted bug cures.
13 * Niibe Yutaka : async I/O support.
14 * Carsten Paeth : PF_UNIX check, address fixes.
15 * Alan Cox : Limit size of allocated blocks.
16 * Alan Cox : Fixed the stupid socketpair bug.
17 * Alan Cox : BSD compatibility fine tuning.
18 * Alan Cox : Fixed a bug in connect when interrupted.
19 * Alan Cox : Sorted out a proper draft version of
20 * file descriptor passing hacked up from
21 * Mike Shaver's work.
22 * Marty Leisner : Fixes to fd passing
23 * Nick Nevin : recvmsg bugfix.
24 * Alan Cox : Started proper garbage collector
25 * Heiko EiBfeldt : Missing verify_area check
26 * Alan Cox : Started POSIXisms
27 * Andreas Schwab : Replace inode by dentry for proper
28 * reference counting
29 * Kirk Petersen : Made this a module
30 * Christoph Rohland : Elegant non-blocking accept/connect algorithm.
31 * Lots of bug fixes.
32 * Alexey Kuznetosv : Repaired (I hope) bugs introduces
33 * by above two patches.
34 * Andrea Arcangeli : If possible we block in connect(2)
35 * if the max backlog of the listen socket
36 * is been reached. This won't break
37 * old apps and it will avoid huge amount
38 * of socks hashed (this for unix_gc()
39 * performances reasons).
40 * Security fix that limits the max
41 * number of socks to 2*max_files and
42 * the number of skb queueable in the
43 * dgram receiver.
44 * Artur Skawina : Hash function optimizations
45 * Alexey Kuznetsov : Full scale SMP. Lot of bugs are introduced 8)
46 * Malcolm Beattie : Set peercred for socketpair
47 * Michal Ostrowski : Module initialization cleanup.
48 * Arnaldo C. Melo : Remove MOD_{INC,DEC}_USE_COUNT,
49 * the core infrastructure is doing that
50 * for all net proto families now (2.5.69+)
51 *
52 *
53 * Known differences from reference BSD that was tested:
54 *
55 * [TO FIX]
56 * ECONNREFUSED is not returned from one end of a connected() socket to the
57 * other the moment one end closes.
58 * fstat() doesn't return st_dev=0, and give the blksize as high water mark
59 * and a fake inode identifier (nor the BSD first socket fstat twice bug).
60 * [NOT TO FIX]
61 * accept() returns a path name even if the connecting socket has closed
62 * in the meantime (BSD loses the path and gives up).
63 * accept() returns 0 length path for an unbound connector. BSD returns 16
64 * and a null first byte in the path (but not for gethost/peername - BSD bug ??)
65 * socketpair(...SOCK_RAW..) doesn't panic the kernel.
66 * BSD af_unix apparently has connect forgetting to block properly.
67 * (need to check this with the POSIX spec in detail)
68 *
69 * Differences from 2.0.0-11-... (ANK)
70 * Bug fixes and improvements.
71 * - client shutdown killed server socket.
72 * - removed all useless cli/sti pairs.
73 *
74 * Semantic changes/extensions.
75 * - generic control message passing.
76 * - SCM_CREDENTIALS control message.
77 * - "Abstract" (not FS based) socket bindings.
78 * Abstract names are sequences of bytes (not zero terminated)
79 * started by 0, so that this name space does not intersect
80 * with BSD names.
81 */
82
83#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070084#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070085#include <linux/signal.h>
86#include <linux/sched.h>
87#include <linux/errno.h>
88#include <linux/string.h>
89#include <linux/stat.h>
90#include <linux/dcache.h>
91#include <linux/namei.h>
92#include <linux/socket.h>
93#include <linux/un.h>
94#include <linux/fcntl.h>
95#include <linux/termios.h>
96#include <linux/sockios.h>
97#include <linux/net.h>
98#include <linux/in.h>
99#include <linux/fs.h>
100#include <linux/slab.h>
101#include <asm/uaccess.h>
102#include <linux/skbuff.h>
103#include <linux/netdevice.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200104#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#include <net/sock.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -0700106#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#include <net/af_unix.h>
108#include <linux/proc_fs.h>
109#include <linux/seq_file.h>
110#include <net/scm.h>
111#include <linux/init.h>
112#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113#include <linux/rtnetlink.h>
114#include <linux/mount.h>
115#include <net/checksum.h>
116#include <linux/security.h>
117
Adrian Bunk13111692007-07-30 18:05:45 -0700118static struct hlist_head unix_socket_table[UNIX_HASH_SIZE + 1];
119static DEFINE_SPINLOCK(unix_table_lock);
Eric Dumazet518de9b2010-10-26 14:22:44 -0700120static atomic_long_t unix_nr_socks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121
122#define unix_sockets_unbound (&unix_socket_table[UNIX_HASH_SIZE])
123
124#define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash != UNIX_HASH_SIZE)
125
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700126#ifdef CONFIG_SECURITY_NETWORK
Catherine Zhangdc49c1f2006-08-02 14:12:06 -0700127static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb)
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700128{
Catherine Zhangdc49c1f2006-08-02 14:12:06 -0700129 memcpy(UNIXSID(skb), &scm->secid, sizeof(u32));
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700130}
131
132static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb)
133{
Catherine Zhangdc49c1f2006-08-02 14:12:06 -0700134 scm->secid = *UNIXSID(skb);
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700135}
136#else
Catherine Zhangdc49c1f2006-08-02 14:12:06 -0700137static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb)
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700138{ }
139
140static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb)
141{ }
142#endif /* CONFIG_SECURITY_NETWORK */
143
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144/*
145 * SMP locking strategy:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800146 * hash table is protected with spinlock unix_table_lock
Stephen Hemminger663717f2010-02-18 14:12:06 -0800147 * each socket state is protected by separate spin lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 */
149
Al Viro44bb9362006-11-14 21:36:14 -0800150static inline unsigned unix_hash_fold(__wsum n)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151{
Al Viro44bb9362006-11-14 21:36:14 -0800152 unsigned hash = (__force unsigned)n;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 hash ^= hash>>16;
154 hash ^= hash>>8;
155 return hash&(UNIX_HASH_SIZE-1);
156}
157
158#define unix_peer(sk) (unix_sk(sk)->peer)
159
160static inline int unix_our_peer(struct sock *sk, struct sock *osk)
161{
162 return unix_peer(osk) == sk;
163}
164
165static inline int unix_may_send(struct sock *sk, struct sock *osk)
166{
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800167 return unix_peer(osk) == NULL || unix_our_peer(sk, osk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168}
169
Rainer Weikusat3c734192008-06-17 22:28:05 -0700170static inline int unix_recvq_full(struct sock const *sk)
171{
172 return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog;
173}
174
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175static struct sock *unix_peer_get(struct sock *s)
176{
177 struct sock *peer;
178
David S. Miller1c92b4e2007-05-31 13:24:26 -0700179 unix_state_lock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 peer = unix_peer(s);
181 if (peer)
182 sock_hold(peer);
David S. Miller1c92b4e2007-05-31 13:24:26 -0700183 unix_state_unlock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 return peer;
185}
186
187static inline void unix_release_addr(struct unix_address *addr)
188{
189 if (atomic_dec_and_test(&addr->refcnt))
190 kfree(addr);
191}
192
193/*
194 * Check unix socket name:
195 * - should be not zero length.
196 * - if started by not zero, should be NULL terminated (FS object)
197 * - if started by zero, it is abstract name.
198 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900199
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800200static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned *hashp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
202 if (len <= sizeof(short) || len > sizeof(*sunaddr))
203 return -EINVAL;
204 if (!sunaddr || sunaddr->sun_family != AF_UNIX)
205 return -EINVAL;
206 if (sunaddr->sun_path[0]) {
207 /*
208 * This may look like an off by one error but it is a bit more
209 * subtle. 108 is the longest valid AF_UNIX path for a binding.
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300210 * sun_path[108] doesn't as such exist. However in kernel space
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 * we are guaranteed that it is a valid memory location in our
212 * kernel address buffer.
213 */
Jianjun Konge27dfce2008-11-01 21:38:31 -0700214 ((char *)sunaddr)[len] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 len = strlen(sunaddr->sun_path)+1+sizeof(short);
216 return len;
217 }
218
Joe Perches07f07572008-11-19 15:44:53 -0800219 *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 return len;
221}
222
223static void __unix_remove_socket(struct sock *sk)
224{
225 sk_del_node_init(sk);
226}
227
228static void __unix_insert_socket(struct hlist_head *list, struct sock *sk)
229{
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700230 WARN_ON(!sk_unhashed(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 sk_add_node(sk, list);
232}
233
234static inline void unix_remove_socket(struct sock *sk)
235{
David S. Millerfbe9cc42005-12-13 23:26:29 -0800236 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 __unix_remove_socket(sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800238 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239}
240
241static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk)
242{
David S. Millerfbe9cc42005-12-13 23:26:29 -0800243 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 __unix_insert_socket(list, sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800245 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246}
247
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800248static struct sock *__unix_find_socket_byname(struct net *net,
249 struct sockaddr_un *sunname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 int len, int type, unsigned hash)
251{
252 struct sock *s;
253 struct hlist_node *node;
254
255 sk_for_each(s, node, &unix_socket_table[hash ^ type]) {
256 struct unix_sock *u = unix_sk(s);
257
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900258 if (!net_eq(sock_net(s), net))
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800259 continue;
260
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 if (u->addr->len == len &&
262 !memcmp(u->addr->name, sunname, len))
263 goto found;
264 }
265 s = NULL;
266found:
267 return s;
268}
269
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800270static inline struct sock *unix_find_socket_byname(struct net *net,
271 struct sockaddr_un *sunname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 int len, int type,
273 unsigned hash)
274{
275 struct sock *s;
276
David S. Millerfbe9cc42005-12-13 23:26:29 -0800277 spin_lock(&unix_table_lock);
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800278 s = __unix_find_socket_byname(net, sunname, len, type, hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 if (s)
280 sock_hold(s);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800281 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 return s;
283}
284
Eric W. Biederman6616f782010-06-13 03:35:48 +0000285static struct sock *unix_find_socket_byinode(struct inode *i)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286{
287 struct sock *s;
288 struct hlist_node *node;
289
David S. Millerfbe9cc42005-12-13 23:26:29 -0800290 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 sk_for_each(s, node,
292 &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) {
293 struct dentry *dentry = unix_sk(s)->dentry;
294
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800295 if (dentry && dentry->d_inode == i) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 sock_hold(s);
297 goto found;
298 }
299 }
300 s = NULL;
301found:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800302 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 return s;
304}
305
306static inline int unix_writable(struct sock *sk)
307{
308 return (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf;
309}
310
311static void unix_write_space(struct sock *sk)
312{
Eric Dumazet43815482010-04-29 11:01:49 +0000313 struct socket_wq *wq;
314
315 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 if (unix_writable(sk)) {
Eric Dumazet43815482010-04-29 11:01:49 +0000317 wq = rcu_dereference(sk->sk_wq);
318 if (wq_has_sleeper(wq))
Eric Dumazet67426b72010-10-29 20:44:44 +0000319 wake_up_interruptible_sync_poll(&wq->wait,
320 POLLOUT | POLLWRNORM | POLLWRBAND);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +0800321 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 }
Eric Dumazet43815482010-04-29 11:01:49 +0000323 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324}
325
326/* When dgram socket disconnects (or changes its peer), we clear its receive
327 * queue of packets arrived from previous peer. First, it allows to do
328 * flow control based only on wmem_alloc; second, sk connected to peer
329 * may receive messages only from that peer. */
330static void unix_dgram_disconnected(struct sock *sk, struct sock *other)
331{
David S. Millerb03efcf2005-07-08 14:57:23 -0700332 if (!skb_queue_empty(&sk->sk_receive_queue)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 skb_queue_purge(&sk->sk_receive_queue);
334 wake_up_interruptible_all(&unix_sk(sk)->peer_wait);
335
336 /* If one link of bidirectional dgram pipe is disconnected,
337 * we signal error. Messages are lost. Do not make this,
338 * when peer was not connected to us.
339 */
340 if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) {
341 other->sk_err = ECONNRESET;
342 other->sk_error_report(other);
343 }
344 }
345}
346
347static void unix_sock_destructor(struct sock *sk)
348{
349 struct unix_sock *u = unix_sk(sk);
350
351 skb_queue_purge(&sk->sk_receive_queue);
352
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700353 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
354 WARN_ON(!sk_unhashed(sk));
355 WARN_ON(sk->sk_socket);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet6b41e7d2008-11-19 15:48:09 -0800357 printk(KERN_INFO "Attempt to release alive unix socket: %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 return;
359 }
360
361 if (u->addr)
362 unix_release_addr(u->addr);
363
Eric Dumazet518de9b2010-10-26 14:22:44 -0700364 atomic_long_dec(&unix_nr_socks);
David S. Miller6f756a82008-11-23 17:34:03 -0800365 local_bh_disable();
Eric Dumazeta8076d82008-11-17 02:38:49 -0800366 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
David S. Miller6f756a82008-11-23 17:34:03 -0800367 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368#ifdef UNIX_REFCNT_DEBUG
Eric Dumazet518de9b2010-10-26 14:22:44 -0700369 printk(KERN_DEBUG "UNIX %p is destroyed, %ld are still alive.\n", sk,
370 atomic_long_read(&unix_nr_socks));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371#endif
372}
373
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800374static int unix_release_sock(struct sock *sk, int embrion)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375{
376 struct unix_sock *u = unix_sk(sk);
377 struct dentry *dentry;
378 struct vfsmount *mnt;
379 struct sock *skpair;
380 struct sk_buff *skb;
381 int state;
382
383 unix_remove_socket(sk);
384
385 /* Clear state */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700386 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 sock_orphan(sk);
388 sk->sk_shutdown = SHUTDOWN_MASK;
389 dentry = u->dentry;
390 u->dentry = NULL;
391 mnt = u->mnt;
392 u->mnt = NULL;
393 state = sk->sk_state;
394 sk->sk_state = TCP_CLOSE;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700395 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
397 wake_up_interruptible_all(&u->peer_wait);
398
Jianjun Konge27dfce2008-11-01 21:38:31 -0700399 skpair = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400
Jianjun Konge27dfce2008-11-01 21:38:31 -0700401 if (skpair != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) {
David S. Miller1c92b4e2007-05-31 13:24:26 -0700403 unix_state_lock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 /* No more writes */
405 skpair->sk_shutdown = SHUTDOWN_MASK;
406 if (!skb_queue_empty(&sk->sk_receive_queue) || embrion)
407 skpair->sk_err = ECONNRESET;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700408 unix_state_unlock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 skpair->sk_state_change(skpair);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +0800410 sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 }
412 sock_put(skpair); /* It may now die */
413 unix_peer(sk) = NULL;
414 }
415
416 /* Try to flush out this socket. Throw out buffers at least */
417
418 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
Jianjun Konge27dfce2008-11-01 21:38:31 -0700419 if (state == TCP_LISTEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 unix_release_sock(skb->sk, 1);
421 /* passed fds are erased in the kfree_skb hook */
422 kfree_skb(skb);
423 }
424
425 if (dentry) {
426 dput(dentry);
427 mntput(mnt);
428 }
429
430 sock_put(sk);
431
432 /* ---- Socket is dead now and most probably destroyed ---- */
433
434 /*
435 * Fixme: BSD difference: In BSD all sockets connected to use get
436 * ECONNRESET and we die on the spot. In Linux we behave
437 * like files and pipes do and wait for the last
438 * dereference.
439 *
440 * Can't we simply set sock->err?
441 *
442 * What the above comment does talk about? --ANK(980817)
443 */
444
Pavel Emelyanov9305cfa2007-11-10 22:06:01 -0800445 if (unix_tot_inflight)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900446 unix_gc(); /* Garbage collect fds */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447
448 return 0;
449}
450
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000451static void init_peercred(struct sock *sk)
452{
453 put_pid(sk->sk_peer_pid);
454 if (sk->sk_peer_cred)
455 put_cred(sk->sk_peer_cred);
456 sk->sk_peer_pid = get_pid(task_tgid(current));
457 sk->sk_peer_cred = get_current_cred();
458}
459
460static void copy_peercred(struct sock *sk, struct sock *peersk)
461{
462 put_pid(sk->sk_peer_pid);
463 if (sk->sk_peer_cred)
464 put_cred(sk->sk_peer_cred);
465 sk->sk_peer_pid = get_pid(peersk->sk_peer_pid);
466 sk->sk_peer_cred = get_cred(peersk->sk_peer_cred);
467}
468
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469static int unix_listen(struct socket *sock, int backlog)
470{
471 int err;
472 struct sock *sk = sock->sk;
473 struct unix_sock *u = unix_sk(sk);
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000474 struct pid *old_pid = NULL;
475 const struct cred *old_cred = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476
477 err = -EOPNOTSUPP;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800478 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
479 goto out; /* Only stream/seqpacket sockets accept */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 err = -EINVAL;
481 if (!u->addr)
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800482 goto out; /* No listens on an unbound socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700483 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN)
485 goto out_unlock;
486 if (backlog > sk->sk_max_ack_backlog)
487 wake_up_interruptible_all(&u->peer_wait);
488 sk->sk_max_ack_backlog = backlog;
489 sk->sk_state = TCP_LISTEN;
490 /* set credentials so connect can copy them */
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000491 init_peercred(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 err = 0;
493
494out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -0700495 unix_state_unlock(sk);
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000496 put_pid(old_pid);
497 if (old_cred)
498 put_cred(old_cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499out:
500 return err;
501}
502
503static int unix_release(struct socket *);
504static int unix_bind(struct socket *, struct sockaddr *, int);
505static int unix_stream_connect(struct socket *, struct sockaddr *,
506 int addr_len, int flags);
507static int unix_socketpair(struct socket *, struct socket *);
508static int unix_accept(struct socket *, struct socket *, int);
509static int unix_getname(struct socket *, struct sockaddr *, int *, int);
510static unsigned int unix_poll(struct file *, struct socket *, poll_table *);
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700511static unsigned int unix_dgram_poll(struct file *, struct socket *,
512 poll_table *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513static int unix_ioctl(struct socket *, unsigned int, unsigned long);
514static int unix_shutdown(struct socket *, int);
515static int unix_stream_sendmsg(struct kiocb *, struct socket *,
516 struct msghdr *, size_t);
517static int unix_stream_recvmsg(struct kiocb *, struct socket *,
518 struct msghdr *, size_t, int);
519static int unix_dgram_sendmsg(struct kiocb *, struct socket *,
520 struct msghdr *, size_t);
521static int unix_dgram_recvmsg(struct kiocb *, struct socket *,
522 struct msghdr *, size_t, int);
523static int unix_dgram_connect(struct socket *, struct sockaddr *,
524 int, int);
525static int unix_seqpacket_sendmsg(struct kiocb *, struct socket *,
526 struct msghdr *, size_t);
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +0000527static int unix_seqpacket_recvmsg(struct kiocb *, struct socket *,
528 struct msghdr *, size_t, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800530static const struct proto_ops unix_stream_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 .family = PF_UNIX,
532 .owner = THIS_MODULE,
533 .release = unix_release,
534 .bind = unix_bind,
535 .connect = unix_stream_connect,
536 .socketpair = unix_socketpair,
537 .accept = unix_accept,
538 .getname = unix_getname,
539 .poll = unix_poll,
540 .ioctl = unix_ioctl,
541 .listen = unix_listen,
542 .shutdown = unix_shutdown,
543 .setsockopt = sock_no_setsockopt,
544 .getsockopt = sock_no_getsockopt,
545 .sendmsg = unix_stream_sendmsg,
546 .recvmsg = unix_stream_recvmsg,
547 .mmap = sock_no_mmap,
548 .sendpage = sock_no_sendpage,
549};
550
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800551static const struct proto_ops unix_dgram_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 .family = PF_UNIX,
553 .owner = THIS_MODULE,
554 .release = unix_release,
555 .bind = unix_bind,
556 .connect = unix_dgram_connect,
557 .socketpair = unix_socketpair,
558 .accept = sock_no_accept,
559 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700560 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 .ioctl = unix_ioctl,
562 .listen = sock_no_listen,
563 .shutdown = unix_shutdown,
564 .setsockopt = sock_no_setsockopt,
565 .getsockopt = sock_no_getsockopt,
566 .sendmsg = unix_dgram_sendmsg,
567 .recvmsg = unix_dgram_recvmsg,
568 .mmap = sock_no_mmap,
569 .sendpage = sock_no_sendpage,
570};
571
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800572static const struct proto_ops unix_seqpacket_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 .family = PF_UNIX,
574 .owner = THIS_MODULE,
575 .release = unix_release,
576 .bind = unix_bind,
577 .connect = unix_stream_connect,
578 .socketpair = unix_socketpair,
579 .accept = unix_accept,
580 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700581 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 .ioctl = unix_ioctl,
583 .listen = unix_listen,
584 .shutdown = unix_shutdown,
585 .setsockopt = sock_no_setsockopt,
586 .getsockopt = sock_no_getsockopt,
587 .sendmsg = unix_seqpacket_sendmsg,
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +0000588 .recvmsg = unix_seqpacket_recvmsg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 .mmap = sock_no_mmap,
590 .sendpage = sock_no_sendpage,
591};
592
593static struct proto unix_proto = {
Eric Dumazet248969a2008-11-17 00:00:30 -0800594 .name = "UNIX",
595 .owner = THIS_MODULE,
Eric Dumazet248969a2008-11-17 00:00:30 -0800596 .obj_size = sizeof(struct unix_sock),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597};
598
Ingo Molnara09785a2006-07-03 00:25:12 -0700599/*
600 * AF_UNIX sockets do not interact with hardware, hence they
601 * dont trigger interrupts - so it's safe for them to have
602 * bh-unsafe locking for their sk_receive_queue.lock. Split off
603 * this special lock-class by reinitializing the spinlock key:
604 */
605static struct lock_class_key af_unix_sk_receive_queue_lock_key;
606
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800607static struct sock *unix_create1(struct net *net, struct socket *sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608{
609 struct sock *sk = NULL;
610 struct unix_sock *u;
611
Eric Dumazet518de9b2010-10-26 14:22:44 -0700612 atomic_long_inc(&unix_nr_socks);
613 if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files())
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 goto out;
615
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700616 sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 if (!sk)
618 goto out;
619
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800620 sock_init_data(sock, sk);
Ingo Molnara09785a2006-07-03 00:25:12 -0700621 lockdep_set_class(&sk->sk_receive_queue.lock,
622 &af_unix_sk_receive_queue_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
624 sk->sk_write_space = unix_write_space;
Denis V. Luneva0a53c82007-12-11 04:19:17 -0800625 sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 sk->sk_destruct = unix_sock_destructor;
627 u = unix_sk(sk);
628 u->dentry = NULL;
629 u->mnt = NULL;
Benjamin LaHaisefd19f322006-01-03 14:10:46 -0800630 spin_lock_init(&u->lock);
Al Viro516e0cc2008-07-26 00:39:17 -0400631 atomic_long_set(&u->inflight, 0);
Miklos Szeredi1fd05ba2007-07-11 14:22:39 -0700632 INIT_LIST_HEAD(&u->link);
Ingo Molnar57b47a52006-03-20 22:35:41 -0800633 mutex_init(&u->readlock); /* single task reading lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 init_waitqueue_head(&u->peer_wait);
635 unix_insert_socket(unix_sockets_unbound, sk);
636out:
Pavel Emelyanov284b3272007-11-10 22:08:30 -0800637 if (sk == NULL)
Eric Dumazet518de9b2010-10-26 14:22:44 -0700638 atomic_long_dec(&unix_nr_socks);
Eric Dumazet920de802008-11-24 00:09:29 -0800639 else {
640 local_bh_disable();
Eric Dumazeta8076d82008-11-17 02:38:49 -0800641 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet920de802008-11-24 00:09:29 -0800642 local_bh_enable();
643 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 return sk;
645}
646
Eric Paris3f378b62009-11-05 22:18:14 -0800647static int unix_create(struct net *net, struct socket *sock, int protocol,
648 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649{
650 if (protocol && protocol != PF_UNIX)
651 return -EPROTONOSUPPORT;
652
653 sock->state = SS_UNCONNECTED;
654
655 switch (sock->type) {
656 case SOCK_STREAM:
657 sock->ops = &unix_stream_ops;
658 break;
659 /*
660 * Believe it or not BSD has AF_UNIX, SOCK_RAW though
661 * nothing uses it.
662 */
663 case SOCK_RAW:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700664 sock->type = SOCK_DGRAM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 case SOCK_DGRAM:
666 sock->ops = &unix_dgram_ops;
667 break;
668 case SOCK_SEQPACKET:
669 sock->ops = &unix_seqpacket_ops;
670 break;
671 default:
672 return -ESOCKTNOSUPPORT;
673 }
674
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700675 return unix_create1(net, sock) ? 0 : -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676}
677
678static int unix_release(struct socket *sock)
679{
680 struct sock *sk = sock->sk;
681
682 if (!sk)
683 return 0;
684
685 sock->sk = NULL;
686
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800687 return unix_release_sock(sk, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688}
689
690static int unix_autobind(struct socket *sock)
691{
692 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900693 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 struct unix_sock *u = unix_sk(sk);
695 static u32 ordernum = 1;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800696 struct unix_address *addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 int err;
Tetsuo Handa8df73ff2010-09-04 01:34:28 +0000698 unsigned int retries = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699
Ingo Molnar57b47a52006-03-20 22:35:41 -0800700 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701
702 err = 0;
703 if (u->addr)
704 goto out;
705
706 err = -ENOMEM;
Panagiotis Issaris0da974f2006-07-21 14:51:30 -0700707 addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 if (!addr)
709 goto out;
710
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 addr->name->sun_family = AF_UNIX;
712 atomic_set(&addr->refcnt, 1);
713
714retry:
715 addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short);
Joe Perches07f07572008-11-19 15:44:53 -0800716 addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717
David S. Millerfbe9cc42005-12-13 23:26:29 -0800718 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 ordernum = (ordernum+1)&0xFFFFF;
720
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800721 if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 addr->hash)) {
David S. Millerfbe9cc42005-12-13 23:26:29 -0800723 spin_unlock(&unix_table_lock);
Tetsuo Handa8df73ff2010-09-04 01:34:28 +0000724 /*
725 * __unix_find_socket_byname() may take long time if many names
726 * are already in use.
727 */
728 cond_resched();
729 /* Give up if all names seems to be in use. */
730 if (retries++ == 0xFFFFF) {
731 err = -ENOSPC;
732 kfree(addr);
733 goto out;
734 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 goto retry;
736 }
737 addr->hash ^= sk->sk_type;
738
739 __unix_remove_socket(sk);
740 u->addr = addr;
741 __unix_insert_socket(&unix_socket_table[addr->hash], sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800742 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 err = 0;
744
Ingo Molnar57b47a52006-03-20 22:35:41 -0800745out: mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 return err;
747}
748
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800749static struct sock *unix_find_other(struct net *net,
750 struct sockaddr_un *sunname, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 int type, unsigned hash, int *error)
752{
753 struct sock *u;
Al Viro421748e2008-08-02 01:04:36 -0400754 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 int err = 0;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900756
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 if (sunname->sun_path[0]) {
Al Viro421748e2008-08-02 01:04:36 -0400758 struct inode *inode;
759 err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 if (err)
761 goto fail;
Al Viro421748e2008-08-02 01:04:36 -0400762 inode = path.dentry->d_inode;
763 err = inode_permission(inode, MAY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 if (err)
765 goto put_fail;
766
767 err = -ECONNREFUSED;
Al Viro421748e2008-08-02 01:04:36 -0400768 if (!S_ISSOCK(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 goto put_fail;
Eric W. Biederman6616f782010-06-13 03:35:48 +0000770 u = unix_find_socket_byinode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 if (!u)
772 goto put_fail;
773
774 if (u->sk_type == type)
Al Viro421748e2008-08-02 01:04:36 -0400775 touch_atime(path.mnt, path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
Al Viro421748e2008-08-02 01:04:36 -0400777 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
Jianjun Konge27dfce2008-11-01 21:38:31 -0700779 err = -EPROTOTYPE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 if (u->sk_type != type) {
781 sock_put(u);
782 goto fail;
783 }
784 } else {
785 err = -ECONNREFUSED;
Jianjun Konge27dfce2008-11-01 21:38:31 -0700786 u = unix_find_socket_byname(net, sunname, len, type, hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 if (u) {
788 struct dentry *dentry;
789 dentry = unix_sk(u)->dentry;
790 if (dentry)
791 touch_atime(unix_sk(u)->mnt, dentry);
792 } else
793 goto fail;
794 }
795 return u;
796
797put_fail:
Al Viro421748e2008-08-02 01:04:36 -0400798 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799fail:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700800 *error = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 return NULL;
802}
803
804
805static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
806{
807 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900808 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700810 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800811 struct dentry *dentry = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 struct nameidata nd;
813 int err;
814 unsigned hash;
815 struct unix_address *addr;
816 struct hlist_head *list;
817
818 err = -EINVAL;
819 if (sunaddr->sun_family != AF_UNIX)
820 goto out;
821
Jianjun Konge27dfce2008-11-01 21:38:31 -0700822 if (addr_len == sizeof(short)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 err = unix_autobind(sock);
824 goto out;
825 }
826
827 err = unix_mkname(sunaddr, addr_len, &hash);
828 if (err < 0)
829 goto out;
830 addr_len = err;
831
Ingo Molnar57b47a52006-03-20 22:35:41 -0800832 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833
834 err = -EINVAL;
835 if (u->addr)
836 goto out_up;
837
838 err = -ENOMEM;
839 addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL);
840 if (!addr)
841 goto out_up;
842
843 memcpy(addr->name, sunaddr, addr_len);
844 addr->len = addr_len;
845 addr->hash = hash ^ sk->sk_type;
846 atomic_set(&addr->refcnt, 1);
847
848 if (sunaddr->sun_path[0]) {
849 unsigned int mode;
850 err = 0;
851 /*
852 * Get the parent directory, calculate the hash for last
853 * component.
854 */
Al Viroc9c6cac2011-02-16 15:15:47 -0500855 err = kern_path_parent(sunaddr->sun_path, &nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 if (err)
857 goto out_mknod_parent;
Christoph Hellwigf81a0bf2005-05-19 12:26:43 -0700858
859 dentry = lookup_create(&nd, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 err = PTR_ERR(dentry);
861 if (IS_ERR(dentry))
862 goto out_mknod_unlock;
Christoph Hellwigf81a0bf2005-05-19 12:26:43 -0700863
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 /*
865 * All right, let's create it.
866 */
867 mode = S_IFSOCK |
Al Viroce3b0f82009-03-29 19:08:22 -0400868 (SOCK_INODE(sock)->i_mode & ~current_umask());
Dave Hansen463c3192008-02-15 14:37:57 -0800869 err = mnt_want_write(nd.path.mnt);
870 if (err)
871 goto out_mknod_dput;
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900872 err = security_path_mknod(&nd.path, dentry, mode, 0);
873 if (err)
874 goto out_mknod_drop_write;
Jan Blunck4ac91372008-02-14 19:34:32 -0800875 err = vfs_mknod(nd.path.dentry->d_inode, dentry, mode, 0);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900876out_mknod_drop_write:
Dave Hansen463c3192008-02-15 14:37:57 -0800877 mnt_drop_write(nd.path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 if (err)
879 goto out_mknod_dput;
Jan Blunck4ac91372008-02-14 19:34:32 -0800880 mutex_unlock(&nd.path.dentry->d_inode->i_mutex);
881 dput(nd.path.dentry);
882 nd.path.dentry = dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883
884 addr->hash = UNIX_HASH_SIZE;
885 }
886
David S. Millerfbe9cc42005-12-13 23:26:29 -0800887 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888
889 if (!sunaddr->sun_path[0]) {
890 err = -EADDRINUSE;
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800891 if (__unix_find_socket_byname(net, sunaddr, addr_len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 sk->sk_type, hash)) {
893 unix_release_addr(addr);
894 goto out_unlock;
895 }
896
897 list = &unix_socket_table[addr->hash];
898 } else {
899 list = &unix_socket_table[dentry->d_inode->i_ino & (UNIX_HASH_SIZE-1)];
Jan Blunck4ac91372008-02-14 19:34:32 -0800900 u->dentry = nd.path.dentry;
901 u->mnt = nd.path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 }
903
904 err = 0;
905 __unix_remove_socket(sk);
906 u->addr = addr;
907 __unix_insert_socket(list, sk);
908
909out_unlock:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800910 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911out_up:
Ingo Molnar57b47a52006-03-20 22:35:41 -0800912 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913out:
914 return err;
915
916out_mknod_dput:
917 dput(dentry);
918out_mknod_unlock:
Jan Blunck4ac91372008-02-14 19:34:32 -0800919 mutex_unlock(&nd.path.dentry->d_inode->i_mutex);
Jan Blunck1d957f92008-02-14 19:34:35 -0800920 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921out_mknod_parent:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700922 if (err == -EEXIST)
923 err = -EADDRINUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 unix_release_addr(addr);
925 goto out_up;
926}
927
David S. Miller278a3de2007-05-31 15:19:20 -0700928static void unix_state_double_lock(struct sock *sk1, struct sock *sk2)
929{
930 if (unlikely(sk1 == sk2) || !sk2) {
931 unix_state_lock(sk1);
932 return;
933 }
934 if (sk1 < sk2) {
935 unix_state_lock(sk1);
936 unix_state_lock_nested(sk2);
937 } else {
938 unix_state_lock(sk2);
939 unix_state_lock_nested(sk1);
940 }
941}
942
943static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2)
944{
945 if (unlikely(sk1 == sk2) || !sk2) {
946 unix_state_unlock(sk1);
947 return;
948 }
949 unix_state_unlock(sk1);
950 unix_state_unlock(sk2);
951}
952
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr,
954 int alen, int flags)
955{
956 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900957 struct net *net = sock_net(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700958 struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 struct sock *other;
960 unsigned hash;
961 int err;
962
963 if (addr->sa_family != AF_UNSPEC) {
964 err = unix_mkname(sunaddr, alen, &hash);
965 if (err < 0)
966 goto out;
967 alen = err;
968
969 if (test_bit(SOCK_PASSCRED, &sock->flags) &&
970 !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0)
971 goto out;
972
David S. Miller278a3de2007-05-31 15:19:20 -0700973restart:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700974 other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 if (!other)
976 goto out;
977
David S. Miller278a3de2007-05-31 15:19:20 -0700978 unix_state_double_lock(sk, other);
979
980 /* Apparently VFS overslept socket death. Retry. */
981 if (sock_flag(other, SOCK_DEAD)) {
982 unix_state_double_unlock(sk, other);
983 sock_put(other);
984 goto restart;
985 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986
987 err = -EPERM;
988 if (!unix_may_send(sk, other))
989 goto out_unlock;
990
991 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
992 if (err)
993 goto out_unlock;
994
995 } else {
996 /*
997 * 1003.1g breaking connected state with AF_UNSPEC
998 */
999 other = NULL;
David S. Miller278a3de2007-05-31 15:19:20 -07001000 unix_state_double_lock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 }
1002
1003 /*
1004 * If it was connected, reconnect.
1005 */
1006 if (unix_peer(sk)) {
1007 struct sock *old_peer = unix_peer(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001008 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -07001009 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
1011 if (other != old_peer)
1012 unix_dgram_disconnected(sk, old_peer);
1013 sock_put(old_peer);
1014 } else {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001015 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -07001016 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 }
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001018 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
1020out_unlock:
David S. Miller278a3de2007-05-31 15:19:20 -07001021 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 sock_put(other);
1023out:
1024 return err;
1025}
1026
1027static long unix_wait_for_peer(struct sock *other, long timeo)
1028{
1029 struct unix_sock *u = unix_sk(other);
1030 int sched;
1031 DEFINE_WAIT(wait);
1032
1033 prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE);
1034
1035 sched = !sock_flag(other, SOCK_DEAD) &&
1036 !(other->sk_shutdown & RCV_SHUTDOWN) &&
Rainer Weikusat3c734192008-06-17 22:28:05 -07001037 unix_recvq_full(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038
David S. Miller1c92b4e2007-05-31 13:24:26 -07001039 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040
1041 if (sched)
1042 timeo = schedule_timeout(timeo);
1043
1044 finish_wait(&u->peer_wait, &wait);
1045 return timeo;
1046}
1047
1048static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr,
1049 int addr_len, int flags)
1050{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001051 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001053 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054 struct unix_sock *u = unix_sk(sk), *newu, *otheru;
1055 struct sock *newsk = NULL;
1056 struct sock *other = NULL;
1057 struct sk_buff *skb = NULL;
1058 unsigned hash;
1059 int st;
1060 int err;
1061 long timeo;
1062
1063 err = unix_mkname(sunaddr, addr_len, &hash);
1064 if (err < 0)
1065 goto out;
1066 addr_len = err;
1067
Joe Perchesf64f9e72009-11-29 16:55:45 -08001068 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr &&
1069 (err = unix_autobind(sock)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 goto out;
1071
1072 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1073
1074 /* First of all allocate resources.
1075 If we will make it after state is locked,
1076 we will have to recheck all again in any case.
1077 */
1078
1079 err = -ENOMEM;
1080
1081 /* create new sock for complete connection */
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001082 newsk = unix_create1(sock_net(sk), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 if (newsk == NULL)
1084 goto out;
1085
1086 /* Allocate skb for sending to listening sock */
1087 skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL);
1088 if (skb == NULL)
1089 goto out;
1090
1091restart:
1092 /* Find listening sock. */
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001093 other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 if (!other)
1095 goto out;
1096
1097 /* Latch state of peer */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001098 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099
1100 /* Apparently VFS overslept socket death. Retry. */
1101 if (sock_flag(other, SOCK_DEAD)) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001102 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 sock_put(other);
1104 goto restart;
1105 }
1106
1107 err = -ECONNREFUSED;
1108 if (other->sk_state != TCP_LISTEN)
1109 goto out_unlock;
Tomoki Sekiyama77238f22009-10-18 23:17:37 -07001110 if (other->sk_shutdown & RCV_SHUTDOWN)
1111 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
Rainer Weikusat3c734192008-06-17 22:28:05 -07001113 if (unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 err = -EAGAIN;
1115 if (!timeo)
1116 goto out_unlock;
1117
1118 timeo = unix_wait_for_peer(other, timeo);
1119
1120 err = sock_intr_errno(timeo);
1121 if (signal_pending(current))
1122 goto out;
1123 sock_put(other);
1124 goto restart;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001125 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126
1127 /* Latch our state.
1128
Daniel Balutae5537bf2011-03-14 15:25:33 -07001129 It is tricky place. We need to grab our state lock and cannot
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 drop lock on peer. It is dangerous because deadlock is
1131 possible. Connect to self case and simultaneous
1132 attempt to connect are eliminated by checking socket
1133 state. other is TCP_LISTEN, if sk is TCP_LISTEN we
1134 check this before attempt to grab lock.
1135
1136 Well, and we have to recheck the state after socket locked.
1137 */
1138 st = sk->sk_state;
1139
1140 switch (st) {
1141 case TCP_CLOSE:
1142 /* This is ok... continue with connect */
1143 break;
1144 case TCP_ESTABLISHED:
1145 /* Socket is already connected */
1146 err = -EISCONN;
1147 goto out_unlock;
1148 default:
1149 err = -EINVAL;
1150 goto out_unlock;
1151 }
1152
David S. Miller1c92b4e2007-05-31 13:24:26 -07001153 unix_state_lock_nested(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
1155 if (sk->sk_state != st) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001156 unix_state_unlock(sk);
1157 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 sock_put(other);
1159 goto restart;
1160 }
1161
David S. Miller3610cda2011-01-05 15:38:53 -08001162 err = security_unix_stream_connect(sk, other, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 if (err) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001164 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 goto out_unlock;
1166 }
1167
1168 /* The way is open! Fastly set all the necessary fields... */
1169
1170 sock_hold(sk);
1171 unix_peer(newsk) = sk;
1172 newsk->sk_state = TCP_ESTABLISHED;
1173 newsk->sk_type = sk->sk_type;
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001174 init_peercred(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 newu = unix_sk(newsk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001176 RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 otheru = unix_sk(other);
1178
1179 /* copy address information from listening to new sock*/
1180 if (otheru->addr) {
1181 atomic_inc(&otheru->addr->refcnt);
1182 newu->addr = otheru->addr;
1183 }
1184 if (otheru->dentry) {
1185 newu->dentry = dget(otheru->dentry);
1186 newu->mnt = mntget(otheru->mnt);
1187 }
1188
1189 /* Set credentials */
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001190 copy_peercred(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 sock->state = SS_CONNECTED;
1193 sk->sk_state = TCP_ESTABLISHED;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001194 sock_hold(newsk);
1195
1196 smp_mb__after_atomic_inc(); /* sock_hold() does an atomic_inc() */
1197 unix_peer(sk) = newsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198
David S. Miller1c92b4e2007-05-31 13:24:26 -07001199 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200
1201 /* take ten and and send info to listening sock */
1202 spin_lock(&other->sk_receive_queue.lock);
1203 __skb_queue_tail(&other->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 spin_unlock(&other->sk_receive_queue.lock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001205 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 other->sk_data_ready(other, 0);
1207 sock_put(other);
1208 return 0;
1209
1210out_unlock:
1211 if (other)
David S. Miller1c92b4e2007-05-31 13:24:26 -07001212 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213
1214out:
Wei Yongjun40d44442009-02-25 00:32:45 +00001215 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 if (newsk)
1217 unix_release_sock(newsk, 0);
1218 if (other)
1219 sock_put(other);
1220 return err;
1221}
1222
1223static int unix_socketpair(struct socket *socka, struct socket *sockb)
1224{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001225 struct sock *ska = socka->sk, *skb = sockb->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226
1227 /* Join our sockets back to back */
1228 sock_hold(ska);
1229 sock_hold(skb);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001230 unix_peer(ska) = skb;
1231 unix_peer(skb) = ska;
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001232 init_peercred(ska);
1233 init_peercred(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234
1235 if (ska->sk_type != SOCK_DGRAM) {
1236 ska->sk_state = TCP_ESTABLISHED;
1237 skb->sk_state = TCP_ESTABLISHED;
1238 socka->state = SS_CONNECTED;
1239 sockb->state = SS_CONNECTED;
1240 }
1241 return 0;
1242}
1243
1244static int unix_accept(struct socket *sock, struct socket *newsock, int flags)
1245{
1246 struct sock *sk = sock->sk;
1247 struct sock *tsk;
1248 struct sk_buff *skb;
1249 int err;
1250
1251 err = -EOPNOTSUPP;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001252 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 goto out;
1254
1255 err = -EINVAL;
1256 if (sk->sk_state != TCP_LISTEN)
1257 goto out;
1258
1259 /* If socket state is TCP_LISTEN it cannot change (for now...),
1260 * so that no locks are necessary.
1261 */
1262
1263 skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err);
1264 if (!skb) {
1265 /* This means receive shutdown. */
1266 if (err == 0)
1267 err = -EINVAL;
1268 goto out;
1269 }
1270
1271 tsk = skb->sk;
1272 skb_free_datagram(sk, skb);
1273 wake_up_interruptible(&unix_sk(sk)->peer_wait);
1274
1275 /* attach accepted sock to socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001276 unix_state_lock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 newsock->state = SS_CONNECTED;
1278 sock_graft(tsk, newsock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001279 unix_state_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 return 0;
1281
1282out:
1283 return err;
1284}
1285
1286
1287static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer)
1288{
1289 struct sock *sk = sock->sk;
1290 struct unix_sock *u;
Cyrill Gorcunov13cfa972009-11-08 05:51:19 +00001291 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 int err = 0;
1293
1294 if (peer) {
1295 sk = unix_peer_get(sk);
1296
1297 err = -ENOTCONN;
1298 if (!sk)
1299 goto out;
1300 err = 0;
1301 } else {
1302 sock_hold(sk);
1303 }
1304
1305 u = unix_sk(sk);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001306 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 if (!u->addr) {
1308 sunaddr->sun_family = AF_UNIX;
1309 sunaddr->sun_path[0] = 0;
1310 *uaddr_len = sizeof(short);
1311 } else {
1312 struct unix_address *addr = u->addr;
1313
1314 *uaddr_len = addr->len;
1315 memcpy(sunaddr, addr->name, *uaddr_len);
1316 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07001317 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 sock_put(sk);
1319out:
1320 return err;
1321}
1322
1323static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb)
1324{
1325 int i;
1326
1327 scm->fp = UNIXCB(skb).fp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 UNIXCB(skb).fp = NULL;
1329
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001330 for (i = scm->fp->count-1; i >= 0; i--)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 unix_notinflight(scm->fp->fp[i]);
1332}
1333
Eric W. Biederman7361c362010-06-13 03:34:33 +00001334static void unix_destruct_scm(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335{
1336 struct scm_cookie scm;
1337 memset(&scm, 0, sizeof(scm));
Eric W. Biederman7361c362010-06-13 03:34:33 +00001338 scm.pid = UNIXCB(skb).pid;
1339 scm.cred = UNIXCB(skb).cred;
1340 if (UNIXCB(skb).fp)
1341 unix_detach_fds(&scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342
1343 /* Alas, it calls VFS */
1344 /* So fscking what? fput() had been SMP-safe since the last Summer */
1345 scm_destroy(&scm);
1346 sock_wfree(skb);
1347}
1348
Eric Dumazet25888e32010-11-25 04:11:39 +00001349#define MAX_RECURSION_LEVEL 4
1350
Miklos Szeredi62093442008-11-09 15:23:57 +01001351static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352{
1353 int i;
Eric Dumazet25888e32010-11-25 04:11:39 +00001354 unsigned char max_level = 0;
1355 int unix_sock_count = 0;
1356
1357 for (i = scm->fp->count - 1; i >= 0; i--) {
1358 struct sock *sk = unix_get_socket(scm->fp->fp[i]);
1359
1360 if (sk) {
1361 unix_sock_count++;
1362 max_level = max(max_level,
1363 unix_sk(sk)->recursion_level);
1364 }
1365 }
1366 if (unlikely(max_level > MAX_RECURSION_LEVEL))
1367 return -ETOOMANYREFS;
Miklos Szeredi62093442008-11-09 15:23:57 +01001368
1369 /*
1370 * Need to duplicate file references for the sake of garbage
1371 * collection. Otherwise a socket in the fps might become a
1372 * candidate for GC while the skb is not yet queued.
1373 */
1374 UNIXCB(skb).fp = scm_fp_dup(scm->fp);
1375 if (!UNIXCB(skb).fp)
1376 return -ENOMEM;
1377
Eric Dumazet25888e32010-11-25 04:11:39 +00001378 if (unix_sock_count) {
1379 for (i = scm->fp->count - 1; i >= 0; i--)
1380 unix_inflight(scm->fp->fp[i]);
1381 }
1382 return max_level;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383}
1384
Eric W. Biederman7361c362010-06-13 03:34:33 +00001385static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds)
1386{
1387 int err = 0;
1388 UNIXCB(skb).pid = get_pid(scm->pid);
1389 UNIXCB(skb).cred = get_cred(scm->cred);
1390 UNIXCB(skb).fp = NULL;
1391 if (scm->fp && send_fds)
1392 err = unix_attach_fds(scm, skb);
1393
1394 skb->destructor = unix_destruct_scm;
1395 return err;
1396}
1397
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398/*
1399 * Send AF_UNIX data.
1400 */
1401
1402static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock,
1403 struct msghdr *msg, size_t len)
1404{
1405 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1406 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001407 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001409 struct sockaddr_un *sunaddr = msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 struct sock *other = NULL;
1411 int namelen = 0; /* fake GCC */
1412 int err;
1413 unsigned hash;
1414 struct sk_buff *skb;
1415 long timeo;
1416 struct scm_cookie tmp_scm;
Eric Dumazet25888e32010-11-25 04:11:39 +00001417 int max_level;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
1419 if (NULL == siocb->scm)
1420 siocb->scm = &tmp_scm;
dann frazier5f23b732008-11-26 15:32:27 -08001421 wait_for_unix_gc();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 err = scm_send(sock, msg, siocb->scm);
1423 if (err < 0)
1424 return err;
1425
1426 err = -EOPNOTSUPP;
1427 if (msg->msg_flags&MSG_OOB)
1428 goto out;
1429
1430 if (msg->msg_namelen) {
1431 err = unix_mkname(sunaddr, msg->msg_namelen, &hash);
1432 if (err < 0)
1433 goto out;
1434 namelen = err;
1435 } else {
1436 sunaddr = NULL;
1437 err = -ENOTCONN;
1438 other = unix_peer_get(sk);
1439 if (!other)
1440 goto out;
1441 }
1442
Joe Perchesf64f9e72009-11-29 16:55:45 -08001443 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr
1444 && (err = unix_autobind(sock)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 goto out;
1446
1447 err = -EMSGSIZE;
1448 if (len > sk->sk_sndbuf - 32)
1449 goto out;
1450
1451 skb = sock_alloc_send_skb(sk, len, msg->msg_flags&MSG_DONTWAIT, &err);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001452 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 goto out;
1454
Eric W. Biederman7361c362010-06-13 03:34:33 +00001455 err = unix_scm_to_skb(siocb->scm, skb, true);
Eric Dumazet25888e32010-11-25 04:11:39 +00001456 if (err < 0)
Eric W. Biederman7361c362010-06-13 03:34:33 +00001457 goto out_free;
Eric Dumazet25888e32010-11-25 04:11:39 +00001458 max_level = err + 1;
Catherine Zhangdc49c1f2006-08-02 14:12:06 -07001459 unix_get_secdata(siocb->scm, skb);
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001460
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001461 skb_reset_transport_header(skb);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001462 err = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 if (err)
1464 goto out_free;
1465
1466 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1467
1468restart:
1469 if (!other) {
1470 err = -ECONNRESET;
1471 if (sunaddr == NULL)
1472 goto out_free;
1473
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001474 other = unix_find_other(net, sunaddr, namelen, sk->sk_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 hash, &err);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001476 if (other == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 goto out_free;
1478 }
1479
Alban Crequyd6ae3ba2011-01-18 06:39:15 +00001480 if (sk_filter(other, skb) < 0) {
1481 /* Toss the packet but do not return any error to the sender */
1482 err = len;
1483 goto out_free;
1484 }
1485
David S. Miller1c92b4e2007-05-31 13:24:26 -07001486 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 err = -EPERM;
1488 if (!unix_may_send(sk, other))
1489 goto out_unlock;
1490
1491 if (sock_flag(other, SOCK_DEAD)) {
1492 /*
1493 * Check with 1003.1g - what should
1494 * datagram error
1495 */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001496 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 sock_put(other);
1498
1499 err = 0;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001500 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 if (unix_peer(sk) == other) {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001502 unix_peer(sk) = NULL;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001503 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504
1505 unix_dgram_disconnected(sk, other);
1506 sock_put(other);
1507 err = -ECONNREFUSED;
1508 } else {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001509 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 }
1511
1512 other = NULL;
1513 if (err)
1514 goto out_free;
1515 goto restart;
1516 }
1517
1518 err = -EPIPE;
1519 if (other->sk_shutdown & RCV_SHUTDOWN)
1520 goto out_unlock;
1521
1522 if (sk->sk_type != SOCK_SEQPACKET) {
1523 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1524 if (err)
1525 goto out_unlock;
1526 }
1527
Rainer Weikusat3c734192008-06-17 22:28:05 -07001528 if (unix_peer(other) != sk && unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 if (!timeo) {
1530 err = -EAGAIN;
1531 goto out_unlock;
1532 }
1533
1534 timeo = unix_wait_for_peer(other, timeo);
1535
1536 err = sock_intr_errno(timeo);
1537 if (signal_pending(current))
1538 goto out_free;
1539
1540 goto restart;
1541 }
1542
Alban Crequy3f661162010-10-04 08:48:28 +00001543 if (sock_flag(other, SOCK_RCVTSTAMP))
1544 __net_timestamp(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 skb_queue_tail(&other->sk_receive_queue, skb);
Eric Dumazet25888e32010-11-25 04:11:39 +00001546 if (max_level > unix_sk(other)->recursion_level)
1547 unix_sk(other)->recursion_level = max_level;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001548 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 other->sk_data_ready(other, len);
1550 sock_put(other);
1551 scm_destroy(siocb->scm);
1552 return len;
1553
1554out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001555 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556out_free:
1557 kfree_skb(skb);
1558out:
1559 if (other)
1560 sock_put(other);
1561 scm_destroy(siocb->scm);
1562 return err;
1563}
1564
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001565
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock,
1567 struct msghdr *msg, size_t len)
1568{
1569 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1570 struct sock *sk = sock->sk;
1571 struct sock *other = NULL;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001572 int err, size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 struct sk_buff *skb;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001574 int sent = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575 struct scm_cookie tmp_scm;
Miklos Szeredi8ba69ba2009-09-11 11:31:45 -07001576 bool fds_sent = false;
Eric Dumazet25888e32010-11-25 04:11:39 +00001577 int max_level;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578
1579 if (NULL == siocb->scm)
1580 siocb->scm = &tmp_scm;
dann frazier5f23b732008-11-26 15:32:27 -08001581 wait_for_unix_gc();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 err = scm_send(sock, msg, siocb->scm);
1583 if (err < 0)
1584 return err;
1585
1586 err = -EOPNOTSUPP;
1587 if (msg->msg_flags&MSG_OOB)
1588 goto out_err;
1589
1590 if (msg->msg_namelen) {
1591 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1592 goto out_err;
1593 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 err = -ENOTCONN;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001595 other = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596 if (!other)
1597 goto out_err;
1598 }
1599
1600 if (sk->sk_shutdown & SEND_SHUTDOWN)
1601 goto pipe_err;
1602
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001603 while (sent < len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 /*
Benjamin LaHaisee9df7d72006-03-20 21:29:05 -08001605 * Optimisation for the fact that under 0.01% of X
1606 * messages typically need breaking up.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607 */
1608
Benjamin LaHaisee9df7d72006-03-20 21:29:05 -08001609 size = len-sent;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610
1611 /* Keep two messages in the pipe so it schedules better */
Benjamin LaHaisee9df7d72006-03-20 21:29:05 -08001612 if (size > ((sk->sk_sndbuf >> 1) - 64))
1613 size = (sk->sk_sndbuf >> 1) - 64;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614
1615 if (size > SKB_MAX_ALLOC)
1616 size = SKB_MAX_ALLOC;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001617
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 /*
1619 * Grab a buffer
1620 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001621
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001622 skb = sock_alloc_send_skb(sk, size, msg->msg_flags&MSG_DONTWAIT,
1623 &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624
Jianjun Konge27dfce2008-11-01 21:38:31 -07001625 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 goto out_err;
1627
1628 /*
1629 * If you pass two values to the sock_alloc_send_skb
1630 * it tries to grab the large buffer with GFP_NOFS
1631 * (which can fail easily), and if it fails grab the
1632 * fallback size buffer which is under a page and will
1633 * succeed. [Alan]
1634 */
1635 size = min_t(int, size, skb_tailroom(skb));
1636
Eric W. Biederman7361c362010-06-13 03:34:33 +00001637
Miklos Szeredi8ba69ba2009-09-11 11:31:45 -07001638 /* Only send the fds in the first buffer */
Eric W. Biederman7361c362010-06-13 03:34:33 +00001639 err = unix_scm_to_skb(siocb->scm, skb, !fds_sent);
Eric Dumazet25888e32010-11-25 04:11:39 +00001640 if (err < 0) {
Eric W. Biederman7361c362010-06-13 03:34:33 +00001641 kfree_skb(skb);
1642 goto out_err;
Miklos Szeredi62093442008-11-09 15:23:57 +01001643 }
Eric Dumazet25888e32010-11-25 04:11:39 +00001644 max_level = err + 1;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001645 fds_sent = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001647 err = memcpy_fromiovec(skb_put(skb, size), msg->msg_iov, size);
1648 if (err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 kfree_skb(skb);
1650 goto out_err;
1651 }
1652
David S. Miller1c92b4e2007-05-31 13:24:26 -07001653 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654
1655 if (sock_flag(other, SOCK_DEAD) ||
1656 (other->sk_shutdown & RCV_SHUTDOWN))
1657 goto pipe_err_free;
1658
1659 skb_queue_tail(&other->sk_receive_queue, skb);
Eric Dumazet25888e32010-11-25 04:11:39 +00001660 if (max_level > unix_sk(other)->recursion_level)
1661 unix_sk(other)->recursion_level = max_level;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001662 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 other->sk_data_ready(other, size);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001664 sent += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666
1667 scm_destroy(siocb->scm);
1668 siocb->scm = NULL;
1669
1670 return sent;
1671
1672pipe_err_free:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001673 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 kfree_skb(skb);
1675pipe_err:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001676 if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL))
1677 send_sig(SIGPIPE, current, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 err = -EPIPE;
1679out_err:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 scm_destroy(siocb->scm);
1681 siocb->scm = NULL;
1682 return sent ? : err;
1683}
1684
1685static int unix_seqpacket_sendmsg(struct kiocb *kiocb, struct socket *sock,
1686 struct msghdr *msg, size_t len)
1687{
1688 int err;
1689 struct sock *sk = sock->sk;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001690
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 err = sock_error(sk);
1692 if (err)
1693 return err;
1694
1695 if (sk->sk_state != TCP_ESTABLISHED)
1696 return -ENOTCONN;
1697
1698 if (msg->msg_namelen)
1699 msg->msg_namelen = 0;
1700
1701 return unix_dgram_sendmsg(kiocb, sock, msg, len);
1702}
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001703
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +00001704static int unix_seqpacket_recvmsg(struct kiocb *iocb, struct socket *sock,
1705 struct msghdr *msg, size_t size,
1706 int flags)
1707{
1708 struct sock *sk = sock->sk;
1709
1710 if (sk->sk_state != TCP_ESTABLISHED)
1711 return -ENOTCONN;
1712
1713 return unix_dgram_recvmsg(iocb, sock, msg, size, flags);
1714}
1715
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716static void unix_copy_addr(struct msghdr *msg, struct sock *sk)
1717{
1718 struct unix_sock *u = unix_sk(sk);
1719
1720 msg->msg_namelen = 0;
1721 if (u->addr) {
1722 msg->msg_namelen = u->addr->len;
1723 memcpy(msg->msg_name, u->addr->name, u->addr->len);
1724 }
1725}
1726
1727static int unix_dgram_recvmsg(struct kiocb *iocb, struct socket *sock,
1728 struct msghdr *msg, size_t size,
1729 int flags)
1730{
1731 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
1732 struct scm_cookie tmp_scm;
1733 struct sock *sk = sock->sk;
1734 struct unix_sock *u = unix_sk(sk);
1735 int noblock = flags & MSG_DONTWAIT;
1736 struct sk_buff *skb;
1737 int err;
1738
1739 err = -EOPNOTSUPP;
1740 if (flags&MSG_OOB)
1741 goto out;
1742
1743 msg->msg_namelen = 0;
1744
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001745 err = mutex_lock_interruptible(&u->readlock);
1746 if (err) {
1747 err = sock_intr_errno(sock_rcvtimeo(sk, noblock));
1748 goto out;
1749 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750
1751 skb = skb_recv_datagram(sk, flags, noblock, &err);
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001752 if (!skb) {
1753 unix_state_lock(sk);
1754 /* Signal EOF on disconnected non-blocking SEQPACKET socket. */
1755 if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN &&
1756 (sk->sk_shutdown & RCV_SHUTDOWN))
1757 err = 0;
1758 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 goto out_unlock;
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001760 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761
Eric Dumazet67426b72010-10-29 20:44:44 +00001762 wake_up_interruptible_sync_poll(&u->peer_wait,
1763 POLLOUT | POLLWRNORM | POLLWRBAND);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764
1765 if (msg->msg_name)
1766 unix_copy_addr(msg, skb->sk);
1767
1768 if (size > skb->len)
1769 size = skb->len;
1770 else if (size < skb->len)
1771 msg->msg_flags |= MSG_TRUNC;
1772
1773 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, size);
1774 if (err)
1775 goto out_free;
1776
Alban Crequy3f661162010-10-04 08:48:28 +00001777 if (sock_flag(sk, SOCK_RCVTSTAMP))
1778 __sock_recv_timestamp(msg, sk, skb);
1779
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 if (!siocb->scm) {
1781 siocb->scm = &tmp_scm;
1782 memset(&tmp_scm, 0, sizeof(tmp_scm));
1783 }
Eric W. Biederman7361c362010-06-13 03:34:33 +00001784 scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).cred);
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001785 unix_set_secdata(siocb->scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001787 if (!(flags & MSG_PEEK)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788 if (UNIXCB(skb).fp)
1789 unix_detach_fds(siocb->scm, skb);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001790 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 /* It is questionable: on PEEK we could:
1792 - do not return fds - good, but too simple 8)
1793 - return fds, and do not return them on read (old strategy,
1794 apparently wrong)
1795 - clone fds (I chose it for now, it is the most universal
1796 solution)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001797
1798 POSIX 1003.1g does not actually define this clearly
1799 at all. POSIX 1003.1g doesn't define a lot of things
1800 clearly however!
1801
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 */
1803 if (UNIXCB(skb).fp)
1804 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
1805 }
1806 err = size;
1807
1808 scm_recv(sock, msg, siocb->scm, flags);
1809
1810out_free:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001811 skb_free_datagram(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812out_unlock:
Ingo Molnar57b47a52006-03-20 22:35:41 -08001813 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814out:
1815 return err;
1816}
1817
1818/*
1819 * Sleep until data has arrive. But check for races..
1820 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001821
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001822static long unix_stream_data_wait(struct sock *sk, long timeo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823{
1824 DEFINE_WAIT(wait);
1825
David S. Miller1c92b4e2007-05-31 13:24:26 -07001826 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827
1828 for (;;) {
Eric Dumazetaa395142010-04-20 13:03:51 +00001829 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830
David S. Millerb03efcf2005-07-08 14:57:23 -07001831 if (!skb_queue_empty(&sk->sk_receive_queue) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832 sk->sk_err ||
1833 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1834 signal_pending(current) ||
1835 !timeo)
1836 break;
1837
1838 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001839 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 timeo = schedule_timeout(timeo);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001841 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1843 }
1844
Eric Dumazetaa395142010-04-20 13:03:51 +00001845 finish_wait(sk_sleep(sk), &wait);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001846 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847 return timeo;
1848}
1849
1850
1851
1852static int unix_stream_recvmsg(struct kiocb *iocb, struct socket *sock,
1853 struct msghdr *msg, size_t size,
1854 int flags)
1855{
1856 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
1857 struct scm_cookie tmp_scm;
1858 struct sock *sk = sock->sk;
1859 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001860 struct sockaddr_un *sunaddr = msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 int copied = 0;
1862 int check_creds = 0;
1863 int target;
1864 int err = 0;
1865 long timeo;
1866
1867 err = -EINVAL;
1868 if (sk->sk_state != TCP_ESTABLISHED)
1869 goto out;
1870
1871 err = -EOPNOTSUPP;
1872 if (flags&MSG_OOB)
1873 goto out;
1874
1875 target = sock_rcvlowat(sk, flags&MSG_WAITALL, size);
1876 timeo = sock_rcvtimeo(sk, flags&MSG_DONTWAIT);
1877
1878 msg->msg_namelen = 0;
1879
1880 /* Lock the socket to prevent queue disordering
1881 * while sleeps in memcpy_tomsg
1882 */
1883
1884 if (!siocb->scm) {
1885 siocb->scm = &tmp_scm;
1886 memset(&tmp_scm, 0, sizeof(tmp_scm));
1887 }
1888
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001889 err = mutex_lock_interruptible(&u->readlock);
1890 if (err) {
1891 err = sock_intr_errno(timeo);
1892 goto out;
1893 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001895 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 int chunk;
1897 struct sk_buff *skb;
1898
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001899 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 skb = skb_dequeue(&sk->sk_receive_queue);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001901 if (skb == NULL) {
Eric Dumazet25888e32010-11-25 04:11:39 +00001902 unix_sk(sk)->recursion_level = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 if (copied >= target)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001904 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905
1906 /*
1907 * POSIX 1003.1g mandates this order.
1908 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001909
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001910 err = sock_error(sk);
1911 if (err)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001912 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 if (sk->sk_shutdown & RCV_SHUTDOWN)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001914 goto unlock;
1915
1916 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 err = -EAGAIN;
1918 if (!timeo)
1919 break;
Ingo Molnar57b47a52006-03-20 22:35:41 -08001920 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921
1922 timeo = unix_stream_data_wait(sk, timeo);
1923
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001924 if (signal_pending(current)
1925 || mutex_lock_interruptible(&u->readlock)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 err = sock_intr_errno(timeo);
1927 goto out;
1928 }
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001929
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 continue;
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001931 unlock:
1932 unix_state_unlock(sk);
1933 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 }
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001935 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936
1937 if (check_creds) {
1938 /* Never glue messages from different writers */
Eric W. Biederman7361c362010-06-13 03:34:33 +00001939 if ((UNIXCB(skb).pid != siocb->scm->pid) ||
1940 (UNIXCB(skb).cred != siocb->scm->cred)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 skb_queue_head(&sk->sk_receive_queue, skb);
1942 break;
1943 }
1944 } else {
1945 /* Copy credentials */
Eric W. Biederman7361c362010-06-13 03:34:33 +00001946 scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947 check_creds = 1;
1948 }
1949
1950 /* Copy address just once */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001951 if (sunaddr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 unix_copy_addr(msg, skb->sk);
1953 sunaddr = NULL;
1954 }
1955
1956 chunk = min_t(unsigned int, skb->len, size);
1957 if (memcpy_toiovec(msg->msg_iov, skb->data, chunk)) {
1958 skb_queue_head(&sk->sk_receive_queue, skb);
1959 if (copied == 0)
1960 copied = -EFAULT;
1961 break;
1962 }
1963 copied += chunk;
1964 size -= chunk;
1965
1966 /* Mark read part of skb as used */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001967 if (!(flags & MSG_PEEK)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968 skb_pull(skb, chunk);
1969
1970 if (UNIXCB(skb).fp)
1971 unix_detach_fds(siocb->scm, skb);
1972
1973 /* put the skb back if we didn't use it up.. */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001974 if (skb->len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975 skb_queue_head(&sk->sk_receive_queue, skb);
1976 break;
1977 }
1978
Neil Horman70d4bf62010-07-20 06:45:56 +00001979 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980
1981 if (siocb->scm->fp)
1982 break;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001983 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 /* It is questionable, see note in unix_dgram_recvmsg.
1985 */
1986 if (UNIXCB(skb).fp)
1987 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
1988
1989 /* put message back and return */
1990 skb_queue_head(&sk->sk_receive_queue, skb);
1991 break;
1992 }
1993 } while (size);
1994
Ingo Molnar57b47a52006-03-20 22:35:41 -08001995 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 scm_recv(sock, msg, siocb->scm, flags);
1997out:
1998 return copied ? : err;
1999}
2000
2001static int unix_shutdown(struct socket *sock, int mode)
2002{
2003 struct sock *sk = sock->sk;
2004 struct sock *other;
2005
2006 mode = (mode+1)&(RCV_SHUTDOWN|SEND_SHUTDOWN);
2007
Alban Crequy7180a032011-01-19 04:56:36 +00002008 if (!mode)
2009 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010
Alban Crequy7180a032011-01-19 04:56:36 +00002011 unix_state_lock(sk);
2012 sk->sk_shutdown |= mode;
2013 other = unix_peer(sk);
2014 if (other)
2015 sock_hold(other);
2016 unix_state_unlock(sk);
2017 sk->sk_state_change(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018
Alban Crequy7180a032011-01-19 04:56:36 +00002019 if (other &&
2020 (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021
Alban Crequy7180a032011-01-19 04:56:36 +00002022 int peer_mode = 0;
2023
2024 if (mode&RCV_SHUTDOWN)
2025 peer_mode |= SEND_SHUTDOWN;
2026 if (mode&SEND_SHUTDOWN)
2027 peer_mode |= RCV_SHUTDOWN;
2028 unix_state_lock(other);
2029 other->sk_shutdown |= peer_mode;
2030 unix_state_unlock(other);
2031 other->sk_state_change(other);
2032 if (peer_mode == SHUTDOWN_MASK)
2033 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP);
2034 else if (peer_mode & RCV_SHUTDOWN)
2035 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036 }
Alban Crequy7180a032011-01-19 04:56:36 +00002037 if (other)
2038 sock_put(other);
2039
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 return 0;
2041}
2042
2043static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2044{
2045 struct sock *sk = sock->sk;
Jianjun Konge27dfce2008-11-01 21:38:31 -07002046 long amount = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 int err;
2048
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002049 switch (cmd) {
2050 case SIOCOUTQ:
Eric Dumazet31e6d362009-06-17 19:05:41 -07002051 amount = sk_wmem_alloc_get(sk);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002052 err = put_user(amount, (int __user *)arg);
2053 break;
2054 case SIOCINQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 {
2056 struct sk_buff *skb;
2057
2058 if (sk->sk_state == TCP_LISTEN) {
2059 err = -EINVAL;
2060 break;
2061 }
2062
2063 spin_lock(&sk->sk_receive_queue.lock);
2064 if (sk->sk_type == SOCK_STREAM ||
2065 sk->sk_type == SOCK_SEQPACKET) {
2066 skb_queue_walk(&sk->sk_receive_queue, skb)
2067 amount += skb->len;
2068 } else {
2069 skb = skb_peek(&sk->sk_receive_queue);
2070 if (skb)
Jianjun Konge27dfce2008-11-01 21:38:31 -07002071 amount = skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072 }
2073 spin_unlock(&sk->sk_receive_queue.lock);
2074 err = put_user(amount, (int __user *)arg);
2075 break;
2076 }
2077
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002078 default:
2079 err = -ENOIOCTLCMD;
2080 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081 }
2082 return err;
2083}
2084
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002085static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086{
2087 struct sock *sk = sock->sk;
2088 unsigned int mask;
2089
Eric Dumazetaa395142010-04-20 13:03:51 +00002090 sock_poll_wait(file, sk_sleep(sk), wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 mask = 0;
2092
2093 /* exceptional events? */
2094 if (sk->sk_err)
2095 mask |= POLLERR;
2096 if (sk->sk_shutdown == SHUTDOWN_MASK)
2097 mask |= POLLHUP;
Davide Libenzif348d702006-03-25 03:07:39 -08002098 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazetdb409802010-09-06 11:13:50 +00002099 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100
2101 /* readable? */
Eric Dumazetdb409802010-09-06 11:13:50 +00002102 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 mask |= POLLIN | POLLRDNORM;
2104
2105 /* Connection-based need to check for termination and startup */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002106 if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) &&
2107 sk->sk_state == TCP_CLOSE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 mask |= POLLHUP;
2109
2110 /*
2111 * we set writable also when the other side has shut down the
2112 * connection. This prevents stuck sockets.
2113 */
2114 if (unix_writable(sk))
2115 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2116
2117 return mask;
2118}
2119
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002120static unsigned int unix_dgram_poll(struct file *file, struct socket *sock,
2121 poll_table *wait)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002122{
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002123 struct sock *sk = sock->sk, *other;
2124 unsigned int mask, writable;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002125
Eric Dumazetaa395142010-04-20 13:03:51 +00002126 sock_poll_wait(file, sk_sleep(sk), wait);
Rainer Weikusat3c734192008-06-17 22:28:05 -07002127 mask = 0;
2128
2129 /* exceptional events? */
2130 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
2131 mask |= POLLERR;
2132 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazet5456f092010-10-31 05:36:23 +00002133 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002134 if (sk->sk_shutdown == SHUTDOWN_MASK)
2135 mask |= POLLHUP;
2136
2137 /* readable? */
Eric Dumazet5456f092010-10-31 05:36:23 +00002138 if (!skb_queue_empty(&sk->sk_receive_queue))
Rainer Weikusat3c734192008-06-17 22:28:05 -07002139 mask |= POLLIN | POLLRDNORM;
2140
2141 /* Connection-based need to check for termination and startup */
2142 if (sk->sk_type == SOCK_SEQPACKET) {
2143 if (sk->sk_state == TCP_CLOSE)
2144 mask |= POLLHUP;
2145 /* connection hasn't started yet? */
2146 if (sk->sk_state == TCP_SYN_SENT)
2147 return mask;
2148 }
2149
Eric Dumazet973a34a2010-10-31 05:38:25 +00002150 /* No write status requested, avoid expensive OUT tests. */
2151 if (wait && !(wait->key & (POLLWRBAND | POLLWRNORM | POLLOUT)))
2152 return mask;
2153
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002154 writable = unix_writable(sk);
Eric Dumazet5456f092010-10-31 05:36:23 +00002155 other = unix_peer_get(sk);
2156 if (other) {
2157 if (unix_peer(other) != sk) {
2158 sock_poll_wait(file, &unix_sk(other)->peer_wait, wait);
2159 if (unix_recvq_full(other))
2160 writable = 0;
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002161 }
Eric Dumazet5456f092010-10-31 05:36:23 +00002162 sock_put(other);
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002163 }
2164
2165 if (writable)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002166 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2167 else
2168 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
2169
Rainer Weikusat3c734192008-06-17 22:28:05 -07002170 return mask;
2171}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002172
2173#ifdef CONFIG_PROC_FS
Pavel Emelyanova53eb3f2007-11-23 20:30:01 +08002174static struct sock *first_unix_socket(int *i)
2175{
2176 for (*i = 0; *i <= UNIX_HASH_SIZE; (*i)++) {
2177 if (!hlist_empty(&unix_socket_table[*i]))
2178 return __sk_head(&unix_socket_table[*i]);
2179 }
2180 return NULL;
2181}
2182
2183static struct sock *next_unix_socket(int *i, struct sock *s)
2184{
2185 struct sock *next = sk_next(s);
2186 /* More in this chain? */
2187 if (next)
2188 return next;
2189 /* Look for next non-empty chain. */
2190 for ((*i)++; *i <= UNIX_HASH_SIZE; (*i)++) {
2191 if (!hlist_empty(&unix_socket_table[*i]))
2192 return __sk_head(&unix_socket_table[*i]);
2193 }
2194 return NULL;
2195}
2196
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002197struct unix_iter_state {
Denis V. Luneve372c412007-11-19 22:31:54 -08002198 struct seq_net_private p;
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002199 int i;
2200};
Jianjun Konge27dfce2008-11-01 21:38:31 -07002201
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002202static struct sock *unix_seq_idx(struct seq_file *seq, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203{
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002204 struct unix_iter_state *iter = seq->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205 loff_t off = 0;
2206 struct sock *s;
2207
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002208 for (s = first_unix_socket(&iter->i); s; s = next_unix_socket(&iter->i, s)) {
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002209 if (sock_net(s) != seq_file_net(seq))
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002210 continue;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002211 if (off == pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 return s;
2213 ++off;
2214 }
2215 return NULL;
2216}
2217
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218static void *unix_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002219 __acquires(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002221 spin_lock(&unix_table_lock);
Joe Perchesb9f31242008-04-12 19:04:38 -07002222 return *pos ? unix_seq_idx(seq, *pos - 1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223}
2224
2225static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2226{
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002227 struct unix_iter_state *iter = seq->private;
2228 struct sock *sk = v;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 ++*pos;
2230
Joe Perchesb9f31242008-04-12 19:04:38 -07002231 if (v == SEQ_START_TOKEN)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002232 sk = first_unix_socket(&iter->i);
2233 else
2234 sk = next_unix_socket(&iter->i, sk);
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002235 while (sk && (sock_net(sk) != seq_file_net(seq)))
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002236 sk = next_unix_socket(&iter->i, sk);
2237 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238}
2239
2240static void unix_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002241 __releases(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002243 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244}
2245
2246static int unix_seq_show(struct seq_file *seq, void *v)
2247{
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002248
Joe Perchesb9f31242008-04-12 19:04:38 -07002249 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250 seq_puts(seq, "Num RefCount Protocol Flags Type St "
2251 "Inode Path\n");
2252 else {
2253 struct sock *s = v;
2254 struct unix_sock *u = unix_sk(s);
David S. Miller1c92b4e2007-05-31 13:24:26 -07002255 unix_state_lock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002256
2257 seq_printf(seq, "%p: %08X %08X %08X %04X %02X %5lu",
2258 s,
2259 atomic_read(&s->sk_refcnt),
2260 0,
2261 s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0,
2262 s->sk_type,
2263 s->sk_socket ?
2264 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) :
2265 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING),
2266 sock_i_ino(s));
2267
2268 if (u->addr) {
2269 int i, len;
2270 seq_putc(seq, ' ');
2271
2272 i = 0;
2273 len = u->addr->len - sizeof(short);
2274 if (!UNIX_ABSTRACT(s))
2275 len--;
2276 else {
2277 seq_putc(seq, '@');
2278 i++;
2279 }
2280 for ( ; i < len; i++)
2281 seq_putc(seq, u->addr->name->sun_path[i]);
2282 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07002283 unix_state_unlock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 seq_putc(seq, '\n');
2285 }
2286
2287 return 0;
2288}
2289
Philippe De Muyter56b3d972007-07-10 23:07:31 -07002290static const struct seq_operations unix_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291 .start = unix_seq_start,
2292 .next = unix_seq_next,
2293 .stop = unix_seq_stop,
2294 .show = unix_seq_show,
2295};
2296
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297static int unix_seq_open(struct inode *inode, struct file *file)
2298{
Denis V. Luneve372c412007-11-19 22:31:54 -08002299 return seq_open_net(inode, file, &unix_seq_ops,
2300 sizeof(struct unix_iter_state));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301}
2302
Arjan van de Venda7071d2007-02-12 00:55:36 -08002303static const struct file_operations unix_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304 .owner = THIS_MODULE,
2305 .open = unix_seq_open,
2306 .read = seq_read,
2307 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08002308 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309};
2310
2311#endif
2312
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00002313static const struct net_proto_family unix_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314 .family = PF_UNIX,
2315 .create = unix_create,
2316 .owner = THIS_MODULE,
2317};
2318
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002319
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002320static int __net_init unix_net_init(struct net *net)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002321{
2322 int error = -ENOMEM;
2323
Denis V. Luneva0a53c82007-12-11 04:19:17 -08002324 net->unx.sysctl_max_dgram_qlen = 10;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002325 if (unix_sysctl_register(net))
2326 goto out;
Pavel Emelyanovd392e492007-12-01 23:44:15 +11002327
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002328#ifdef CONFIG_PROC_FS
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002329 if (!proc_net_fops_create(net, "unix", 0, &unix_seq_fops)) {
2330 unix_sysctl_unregister(net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002331 goto out;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002332 }
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002333#endif
2334 error = 0;
2335out:
Jianjun Kong48dcc332008-11-01 21:37:27 -07002336 return error;
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002337}
2338
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002339static void __net_exit unix_net_exit(struct net *net)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002340{
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002341 unix_sysctl_unregister(net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002342 proc_net_remove(net, "unix");
2343}
2344
2345static struct pernet_operations unix_net_ops = {
2346 .init = unix_net_init,
2347 .exit = unix_net_exit,
2348};
2349
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350static int __init af_unix_init(void)
2351{
2352 int rc = -1;
2353 struct sk_buff *dummy_skb;
2354
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07002355 BUILD_BUG_ON(sizeof(struct unix_skb_parms) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356
2357 rc = proto_register(&unix_proto, 1);
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002358 if (rc != 0) {
2359 printk(KERN_CRIT "%s: Cannot create unix_sock SLAB cache!\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08002360 __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002361 goto out;
2362 }
2363
2364 sock_register(&unix_family_ops);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002365 register_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366out:
2367 return rc;
2368}
2369
2370static void __exit af_unix_exit(void)
2371{
2372 sock_unregister(PF_UNIX);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373 proto_unregister(&unix_proto);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002374 unregister_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375}
2376
David Woodhouse3d366962008-04-24 00:59:25 -07002377/* Earlier than device_initcall() so that other drivers invoking
2378 request_module() don't end up in a loop when modprobe tries
2379 to use a UNIX socket. But later than subsys_initcall() because
2380 we depend on stuff initialised there */
2381fs_initcall(af_unix_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382module_exit(af_unix_exit);
2383
2384MODULE_LICENSE("GPL");
2385MODULE_ALIAS_NETPROTO(PF_UNIX);