blob: 9dcc6e7f96ec2126039ce6b57768b7274bf5f19e [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);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120static atomic_t unix_nr_socks = ATOMIC_INIT(0);
121
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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 * each socket state is protected by separate rwlock.
148 */
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.
210 * sun_path[108] doesnt as such exist. However in kernel space
211 * 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
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800285static struct sock *unix_find_socket_byinode(struct net *net, 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
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900295 if (!net_eq(sock_net(s), net))
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800296 continue;
297
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800298 if (dentry && dentry->d_inode == i) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 sock_hold(s);
300 goto found;
301 }
302 }
303 s = NULL;
304found:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800305 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 return s;
307}
308
309static inline int unix_writable(struct sock *sk)
310{
311 return (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf;
312}
313
314static void unix_write_space(struct sock *sk)
315{
316 read_lock(&sk->sk_callback_lock);
317 if (unix_writable(sk)) {
318 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
Ingo Molnar71e20f12007-10-15 17:00:19 +0200319 wake_up_interruptible_sync(sk->sk_sleep);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +0800320 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 }
322 read_unlock(&sk->sk_callback_lock);
323}
324
325/* When dgram socket disconnects (or changes its peer), we clear its receive
326 * queue of packets arrived from previous peer. First, it allows to do
327 * flow control based only on wmem_alloc; second, sk connected to peer
328 * may receive messages only from that peer. */
329static void unix_dgram_disconnected(struct sock *sk, struct sock *other)
330{
David S. Millerb03efcf2005-07-08 14:57:23 -0700331 if (!skb_queue_empty(&sk->sk_receive_queue)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 skb_queue_purge(&sk->sk_receive_queue);
333 wake_up_interruptible_all(&unix_sk(sk)->peer_wait);
334
335 /* If one link of bidirectional dgram pipe is disconnected,
336 * we signal error. Messages are lost. Do not make this,
337 * when peer was not connected to us.
338 */
339 if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) {
340 other->sk_err = ECONNRESET;
341 other->sk_error_report(other);
342 }
343 }
344}
345
346static void unix_sock_destructor(struct sock *sk)
347{
348 struct unix_sock *u = unix_sk(sk);
349
350 skb_queue_purge(&sk->sk_receive_queue);
351
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700352 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
353 WARN_ON(!sk_unhashed(sk));
354 WARN_ON(sk->sk_socket);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet6b41e7d2008-11-19 15:48:09 -0800356 printk(KERN_INFO "Attempt to release alive unix socket: %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 return;
358 }
359
360 if (u->addr)
361 unix_release_addr(u->addr);
362
363 atomic_dec(&unix_nr_socks);
David S. Miller6f756a82008-11-23 17:34:03 -0800364 local_bh_disable();
Eric Dumazeta8076d82008-11-17 02:38:49 -0800365 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
David S. Miller6f756a82008-11-23 17:34:03 -0800366 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367#ifdef UNIX_REFCNT_DEBUG
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800368 printk(KERN_DEBUG "UNIX %p is destroyed, %d are still alive.\n", sk,
369 atomic_read(&unix_nr_socks));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370#endif
371}
372
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800373static int unix_release_sock(struct sock *sk, int embrion)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374{
375 struct unix_sock *u = unix_sk(sk);
376 struct dentry *dentry;
377 struct vfsmount *mnt;
378 struct sock *skpair;
379 struct sk_buff *skb;
380 int state;
381
382 unix_remove_socket(sk);
383
384 /* Clear state */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700385 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 sock_orphan(sk);
387 sk->sk_shutdown = SHUTDOWN_MASK;
388 dentry = u->dentry;
389 u->dentry = NULL;
390 mnt = u->mnt;
391 u->mnt = NULL;
392 state = sk->sk_state;
393 sk->sk_state = TCP_CLOSE;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700394 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395
396 wake_up_interruptible_all(&u->peer_wait);
397
Jianjun Konge27dfce2008-11-01 21:38:31 -0700398 skpair = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
Jianjun Konge27dfce2008-11-01 21:38:31 -0700400 if (skpair != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) {
David S. Miller1c92b4e2007-05-31 13:24:26 -0700402 unix_state_lock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 /* No more writes */
404 skpair->sk_shutdown = SHUTDOWN_MASK;
405 if (!skb_queue_empty(&sk->sk_receive_queue) || embrion)
406 skpair->sk_err = ECONNRESET;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700407 unix_state_unlock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 skpair->sk_state_change(skpair);
409 read_lock(&skpair->sk_callback_lock);
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 read_unlock(&skpair->sk_callback_lock);
412 }
413 sock_put(skpair); /* It may now die */
414 unix_peer(sk) = NULL;
415 }
416
417 /* Try to flush out this socket. Throw out buffers at least */
418
419 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
Jianjun Konge27dfce2008-11-01 21:38:31 -0700420 if (state == TCP_LISTEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 unix_release_sock(skb->sk, 1);
422 /* passed fds are erased in the kfree_skb hook */
423 kfree_skb(skb);
424 }
425
426 if (dentry) {
427 dput(dentry);
428 mntput(mnt);
429 }
430
431 sock_put(sk);
432
433 /* ---- Socket is dead now and most probably destroyed ---- */
434
435 /*
436 * Fixme: BSD difference: In BSD all sockets connected to use get
437 * ECONNRESET and we die on the spot. In Linux we behave
438 * like files and pipes do and wait for the last
439 * dereference.
440 *
441 * Can't we simply set sock->err?
442 *
443 * What the above comment does talk about? --ANK(980817)
444 */
445
Pavel Emelyanov9305cfa2007-11-10 22:06:01 -0800446 if (unix_tot_inflight)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900447 unix_gc(); /* Garbage collect fds */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448
449 return 0;
450}
451
452static int unix_listen(struct socket *sock, int backlog)
453{
454 int err;
455 struct sock *sk = sock->sk;
456 struct unix_sock *u = unix_sk(sk);
457
458 err = -EOPNOTSUPP;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800459 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
460 goto out; /* Only stream/seqpacket sockets accept */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 err = -EINVAL;
462 if (!u->addr)
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800463 goto out; /* No listens on an unbound socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700464 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN)
466 goto out_unlock;
467 if (backlog > sk->sk_max_ack_backlog)
468 wake_up_interruptible_all(&u->peer_wait);
469 sk->sk_max_ack_backlog = backlog;
470 sk->sk_state = TCP_LISTEN;
471 /* set credentials so connect can copy them */
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700472 sk->sk_peercred.pid = task_tgid_vnr(current);
David Howells19d65622008-11-14 10:39:10 +1100473 current_euid_egid(&sk->sk_peercred.uid, &sk->sk_peercred.gid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 err = 0;
475
476out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -0700477 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478out:
479 return err;
480}
481
482static int unix_release(struct socket *);
483static int unix_bind(struct socket *, struct sockaddr *, int);
484static int unix_stream_connect(struct socket *, struct sockaddr *,
485 int addr_len, int flags);
486static int unix_socketpair(struct socket *, struct socket *);
487static int unix_accept(struct socket *, struct socket *, int);
488static int unix_getname(struct socket *, struct sockaddr *, int *, int);
489static unsigned int unix_poll(struct file *, struct socket *, poll_table *);
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700490static unsigned int unix_dgram_poll(struct file *, struct socket *,
491 poll_table *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492static int unix_ioctl(struct socket *, unsigned int, unsigned long);
493static int unix_shutdown(struct socket *, int);
494static int unix_stream_sendmsg(struct kiocb *, struct socket *,
495 struct msghdr *, size_t);
496static int unix_stream_recvmsg(struct kiocb *, struct socket *,
497 struct msghdr *, size_t, int);
498static int unix_dgram_sendmsg(struct kiocb *, struct socket *,
499 struct msghdr *, size_t);
500static int unix_dgram_recvmsg(struct kiocb *, struct socket *,
501 struct msghdr *, size_t, int);
502static int unix_dgram_connect(struct socket *, struct sockaddr *,
503 int, int);
504static int unix_seqpacket_sendmsg(struct kiocb *, struct socket *,
505 struct msghdr *, size_t);
506
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800507static const struct proto_ops unix_stream_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 .family = PF_UNIX,
509 .owner = THIS_MODULE,
510 .release = unix_release,
511 .bind = unix_bind,
512 .connect = unix_stream_connect,
513 .socketpair = unix_socketpair,
514 .accept = unix_accept,
515 .getname = unix_getname,
516 .poll = unix_poll,
517 .ioctl = unix_ioctl,
518 .listen = unix_listen,
519 .shutdown = unix_shutdown,
520 .setsockopt = sock_no_setsockopt,
521 .getsockopt = sock_no_getsockopt,
522 .sendmsg = unix_stream_sendmsg,
523 .recvmsg = unix_stream_recvmsg,
524 .mmap = sock_no_mmap,
525 .sendpage = sock_no_sendpage,
526};
527
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800528static const struct proto_ops unix_dgram_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 .family = PF_UNIX,
530 .owner = THIS_MODULE,
531 .release = unix_release,
532 .bind = unix_bind,
533 .connect = unix_dgram_connect,
534 .socketpair = unix_socketpair,
535 .accept = sock_no_accept,
536 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700537 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 .ioctl = unix_ioctl,
539 .listen = sock_no_listen,
540 .shutdown = unix_shutdown,
541 .setsockopt = sock_no_setsockopt,
542 .getsockopt = sock_no_getsockopt,
543 .sendmsg = unix_dgram_sendmsg,
544 .recvmsg = unix_dgram_recvmsg,
545 .mmap = sock_no_mmap,
546 .sendpage = sock_no_sendpage,
547};
548
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800549static const struct proto_ops unix_seqpacket_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 .family = PF_UNIX,
551 .owner = THIS_MODULE,
552 .release = unix_release,
553 .bind = unix_bind,
554 .connect = unix_stream_connect,
555 .socketpair = unix_socketpair,
556 .accept = unix_accept,
557 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700558 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 .ioctl = unix_ioctl,
560 .listen = unix_listen,
561 .shutdown = unix_shutdown,
562 .setsockopt = sock_no_setsockopt,
563 .getsockopt = sock_no_getsockopt,
564 .sendmsg = unix_seqpacket_sendmsg,
565 .recvmsg = unix_dgram_recvmsg,
566 .mmap = sock_no_mmap,
567 .sendpage = sock_no_sendpage,
568};
569
570static struct proto unix_proto = {
Eric Dumazet248969a2008-11-17 00:00:30 -0800571 .name = "UNIX",
572 .owner = THIS_MODULE,
Eric Dumazet248969a2008-11-17 00:00:30 -0800573 .obj_size = sizeof(struct unix_sock),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574};
575
Ingo Molnara09785a2006-07-03 00:25:12 -0700576/*
577 * AF_UNIX sockets do not interact with hardware, hence they
578 * dont trigger interrupts - so it's safe for them to have
579 * bh-unsafe locking for their sk_receive_queue.lock. Split off
580 * this special lock-class by reinitializing the spinlock key:
581 */
582static struct lock_class_key af_unix_sk_receive_queue_lock_key;
583
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800584static struct sock *unix_create1(struct net *net, struct socket *sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585{
586 struct sock *sk = NULL;
587 struct unix_sock *u;
588
Pavel Emelyanov284b3272007-11-10 22:08:30 -0800589 atomic_inc(&unix_nr_socks);
590 if (atomic_read(&unix_nr_socks) > 2 * get_max_files())
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 goto out;
592
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700593 sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 if (!sk)
595 goto out;
596
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800597 sock_init_data(sock, sk);
Ingo Molnara09785a2006-07-03 00:25:12 -0700598 lockdep_set_class(&sk->sk_receive_queue.lock,
599 &af_unix_sk_receive_queue_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601 sk->sk_write_space = unix_write_space;
Denis V. Luneva0a53c82007-12-11 04:19:17 -0800602 sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 sk->sk_destruct = unix_sock_destructor;
604 u = unix_sk(sk);
605 u->dentry = NULL;
606 u->mnt = NULL;
Benjamin LaHaisefd19f322006-01-03 14:10:46 -0800607 spin_lock_init(&u->lock);
Al Viro516e0cc2008-07-26 00:39:17 -0400608 atomic_long_set(&u->inflight, 0);
Miklos Szeredi1fd05ba2007-07-11 14:22:39 -0700609 INIT_LIST_HEAD(&u->link);
Ingo Molnar57b47a52006-03-20 22:35:41 -0800610 mutex_init(&u->readlock); /* single task reading lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 init_waitqueue_head(&u->peer_wait);
612 unix_insert_socket(unix_sockets_unbound, sk);
613out:
Pavel Emelyanov284b3272007-11-10 22:08:30 -0800614 if (sk == NULL)
615 atomic_dec(&unix_nr_socks);
Eric Dumazet920de802008-11-24 00:09:29 -0800616 else {
617 local_bh_disable();
Eric Dumazeta8076d82008-11-17 02:38:49 -0800618 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet920de802008-11-24 00:09:29 -0800619 local_bh_enable();
620 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 return sk;
622}
623
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700624static int unix_create(struct net *net, struct socket *sock, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625{
626 if (protocol && protocol != PF_UNIX)
627 return -EPROTONOSUPPORT;
628
629 sock->state = SS_UNCONNECTED;
630
631 switch (sock->type) {
632 case SOCK_STREAM:
633 sock->ops = &unix_stream_ops;
634 break;
635 /*
636 * Believe it or not BSD has AF_UNIX, SOCK_RAW though
637 * nothing uses it.
638 */
639 case SOCK_RAW:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700640 sock->type = SOCK_DGRAM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 case SOCK_DGRAM:
642 sock->ops = &unix_dgram_ops;
643 break;
644 case SOCK_SEQPACKET:
645 sock->ops = &unix_seqpacket_ops;
646 break;
647 default:
648 return -ESOCKTNOSUPPORT;
649 }
650
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700651 return unix_create1(net, sock) ? 0 : -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652}
653
654static int unix_release(struct socket *sock)
655{
656 struct sock *sk = sock->sk;
657
658 if (!sk)
659 return 0;
660
661 sock->sk = NULL;
662
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800663 return unix_release_sock(sk, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664}
665
666static int unix_autobind(struct socket *sock)
667{
668 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900669 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 struct unix_sock *u = unix_sk(sk);
671 static u32 ordernum = 1;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800672 struct unix_address *addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 int err;
674
Ingo Molnar57b47a52006-03-20 22:35:41 -0800675 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
677 err = 0;
678 if (u->addr)
679 goto out;
680
681 err = -ENOMEM;
Panagiotis Issaris0da974f2006-07-21 14:51:30 -0700682 addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 if (!addr)
684 goto out;
685
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 addr->name->sun_family = AF_UNIX;
687 atomic_set(&addr->refcnt, 1);
688
689retry:
690 addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short);
Joe Perches07f07572008-11-19 15:44:53 -0800691 addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
David S. Millerfbe9cc42005-12-13 23:26:29 -0800693 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 ordernum = (ordernum+1)&0xFFFFF;
695
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800696 if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 addr->hash)) {
David S. Millerfbe9cc42005-12-13 23:26:29 -0800698 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 /* Sanity yield. It is unusual case, but yet... */
700 if (!(ordernum&0xFF))
701 yield();
702 goto retry;
703 }
704 addr->hash ^= sk->sk_type;
705
706 __unix_remove_socket(sk);
707 u->addr = addr;
708 __unix_insert_socket(&unix_socket_table[addr->hash], sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800709 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 err = 0;
711
Ingo Molnar57b47a52006-03-20 22:35:41 -0800712out: mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 return err;
714}
715
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800716static struct sock *unix_find_other(struct net *net,
717 struct sockaddr_un *sunname, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 int type, unsigned hash, int *error)
719{
720 struct sock *u;
Al Viro421748e2008-08-02 01:04:36 -0400721 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 int err = 0;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900723
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 if (sunname->sun_path[0]) {
Al Viro421748e2008-08-02 01:04:36 -0400725 struct inode *inode;
726 err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 if (err)
728 goto fail;
Al Viro421748e2008-08-02 01:04:36 -0400729 inode = path.dentry->d_inode;
730 err = inode_permission(inode, MAY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 if (err)
732 goto put_fail;
733
734 err = -ECONNREFUSED;
Al Viro421748e2008-08-02 01:04:36 -0400735 if (!S_ISSOCK(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 goto put_fail;
Al Viro421748e2008-08-02 01:04:36 -0400737 u = unix_find_socket_byinode(net, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 if (!u)
739 goto put_fail;
740
741 if (u->sk_type == type)
Al Viro421748e2008-08-02 01:04:36 -0400742 touch_atime(path.mnt, path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
Al Viro421748e2008-08-02 01:04:36 -0400744 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745
Jianjun Konge27dfce2008-11-01 21:38:31 -0700746 err = -EPROTOTYPE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 if (u->sk_type != type) {
748 sock_put(u);
749 goto fail;
750 }
751 } else {
752 err = -ECONNREFUSED;
Jianjun Konge27dfce2008-11-01 21:38:31 -0700753 u = unix_find_socket_byname(net, sunname, len, type, hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 if (u) {
755 struct dentry *dentry;
756 dentry = unix_sk(u)->dentry;
757 if (dentry)
758 touch_atime(unix_sk(u)->mnt, dentry);
759 } else
760 goto fail;
761 }
762 return u;
763
764put_fail:
Al Viro421748e2008-08-02 01:04:36 -0400765 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766fail:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700767 *error = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 return NULL;
769}
770
771
772static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
773{
774 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900775 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700777 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800778 struct dentry *dentry = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 struct nameidata nd;
780 int err;
781 unsigned hash;
782 struct unix_address *addr;
783 struct hlist_head *list;
784
785 err = -EINVAL;
786 if (sunaddr->sun_family != AF_UNIX)
787 goto out;
788
Jianjun Konge27dfce2008-11-01 21:38:31 -0700789 if (addr_len == sizeof(short)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 err = unix_autobind(sock);
791 goto out;
792 }
793
794 err = unix_mkname(sunaddr, addr_len, &hash);
795 if (err < 0)
796 goto out;
797 addr_len = err;
798
Ingo Molnar57b47a52006-03-20 22:35:41 -0800799 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
801 err = -EINVAL;
802 if (u->addr)
803 goto out_up;
804
805 err = -ENOMEM;
806 addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL);
807 if (!addr)
808 goto out_up;
809
810 memcpy(addr->name, sunaddr, addr_len);
811 addr->len = addr_len;
812 addr->hash = hash ^ sk->sk_type;
813 atomic_set(&addr->refcnt, 1);
814
815 if (sunaddr->sun_path[0]) {
816 unsigned int mode;
817 err = 0;
818 /*
819 * Get the parent directory, calculate the hash for last
820 * component.
821 */
822 err = path_lookup(sunaddr->sun_path, LOOKUP_PARENT, &nd);
823 if (err)
824 goto out_mknod_parent;
Christoph Hellwigf81a0bf2005-05-19 12:26:43 -0700825
826 dentry = lookup_create(&nd, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 err = PTR_ERR(dentry);
828 if (IS_ERR(dentry))
829 goto out_mknod_unlock;
Christoph Hellwigf81a0bf2005-05-19 12:26:43 -0700830
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 /*
832 * All right, let's create it.
833 */
834 mode = S_IFSOCK |
Al Viroce3b0f82009-03-29 19:08:22 -0400835 (SOCK_INODE(sock)->i_mode & ~current_umask());
Dave Hansen463c3192008-02-15 14:37:57 -0800836 err = mnt_want_write(nd.path.mnt);
837 if (err)
838 goto out_mknod_dput;
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900839 err = security_path_mknod(&nd.path, dentry, mode, 0);
840 if (err)
841 goto out_mknod_drop_write;
Jan Blunck4ac91372008-02-14 19:34:32 -0800842 err = vfs_mknod(nd.path.dentry->d_inode, dentry, mode, 0);
Kentaro Takedabe6d3e52008-12-17 13:24:15 +0900843out_mknod_drop_write:
Dave Hansen463c3192008-02-15 14:37:57 -0800844 mnt_drop_write(nd.path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 if (err)
846 goto out_mknod_dput;
Jan Blunck4ac91372008-02-14 19:34:32 -0800847 mutex_unlock(&nd.path.dentry->d_inode->i_mutex);
848 dput(nd.path.dentry);
849 nd.path.dentry = dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850
851 addr->hash = UNIX_HASH_SIZE;
852 }
853
David S. Millerfbe9cc42005-12-13 23:26:29 -0800854 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855
856 if (!sunaddr->sun_path[0]) {
857 err = -EADDRINUSE;
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800858 if (__unix_find_socket_byname(net, sunaddr, addr_len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 sk->sk_type, hash)) {
860 unix_release_addr(addr);
861 goto out_unlock;
862 }
863
864 list = &unix_socket_table[addr->hash];
865 } else {
866 list = &unix_socket_table[dentry->d_inode->i_ino & (UNIX_HASH_SIZE-1)];
Jan Blunck4ac91372008-02-14 19:34:32 -0800867 u->dentry = nd.path.dentry;
868 u->mnt = nd.path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 }
870
871 err = 0;
872 __unix_remove_socket(sk);
873 u->addr = addr;
874 __unix_insert_socket(list, sk);
875
876out_unlock:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800877 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878out_up:
Ingo Molnar57b47a52006-03-20 22:35:41 -0800879 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880out:
881 return err;
882
883out_mknod_dput:
884 dput(dentry);
885out_mknod_unlock:
Jan Blunck4ac91372008-02-14 19:34:32 -0800886 mutex_unlock(&nd.path.dentry->d_inode->i_mutex);
Jan Blunck1d957f92008-02-14 19:34:35 -0800887 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888out_mknod_parent:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700889 if (err == -EEXIST)
890 err = -EADDRINUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 unix_release_addr(addr);
892 goto out_up;
893}
894
David S. Miller278a3de2007-05-31 15:19:20 -0700895static void unix_state_double_lock(struct sock *sk1, struct sock *sk2)
896{
897 if (unlikely(sk1 == sk2) || !sk2) {
898 unix_state_lock(sk1);
899 return;
900 }
901 if (sk1 < sk2) {
902 unix_state_lock(sk1);
903 unix_state_lock_nested(sk2);
904 } else {
905 unix_state_lock(sk2);
906 unix_state_lock_nested(sk1);
907 }
908}
909
910static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2)
911{
912 if (unlikely(sk1 == sk2) || !sk2) {
913 unix_state_unlock(sk1);
914 return;
915 }
916 unix_state_unlock(sk1);
917 unix_state_unlock(sk2);
918}
919
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr,
921 int alen, int flags)
922{
923 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900924 struct net *net = sock_net(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700925 struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 struct sock *other;
927 unsigned hash;
928 int err;
929
930 if (addr->sa_family != AF_UNSPEC) {
931 err = unix_mkname(sunaddr, alen, &hash);
932 if (err < 0)
933 goto out;
934 alen = err;
935
936 if (test_bit(SOCK_PASSCRED, &sock->flags) &&
937 !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0)
938 goto out;
939
David S. Miller278a3de2007-05-31 15:19:20 -0700940restart:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700941 other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 if (!other)
943 goto out;
944
David S. Miller278a3de2007-05-31 15:19:20 -0700945 unix_state_double_lock(sk, other);
946
947 /* Apparently VFS overslept socket death. Retry. */
948 if (sock_flag(other, SOCK_DEAD)) {
949 unix_state_double_unlock(sk, other);
950 sock_put(other);
951 goto restart;
952 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953
954 err = -EPERM;
955 if (!unix_may_send(sk, other))
956 goto out_unlock;
957
958 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
959 if (err)
960 goto out_unlock;
961
962 } else {
963 /*
964 * 1003.1g breaking connected state with AF_UNSPEC
965 */
966 other = NULL;
David S. Miller278a3de2007-05-31 15:19:20 -0700967 unix_state_double_lock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 }
969
970 /*
971 * If it was connected, reconnect.
972 */
973 if (unix_peer(sk)) {
974 struct sock *old_peer = unix_peer(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700975 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -0700976 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977
978 if (other != old_peer)
979 unix_dgram_disconnected(sk, old_peer);
980 sock_put(old_peer);
981 } else {
Jianjun Konge27dfce2008-11-01 21:38:31 -0700982 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -0700983 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 }
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900985 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986
987out_unlock:
David S. Miller278a3de2007-05-31 15:19:20 -0700988 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 sock_put(other);
990out:
991 return err;
992}
993
994static long unix_wait_for_peer(struct sock *other, long timeo)
995{
996 struct unix_sock *u = unix_sk(other);
997 int sched;
998 DEFINE_WAIT(wait);
999
1000 prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE);
1001
1002 sched = !sock_flag(other, SOCK_DEAD) &&
1003 !(other->sk_shutdown & RCV_SHUTDOWN) &&
Rainer Weikusat3c734192008-06-17 22:28:05 -07001004 unix_recvq_full(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005
David S. Miller1c92b4e2007-05-31 13:24:26 -07001006 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007
1008 if (sched)
1009 timeo = schedule_timeout(timeo);
1010
1011 finish_wait(&u->peer_wait, &wait);
1012 return timeo;
1013}
1014
1015static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr,
1016 int addr_len, int flags)
1017{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001018 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001020 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 struct unix_sock *u = unix_sk(sk), *newu, *otheru;
1022 struct sock *newsk = NULL;
1023 struct sock *other = NULL;
1024 struct sk_buff *skb = NULL;
1025 unsigned hash;
1026 int st;
1027 int err;
1028 long timeo;
1029
1030 err = unix_mkname(sunaddr, addr_len, &hash);
1031 if (err < 0)
1032 goto out;
1033 addr_len = err;
1034
1035 if (test_bit(SOCK_PASSCRED, &sock->flags)
1036 && !u->addr && (err = unix_autobind(sock)) != 0)
1037 goto out;
1038
1039 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1040
1041 /* First of all allocate resources.
1042 If we will make it after state is locked,
1043 we will have to recheck all again in any case.
1044 */
1045
1046 err = -ENOMEM;
1047
1048 /* create new sock for complete connection */
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001049 newsk = unix_create1(sock_net(sk), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 if (newsk == NULL)
1051 goto out;
1052
1053 /* Allocate skb for sending to listening sock */
1054 skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL);
1055 if (skb == NULL)
1056 goto out;
1057
1058restart:
1059 /* Find listening sock. */
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001060 other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 if (!other)
1062 goto out;
1063
1064 /* Latch state of peer */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001065 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066
1067 /* Apparently VFS overslept socket death. Retry. */
1068 if (sock_flag(other, SOCK_DEAD)) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001069 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 sock_put(other);
1071 goto restart;
1072 }
1073
1074 err = -ECONNREFUSED;
1075 if (other->sk_state != TCP_LISTEN)
1076 goto out_unlock;
1077
Rainer Weikusat3c734192008-06-17 22:28:05 -07001078 if (unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 err = -EAGAIN;
1080 if (!timeo)
1081 goto out_unlock;
1082
1083 timeo = unix_wait_for_peer(other, timeo);
1084
1085 err = sock_intr_errno(timeo);
1086 if (signal_pending(current))
1087 goto out;
1088 sock_put(other);
1089 goto restart;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001090 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091
1092 /* Latch our state.
1093
1094 It is tricky place. We need to grab write lock and cannot
1095 drop lock on peer. It is dangerous because deadlock is
1096 possible. Connect to self case and simultaneous
1097 attempt to connect are eliminated by checking socket
1098 state. other is TCP_LISTEN, if sk is TCP_LISTEN we
1099 check this before attempt to grab lock.
1100
1101 Well, and we have to recheck the state after socket locked.
1102 */
1103 st = sk->sk_state;
1104
1105 switch (st) {
1106 case TCP_CLOSE:
1107 /* This is ok... continue with connect */
1108 break;
1109 case TCP_ESTABLISHED:
1110 /* Socket is already connected */
1111 err = -EISCONN;
1112 goto out_unlock;
1113 default:
1114 err = -EINVAL;
1115 goto out_unlock;
1116 }
1117
David S. Miller1c92b4e2007-05-31 13:24:26 -07001118 unix_state_lock_nested(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119
1120 if (sk->sk_state != st) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001121 unix_state_unlock(sk);
1122 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 sock_put(other);
1124 goto restart;
1125 }
1126
1127 err = security_unix_stream_connect(sock, other->sk_socket, newsk);
1128 if (err) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001129 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 goto out_unlock;
1131 }
1132
1133 /* The way is open! Fastly set all the necessary fields... */
1134
1135 sock_hold(sk);
1136 unix_peer(newsk) = sk;
1137 newsk->sk_state = TCP_ESTABLISHED;
1138 newsk->sk_type = sk->sk_type;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001139 newsk->sk_peercred.pid = task_tgid_vnr(current);
David Howells19d65622008-11-14 10:39:10 +11001140 current_euid_egid(&newsk->sk_peercred.uid, &newsk->sk_peercred.gid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 newu = unix_sk(newsk);
1142 newsk->sk_sleep = &newu->peer_wait;
1143 otheru = unix_sk(other);
1144
1145 /* copy address information from listening to new sock*/
1146 if (otheru->addr) {
1147 atomic_inc(&otheru->addr->refcnt);
1148 newu->addr = otheru->addr;
1149 }
1150 if (otheru->dentry) {
1151 newu->dentry = dget(otheru->dentry);
1152 newu->mnt = mntget(otheru->mnt);
1153 }
1154
1155 /* Set credentials */
1156 sk->sk_peercred = other->sk_peercred;
1157
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 sock->state = SS_CONNECTED;
1159 sk->sk_state = TCP_ESTABLISHED;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001160 sock_hold(newsk);
1161
1162 smp_mb__after_atomic_inc(); /* sock_hold() does an atomic_inc() */
1163 unix_peer(sk) = newsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164
David S. Miller1c92b4e2007-05-31 13:24:26 -07001165 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166
1167 /* take ten and and send info to listening sock */
1168 spin_lock(&other->sk_receive_queue.lock);
1169 __skb_queue_tail(&other->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 spin_unlock(&other->sk_receive_queue.lock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001171 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 other->sk_data_ready(other, 0);
1173 sock_put(other);
1174 return 0;
1175
1176out_unlock:
1177 if (other)
David S. Miller1c92b4e2007-05-31 13:24:26 -07001178 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179
1180out:
Wei Yongjun40d44442009-02-25 00:32:45 +00001181 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 if (newsk)
1183 unix_release_sock(newsk, 0);
1184 if (other)
1185 sock_put(other);
1186 return err;
1187}
1188
1189static int unix_socketpair(struct socket *socka, struct socket *sockb)
1190{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001191 struct sock *ska = socka->sk, *skb = sockb->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192
1193 /* Join our sockets back to back */
1194 sock_hold(ska);
1195 sock_hold(skb);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001196 unix_peer(ska) = skb;
1197 unix_peer(skb) = ska;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001198 ska->sk_peercred.pid = skb->sk_peercred.pid = task_tgid_vnr(current);
David Howells19d65622008-11-14 10:39:10 +11001199 current_euid_egid(&skb->sk_peercred.uid, &skb->sk_peercred.gid);
1200 ska->sk_peercred.uid = skb->sk_peercred.uid;
1201 ska->sk_peercred.gid = skb->sk_peercred.gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202
1203 if (ska->sk_type != SOCK_DGRAM) {
1204 ska->sk_state = TCP_ESTABLISHED;
1205 skb->sk_state = TCP_ESTABLISHED;
1206 socka->state = SS_CONNECTED;
1207 sockb->state = SS_CONNECTED;
1208 }
1209 return 0;
1210}
1211
1212static int unix_accept(struct socket *sock, struct socket *newsock, int flags)
1213{
1214 struct sock *sk = sock->sk;
1215 struct sock *tsk;
1216 struct sk_buff *skb;
1217 int err;
1218
1219 err = -EOPNOTSUPP;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001220 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 goto out;
1222
1223 err = -EINVAL;
1224 if (sk->sk_state != TCP_LISTEN)
1225 goto out;
1226
1227 /* If socket state is TCP_LISTEN it cannot change (for now...),
1228 * so that no locks are necessary.
1229 */
1230
1231 skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err);
1232 if (!skb) {
1233 /* This means receive shutdown. */
1234 if (err == 0)
1235 err = -EINVAL;
1236 goto out;
1237 }
1238
1239 tsk = skb->sk;
1240 skb_free_datagram(sk, skb);
1241 wake_up_interruptible(&unix_sk(sk)->peer_wait);
1242
1243 /* attach accepted sock to socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001244 unix_state_lock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 newsock->state = SS_CONNECTED;
1246 sock_graft(tsk, newsock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001247 unix_state_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 return 0;
1249
1250out:
1251 return err;
1252}
1253
1254
1255static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer)
1256{
1257 struct sock *sk = sock->sk;
1258 struct unix_sock *u;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001259 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 int err = 0;
1261
1262 if (peer) {
1263 sk = unix_peer_get(sk);
1264
1265 err = -ENOTCONN;
1266 if (!sk)
1267 goto out;
1268 err = 0;
1269 } else {
1270 sock_hold(sk);
1271 }
1272
1273 u = unix_sk(sk);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001274 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 if (!u->addr) {
1276 sunaddr->sun_family = AF_UNIX;
1277 sunaddr->sun_path[0] = 0;
1278 *uaddr_len = sizeof(short);
1279 } else {
1280 struct unix_address *addr = u->addr;
1281
1282 *uaddr_len = addr->len;
1283 memcpy(sunaddr, addr->name, *uaddr_len);
1284 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07001285 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 sock_put(sk);
1287out:
1288 return err;
1289}
1290
1291static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb)
1292{
1293 int i;
1294
1295 scm->fp = UNIXCB(skb).fp;
1296 skb->destructor = sock_wfree;
1297 UNIXCB(skb).fp = NULL;
1298
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001299 for (i = scm->fp->count-1; i >= 0; i--)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 unix_notinflight(scm->fp->fp[i]);
1301}
1302
1303static void unix_destruct_fds(struct sk_buff *skb)
1304{
1305 struct scm_cookie scm;
1306 memset(&scm, 0, sizeof(scm));
1307 unix_detach_fds(&scm, skb);
1308
1309 /* Alas, it calls VFS */
1310 /* So fscking what? fput() had been SMP-safe since the last Summer */
1311 scm_destroy(&scm);
1312 sock_wfree(skb);
1313}
1314
Miklos Szeredi62093442008-11-09 15:23:57 +01001315static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316{
1317 int i;
Miklos Szeredi62093442008-11-09 15:23:57 +01001318
1319 /*
1320 * Need to duplicate file references for the sake of garbage
1321 * collection. Otherwise a socket in the fps might become a
1322 * candidate for GC while the skb is not yet queued.
1323 */
1324 UNIXCB(skb).fp = scm_fp_dup(scm->fp);
1325 if (!UNIXCB(skb).fp)
1326 return -ENOMEM;
1327
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001328 for (i = scm->fp->count-1; i >= 0; i--)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 unix_inflight(scm->fp->fp[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 skb->destructor = unix_destruct_fds;
Miklos Szeredi62093442008-11-09 15:23:57 +01001331 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332}
1333
1334/*
1335 * Send AF_UNIX data.
1336 */
1337
1338static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock,
1339 struct msghdr *msg, size_t len)
1340{
1341 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1342 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001343 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001345 struct sockaddr_un *sunaddr = msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 struct sock *other = NULL;
1347 int namelen = 0; /* fake GCC */
1348 int err;
1349 unsigned hash;
1350 struct sk_buff *skb;
1351 long timeo;
1352 struct scm_cookie tmp_scm;
1353
1354 if (NULL == siocb->scm)
1355 siocb->scm = &tmp_scm;
dann frazier5f23b732008-11-26 15:32:27 -08001356 wait_for_unix_gc();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357 err = scm_send(sock, msg, siocb->scm);
1358 if (err < 0)
1359 return err;
1360
1361 err = -EOPNOTSUPP;
1362 if (msg->msg_flags&MSG_OOB)
1363 goto out;
1364
1365 if (msg->msg_namelen) {
1366 err = unix_mkname(sunaddr, msg->msg_namelen, &hash);
1367 if (err < 0)
1368 goto out;
1369 namelen = err;
1370 } else {
1371 sunaddr = NULL;
1372 err = -ENOTCONN;
1373 other = unix_peer_get(sk);
1374 if (!other)
1375 goto out;
1376 }
1377
1378 if (test_bit(SOCK_PASSCRED, &sock->flags)
1379 && !u->addr && (err = unix_autobind(sock)) != 0)
1380 goto out;
1381
1382 err = -EMSGSIZE;
1383 if (len > sk->sk_sndbuf - 32)
1384 goto out;
1385
1386 skb = sock_alloc_send_skb(sk, len, msg->msg_flags&MSG_DONTWAIT, &err);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001387 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 goto out;
1389
1390 memcpy(UNIXCREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
Miklos Szeredi62093442008-11-09 15:23:57 +01001391 if (siocb->scm->fp) {
1392 err = unix_attach_fds(siocb->scm, skb);
1393 if (err)
1394 goto out_free;
1395 }
Catherine Zhangdc49c1f2006-08-02 14:12:06 -07001396 unix_get_secdata(siocb->scm, skb);
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001397
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001398 skb_reset_transport_header(skb);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001399 err = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 if (err)
1401 goto out_free;
1402
1403 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1404
1405restart:
1406 if (!other) {
1407 err = -ECONNRESET;
1408 if (sunaddr == NULL)
1409 goto out_free;
1410
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001411 other = unix_find_other(net, sunaddr, namelen, sk->sk_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 hash, &err);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001413 if (other == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 goto out_free;
1415 }
1416
David S. Miller1c92b4e2007-05-31 13:24:26 -07001417 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 err = -EPERM;
1419 if (!unix_may_send(sk, other))
1420 goto out_unlock;
1421
1422 if (sock_flag(other, SOCK_DEAD)) {
1423 /*
1424 * Check with 1003.1g - what should
1425 * datagram error
1426 */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001427 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 sock_put(other);
1429
1430 err = 0;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001431 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 if (unix_peer(sk) == other) {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001433 unix_peer(sk) = NULL;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001434 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435
1436 unix_dgram_disconnected(sk, other);
1437 sock_put(other);
1438 err = -ECONNREFUSED;
1439 } else {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001440 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 }
1442
1443 other = NULL;
1444 if (err)
1445 goto out_free;
1446 goto restart;
1447 }
1448
1449 err = -EPIPE;
1450 if (other->sk_shutdown & RCV_SHUTDOWN)
1451 goto out_unlock;
1452
1453 if (sk->sk_type != SOCK_SEQPACKET) {
1454 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1455 if (err)
1456 goto out_unlock;
1457 }
1458
Rainer Weikusat3c734192008-06-17 22:28:05 -07001459 if (unix_peer(other) != sk && unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 if (!timeo) {
1461 err = -EAGAIN;
1462 goto out_unlock;
1463 }
1464
1465 timeo = unix_wait_for_peer(other, timeo);
1466
1467 err = sock_intr_errno(timeo);
1468 if (signal_pending(current))
1469 goto out_free;
1470
1471 goto restart;
1472 }
1473
1474 skb_queue_tail(&other->sk_receive_queue, skb);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001475 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 other->sk_data_ready(other, len);
1477 sock_put(other);
1478 scm_destroy(siocb->scm);
1479 return len;
1480
1481out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001482 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483out_free:
1484 kfree_skb(skb);
1485out:
1486 if (other)
1487 sock_put(other);
1488 scm_destroy(siocb->scm);
1489 return err;
1490}
1491
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001492
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock,
1494 struct msghdr *msg, size_t len)
1495{
1496 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1497 struct sock *sk = sock->sk;
1498 struct sock *other = NULL;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001499 struct sockaddr_un *sunaddr = msg->msg_name;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001500 int err, size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 struct sk_buff *skb;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001502 int sent = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 struct scm_cookie tmp_scm;
1504
1505 if (NULL == siocb->scm)
1506 siocb->scm = &tmp_scm;
dann frazier5f23b732008-11-26 15:32:27 -08001507 wait_for_unix_gc();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508 err = scm_send(sock, msg, siocb->scm);
1509 if (err < 0)
1510 return err;
1511
1512 err = -EOPNOTSUPP;
1513 if (msg->msg_flags&MSG_OOB)
1514 goto out_err;
1515
1516 if (msg->msg_namelen) {
1517 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1518 goto out_err;
1519 } else {
1520 sunaddr = NULL;
1521 err = -ENOTCONN;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001522 other = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 if (!other)
1524 goto out_err;
1525 }
1526
1527 if (sk->sk_shutdown & SEND_SHUTDOWN)
1528 goto pipe_err;
1529
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001530 while (sent < len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 /*
Benjamin LaHaisee9df7d72006-03-20 21:29:05 -08001532 * Optimisation for the fact that under 0.01% of X
1533 * messages typically need breaking up.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 */
1535
Benjamin LaHaisee9df7d72006-03-20 21:29:05 -08001536 size = len-sent;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537
1538 /* Keep two messages in the pipe so it schedules better */
Benjamin LaHaisee9df7d72006-03-20 21:29:05 -08001539 if (size > ((sk->sk_sndbuf >> 1) - 64))
1540 size = (sk->sk_sndbuf >> 1) - 64;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541
1542 if (size > SKB_MAX_ALLOC)
1543 size = SKB_MAX_ALLOC;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001544
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 /*
1546 * Grab a buffer
1547 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001548
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001549 skb = sock_alloc_send_skb(sk, size, msg->msg_flags&MSG_DONTWAIT,
1550 &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551
Jianjun Konge27dfce2008-11-01 21:38:31 -07001552 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553 goto out_err;
1554
1555 /*
1556 * If you pass two values to the sock_alloc_send_skb
1557 * it tries to grab the large buffer with GFP_NOFS
1558 * (which can fail easily), and if it fails grab the
1559 * fallback size buffer which is under a page and will
1560 * succeed. [Alan]
1561 */
1562 size = min_t(int, size, skb_tailroom(skb));
1563
1564 memcpy(UNIXCREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
Miklos Szeredi62093442008-11-09 15:23:57 +01001565 if (siocb->scm->fp) {
1566 err = unix_attach_fds(siocb->scm, skb);
1567 if (err) {
1568 kfree_skb(skb);
1569 goto out_err;
1570 }
1571 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001573 err = memcpy_fromiovec(skb_put(skb, size), msg->msg_iov, size);
1574 if (err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575 kfree_skb(skb);
1576 goto out_err;
1577 }
1578
David S. Miller1c92b4e2007-05-31 13:24:26 -07001579 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580
1581 if (sock_flag(other, SOCK_DEAD) ||
1582 (other->sk_shutdown & RCV_SHUTDOWN))
1583 goto pipe_err_free;
1584
1585 skb_queue_tail(&other->sk_receive_queue, skb);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001586 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 other->sk_data_ready(other, size);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001588 sent += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590
1591 scm_destroy(siocb->scm);
1592 siocb->scm = NULL;
1593
1594 return sent;
1595
1596pipe_err_free:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001597 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 kfree_skb(skb);
1599pipe_err:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001600 if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL))
1601 send_sig(SIGPIPE, current, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 err = -EPIPE;
1603out_err:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 scm_destroy(siocb->scm);
1605 siocb->scm = NULL;
1606 return sent ? : err;
1607}
1608
1609static int unix_seqpacket_sendmsg(struct kiocb *kiocb, struct socket *sock,
1610 struct msghdr *msg, size_t len)
1611{
1612 int err;
1613 struct sock *sk = sock->sk;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001614
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 err = sock_error(sk);
1616 if (err)
1617 return err;
1618
1619 if (sk->sk_state != TCP_ESTABLISHED)
1620 return -ENOTCONN;
1621
1622 if (msg->msg_namelen)
1623 msg->msg_namelen = 0;
1624
1625 return unix_dgram_sendmsg(kiocb, sock, msg, len);
1626}
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001627
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628static void unix_copy_addr(struct msghdr *msg, struct sock *sk)
1629{
1630 struct unix_sock *u = unix_sk(sk);
1631
1632 msg->msg_namelen = 0;
1633 if (u->addr) {
1634 msg->msg_namelen = u->addr->len;
1635 memcpy(msg->msg_name, u->addr->name, u->addr->len);
1636 }
1637}
1638
1639static int unix_dgram_recvmsg(struct kiocb *iocb, struct socket *sock,
1640 struct msghdr *msg, size_t size,
1641 int flags)
1642{
1643 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
1644 struct scm_cookie tmp_scm;
1645 struct sock *sk = sock->sk;
1646 struct unix_sock *u = unix_sk(sk);
1647 int noblock = flags & MSG_DONTWAIT;
1648 struct sk_buff *skb;
1649 int err;
1650
1651 err = -EOPNOTSUPP;
1652 if (flags&MSG_OOB)
1653 goto out;
1654
1655 msg->msg_namelen = 0;
1656
Ingo Molnar57b47a52006-03-20 22:35:41 -08001657 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658
1659 skb = skb_recv_datagram(sk, flags, noblock, &err);
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001660 if (!skb) {
1661 unix_state_lock(sk);
1662 /* Signal EOF on disconnected non-blocking SEQPACKET socket. */
1663 if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN &&
1664 (sk->sk_shutdown & RCV_SHUTDOWN))
1665 err = 0;
1666 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 goto out_unlock;
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001668 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669
Ingo Molnar71e20f12007-10-15 17:00:19 +02001670 wake_up_interruptible_sync(&u->peer_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671
1672 if (msg->msg_name)
1673 unix_copy_addr(msg, skb->sk);
1674
1675 if (size > skb->len)
1676 size = skb->len;
1677 else if (size < skb->len)
1678 msg->msg_flags |= MSG_TRUNC;
1679
1680 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, size);
1681 if (err)
1682 goto out_free;
1683
1684 if (!siocb->scm) {
1685 siocb->scm = &tmp_scm;
1686 memset(&tmp_scm, 0, sizeof(tmp_scm));
1687 }
1688 siocb->scm->creds = *UNIXCREDS(skb);
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001689 unix_set_secdata(siocb->scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001691 if (!(flags & MSG_PEEK)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 if (UNIXCB(skb).fp)
1693 unix_detach_fds(siocb->scm, skb);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001694 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 /* It is questionable: on PEEK we could:
1696 - do not return fds - good, but too simple 8)
1697 - return fds, and do not return them on read (old strategy,
1698 apparently wrong)
1699 - clone fds (I chose it for now, it is the most universal
1700 solution)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001701
1702 POSIX 1003.1g does not actually define this clearly
1703 at all. POSIX 1003.1g doesn't define a lot of things
1704 clearly however!
1705
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 */
1707 if (UNIXCB(skb).fp)
1708 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
1709 }
1710 err = size;
1711
1712 scm_recv(sock, msg, siocb->scm, flags);
1713
1714out_free:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001715 skb_free_datagram(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716out_unlock:
Ingo Molnar57b47a52006-03-20 22:35:41 -08001717 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718out:
1719 return err;
1720}
1721
1722/*
1723 * Sleep until data has arrive. But check for races..
1724 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001725
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001726static long unix_stream_data_wait(struct sock *sk, long timeo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727{
1728 DEFINE_WAIT(wait);
1729
David S. Miller1c92b4e2007-05-31 13:24:26 -07001730 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
1732 for (;;) {
1733 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1734
David S. Millerb03efcf2005-07-08 14:57:23 -07001735 if (!skb_queue_empty(&sk->sk_receive_queue) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 sk->sk_err ||
1737 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1738 signal_pending(current) ||
1739 !timeo)
1740 break;
1741
1742 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001743 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 timeo = schedule_timeout(timeo);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001745 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1747 }
1748
1749 finish_wait(sk->sk_sleep, &wait);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001750 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 return timeo;
1752}
1753
1754
1755
1756static int unix_stream_recvmsg(struct kiocb *iocb, struct socket *sock,
1757 struct msghdr *msg, size_t size,
1758 int flags)
1759{
1760 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
1761 struct scm_cookie tmp_scm;
1762 struct sock *sk = sock->sk;
1763 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001764 struct sockaddr_un *sunaddr = msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 int copied = 0;
1766 int check_creds = 0;
1767 int target;
1768 int err = 0;
1769 long timeo;
1770
1771 err = -EINVAL;
1772 if (sk->sk_state != TCP_ESTABLISHED)
1773 goto out;
1774
1775 err = -EOPNOTSUPP;
1776 if (flags&MSG_OOB)
1777 goto out;
1778
1779 target = sock_rcvlowat(sk, flags&MSG_WAITALL, size);
1780 timeo = sock_rcvtimeo(sk, flags&MSG_DONTWAIT);
1781
1782 msg->msg_namelen = 0;
1783
1784 /* Lock the socket to prevent queue disordering
1785 * while sleeps in memcpy_tomsg
1786 */
1787
1788 if (!siocb->scm) {
1789 siocb->scm = &tmp_scm;
1790 memset(&tmp_scm, 0, sizeof(tmp_scm));
1791 }
1792
Ingo Molnar57b47a52006-03-20 22:35:41 -08001793 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001795 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 int chunk;
1797 struct sk_buff *skb;
1798
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001799 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 skb = skb_dequeue(&sk->sk_receive_queue);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001801 if (skb == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 if (copied >= target)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001803 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804
1805 /*
1806 * POSIX 1003.1g mandates this order.
1807 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001808
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001809 err = sock_error(sk);
1810 if (err)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001811 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 if (sk->sk_shutdown & RCV_SHUTDOWN)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001813 goto unlock;
1814
1815 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 err = -EAGAIN;
1817 if (!timeo)
1818 break;
Ingo Molnar57b47a52006-03-20 22:35:41 -08001819 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820
1821 timeo = unix_stream_data_wait(sk, timeo);
1822
1823 if (signal_pending(current)) {
1824 err = sock_intr_errno(timeo);
1825 goto out;
1826 }
Ingo Molnar57b47a52006-03-20 22:35:41 -08001827 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828 continue;
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001829 unlock:
1830 unix_state_unlock(sk);
1831 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832 }
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001833 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834
1835 if (check_creds) {
1836 /* Never glue messages from different writers */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001837 if (memcmp(UNIXCREDS(skb), &siocb->scm->creds,
1838 sizeof(siocb->scm->creds)) != 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 skb_queue_head(&sk->sk_receive_queue, skb);
1840 break;
1841 }
1842 } else {
1843 /* Copy credentials */
1844 siocb->scm->creds = *UNIXCREDS(skb);
1845 check_creds = 1;
1846 }
1847
1848 /* Copy address just once */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001849 if (sunaddr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 unix_copy_addr(msg, skb->sk);
1851 sunaddr = NULL;
1852 }
1853
1854 chunk = min_t(unsigned int, skb->len, size);
1855 if (memcpy_toiovec(msg->msg_iov, skb->data, chunk)) {
1856 skb_queue_head(&sk->sk_receive_queue, skb);
1857 if (copied == 0)
1858 copied = -EFAULT;
1859 break;
1860 }
1861 copied += chunk;
1862 size -= chunk;
1863
1864 /* Mark read part of skb as used */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001865 if (!(flags & MSG_PEEK)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866 skb_pull(skb, chunk);
1867
1868 if (UNIXCB(skb).fp)
1869 unix_detach_fds(siocb->scm, skb);
1870
1871 /* put the skb back if we didn't use it up.. */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001872 if (skb->len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 skb_queue_head(&sk->sk_receive_queue, skb);
1874 break;
1875 }
1876
1877 kfree_skb(skb);
1878
1879 if (siocb->scm->fp)
1880 break;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001881 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 /* It is questionable, see note in unix_dgram_recvmsg.
1883 */
1884 if (UNIXCB(skb).fp)
1885 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
1886
1887 /* put message back and return */
1888 skb_queue_head(&sk->sk_receive_queue, skb);
1889 break;
1890 }
1891 } while (size);
1892
Ingo Molnar57b47a52006-03-20 22:35:41 -08001893 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894 scm_recv(sock, msg, siocb->scm, flags);
1895out:
1896 return copied ? : err;
1897}
1898
1899static int unix_shutdown(struct socket *sock, int mode)
1900{
1901 struct sock *sk = sock->sk;
1902 struct sock *other;
1903
1904 mode = (mode+1)&(RCV_SHUTDOWN|SEND_SHUTDOWN);
1905
1906 if (mode) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001907 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908 sk->sk_shutdown |= mode;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001909 other = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 if (other)
1911 sock_hold(other);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001912 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 sk->sk_state_change(sk);
1914
1915 if (other &&
1916 (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) {
1917
1918 int peer_mode = 0;
1919
1920 if (mode&RCV_SHUTDOWN)
1921 peer_mode |= SEND_SHUTDOWN;
1922 if (mode&SEND_SHUTDOWN)
1923 peer_mode |= RCV_SHUTDOWN;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001924 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925 other->sk_shutdown |= peer_mode;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001926 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 other->sk_state_change(other);
1928 read_lock(&other->sk_callback_lock);
1929 if (peer_mode == SHUTDOWN_MASK)
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001930 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 else if (peer_mode & RCV_SHUTDOWN)
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001932 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 read_unlock(&other->sk_callback_lock);
1934 }
1935 if (other)
1936 sock_put(other);
1937 }
1938 return 0;
1939}
1940
1941static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1942{
1943 struct sock *sk = sock->sk;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001944 long amount = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 int err;
1946
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001947 switch (cmd) {
1948 case SIOCOUTQ:
1949 amount = atomic_read(&sk->sk_wmem_alloc);
1950 err = put_user(amount, (int __user *)arg);
1951 break;
1952 case SIOCINQ:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953 {
1954 struct sk_buff *skb;
1955
1956 if (sk->sk_state == TCP_LISTEN) {
1957 err = -EINVAL;
1958 break;
1959 }
1960
1961 spin_lock(&sk->sk_receive_queue.lock);
1962 if (sk->sk_type == SOCK_STREAM ||
1963 sk->sk_type == SOCK_SEQPACKET) {
1964 skb_queue_walk(&sk->sk_receive_queue, skb)
1965 amount += skb->len;
1966 } else {
1967 skb = skb_peek(&sk->sk_receive_queue);
1968 if (skb)
Jianjun Konge27dfce2008-11-01 21:38:31 -07001969 amount = skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970 }
1971 spin_unlock(&sk->sk_receive_queue.lock);
1972 err = put_user(amount, (int __user *)arg);
1973 break;
1974 }
1975
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001976 default:
1977 err = -ENOIOCTLCMD;
1978 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 }
1980 return err;
1981}
1982
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001983static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984{
1985 struct sock *sk = sock->sk;
1986 unsigned int mask;
1987
1988 poll_wait(file, sk->sk_sleep, wait);
1989 mask = 0;
1990
1991 /* exceptional events? */
1992 if (sk->sk_err)
1993 mask |= POLLERR;
1994 if (sk->sk_shutdown == SHUTDOWN_MASK)
1995 mask |= POLLHUP;
Davide Libenzif348d702006-03-25 03:07:39 -08001996 if (sk->sk_shutdown & RCV_SHUTDOWN)
1997 mask |= POLLRDHUP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998
1999 /* readable? */
2000 if (!skb_queue_empty(&sk->sk_receive_queue) ||
2001 (sk->sk_shutdown & RCV_SHUTDOWN))
2002 mask |= POLLIN | POLLRDNORM;
2003
2004 /* Connection-based need to check for termination and startup */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002005 if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) &&
2006 sk->sk_state == TCP_CLOSE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 mask |= POLLHUP;
2008
2009 /*
2010 * we set writable also when the other side has shut down the
2011 * connection. This prevents stuck sockets.
2012 */
2013 if (unix_writable(sk))
2014 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2015
2016 return mask;
2017}
2018
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002019static unsigned int unix_dgram_poll(struct file *file, struct socket *sock,
2020 poll_table *wait)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002021{
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002022 struct sock *sk = sock->sk, *other;
2023 unsigned int mask, writable;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002024
2025 poll_wait(file, sk->sk_sleep, wait);
Rainer Weikusat3c734192008-06-17 22:28:05 -07002026 mask = 0;
2027
2028 /* exceptional events? */
2029 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
2030 mask |= POLLERR;
2031 if (sk->sk_shutdown & RCV_SHUTDOWN)
2032 mask |= POLLRDHUP;
2033 if (sk->sk_shutdown == SHUTDOWN_MASK)
2034 mask |= POLLHUP;
2035
2036 /* readable? */
2037 if (!skb_queue_empty(&sk->sk_receive_queue) ||
2038 (sk->sk_shutdown & RCV_SHUTDOWN))
2039 mask |= POLLIN | POLLRDNORM;
2040
2041 /* Connection-based need to check for termination and startup */
2042 if (sk->sk_type == SOCK_SEQPACKET) {
2043 if (sk->sk_state == TCP_CLOSE)
2044 mask |= POLLHUP;
2045 /* connection hasn't started yet? */
2046 if (sk->sk_state == TCP_SYN_SENT)
2047 return mask;
2048 }
2049
2050 /* writable? */
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002051 writable = unix_writable(sk);
2052 if (writable) {
2053 other = unix_peer_get(sk);
2054 if (other) {
2055 if (unix_peer(other) != sk) {
2056 poll_wait(file, &unix_sk(other)->peer_wait,
2057 wait);
2058 if (unix_recvq_full(other))
2059 writable = 0;
2060 }
2061
2062 sock_put(other);
2063 }
2064 }
2065
2066 if (writable)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002067 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2068 else
2069 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
2070
Rainer Weikusat3c734192008-06-17 22:28:05 -07002071 return mask;
2072}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073
2074#ifdef CONFIG_PROC_FS
Pavel Emelyanova53eb3f2007-11-23 20:30:01 +08002075static struct sock *first_unix_socket(int *i)
2076{
2077 for (*i = 0; *i <= UNIX_HASH_SIZE; (*i)++) {
2078 if (!hlist_empty(&unix_socket_table[*i]))
2079 return __sk_head(&unix_socket_table[*i]);
2080 }
2081 return NULL;
2082}
2083
2084static struct sock *next_unix_socket(int *i, struct sock *s)
2085{
2086 struct sock *next = sk_next(s);
2087 /* More in this chain? */
2088 if (next)
2089 return next;
2090 /* Look for next non-empty chain. */
2091 for ((*i)++; *i <= UNIX_HASH_SIZE; (*i)++) {
2092 if (!hlist_empty(&unix_socket_table[*i]))
2093 return __sk_head(&unix_socket_table[*i]);
2094 }
2095 return NULL;
2096}
2097
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002098struct unix_iter_state {
Denis V. Luneve372c412007-11-19 22:31:54 -08002099 struct seq_net_private p;
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002100 int i;
2101};
Jianjun Konge27dfce2008-11-01 21:38:31 -07002102
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002103static struct sock *unix_seq_idx(struct seq_file *seq, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104{
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002105 struct unix_iter_state *iter = seq->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106 loff_t off = 0;
2107 struct sock *s;
2108
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002109 for (s = first_unix_socket(&iter->i); s; s = next_unix_socket(&iter->i, s)) {
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002110 if (sock_net(s) != seq_file_net(seq))
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002111 continue;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002112 if (off == pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113 return s;
2114 ++off;
2115 }
2116 return NULL;
2117}
2118
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119static void *unix_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002120 __acquires(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002122 spin_lock(&unix_table_lock);
Joe Perchesb9f31242008-04-12 19:04:38 -07002123 return *pos ? unix_seq_idx(seq, *pos - 1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124}
2125
2126static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2127{
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002128 struct unix_iter_state *iter = seq->private;
2129 struct sock *sk = v;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130 ++*pos;
2131
Joe Perchesb9f31242008-04-12 19:04:38 -07002132 if (v == SEQ_START_TOKEN)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002133 sk = first_unix_socket(&iter->i);
2134 else
2135 sk = next_unix_socket(&iter->i, sk);
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002136 while (sk && (sock_net(sk) != seq_file_net(seq)))
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002137 sk = next_unix_socket(&iter->i, sk);
2138 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139}
2140
2141static void unix_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002142 __releases(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002144 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145}
2146
2147static int unix_seq_show(struct seq_file *seq, void *v)
2148{
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002149
Joe Perchesb9f31242008-04-12 19:04:38 -07002150 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151 seq_puts(seq, "Num RefCount Protocol Flags Type St "
2152 "Inode Path\n");
2153 else {
2154 struct sock *s = v;
2155 struct unix_sock *u = unix_sk(s);
David S. Miller1c92b4e2007-05-31 13:24:26 -07002156 unix_state_lock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157
2158 seq_printf(seq, "%p: %08X %08X %08X %04X %02X %5lu",
2159 s,
2160 atomic_read(&s->sk_refcnt),
2161 0,
2162 s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0,
2163 s->sk_type,
2164 s->sk_socket ?
2165 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) :
2166 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING),
2167 sock_i_ino(s));
2168
2169 if (u->addr) {
2170 int i, len;
2171 seq_putc(seq, ' ');
2172
2173 i = 0;
2174 len = u->addr->len - sizeof(short);
2175 if (!UNIX_ABSTRACT(s))
2176 len--;
2177 else {
2178 seq_putc(seq, '@');
2179 i++;
2180 }
2181 for ( ; i < len; i++)
2182 seq_putc(seq, u->addr->name->sun_path[i]);
2183 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07002184 unix_state_unlock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185 seq_putc(seq, '\n');
2186 }
2187
2188 return 0;
2189}
2190
Philippe De Muyter56b3d972007-07-10 23:07:31 -07002191static const struct seq_operations unix_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192 .start = unix_seq_start,
2193 .next = unix_seq_next,
2194 .stop = unix_seq_stop,
2195 .show = unix_seq_show,
2196};
2197
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198static int unix_seq_open(struct inode *inode, struct file *file)
2199{
Denis V. Luneve372c412007-11-19 22:31:54 -08002200 return seq_open_net(inode, file, &unix_seq_ops,
2201 sizeof(struct unix_iter_state));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202}
2203
Arjan van de Venda7071d2007-02-12 00:55:36 -08002204static const struct file_operations unix_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205 .owner = THIS_MODULE,
2206 .open = unix_seq_open,
2207 .read = seq_read,
2208 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08002209 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210};
2211
2212#endif
2213
2214static struct net_proto_family unix_family_ops = {
2215 .family = PF_UNIX,
2216 .create = unix_create,
2217 .owner = THIS_MODULE,
2218};
2219
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002220
2221static int unix_net_init(struct net *net)
2222{
2223 int error = -ENOMEM;
2224
Denis V. Luneva0a53c82007-12-11 04:19:17 -08002225 net->unx.sysctl_max_dgram_qlen = 10;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002226 if (unix_sysctl_register(net))
2227 goto out;
Pavel Emelyanovd392e492007-12-01 23:44:15 +11002228
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002229#ifdef CONFIG_PROC_FS
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002230 if (!proc_net_fops_create(net, "unix", 0, &unix_seq_fops)) {
2231 unix_sysctl_unregister(net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002232 goto out;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002233 }
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002234#endif
2235 error = 0;
2236out:
Jianjun Kong48dcc332008-11-01 21:37:27 -07002237 return error;
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002238}
2239
2240static void unix_net_exit(struct net *net)
2241{
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002242 unix_sysctl_unregister(net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002243 proc_net_remove(net, "unix");
2244}
2245
2246static struct pernet_operations unix_net_ops = {
2247 .init = unix_net_init,
2248 .exit = unix_net_exit,
2249};
2250
Linus Torvalds1da177e2005-04-16 15:20:36 -07002251static int __init af_unix_init(void)
2252{
2253 int rc = -1;
2254 struct sk_buff *dummy_skb;
2255
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07002256 BUILD_BUG_ON(sizeof(struct unix_skb_parms) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257
2258 rc = proto_register(&unix_proto, 1);
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002259 if (rc != 0) {
2260 printk(KERN_CRIT "%s: Cannot create unix_sock SLAB cache!\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08002261 __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002262 goto out;
2263 }
2264
2265 sock_register(&unix_family_ops);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002266 register_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267out:
2268 return rc;
2269}
2270
2271static void __exit af_unix_exit(void)
2272{
2273 sock_unregister(PF_UNIX);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274 proto_unregister(&unix_proto);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002275 unregister_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276}
2277
David Woodhouse3d366962008-04-24 00:59:25 -07002278/* Earlier than device_initcall() so that other drivers invoking
2279 request_module() don't end up in a loop when modprobe tries
2280 to use a UNIX socket. But later than subsys_initcall() because
2281 we depend on stuff initialised there */
2282fs_initcall(af_unix_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283module_exit(af_unix_exit);
2284
2285MODULE_LICENSE("GPL");
2286MODULE_ALIAS_NETPROTO(PF_UNIX);