blob: 0b80634b2b72da47a1716c987532bd870672d413 [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{
167 return (unix_peer(osk) == NULL || unix_our_peer(sk, osk));
168}
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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200static int unix_mkname(struct sockaddr_un * sunaddr, int len, unsigned *hashp)
201{
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
219 *hashp = unix_hash_fold(csum_partial((char*)sunaddr, len, 0));
220 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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 if(dentry && dentry->d_inode == i)
299 {
300 sock_hold(s);
301 goto found;
302 }
303 }
304 s = NULL;
305found:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800306 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 return s;
308}
309
310static inline int unix_writable(struct sock *sk)
311{
312 return (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf;
313}
314
315static void unix_write_space(struct sock *sk)
316{
317 read_lock(&sk->sk_callback_lock);
318 if (unix_writable(sk)) {
319 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
Ingo Molnar71e20f12007-10-15 17:00:19 +0200320 wake_up_interruptible_sync(sk->sk_sleep);
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 }
323 read_unlock(&sk->sk_callback_lock);
324}
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)) {
357 printk("Attempt to release alive unix socket: %p\n", sk);
358 return;
359 }
360
361 if (u->addr)
362 unix_release_addr(u->addr);
363
364 atomic_dec(&unix_nr_socks);
365#ifdef UNIX_REFCNT_DEBUG
366 printk(KERN_DEBUG "UNIX %p is destroyed, %d are still alive.\n", sk, atomic_read(&unix_nr_socks));
367#endif
368}
369
370static int unix_release_sock (struct sock *sk, int embrion)
371{
372 struct unix_sock *u = unix_sk(sk);
373 struct dentry *dentry;
374 struct vfsmount *mnt;
375 struct sock *skpair;
376 struct sk_buff *skb;
377 int state;
378
379 unix_remove_socket(sk);
380
381 /* Clear state */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700382 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 sock_orphan(sk);
384 sk->sk_shutdown = SHUTDOWN_MASK;
385 dentry = u->dentry;
386 u->dentry = NULL;
387 mnt = u->mnt;
388 u->mnt = NULL;
389 state = sk->sk_state;
390 sk->sk_state = TCP_CLOSE;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700391 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
393 wake_up_interruptible_all(&u->peer_wait);
394
Jianjun Konge27dfce2008-11-01 21:38:31 -0700395 skpair = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
Jianjun Konge27dfce2008-11-01 21:38:31 -0700397 if (skpair != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) {
David S. Miller1c92b4e2007-05-31 13:24:26 -0700399 unix_state_lock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 /* No more writes */
401 skpair->sk_shutdown = SHUTDOWN_MASK;
402 if (!skb_queue_empty(&sk->sk_receive_queue) || embrion)
403 skpair->sk_err = ECONNRESET;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700404 unix_state_unlock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 skpair->sk_state_change(skpair);
406 read_lock(&skpair->sk_callback_lock);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +0800407 sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 read_unlock(&skpair->sk_callback_lock);
409 }
410 sock_put(skpair); /* It may now die */
411 unix_peer(sk) = NULL;
412 }
413
414 /* Try to flush out this socket. Throw out buffers at least */
415
416 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
Jianjun Konge27dfce2008-11-01 21:38:31 -0700417 if (state == TCP_LISTEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 unix_release_sock(skb->sk, 1);
419 /* passed fds are erased in the kfree_skb hook */
420 kfree_skb(skb);
421 }
422
423 if (dentry) {
424 dput(dentry);
425 mntput(mnt);
426 }
427
428 sock_put(sk);
429
430 /* ---- Socket is dead now and most probably destroyed ---- */
431
432 /*
433 * Fixme: BSD difference: In BSD all sockets connected to use get
434 * ECONNRESET and we die on the spot. In Linux we behave
435 * like files and pipes do and wait for the last
436 * dereference.
437 *
438 * Can't we simply set sock->err?
439 *
440 * What the above comment does talk about? --ANK(980817)
441 */
442
Pavel Emelyanov9305cfa2007-11-10 22:06:01 -0800443 if (unix_tot_inflight)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900444 unix_gc(); /* Garbage collect fds */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
446 return 0;
447}
448
449static int unix_listen(struct socket *sock, int backlog)
450{
451 int err;
452 struct sock *sk = sock->sk;
453 struct unix_sock *u = unix_sk(sk);
454
455 err = -EOPNOTSUPP;
456 if (sock->type!=SOCK_STREAM && sock->type!=SOCK_SEQPACKET)
457 goto out; /* Only stream/seqpacket sockets accept */
458 err = -EINVAL;
459 if (!u->addr)
460 goto out; /* No listens on an unbound socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700461 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN)
463 goto out_unlock;
464 if (backlog > sk->sk_max_ack_backlog)
465 wake_up_interruptible_all(&u->peer_wait);
466 sk->sk_max_ack_backlog = backlog;
467 sk->sk_state = TCP_LISTEN;
468 /* set credentials so connect can copy them */
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700469 sk->sk_peercred.pid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 sk->sk_peercred.uid = current->euid;
471 sk->sk_peercred.gid = current->egid;
472 err = 0;
473
474out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -0700475 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476out:
477 return err;
478}
479
480static int unix_release(struct socket *);
481static int unix_bind(struct socket *, struct sockaddr *, int);
482static int unix_stream_connect(struct socket *, struct sockaddr *,
483 int addr_len, int flags);
484static int unix_socketpair(struct socket *, struct socket *);
485static int unix_accept(struct socket *, struct socket *, int);
486static int unix_getname(struct socket *, struct sockaddr *, int *, int);
487static unsigned int unix_poll(struct file *, struct socket *, poll_table *);
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700488static unsigned int unix_dgram_poll(struct file *, struct socket *,
489 poll_table *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490static int unix_ioctl(struct socket *, unsigned int, unsigned long);
491static int unix_shutdown(struct socket *, int);
492static int unix_stream_sendmsg(struct kiocb *, struct socket *,
493 struct msghdr *, size_t);
494static int unix_stream_recvmsg(struct kiocb *, struct socket *,
495 struct msghdr *, size_t, int);
496static int unix_dgram_sendmsg(struct kiocb *, struct socket *,
497 struct msghdr *, size_t);
498static int unix_dgram_recvmsg(struct kiocb *, struct socket *,
499 struct msghdr *, size_t, int);
500static int unix_dgram_connect(struct socket *, struct sockaddr *,
501 int, int);
502static int unix_seqpacket_sendmsg(struct kiocb *, struct socket *,
503 struct msghdr *, size_t);
504
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800505static const struct proto_ops unix_stream_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 .family = PF_UNIX,
507 .owner = THIS_MODULE,
508 .release = unix_release,
509 .bind = unix_bind,
510 .connect = unix_stream_connect,
511 .socketpair = unix_socketpair,
512 .accept = unix_accept,
513 .getname = unix_getname,
514 .poll = unix_poll,
515 .ioctl = unix_ioctl,
516 .listen = unix_listen,
517 .shutdown = unix_shutdown,
518 .setsockopt = sock_no_setsockopt,
519 .getsockopt = sock_no_getsockopt,
520 .sendmsg = unix_stream_sendmsg,
521 .recvmsg = unix_stream_recvmsg,
522 .mmap = sock_no_mmap,
523 .sendpage = sock_no_sendpage,
524};
525
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800526static const struct proto_ops unix_dgram_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 .family = PF_UNIX,
528 .owner = THIS_MODULE,
529 .release = unix_release,
530 .bind = unix_bind,
531 .connect = unix_dgram_connect,
532 .socketpair = unix_socketpair,
533 .accept = sock_no_accept,
534 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700535 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 .ioctl = unix_ioctl,
537 .listen = sock_no_listen,
538 .shutdown = unix_shutdown,
539 .setsockopt = sock_no_setsockopt,
540 .getsockopt = sock_no_getsockopt,
541 .sendmsg = unix_dgram_sendmsg,
542 .recvmsg = unix_dgram_recvmsg,
543 .mmap = sock_no_mmap,
544 .sendpage = sock_no_sendpage,
545};
546
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800547static const struct proto_ops unix_seqpacket_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 .family = PF_UNIX,
549 .owner = THIS_MODULE,
550 .release = unix_release,
551 .bind = unix_bind,
552 .connect = unix_stream_connect,
553 .socketpair = unix_socketpair,
554 .accept = unix_accept,
555 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700556 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 .ioctl = unix_ioctl,
558 .listen = unix_listen,
559 .shutdown = unix_shutdown,
560 .setsockopt = sock_no_setsockopt,
561 .getsockopt = sock_no_getsockopt,
562 .sendmsg = unix_seqpacket_sendmsg,
563 .recvmsg = unix_dgram_recvmsg,
564 .mmap = sock_no_mmap,
565 .sendpage = sock_no_sendpage,
566};
567
568static struct proto unix_proto = {
569 .name = "UNIX",
570 .owner = THIS_MODULE,
571 .obj_size = sizeof(struct unix_sock),
572};
573
Ingo Molnara09785a2006-07-03 00:25:12 -0700574/*
575 * AF_UNIX sockets do not interact with hardware, hence they
576 * dont trigger interrupts - so it's safe for them to have
577 * bh-unsafe locking for their sk_receive_queue.lock. Split off
578 * this special lock-class by reinitializing the spinlock key:
579 */
580static struct lock_class_key af_unix_sk_receive_queue_lock_key;
581
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700582static struct sock * unix_create1(struct net *net, struct socket *sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583{
584 struct sock *sk = NULL;
585 struct unix_sock *u;
586
Pavel Emelyanov284b3272007-11-10 22:08:30 -0800587 atomic_inc(&unix_nr_socks);
588 if (atomic_read(&unix_nr_socks) > 2 * get_max_files())
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 goto out;
590
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700591 sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 if (!sk)
593 goto out;
594
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 sock_init_data(sock,sk);
Ingo Molnara09785a2006-07-03 00:25:12 -0700596 lockdep_set_class(&sk->sk_receive_queue.lock,
597 &af_unix_sk_receive_queue_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598
599 sk->sk_write_space = unix_write_space;
Denis V. Luneva0a53c82007-12-11 04:19:17 -0800600 sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 sk->sk_destruct = unix_sock_destructor;
602 u = unix_sk(sk);
603 u->dentry = NULL;
604 u->mnt = NULL;
Benjamin LaHaisefd19f322006-01-03 14:10:46 -0800605 spin_lock_init(&u->lock);
Al Viro516e0cc2008-07-26 00:39:17 -0400606 atomic_long_set(&u->inflight, 0);
Miklos Szeredi1fd05ba2007-07-11 14:22:39 -0700607 INIT_LIST_HEAD(&u->link);
Ingo Molnar57b47a52006-03-20 22:35:41 -0800608 mutex_init(&u->readlock); /* single task reading lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 init_waitqueue_head(&u->peer_wait);
610 unix_insert_socket(unix_sockets_unbound, sk);
611out:
Pavel Emelyanov284b3272007-11-10 22:08:30 -0800612 if (sk == NULL)
613 atomic_dec(&unix_nr_socks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 return sk;
615}
616
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700617static int unix_create(struct net *net, struct socket *sock, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618{
619 if (protocol && protocol != PF_UNIX)
620 return -EPROTONOSUPPORT;
621
622 sock->state = SS_UNCONNECTED;
623
624 switch (sock->type) {
625 case SOCK_STREAM:
626 sock->ops = &unix_stream_ops;
627 break;
628 /*
629 * Believe it or not BSD has AF_UNIX, SOCK_RAW though
630 * nothing uses it.
631 */
632 case SOCK_RAW:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700633 sock->type = SOCK_DGRAM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 case SOCK_DGRAM:
635 sock->ops = &unix_dgram_ops;
636 break;
637 case SOCK_SEQPACKET:
638 sock->ops = &unix_seqpacket_ops;
639 break;
640 default:
641 return -ESOCKTNOSUPPORT;
642 }
643
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700644 return unix_create1(net, sock) ? 0 : -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645}
646
647static int unix_release(struct socket *sock)
648{
649 struct sock *sk = sock->sk;
650
651 if (!sk)
652 return 0;
653
654 sock->sk = NULL;
655
656 return unix_release_sock (sk, 0);
657}
658
659static int unix_autobind(struct socket *sock)
660{
661 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900662 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 struct unix_sock *u = unix_sk(sk);
664 static u32 ordernum = 1;
665 struct unix_address * addr;
666 int err;
667
Ingo Molnar57b47a52006-03-20 22:35:41 -0800668 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
670 err = 0;
671 if (u->addr)
672 goto out;
673
674 err = -ENOMEM;
Panagiotis Issaris0da974f2006-07-21 14:51:30 -0700675 addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 if (!addr)
677 goto out;
678
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 addr->name->sun_family = AF_UNIX;
680 atomic_set(&addr->refcnt, 1);
681
682retry:
683 addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short);
684 addr->hash = unix_hash_fold(csum_partial((void*)addr->name, addr->len, 0));
685
David S. Millerfbe9cc42005-12-13 23:26:29 -0800686 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 ordernum = (ordernum+1)&0xFFFFF;
688
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800689 if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 addr->hash)) {
David S. Millerfbe9cc42005-12-13 23:26:29 -0800691 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 /* Sanity yield. It is unusual case, but yet... */
693 if (!(ordernum&0xFF))
694 yield();
695 goto retry;
696 }
697 addr->hash ^= sk->sk_type;
698
699 __unix_remove_socket(sk);
700 u->addr = addr;
701 __unix_insert_socket(&unix_socket_table[addr->hash], sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800702 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 err = 0;
704
Ingo Molnar57b47a52006-03-20 22:35:41 -0800705out: mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 return err;
707}
708
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800709static struct sock *unix_find_other(struct net *net,
710 struct sockaddr_un *sunname, int len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 int type, unsigned hash, int *error)
712{
713 struct sock *u;
Al Viro421748e2008-08-02 01:04:36 -0400714 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 int err = 0;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900716
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 if (sunname->sun_path[0]) {
Al Viro421748e2008-08-02 01:04:36 -0400718 struct inode *inode;
719 err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 if (err)
721 goto fail;
Al Viro421748e2008-08-02 01:04:36 -0400722 inode = path.dentry->d_inode;
723 err = inode_permission(inode, MAY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 if (err)
725 goto put_fail;
726
727 err = -ECONNREFUSED;
Al Viro421748e2008-08-02 01:04:36 -0400728 if (!S_ISSOCK(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 goto put_fail;
Al Viro421748e2008-08-02 01:04:36 -0400730 u = unix_find_socket_byinode(net, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 if (!u)
732 goto put_fail;
733
734 if (u->sk_type == type)
Al Viro421748e2008-08-02 01:04:36 -0400735 touch_atime(path.mnt, path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736
Al Viro421748e2008-08-02 01:04:36 -0400737 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738
Jianjun Konge27dfce2008-11-01 21:38:31 -0700739 err = -EPROTOTYPE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 if (u->sk_type != type) {
741 sock_put(u);
742 goto fail;
743 }
744 } else {
745 err = -ECONNREFUSED;
Jianjun Konge27dfce2008-11-01 21:38:31 -0700746 u = unix_find_socket_byname(net, sunname, len, type, hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 if (u) {
748 struct dentry *dentry;
749 dentry = unix_sk(u)->dentry;
750 if (dentry)
751 touch_atime(unix_sk(u)->mnt, dentry);
752 } else
753 goto fail;
754 }
755 return u;
756
757put_fail:
Al Viro421748e2008-08-02 01:04:36 -0400758 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759fail:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700760 *error = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 return NULL;
762}
763
764
765static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
766{
767 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900768 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700770 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 struct dentry * dentry = NULL;
772 struct nameidata nd;
773 int err;
774 unsigned hash;
775 struct unix_address *addr;
776 struct hlist_head *list;
777
778 err = -EINVAL;
779 if (sunaddr->sun_family != AF_UNIX)
780 goto out;
781
Jianjun Konge27dfce2008-11-01 21:38:31 -0700782 if (addr_len == sizeof(short)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 err = unix_autobind(sock);
784 goto out;
785 }
786
787 err = unix_mkname(sunaddr, addr_len, &hash);
788 if (err < 0)
789 goto out;
790 addr_len = err;
791
Ingo Molnar57b47a52006-03-20 22:35:41 -0800792 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
794 err = -EINVAL;
795 if (u->addr)
796 goto out_up;
797
798 err = -ENOMEM;
799 addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL);
800 if (!addr)
801 goto out_up;
802
803 memcpy(addr->name, sunaddr, addr_len);
804 addr->len = addr_len;
805 addr->hash = hash ^ sk->sk_type;
806 atomic_set(&addr->refcnt, 1);
807
808 if (sunaddr->sun_path[0]) {
809 unsigned int mode;
810 err = 0;
811 /*
812 * Get the parent directory, calculate the hash for last
813 * component.
814 */
815 err = path_lookup(sunaddr->sun_path, LOOKUP_PARENT, &nd);
816 if (err)
817 goto out_mknod_parent;
Christoph Hellwigf81a0bf2005-05-19 12:26:43 -0700818
819 dentry = lookup_create(&nd, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 err = PTR_ERR(dentry);
821 if (IS_ERR(dentry))
822 goto out_mknod_unlock;
Christoph Hellwigf81a0bf2005-05-19 12:26:43 -0700823
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 /*
825 * All right, let's create it.
826 */
827 mode = S_IFSOCK |
828 (SOCK_INODE(sock)->i_mode & ~current->fs->umask);
Dave Hansen463c3192008-02-15 14:37:57 -0800829 err = mnt_want_write(nd.path.mnt);
830 if (err)
831 goto out_mknod_dput;
Jan Blunck4ac91372008-02-14 19:34:32 -0800832 err = vfs_mknod(nd.path.dentry->d_inode, dentry, mode, 0);
Dave Hansen463c3192008-02-15 14:37:57 -0800833 mnt_drop_write(nd.path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 if (err)
835 goto out_mknod_dput;
Jan Blunck4ac91372008-02-14 19:34:32 -0800836 mutex_unlock(&nd.path.dentry->d_inode->i_mutex);
837 dput(nd.path.dentry);
838 nd.path.dentry = dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839
840 addr->hash = UNIX_HASH_SIZE;
841 }
842
David S. Millerfbe9cc42005-12-13 23:26:29 -0800843 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844
845 if (!sunaddr->sun_path[0]) {
846 err = -EADDRINUSE;
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800847 if (__unix_find_socket_byname(net, sunaddr, addr_len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 sk->sk_type, hash)) {
849 unix_release_addr(addr);
850 goto out_unlock;
851 }
852
853 list = &unix_socket_table[addr->hash];
854 } else {
855 list = &unix_socket_table[dentry->d_inode->i_ino & (UNIX_HASH_SIZE-1)];
Jan Blunck4ac91372008-02-14 19:34:32 -0800856 u->dentry = nd.path.dentry;
857 u->mnt = nd.path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 }
859
860 err = 0;
861 __unix_remove_socket(sk);
862 u->addr = addr;
863 __unix_insert_socket(list, sk);
864
865out_unlock:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800866 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867out_up:
Ingo Molnar57b47a52006-03-20 22:35:41 -0800868 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869out:
870 return err;
871
872out_mknod_dput:
873 dput(dentry);
874out_mknod_unlock:
Jan Blunck4ac91372008-02-14 19:34:32 -0800875 mutex_unlock(&nd.path.dentry->d_inode->i_mutex);
Jan Blunck1d957f92008-02-14 19:34:35 -0800876 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877out_mknod_parent:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700878 if (err == -EEXIST)
879 err = -EADDRINUSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 unix_release_addr(addr);
881 goto out_up;
882}
883
David S. Miller278a3de2007-05-31 15:19:20 -0700884static void unix_state_double_lock(struct sock *sk1, struct sock *sk2)
885{
886 if (unlikely(sk1 == sk2) || !sk2) {
887 unix_state_lock(sk1);
888 return;
889 }
890 if (sk1 < sk2) {
891 unix_state_lock(sk1);
892 unix_state_lock_nested(sk2);
893 } else {
894 unix_state_lock(sk2);
895 unix_state_lock_nested(sk1);
896 }
897}
898
899static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2)
900{
901 if (unlikely(sk1 == sk2) || !sk2) {
902 unix_state_unlock(sk1);
903 return;
904 }
905 unix_state_unlock(sk1);
906 unix_state_unlock(sk2);
907}
908
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr,
910 int alen, int flags)
911{
912 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900913 struct net *net = sock_net(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700914 struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 struct sock *other;
916 unsigned hash;
917 int err;
918
919 if (addr->sa_family != AF_UNSPEC) {
920 err = unix_mkname(sunaddr, alen, &hash);
921 if (err < 0)
922 goto out;
923 alen = err;
924
925 if (test_bit(SOCK_PASSCRED, &sock->flags) &&
926 !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0)
927 goto out;
928
David S. Miller278a3de2007-05-31 15:19:20 -0700929restart:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700930 other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 if (!other)
932 goto out;
933
David S. Miller278a3de2007-05-31 15:19:20 -0700934 unix_state_double_lock(sk, other);
935
936 /* Apparently VFS overslept socket death. Retry. */
937 if (sock_flag(other, SOCK_DEAD)) {
938 unix_state_double_unlock(sk, other);
939 sock_put(other);
940 goto restart;
941 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942
943 err = -EPERM;
944 if (!unix_may_send(sk, other))
945 goto out_unlock;
946
947 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
948 if (err)
949 goto out_unlock;
950
951 } else {
952 /*
953 * 1003.1g breaking connected state with AF_UNSPEC
954 */
955 other = NULL;
David S. Miller278a3de2007-05-31 15:19:20 -0700956 unix_state_double_lock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 }
958
959 /*
960 * If it was connected, reconnect.
961 */
962 if (unix_peer(sk)) {
963 struct sock *old_peer = unix_peer(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700964 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -0700965 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
967 if (other != old_peer)
968 unix_dgram_disconnected(sk, old_peer);
969 sock_put(old_peer);
970 } else {
Jianjun Konge27dfce2008-11-01 21:38:31 -0700971 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -0700972 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 }
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900974 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975
976out_unlock:
David S. Miller278a3de2007-05-31 15:19:20 -0700977 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 sock_put(other);
979out:
980 return err;
981}
982
983static long unix_wait_for_peer(struct sock *other, long timeo)
984{
985 struct unix_sock *u = unix_sk(other);
986 int sched;
987 DEFINE_WAIT(wait);
988
989 prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE);
990
991 sched = !sock_flag(other, SOCK_DEAD) &&
992 !(other->sk_shutdown & RCV_SHUTDOWN) &&
Rainer Weikusat3c734192008-06-17 22:28:05 -0700993 unix_recvq_full(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
David S. Miller1c92b4e2007-05-31 13:24:26 -0700995 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996
997 if (sched)
998 timeo = schedule_timeout(timeo);
999
1000 finish_wait(&u->peer_wait, &wait);
1001 return timeo;
1002}
1003
1004static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr,
1005 int addr_len, int flags)
1006{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001007 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001009 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 struct unix_sock *u = unix_sk(sk), *newu, *otheru;
1011 struct sock *newsk = NULL;
1012 struct sock *other = NULL;
1013 struct sk_buff *skb = NULL;
1014 unsigned hash;
1015 int st;
1016 int err;
1017 long timeo;
1018
1019 err = unix_mkname(sunaddr, addr_len, &hash);
1020 if (err < 0)
1021 goto out;
1022 addr_len = err;
1023
1024 if (test_bit(SOCK_PASSCRED, &sock->flags)
1025 && !u->addr && (err = unix_autobind(sock)) != 0)
1026 goto out;
1027
1028 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1029
1030 /* First of all allocate resources.
1031 If we will make it after state is locked,
1032 we will have to recheck all again in any case.
1033 */
1034
1035 err = -ENOMEM;
1036
1037 /* create new sock for complete connection */
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001038 newsk = unix_create1(sock_net(sk), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 if (newsk == NULL)
1040 goto out;
1041
1042 /* Allocate skb for sending to listening sock */
1043 skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL);
1044 if (skb == NULL)
1045 goto out;
1046
1047restart:
1048 /* Find listening sock. */
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001049 other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 if (!other)
1051 goto out;
1052
1053 /* Latch state of peer */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001054 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055
1056 /* Apparently VFS overslept socket death. Retry. */
1057 if (sock_flag(other, SOCK_DEAD)) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001058 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 sock_put(other);
1060 goto restart;
1061 }
1062
1063 err = -ECONNREFUSED;
1064 if (other->sk_state != TCP_LISTEN)
1065 goto out_unlock;
1066
Rainer Weikusat3c734192008-06-17 22:28:05 -07001067 if (unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 err = -EAGAIN;
1069 if (!timeo)
1070 goto out_unlock;
1071
1072 timeo = unix_wait_for_peer(other, timeo);
1073
1074 err = sock_intr_errno(timeo);
1075 if (signal_pending(current))
1076 goto out;
1077 sock_put(other);
1078 goto restart;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001079 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080
1081 /* Latch our state.
1082
1083 It is tricky place. We need to grab write lock and cannot
1084 drop lock on peer. It is dangerous because deadlock is
1085 possible. Connect to self case and simultaneous
1086 attempt to connect are eliminated by checking socket
1087 state. other is TCP_LISTEN, if sk is TCP_LISTEN we
1088 check this before attempt to grab lock.
1089
1090 Well, and we have to recheck the state after socket locked.
1091 */
1092 st = sk->sk_state;
1093
1094 switch (st) {
1095 case TCP_CLOSE:
1096 /* This is ok... continue with connect */
1097 break;
1098 case TCP_ESTABLISHED:
1099 /* Socket is already connected */
1100 err = -EISCONN;
1101 goto out_unlock;
1102 default:
1103 err = -EINVAL;
1104 goto out_unlock;
1105 }
1106
David S. Miller1c92b4e2007-05-31 13:24:26 -07001107 unix_state_lock_nested(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108
1109 if (sk->sk_state != st) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001110 unix_state_unlock(sk);
1111 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 sock_put(other);
1113 goto restart;
1114 }
1115
1116 err = security_unix_stream_connect(sock, other->sk_socket, newsk);
1117 if (err) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001118 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 goto out_unlock;
1120 }
1121
1122 /* The way is open! Fastly set all the necessary fields... */
1123
1124 sock_hold(sk);
1125 unix_peer(newsk) = sk;
1126 newsk->sk_state = TCP_ESTABLISHED;
1127 newsk->sk_type = sk->sk_type;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001128 newsk->sk_peercred.pid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 newsk->sk_peercred.uid = current->euid;
1130 newsk->sk_peercred.gid = current->egid;
1131 newu = unix_sk(newsk);
1132 newsk->sk_sleep = &newu->peer_wait;
1133 otheru = unix_sk(other);
1134
1135 /* copy address information from listening to new sock*/
1136 if (otheru->addr) {
1137 atomic_inc(&otheru->addr->refcnt);
1138 newu->addr = otheru->addr;
1139 }
1140 if (otheru->dentry) {
1141 newu->dentry = dget(otheru->dentry);
1142 newu->mnt = mntget(otheru->mnt);
1143 }
1144
1145 /* Set credentials */
1146 sk->sk_peercred = other->sk_peercred;
1147
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 sock->state = SS_CONNECTED;
1149 sk->sk_state = TCP_ESTABLISHED;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001150 sock_hold(newsk);
1151
1152 smp_mb__after_atomic_inc(); /* sock_hold() does an atomic_inc() */
1153 unix_peer(sk) = newsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
David S. Miller1c92b4e2007-05-31 13:24:26 -07001155 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
1157 /* take ten and and send info to listening sock */
1158 spin_lock(&other->sk_receive_queue.lock);
1159 __skb_queue_tail(&other->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 spin_unlock(&other->sk_receive_queue.lock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001161 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 other->sk_data_ready(other, 0);
1163 sock_put(other);
1164 return 0;
1165
1166out_unlock:
1167 if (other)
David S. Miller1c92b4e2007-05-31 13:24:26 -07001168 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169
1170out:
1171 if (skb)
1172 kfree_skb(skb);
1173 if (newsk)
1174 unix_release_sock(newsk, 0);
1175 if (other)
1176 sock_put(other);
1177 return err;
1178}
1179
1180static int unix_socketpair(struct socket *socka, struct socket *sockb)
1181{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001182 struct sock *ska = socka->sk, *skb = sockb->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
1184 /* Join our sockets back to back */
1185 sock_hold(ska);
1186 sock_hold(skb);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001187 unix_peer(ska) = skb;
1188 unix_peer(skb) = ska;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001189 ska->sk_peercred.pid = skb->sk_peercred.pid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190 ska->sk_peercred.uid = skb->sk_peercred.uid = current->euid;
1191 ska->sk_peercred.gid = skb->sk_peercred.gid = current->egid;
1192
1193 if (ska->sk_type != SOCK_DGRAM) {
1194 ska->sk_state = TCP_ESTABLISHED;
1195 skb->sk_state = TCP_ESTABLISHED;
1196 socka->state = SS_CONNECTED;
1197 sockb->state = SS_CONNECTED;
1198 }
1199 return 0;
1200}
1201
1202static int unix_accept(struct socket *sock, struct socket *newsock, int flags)
1203{
1204 struct sock *sk = sock->sk;
1205 struct sock *tsk;
1206 struct sk_buff *skb;
1207 int err;
1208
1209 err = -EOPNOTSUPP;
1210 if (sock->type!=SOCK_STREAM && sock->type!=SOCK_SEQPACKET)
1211 goto out;
1212
1213 err = -EINVAL;
1214 if (sk->sk_state != TCP_LISTEN)
1215 goto out;
1216
1217 /* If socket state is TCP_LISTEN it cannot change (for now...),
1218 * so that no locks are necessary.
1219 */
1220
1221 skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err);
1222 if (!skb) {
1223 /* This means receive shutdown. */
1224 if (err == 0)
1225 err = -EINVAL;
1226 goto out;
1227 }
1228
1229 tsk = skb->sk;
1230 skb_free_datagram(sk, skb);
1231 wake_up_interruptible(&unix_sk(sk)->peer_wait);
1232
1233 /* attach accepted sock to socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001234 unix_state_lock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 newsock->state = SS_CONNECTED;
1236 sock_graft(tsk, newsock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001237 unix_state_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238 return 0;
1239
1240out:
1241 return err;
1242}
1243
1244
1245static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer)
1246{
1247 struct sock *sk = sock->sk;
1248 struct unix_sock *u;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001249 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 int err = 0;
1251
1252 if (peer) {
1253 sk = unix_peer_get(sk);
1254
1255 err = -ENOTCONN;
1256 if (!sk)
1257 goto out;
1258 err = 0;
1259 } else {
1260 sock_hold(sk);
1261 }
1262
1263 u = unix_sk(sk);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001264 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 if (!u->addr) {
1266 sunaddr->sun_family = AF_UNIX;
1267 sunaddr->sun_path[0] = 0;
1268 *uaddr_len = sizeof(short);
1269 } else {
1270 struct unix_address *addr = u->addr;
1271
1272 *uaddr_len = addr->len;
1273 memcpy(sunaddr, addr->name, *uaddr_len);
1274 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07001275 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 sock_put(sk);
1277out:
1278 return err;
1279}
1280
1281static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb)
1282{
1283 int i;
1284
1285 scm->fp = UNIXCB(skb).fp;
1286 skb->destructor = sock_wfree;
1287 UNIXCB(skb).fp = NULL;
1288
1289 for (i=scm->fp->count-1; i>=0; i--)
1290 unix_notinflight(scm->fp->fp[i]);
1291}
1292
1293static void unix_destruct_fds(struct sk_buff *skb)
1294{
1295 struct scm_cookie scm;
1296 memset(&scm, 0, sizeof(scm));
1297 unix_detach_fds(&scm, skb);
1298
1299 /* Alas, it calls VFS */
1300 /* So fscking what? fput() had been SMP-safe since the last Summer */
1301 scm_destroy(&scm);
1302 sock_wfree(skb);
1303}
1304
1305static void unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
1306{
1307 int i;
1308 for (i=scm->fp->count-1; i>=0; i--)
1309 unix_inflight(scm->fp->fp[i]);
1310 UNIXCB(skb).fp = scm->fp;
1311 skb->destructor = unix_destruct_fds;
1312 scm->fp = NULL;
1313}
1314
1315/*
1316 * Send AF_UNIX data.
1317 */
1318
1319static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock,
1320 struct msghdr *msg, size_t len)
1321{
1322 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1323 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001324 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001326 struct sockaddr_un *sunaddr = msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 struct sock *other = NULL;
1328 int namelen = 0; /* fake GCC */
1329 int err;
1330 unsigned hash;
1331 struct sk_buff *skb;
1332 long timeo;
1333 struct scm_cookie tmp_scm;
1334
1335 if (NULL == siocb->scm)
1336 siocb->scm = &tmp_scm;
1337 err = scm_send(sock, msg, siocb->scm);
1338 if (err < 0)
1339 return err;
1340
1341 err = -EOPNOTSUPP;
1342 if (msg->msg_flags&MSG_OOB)
1343 goto out;
1344
1345 if (msg->msg_namelen) {
1346 err = unix_mkname(sunaddr, msg->msg_namelen, &hash);
1347 if (err < 0)
1348 goto out;
1349 namelen = err;
1350 } else {
1351 sunaddr = NULL;
1352 err = -ENOTCONN;
1353 other = unix_peer_get(sk);
1354 if (!other)
1355 goto out;
1356 }
1357
1358 if (test_bit(SOCK_PASSCRED, &sock->flags)
1359 && !u->addr && (err = unix_autobind(sock)) != 0)
1360 goto out;
1361
1362 err = -EMSGSIZE;
1363 if (len > sk->sk_sndbuf - 32)
1364 goto out;
1365
1366 skb = sock_alloc_send_skb(sk, len, msg->msg_flags&MSG_DONTWAIT, &err);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001367 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 goto out;
1369
1370 memcpy(UNIXCREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
1371 if (siocb->scm->fp)
1372 unix_attach_fds(siocb->scm, skb);
Catherine Zhangdc49c1f2006-08-02 14:12:06 -07001373 unix_get_secdata(siocb->scm, skb);
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001374
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001375 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 err = memcpy_fromiovec(skb_put(skb,len), msg->msg_iov, len);
1377 if (err)
1378 goto out_free;
1379
1380 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1381
1382restart:
1383 if (!other) {
1384 err = -ECONNRESET;
1385 if (sunaddr == NULL)
1386 goto out_free;
1387
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001388 other = unix_find_other(net, sunaddr, namelen, sk->sk_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 hash, &err);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001390 if (other == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 goto out_free;
1392 }
1393
David S. Miller1c92b4e2007-05-31 13:24:26 -07001394 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 err = -EPERM;
1396 if (!unix_may_send(sk, other))
1397 goto out_unlock;
1398
1399 if (sock_flag(other, SOCK_DEAD)) {
1400 /*
1401 * Check with 1003.1g - what should
1402 * datagram error
1403 */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001404 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 sock_put(other);
1406
1407 err = 0;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001408 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 if (unix_peer(sk) == other) {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001410 unix_peer(sk) = NULL;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001411 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412
1413 unix_dgram_disconnected(sk, other);
1414 sock_put(other);
1415 err = -ECONNREFUSED;
1416 } else {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001417 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 }
1419
1420 other = NULL;
1421 if (err)
1422 goto out_free;
1423 goto restart;
1424 }
1425
1426 err = -EPIPE;
1427 if (other->sk_shutdown & RCV_SHUTDOWN)
1428 goto out_unlock;
1429
1430 if (sk->sk_type != SOCK_SEQPACKET) {
1431 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1432 if (err)
1433 goto out_unlock;
1434 }
1435
Rainer Weikusat3c734192008-06-17 22:28:05 -07001436 if (unix_peer(other) != sk && unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 if (!timeo) {
1438 err = -EAGAIN;
1439 goto out_unlock;
1440 }
1441
1442 timeo = unix_wait_for_peer(other, timeo);
1443
1444 err = sock_intr_errno(timeo);
1445 if (signal_pending(current))
1446 goto out_free;
1447
1448 goto restart;
1449 }
1450
1451 skb_queue_tail(&other->sk_receive_queue, skb);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001452 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 other->sk_data_ready(other, len);
1454 sock_put(other);
1455 scm_destroy(siocb->scm);
1456 return len;
1457
1458out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001459 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460out_free:
1461 kfree_skb(skb);
1462out:
1463 if (other)
1464 sock_put(other);
1465 scm_destroy(siocb->scm);
1466 return err;
1467}
1468
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001469
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock,
1471 struct msghdr *msg, size_t len)
1472{
1473 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1474 struct sock *sk = sock->sk;
1475 struct sock *other = NULL;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001476 struct sockaddr_un *sunaddr = msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 int err,size;
1478 struct sk_buff *skb;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001479 int sent = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 struct scm_cookie tmp_scm;
1481
1482 if (NULL == siocb->scm)
1483 siocb->scm = &tmp_scm;
1484 err = scm_send(sock, msg, siocb->scm);
1485 if (err < 0)
1486 return err;
1487
1488 err = -EOPNOTSUPP;
1489 if (msg->msg_flags&MSG_OOB)
1490 goto out_err;
1491
1492 if (msg->msg_namelen) {
1493 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1494 goto out_err;
1495 } else {
1496 sunaddr = NULL;
1497 err = -ENOTCONN;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001498 other = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 if (!other)
1500 goto out_err;
1501 }
1502
1503 if (sk->sk_shutdown & SEND_SHUTDOWN)
1504 goto pipe_err;
1505
1506 while(sent < len)
1507 {
1508 /*
Benjamin LaHaisee9df7d72006-03-20 21:29:05 -08001509 * Optimisation for the fact that under 0.01% of X
1510 * messages typically need breaking up.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 */
1512
Benjamin LaHaisee9df7d72006-03-20 21:29:05 -08001513 size = len-sent;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514
1515 /* Keep two messages in the pipe so it schedules better */
Benjamin LaHaisee9df7d72006-03-20 21:29:05 -08001516 if (size > ((sk->sk_sndbuf >> 1) - 64))
1517 size = (sk->sk_sndbuf >> 1) - 64;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518
1519 if (size > SKB_MAX_ALLOC)
1520 size = SKB_MAX_ALLOC;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001521
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 /*
1523 * Grab a buffer
1524 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001525
Jianjun Konge27dfce2008-11-01 21:38:31 -07001526 skb = sock_alloc_send_skb(sk,size,msg->msg_flags&MSG_DONTWAIT, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527
Jianjun Konge27dfce2008-11-01 21:38:31 -07001528 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 goto out_err;
1530
1531 /*
1532 * If you pass two values to the sock_alloc_send_skb
1533 * it tries to grab the large buffer with GFP_NOFS
1534 * (which can fail easily), and if it fails grab the
1535 * fallback size buffer which is under a page and will
1536 * succeed. [Alan]
1537 */
1538 size = min_t(int, size, skb_tailroom(skb));
1539
1540 memcpy(UNIXCREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
1541 if (siocb->scm->fp)
1542 unix_attach_fds(siocb->scm, skb);
1543
1544 if ((err = memcpy_fromiovec(skb_put(skb,size), msg->msg_iov, size)) != 0) {
1545 kfree_skb(skb);
1546 goto out_err;
1547 }
1548
David S. Miller1c92b4e2007-05-31 13:24:26 -07001549 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550
1551 if (sock_flag(other, SOCK_DEAD) ||
1552 (other->sk_shutdown & RCV_SHUTDOWN))
1553 goto pipe_err_free;
1554
1555 skb_queue_tail(&other->sk_receive_queue, skb);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001556 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 other->sk_data_ready(other, size);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001558 sent += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560
1561 scm_destroy(siocb->scm);
1562 siocb->scm = NULL;
1563
1564 return sent;
1565
1566pipe_err_free:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001567 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 kfree_skb(skb);
1569pipe_err:
1570 if (sent==0 && !(msg->msg_flags&MSG_NOSIGNAL))
1571 send_sig(SIGPIPE,current,0);
1572 err = -EPIPE;
1573out_err:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 scm_destroy(siocb->scm);
1575 siocb->scm = NULL;
1576 return sent ? : err;
1577}
1578
1579static int unix_seqpacket_sendmsg(struct kiocb *kiocb, struct socket *sock,
1580 struct msghdr *msg, size_t len)
1581{
1582 int err;
1583 struct sock *sk = sock->sk;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001584
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 err = sock_error(sk);
1586 if (err)
1587 return err;
1588
1589 if (sk->sk_state != TCP_ESTABLISHED)
1590 return -ENOTCONN;
1591
1592 if (msg->msg_namelen)
1593 msg->msg_namelen = 0;
1594
1595 return unix_dgram_sendmsg(kiocb, sock, msg, len);
1596}
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001597
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598static void unix_copy_addr(struct msghdr *msg, struct sock *sk)
1599{
1600 struct unix_sock *u = unix_sk(sk);
1601
1602 msg->msg_namelen = 0;
1603 if (u->addr) {
1604 msg->msg_namelen = u->addr->len;
1605 memcpy(msg->msg_name, u->addr->name, u->addr->len);
1606 }
1607}
1608
1609static int unix_dgram_recvmsg(struct kiocb *iocb, struct socket *sock,
1610 struct msghdr *msg, size_t size,
1611 int flags)
1612{
1613 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
1614 struct scm_cookie tmp_scm;
1615 struct sock *sk = sock->sk;
1616 struct unix_sock *u = unix_sk(sk);
1617 int noblock = flags & MSG_DONTWAIT;
1618 struct sk_buff *skb;
1619 int err;
1620
1621 err = -EOPNOTSUPP;
1622 if (flags&MSG_OOB)
1623 goto out;
1624
1625 msg->msg_namelen = 0;
1626
Ingo Molnar57b47a52006-03-20 22:35:41 -08001627 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628
1629 skb = skb_recv_datagram(sk, flags, noblock, &err);
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001630 if (!skb) {
1631 unix_state_lock(sk);
1632 /* Signal EOF on disconnected non-blocking SEQPACKET socket. */
1633 if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN &&
1634 (sk->sk_shutdown & RCV_SHUTDOWN))
1635 err = 0;
1636 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 goto out_unlock;
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001638 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639
Ingo Molnar71e20f12007-10-15 17:00:19 +02001640 wake_up_interruptible_sync(&u->peer_wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641
1642 if (msg->msg_name)
1643 unix_copy_addr(msg, skb->sk);
1644
1645 if (size > skb->len)
1646 size = skb->len;
1647 else if (size < skb->len)
1648 msg->msg_flags |= MSG_TRUNC;
1649
1650 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, size);
1651 if (err)
1652 goto out_free;
1653
1654 if (!siocb->scm) {
1655 siocb->scm = &tmp_scm;
1656 memset(&tmp_scm, 0, sizeof(tmp_scm));
1657 }
1658 siocb->scm->creds = *UNIXCREDS(skb);
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001659 unix_set_secdata(siocb->scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660
1661 if (!(flags & MSG_PEEK))
1662 {
1663 if (UNIXCB(skb).fp)
1664 unix_detach_fds(siocb->scm, skb);
1665 }
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001666 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 {
1668 /* It is questionable: on PEEK we could:
1669 - do not return fds - good, but too simple 8)
1670 - return fds, and do not return them on read (old strategy,
1671 apparently wrong)
1672 - clone fds (I chose it for now, it is the most universal
1673 solution)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001674
1675 POSIX 1003.1g does not actually define this clearly
1676 at all. POSIX 1003.1g doesn't define a lot of things
1677 clearly however!
1678
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679 */
1680 if (UNIXCB(skb).fp)
1681 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
1682 }
1683 err = size;
1684
1685 scm_recv(sock, msg, siocb->scm, flags);
1686
1687out_free:
1688 skb_free_datagram(sk,skb);
1689out_unlock:
Ingo Molnar57b47a52006-03-20 22:35:41 -08001690 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691out:
1692 return err;
1693}
1694
1695/*
1696 * Sleep until data has arrive. But check for races..
1697 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001698
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699static long unix_stream_data_wait(struct sock * sk, long timeo)
1700{
1701 DEFINE_WAIT(wait);
1702
David S. Miller1c92b4e2007-05-31 13:24:26 -07001703 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704
1705 for (;;) {
1706 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1707
David S. Millerb03efcf2005-07-08 14:57:23 -07001708 if (!skb_queue_empty(&sk->sk_receive_queue) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709 sk->sk_err ||
1710 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1711 signal_pending(current) ||
1712 !timeo)
1713 break;
1714
1715 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001716 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 timeo = schedule_timeout(timeo);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001718 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1720 }
1721
1722 finish_wait(sk->sk_sleep, &wait);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001723 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 return timeo;
1725}
1726
1727
1728
1729static int unix_stream_recvmsg(struct kiocb *iocb, struct socket *sock,
1730 struct msghdr *msg, size_t size,
1731 int flags)
1732{
1733 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
1734 struct scm_cookie tmp_scm;
1735 struct sock *sk = sock->sk;
1736 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001737 struct sockaddr_un *sunaddr = msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 int copied = 0;
1739 int check_creds = 0;
1740 int target;
1741 int err = 0;
1742 long timeo;
1743
1744 err = -EINVAL;
1745 if (sk->sk_state != TCP_ESTABLISHED)
1746 goto out;
1747
1748 err = -EOPNOTSUPP;
1749 if (flags&MSG_OOB)
1750 goto out;
1751
1752 target = sock_rcvlowat(sk, flags&MSG_WAITALL, size);
1753 timeo = sock_rcvtimeo(sk, flags&MSG_DONTWAIT);
1754
1755 msg->msg_namelen = 0;
1756
1757 /* Lock the socket to prevent queue disordering
1758 * while sleeps in memcpy_tomsg
1759 */
1760
1761 if (!siocb->scm) {
1762 siocb->scm = &tmp_scm;
1763 memset(&tmp_scm, 0, sizeof(tmp_scm));
1764 }
1765
Ingo Molnar57b47a52006-03-20 22:35:41 -08001766 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767
1768 do
1769 {
1770 int chunk;
1771 struct sk_buff *skb;
1772
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001773 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 skb = skb_dequeue(&sk->sk_receive_queue);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001775 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 {
1777 if (copied >= target)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001778 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779
1780 /*
1781 * POSIX 1003.1g mandates this order.
1782 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001783
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784 if ((err = sock_error(sk)) != 0)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001785 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 if (sk->sk_shutdown & RCV_SHUTDOWN)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001787 goto unlock;
1788
1789 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 err = -EAGAIN;
1791 if (!timeo)
1792 break;
Ingo Molnar57b47a52006-03-20 22:35:41 -08001793 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794
1795 timeo = unix_stream_data_wait(sk, timeo);
1796
1797 if (signal_pending(current)) {
1798 err = sock_intr_errno(timeo);
1799 goto out;
1800 }
Ingo Molnar57b47a52006-03-20 22:35:41 -08001801 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 continue;
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001803 unlock:
1804 unix_state_unlock(sk);
1805 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 }
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001807 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808
1809 if (check_creds) {
1810 /* Never glue messages from different writers */
1811 if (memcmp(UNIXCREDS(skb), &siocb->scm->creds, sizeof(siocb->scm->creds)) != 0) {
1812 skb_queue_head(&sk->sk_receive_queue, skb);
1813 break;
1814 }
1815 } else {
1816 /* Copy credentials */
1817 siocb->scm->creds = *UNIXCREDS(skb);
1818 check_creds = 1;
1819 }
1820
1821 /* Copy address just once */
1822 if (sunaddr)
1823 {
1824 unix_copy_addr(msg, skb->sk);
1825 sunaddr = NULL;
1826 }
1827
1828 chunk = min_t(unsigned int, skb->len, size);
1829 if (memcpy_toiovec(msg->msg_iov, skb->data, chunk)) {
1830 skb_queue_head(&sk->sk_receive_queue, skb);
1831 if (copied == 0)
1832 copied = -EFAULT;
1833 break;
1834 }
1835 copied += chunk;
1836 size -= chunk;
1837
1838 /* Mark read part of skb as used */
1839 if (!(flags & MSG_PEEK))
1840 {
1841 skb_pull(skb, chunk);
1842
1843 if (UNIXCB(skb).fp)
1844 unix_detach_fds(siocb->scm, skb);
1845
1846 /* put the skb back if we didn't use it up.. */
1847 if (skb->len)
1848 {
1849 skb_queue_head(&sk->sk_receive_queue, skb);
1850 break;
1851 }
1852
1853 kfree_skb(skb);
1854
1855 if (siocb->scm->fp)
1856 break;
1857 }
1858 else
1859 {
1860 /* It is questionable, see note in unix_dgram_recvmsg.
1861 */
1862 if (UNIXCB(skb).fp)
1863 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
1864
1865 /* put message back and return */
1866 skb_queue_head(&sk->sk_receive_queue, skb);
1867 break;
1868 }
1869 } while (size);
1870
Ingo Molnar57b47a52006-03-20 22:35:41 -08001871 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 scm_recv(sock, msg, siocb->scm, flags);
1873out:
1874 return copied ? : err;
1875}
1876
1877static int unix_shutdown(struct socket *sock, int mode)
1878{
1879 struct sock *sk = sock->sk;
1880 struct sock *other;
1881
1882 mode = (mode+1)&(RCV_SHUTDOWN|SEND_SHUTDOWN);
1883
1884 if (mode) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001885 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886 sk->sk_shutdown |= mode;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001887 other = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888 if (other)
1889 sock_hold(other);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001890 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 sk->sk_state_change(sk);
1892
1893 if (other &&
1894 (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) {
1895
1896 int peer_mode = 0;
1897
1898 if (mode&RCV_SHUTDOWN)
1899 peer_mode |= SEND_SHUTDOWN;
1900 if (mode&SEND_SHUTDOWN)
1901 peer_mode |= RCV_SHUTDOWN;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001902 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 other->sk_shutdown |= peer_mode;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001904 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905 other->sk_state_change(other);
1906 read_lock(&other->sk_callback_lock);
1907 if (peer_mode == SHUTDOWN_MASK)
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001908 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 else if (peer_mode & RCV_SHUTDOWN)
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001910 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 read_unlock(&other->sk_callback_lock);
1912 }
1913 if (other)
1914 sock_put(other);
1915 }
1916 return 0;
1917}
1918
1919static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1920{
1921 struct sock *sk = sock->sk;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001922 long amount = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 int err;
1924
1925 switch(cmd)
1926 {
1927 case SIOCOUTQ:
1928 amount = atomic_read(&sk->sk_wmem_alloc);
1929 err = put_user(amount, (int __user *)arg);
1930 break;
1931 case SIOCINQ:
1932 {
1933 struct sk_buff *skb;
1934
1935 if (sk->sk_state == TCP_LISTEN) {
1936 err = -EINVAL;
1937 break;
1938 }
1939
1940 spin_lock(&sk->sk_receive_queue.lock);
1941 if (sk->sk_type == SOCK_STREAM ||
1942 sk->sk_type == SOCK_SEQPACKET) {
1943 skb_queue_walk(&sk->sk_receive_queue, skb)
1944 amount += skb->len;
1945 } else {
1946 skb = skb_peek(&sk->sk_receive_queue);
1947 if (skb)
Jianjun Konge27dfce2008-11-01 21:38:31 -07001948 amount = skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 }
1950 spin_unlock(&sk->sk_receive_queue.lock);
1951 err = put_user(amount, (int __user *)arg);
1952 break;
1953 }
1954
1955 default:
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -08001956 err = -ENOIOCTLCMD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 break;
1958 }
1959 return err;
1960}
1961
1962static unsigned int unix_poll(struct file * file, struct socket *sock, poll_table *wait)
1963{
1964 struct sock *sk = sock->sk;
1965 unsigned int mask;
1966
1967 poll_wait(file, sk->sk_sleep, wait);
1968 mask = 0;
1969
1970 /* exceptional events? */
1971 if (sk->sk_err)
1972 mask |= POLLERR;
1973 if (sk->sk_shutdown == SHUTDOWN_MASK)
1974 mask |= POLLHUP;
Davide Libenzif348d702006-03-25 03:07:39 -08001975 if (sk->sk_shutdown & RCV_SHUTDOWN)
1976 mask |= POLLRDHUP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977
1978 /* readable? */
1979 if (!skb_queue_empty(&sk->sk_receive_queue) ||
1980 (sk->sk_shutdown & RCV_SHUTDOWN))
1981 mask |= POLLIN | POLLRDNORM;
1982
1983 /* Connection-based need to check for termination and startup */
1984 if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) && sk->sk_state == TCP_CLOSE)
1985 mask |= POLLHUP;
1986
1987 /*
1988 * we set writable also when the other side has shut down the
1989 * connection. This prevents stuck sockets.
1990 */
1991 if (unix_writable(sk))
1992 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
1993
1994 return mask;
1995}
1996
Rainer Weikusatec0d2152008-06-27 19:34:18 -07001997static unsigned int unix_dgram_poll(struct file *file, struct socket *sock,
1998 poll_table *wait)
Rainer Weikusat3c734192008-06-17 22:28:05 -07001999{
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002000 struct sock *sk = sock->sk, *other;
2001 unsigned int mask, writable;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002002
2003 poll_wait(file, sk->sk_sleep, wait);
Rainer Weikusat3c734192008-06-17 22:28:05 -07002004 mask = 0;
2005
2006 /* exceptional events? */
2007 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
2008 mask |= POLLERR;
2009 if (sk->sk_shutdown & RCV_SHUTDOWN)
2010 mask |= POLLRDHUP;
2011 if (sk->sk_shutdown == SHUTDOWN_MASK)
2012 mask |= POLLHUP;
2013
2014 /* readable? */
2015 if (!skb_queue_empty(&sk->sk_receive_queue) ||
2016 (sk->sk_shutdown & RCV_SHUTDOWN))
2017 mask |= POLLIN | POLLRDNORM;
2018
2019 /* Connection-based need to check for termination and startup */
2020 if (sk->sk_type == SOCK_SEQPACKET) {
2021 if (sk->sk_state == TCP_CLOSE)
2022 mask |= POLLHUP;
2023 /* connection hasn't started yet? */
2024 if (sk->sk_state == TCP_SYN_SENT)
2025 return mask;
2026 }
2027
2028 /* writable? */
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002029 writable = unix_writable(sk);
2030 if (writable) {
2031 other = unix_peer_get(sk);
2032 if (other) {
2033 if (unix_peer(other) != sk) {
2034 poll_wait(file, &unix_sk(other)->peer_wait,
2035 wait);
2036 if (unix_recvq_full(other))
2037 writable = 0;
2038 }
2039
2040 sock_put(other);
2041 }
2042 }
2043
2044 if (writable)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002045 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2046 else
2047 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
2048
Rainer Weikusat3c734192008-06-17 22:28:05 -07002049 return mask;
2050}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051
2052#ifdef CONFIG_PROC_FS
Pavel Emelyanova53eb3f2007-11-23 20:30:01 +08002053static struct sock *first_unix_socket(int *i)
2054{
2055 for (*i = 0; *i <= UNIX_HASH_SIZE; (*i)++) {
2056 if (!hlist_empty(&unix_socket_table[*i]))
2057 return __sk_head(&unix_socket_table[*i]);
2058 }
2059 return NULL;
2060}
2061
2062static struct sock *next_unix_socket(int *i, struct sock *s)
2063{
2064 struct sock *next = sk_next(s);
2065 /* More in this chain? */
2066 if (next)
2067 return next;
2068 /* Look for next non-empty chain. */
2069 for ((*i)++; *i <= UNIX_HASH_SIZE; (*i)++) {
2070 if (!hlist_empty(&unix_socket_table[*i]))
2071 return __sk_head(&unix_socket_table[*i]);
2072 }
2073 return NULL;
2074}
2075
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002076struct unix_iter_state {
Denis V. Luneve372c412007-11-19 22:31:54 -08002077 struct seq_net_private p;
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002078 int i;
2079};
Jianjun Konge27dfce2008-11-01 21:38:31 -07002080
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002081static struct sock *unix_seq_idx(struct seq_file *seq, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082{
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002083 struct unix_iter_state *iter = seq->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084 loff_t off = 0;
2085 struct sock *s;
2086
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002087 for (s = first_unix_socket(&iter->i); s; s = next_unix_socket(&iter->i, s)) {
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002088 if (sock_net(s) != seq_file_net(seq))
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002089 continue;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002090 if (off == pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 return s;
2092 ++off;
2093 }
2094 return NULL;
2095}
2096
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097static void *unix_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002098 __acquires(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002100 spin_lock(&unix_table_lock);
Joe Perchesb9f31242008-04-12 19:04:38 -07002101 return *pos ? unix_seq_idx(seq, *pos - 1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102}
2103
2104static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2105{
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002106 struct unix_iter_state *iter = seq->private;
2107 struct sock *sk = v;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 ++*pos;
2109
Joe Perchesb9f31242008-04-12 19:04:38 -07002110 if (v == SEQ_START_TOKEN)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002111 sk = first_unix_socket(&iter->i);
2112 else
2113 sk = next_unix_socket(&iter->i, sk);
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002114 while (sk && (sock_net(sk) != seq_file_net(seq)))
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002115 sk = next_unix_socket(&iter->i, sk);
2116 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117}
2118
2119static void unix_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002120 __releases(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002122 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123}
2124
2125static int unix_seq_show(struct seq_file *seq, void *v)
2126{
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002127
Joe Perchesb9f31242008-04-12 19:04:38 -07002128 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129 seq_puts(seq, "Num RefCount Protocol Flags Type St "
2130 "Inode Path\n");
2131 else {
2132 struct sock *s = v;
2133 struct unix_sock *u = unix_sk(s);
David S. Miller1c92b4e2007-05-31 13:24:26 -07002134 unix_state_lock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135
2136 seq_printf(seq, "%p: %08X %08X %08X %04X %02X %5lu",
2137 s,
2138 atomic_read(&s->sk_refcnt),
2139 0,
2140 s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0,
2141 s->sk_type,
2142 s->sk_socket ?
2143 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) :
2144 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING),
2145 sock_i_ino(s));
2146
2147 if (u->addr) {
2148 int i, len;
2149 seq_putc(seq, ' ');
2150
2151 i = 0;
2152 len = u->addr->len - sizeof(short);
2153 if (!UNIX_ABSTRACT(s))
2154 len--;
2155 else {
2156 seq_putc(seq, '@');
2157 i++;
2158 }
2159 for ( ; i < len; i++)
2160 seq_putc(seq, u->addr->name->sun_path[i]);
2161 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07002162 unix_state_unlock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002163 seq_putc(seq, '\n');
2164 }
2165
2166 return 0;
2167}
2168
Philippe De Muyter56b3d972007-07-10 23:07:31 -07002169static const struct seq_operations unix_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170 .start = unix_seq_start,
2171 .next = unix_seq_next,
2172 .stop = unix_seq_stop,
2173 .show = unix_seq_show,
2174};
2175
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176static int unix_seq_open(struct inode *inode, struct file *file)
2177{
Denis V. Luneve372c412007-11-19 22:31:54 -08002178 return seq_open_net(inode, file, &unix_seq_ops,
2179 sizeof(struct unix_iter_state));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002180}
2181
Arjan van de Venda7071d2007-02-12 00:55:36 -08002182static const struct file_operations unix_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183 .owner = THIS_MODULE,
2184 .open = unix_seq_open,
2185 .read = seq_read,
2186 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08002187 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188};
2189
2190#endif
2191
2192static struct net_proto_family unix_family_ops = {
2193 .family = PF_UNIX,
2194 .create = unix_create,
2195 .owner = THIS_MODULE,
2196};
2197
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002198
2199static int unix_net_init(struct net *net)
2200{
2201 int error = -ENOMEM;
2202
Denis V. Luneva0a53c82007-12-11 04:19:17 -08002203 net->unx.sysctl_max_dgram_qlen = 10;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002204 if (unix_sysctl_register(net))
2205 goto out;
Pavel Emelyanovd392e492007-12-01 23:44:15 +11002206
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002207#ifdef CONFIG_PROC_FS
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002208 if (!proc_net_fops_create(net, "unix", 0, &unix_seq_fops)) {
2209 unix_sysctl_unregister(net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002210 goto out;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002211 }
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002212#endif
2213 error = 0;
2214out:
2215 return 0;
2216}
2217
2218static void unix_net_exit(struct net *net)
2219{
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002220 unix_sysctl_unregister(net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002221 proc_net_remove(net, "unix");
2222}
2223
2224static struct pernet_operations unix_net_ops = {
2225 .init = unix_net_init,
2226 .exit = unix_net_exit,
2227};
2228
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229static int __init af_unix_init(void)
2230{
2231 int rc = -1;
2232 struct sk_buff *dummy_skb;
2233
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07002234 BUILD_BUG_ON(sizeof(struct unix_skb_parms) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235
2236 rc = proto_register(&unix_proto, 1);
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002237 if (rc != 0) {
2238 printk(KERN_CRIT "%s: Cannot create unix_sock SLAB cache!\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08002239 __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002240 goto out;
2241 }
2242
2243 sock_register(&unix_family_ops);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002244 register_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245out:
2246 return rc;
2247}
2248
2249static void __exit af_unix_exit(void)
2250{
2251 sock_unregister(PF_UNIX);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252 proto_unregister(&unix_proto);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002253 unregister_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254}
2255
David Woodhouse3d366962008-04-24 00:59:25 -07002256/* Earlier than device_initcall() so that other drivers invoking
2257 request_module() don't end up in a loop when modprobe tries
2258 to use a UNIX socket. But later than subsys_initcall() because
2259 we depend on stuff initialised there */
2260fs_initcall(af_unix_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002261module_exit(af_unix_exit);
2262
2263MODULE_LICENSE("GPL");
2264MODULE_ALIAS_NETPROTO(PF_UNIX);