blob: bb7e8ba821f44014d65f3669e017b8c9cca54f53 [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
wangweidong5cc208b2013-12-06 18:03:36 +080083#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
84
Linus Torvalds1da177e2005-04-16 15:20:36 -070085#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070086#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070087#include <linux/signal.h>
88#include <linux/sched.h>
89#include <linux/errno.h>
90#include <linux/string.h>
91#include <linux/stat.h>
92#include <linux/dcache.h>
93#include <linux/namei.h>
94#include <linux/socket.h>
95#include <linux/un.h>
96#include <linux/fcntl.h>
97#include <linux/termios.h>
98#include <linux/sockios.h>
99#include <linux/net.h>
100#include <linux/in.h>
101#include <linux/fs.h>
102#include <linux/slab.h>
103#include <asm/uaccess.h>
104#include <linux/skbuff.h>
105#include <linux/netdevice.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200106#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#include <net/sock.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -0700108#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109#include <net/af_unix.h>
110#include <linux/proc_fs.h>
111#include <linux/seq_file.h>
112#include <net/scm.h>
113#include <linux/init.h>
114#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115#include <linux/rtnetlink.h>
116#include <linux/mount.h>
117#include <net/checksum.h>
118#include <linux/security.h>
Colin Cross2b15af62013-05-06 23:50:21 +0000119#include <linux/freezer.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
Eric Dumazet7123aaa2012-06-08 05:03:21 +0000121struct hlist_head unix_socket_table[2 * UNIX_HASH_SIZE];
Pavel Emelyanovfa7ff562011-12-15 02:44:03 +0000122EXPORT_SYMBOL_GPL(unix_socket_table);
123DEFINE_SPINLOCK(unix_table_lock);
124EXPORT_SYMBOL_GPL(unix_table_lock);
Eric Dumazet518de9b2010-10-26 14:22:44 -0700125static atomic_long_t unix_nr_socks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
Eric Dumazet7123aaa2012-06-08 05:03:21 +0000128static struct hlist_head *unix_sockets_unbound(void *addr)
129{
130 unsigned long hash = (unsigned long)addr;
131
132 hash ^= hash >> 16;
133 hash ^= hash >> 8;
134 hash %= UNIX_HASH_SIZE;
135 return &unix_socket_table[UNIX_HASH_SIZE + hash];
136}
137
138#define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash < UNIX_HASH_SIZE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700140#ifdef CONFIG_SECURITY_NETWORK
Catherine Zhangdc49c1f2006-08-02 14:12:06 -0700141static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb)
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700142{
Catherine Zhangdc49c1f2006-08-02 14:12:06 -0700143 memcpy(UNIXSID(skb), &scm->secid, sizeof(u32));
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700144}
145
146static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb)
147{
Catherine Zhangdc49c1f2006-08-02 14:12:06 -0700148 scm->secid = *UNIXSID(skb);
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700149}
150#else
Catherine Zhangdc49c1f2006-08-02 14:12:06 -0700151static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb)
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700152{ }
153
154static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb)
155{ }
156#endif /* CONFIG_SECURITY_NETWORK */
157
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158/*
159 * SMP locking strategy:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800160 * hash table is protected with spinlock unix_table_lock
Stephen Hemminger663717f2010-02-18 14:12:06 -0800161 * each socket state is protected by separate spin lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 */
163
Eric Dumazet95c96172012-04-15 05:58:06 +0000164static inline unsigned int unix_hash_fold(__wsum n)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165{
Anton Blanchard0a134042014-03-05 14:29:58 +1100166 unsigned int hash = (__force unsigned int)csum_fold(n);
Eric Dumazet95c96172012-04-15 05:58:06 +0000167
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 hash ^= hash>>8;
169 return hash&(UNIX_HASH_SIZE-1);
170}
171
172#define unix_peer(sk) (unix_sk(sk)->peer)
173
174static inline int unix_our_peer(struct sock *sk, struct sock *osk)
175{
176 return unix_peer(osk) == sk;
177}
178
179static inline int unix_may_send(struct sock *sk, struct sock *osk)
180{
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800181 return unix_peer(osk) == NULL || unix_our_peer(sk, osk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182}
183
Rainer Weikusat3c734192008-06-17 22:28:05 -0700184static inline int unix_recvq_full(struct sock const *sk)
185{
186 return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog;
187}
188
Pavel Emelyanovfa7ff562011-12-15 02:44:03 +0000189struct sock *unix_peer_get(struct sock *s)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190{
191 struct sock *peer;
192
David S. Miller1c92b4e2007-05-31 13:24:26 -0700193 unix_state_lock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 peer = unix_peer(s);
195 if (peer)
196 sock_hold(peer);
David S. Miller1c92b4e2007-05-31 13:24:26 -0700197 unix_state_unlock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 return peer;
199}
Pavel Emelyanovfa7ff562011-12-15 02:44:03 +0000200EXPORT_SYMBOL_GPL(unix_peer_get);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
202static inline void unix_release_addr(struct unix_address *addr)
203{
204 if (atomic_dec_and_test(&addr->refcnt))
205 kfree(addr);
206}
207
208/*
209 * Check unix socket name:
210 * - should be not zero length.
211 * - if started by not zero, should be NULL terminated (FS object)
212 * - if started by zero, it is abstract name.
213 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900214
Eric Dumazet95c96172012-04-15 05:58:06 +0000215static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216{
217 if (len <= sizeof(short) || len > sizeof(*sunaddr))
218 return -EINVAL;
219 if (!sunaddr || sunaddr->sun_family != AF_UNIX)
220 return -EINVAL;
221 if (sunaddr->sun_path[0]) {
222 /*
223 * This may look like an off by one error but it is a bit more
224 * subtle. 108 is the longest valid AF_UNIX path for a binding.
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300225 * sun_path[108] doesn't as such exist. However in kernel space
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 * we are guaranteed that it is a valid memory location in our
227 * kernel address buffer.
228 */
Jianjun Konge27dfce2008-11-01 21:38:31 -0700229 ((char *)sunaddr)[len] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 len = strlen(sunaddr->sun_path)+1+sizeof(short);
231 return len;
232 }
233
Joe Perches07f07572008-11-19 15:44:53 -0800234 *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 return len;
236}
237
238static void __unix_remove_socket(struct sock *sk)
239{
240 sk_del_node_init(sk);
241}
242
243static void __unix_insert_socket(struct hlist_head *list, struct sock *sk)
244{
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700245 WARN_ON(!sk_unhashed(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 sk_add_node(sk, list);
247}
248
249static inline void unix_remove_socket(struct sock *sk)
250{
David S. Millerfbe9cc42005-12-13 23:26:29 -0800251 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 __unix_remove_socket(sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800253 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254}
255
256static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk)
257{
David S. Millerfbe9cc42005-12-13 23:26:29 -0800258 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 __unix_insert_socket(list, sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800260 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261}
262
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800263static struct sock *__unix_find_socket_byname(struct net *net,
264 struct sockaddr_un *sunname,
Eric Dumazet95c96172012-04-15 05:58:06 +0000265 int len, int type, unsigned int hash)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266{
267 struct sock *s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
Sasha Levinb67bfe02013-02-27 17:06:00 -0800269 sk_for_each(s, &unix_socket_table[hash ^ type]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 struct unix_sock *u = unix_sk(s);
271
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900272 if (!net_eq(sock_net(s), net))
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800273 continue;
274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 if (u->addr->len == len &&
276 !memcmp(u->addr->name, sunname, len))
277 goto found;
278 }
279 s = NULL;
280found:
281 return s;
282}
283
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800284static inline struct sock *unix_find_socket_byname(struct net *net,
285 struct sockaddr_un *sunname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 int len, int type,
Eric Dumazet95c96172012-04-15 05:58:06 +0000287 unsigned int hash)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288{
289 struct sock *s;
290
David S. Millerfbe9cc42005-12-13 23:26:29 -0800291 spin_lock(&unix_table_lock);
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800292 s = __unix_find_socket_byname(net, sunname, len, type, hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 if (s)
294 sock_hold(s);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800295 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 return s;
297}
298
Eric W. Biederman6616f782010-06-13 03:35:48 +0000299static struct sock *unix_find_socket_byinode(struct inode *i)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300{
301 struct sock *s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
David S. Millerfbe9cc42005-12-13 23:26:29 -0800303 spin_lock(&unix_table_lock);
Sasha Levinb67bfe02013-02-27 17:06:00 -0800304 sk_for_each(s,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) {
Al Viro40ffe672012-03-14 21:54:32 -0400306 struct dentry *dentry = unix_sk(s)->path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800308 if (dentry && dentry->d_inode == i) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 sock_hold(s);
310 goto found;
311 }
312 }
313 s = NULL;
314found:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800315 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 return s;
317}
318
319static inline int unix_writable(struct sock *sk)
320{
321 return (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf;
322}
323
324static void unix_write_space(struct sock *sk)
325{
Eric Dumazet43815482010-04-29 11:01:49 +0000326 struct socket_wq *wq;
327
328 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 if (unix_writable(sk)) {
Eric Dumazet43815482010-04-29 11:01:49 +0000330 wq = rcu_dereference(sk->sk_wq);
331 if (wq_has_sleeper(wq))
Eric Dumazet67426b72010-10-29 20:44:44 +0000332 wake_up_interruptible_sync_poll(&wq->wait,
333 POLLOUT | POLLWRNORM | POLLWRBAND);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +0800334 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 }
Eric Dumazet43815482010-04-29 11:01:49 +0000336 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337}
338
339/* When dgram socket disconnects (or changes its peer), we clear its receive
340 * queue of packets arrived from previous peer. First, it allows to do
341 * flow control based only on wmem_alloc; second, sk connected to peer
342 * may receive messages only from that peer. */
343static void unix_dgram_disconnected(struct sock *sk, struct sock *other)
344{
David S. Millerb03efcf2005-07-08 14:57:23 -0700345 if (!skb_queue_empty(&sk->sk_receive_queue)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 skb_queue_purge(&sk->sk_receive_queue);
347 wake_up_interruptible_all(&unix_sk(sk)->peer_wait);
348
349 /* If one link of bidirectional dgram pipe is disconnected,
350 * we signal error. Messages are lost. Do not make this,
351 * when peer was not connected to us.
352 */
353 if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) {
354 other->sk_err = ECONNRESET;
355 other->sk_error_report(other);
356 }
357 }
358}
359
360static void unix_sock_destructor(struct sock *sk)
361{
362 struct unix_sock *u = unix_sk(sk);
363
364 skb_queue_purge(&sk->sk_receive_queue);
365
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700366 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
367 WARN_ON(!sk_unhashed(sk));
368 WARN_ON(sk->sk_socket);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 if (!sock_flag(sk, SOCK_DEAD)) {
wangweidong5cc208b2013-12-06 18:03:36 +0800370 pr_info("Attempt to release alive unix socket: %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 return;
372 }
373
374 if (u->addr)
375 unix_release_addr(u->addr);
376
Eric Dumazet518de9b2010-10-26 14:22:44 -0700377 atomic_long_dec(&unix_nr_socks);
David S. Miller6f756a82008-11-23 17:34:03 -0800378 local_bh_disable();
Eric Dumazeta8076d82008-11-17 02:38:49 -0800379 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
David S. Miller6f756a82008-11-23 17:34:03 -0800380 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381#ifdef UNIX_REFCNT_DEBUG
wangweidong5cc208b2013-12-06 18:03:36 +0800382 pr_debug("UNIX %p is destroyed, %ld are still alive.\n", sk,
Eric Dumazet518de9b2010-10-26 14:22:44 -0700383 atomic_long_read(&unix_nr_socks));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384#endif
385}
386
Paul Mooreded34e02013-03-25 03:18:33 +0000387static void unix_release_sock(struct sock *sk, int embrion)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388{
389 struct unix_sock *u = unix_sk(sk);
Al Viro40ffe672012-03-14 21:54:32 -0400390 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 struct sock *skpair;
392 struct sk_buff *skb;
393 int state;
394
395 unix_remove_socket(sk);
396
397 /* Clear state */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700398 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 sock_orphan(sk);
400 sk->sk_shutdown = SHUTDOWN_MASK;
Al Viro40ffe672012-03-14 21:54:32 -0400401 path = u->path;
402 u->path.dentry = NULL;
403 u->path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 state = sk->sk_state;
405 sk->sk_state = TCP_CLOSE;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700406 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
408 wake_up_interruptible_all(&u->peer_wait);
409
Jianjun Konge27dfce2008-11-01 21:38:31 -0700410 skpair = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411
Jianjun Konge27dfce2008-11-01 21:38:31 -0700412 if (skpair != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) {
David S. Miller1c92b4e2007-05-31 13:24:26 -0700414 unix_state_lock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 /* No more writes */
416 skpair->sk_shutdown = SHUTDOWN_MASK;
417 if (!skb_queue_empty(&sk->sk_receive_queue) || embrion)
418 skpair->sk_err = ECONNRESET;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700419 unix_state_unlock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 skpair->sk_state_change(skpair);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +0800421 sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 }
423 sock_put(skpair); /* It may now die */
424 unix_peer(sk) = NULL;
425 }
426
427 /* Try to flush out this socket. Throw out buffers at least */
428
429 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
Jianjun Konge27dfce2008-11-01 21:38:31 -0700430 if (state == TCP_LISTEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 unix_release_sock(skb->sk, 1);
432 /* passed fds are erased in the kfree_skb hook */
433 kfree_skb(skb);
434 }
435
Al Viro40ffe672012-03-14 21:54:32 -0400436 if (path.dentry)
437 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438
439 sock_put(sk);
440
441 /* ---- Socket is dead now and most probably destroyed ---- */
442
443 /*
Alan Coxe04dae82012-09-17 00:52:41 +0000444 * Fixme: BSD difference: In BSD all sockets connected to us get
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 * ECONNRESET and we die on the spot. In Linux we behave
446 * like files and pipes do and wait for the last
447 * dereference.
448 *
449 * Can't we simply set sock->err?
450 *
451 * What the above comment does talk about? --ANK(980817)
452 */
453
Pavel Emelyanov9305cfa2007-11-10 22:06:01 -0800454 if (unix_tot_inflight)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900455 unix_gc(); /* Garbage collect fds */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456}
457
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000458static void init_peercred(struct sock *sk)
459{
460 put_pid(sk->sk_peer_pid);
461 if (sk->sk_peer_cred)
462 put_cred(sk->sk_peer_cred);
463 sk->sk_peer_pid = get_pid(task_tgid(current));
464 sk->sk_peer_cred = get_current_cred();
465}
466
467static void copy_peercred(struct sock *sk, struct sock *peersk)
468{
469 put_pid(sk->sk_peer_pid);
470 if (sk->sk_peer_cred)
471 put_cred(sk->sk_peer_cred);
472 sk->sk_peer_pid = get_pid(peersk->sk_peer_pid);
473 sk->sk_peer_cred = get_cred(peersk->sk_peer_cred);
474}
475
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476static int unix_listen(struct socket *sock, int backlog)
477{
478 int err;
479 struct sock *sk = sock->sk;
480 struct unix_sock *u = unix_sk(sk);
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000481 struct pid *old_pid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482
483 err = -EOPNOTSUPP;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800484 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
485 goto out; /* Only stream/seqpacket sockets accept */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 err = -EINVAL;
487 if (!u->addr)
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800488 goto out; /* No listens on an unbound socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700489 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN)
491 goto out_unlock;
492 if (backlog > sk->sk_max_ack_backlog)
493 wake_up_interruptible_all(&u->peer_wait);
494 sk->sk_max_ack_backlog = backlog;
495 sk->sk_state = TCP_LISTEN;
496 /* set credentials so connect can copy them */
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000497 init_peercred(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 err = 0;
499
500out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -0700501 unix_state_unlock(sk);
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000502 put_pid(old_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503out:
504 return err;
505}
506
507static int unix_release(struct socket *);
508static int unix_bind(struct socket *, struct sockaddr *, int);
509static int unix_stream_connect(struct socket *, struct sockaddr *,
510 int addr_len, int flags);
511static int unix_socketpair(struct socket *, struct socket *);
512static int unix_accept(struct socket *, struct socket *, int);
513static int unix_getname(struct socket *, struct sockaddr *, int *, int);
514static unsigned int unix_poll(struct file *, struct socket *, poll_table *);
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700515static unsigned int unix_dgram_poll(struct file *, struct socket *,
516 poll_table *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517static int unix_ioctl(struct socket *, unsigned int, unsigned long);
518static int unix_shutdown(struct socket *, int);
519static int unix_stream_sendmsg(struct kiocb *, struct socket *,
520 struct msghdr *, size_t);
521static int unix_stream_recvmsg(struct kiocb *, struct socket *,
522 struct msghdr *, size_t, int);
523static int unix_dgram_sendmsg(struct kiocb *, struct socket *,
524 struct msghdr *, size_t);
525static int unix_dgram_recvmsg(struct kiocb *, struct socket *,
526 struct msghdr *, size_t, int);
527static int unix_dgram_connect(struct socket *, struct sockaddr *,
528 int, int);
529static int unix_seqpacket_sendmsg(struct kiocb *, struct socket *,
530 struct msghdr *, size_t);
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +0000531static int unix_seqpacket_recvmsg(struct kiocb *, struct socket *,
532 struct msghdr *, size_t, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
Sasha Levin12663bf2013-12-07 17:26:27 -0500534static int unix_set_peek_off(struct sock *sk, int val)
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000535{
536 struct unix_sock *u = unix_sk(sk);
537
Sasha Levin12663bf2013-12-07 17:26:27 -0500538 if (mutex_lock_interruptible(&u->readlock))
539 return -EINTR;
540
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000541 sk->sk_peek_off = val;
542 mutex_unlock(&u->readlock);
Sasha Levin12663bf2013-12-07 17:26:27 -0500543
544 return 0;
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000545}
546
547
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800548static const struct proto_ops unix_stream_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 .family = PF_UNIX,
550 .owner = THIS_MODULE,
551 .release = unix_release,
552 .bind = unix_bind,
553 .connect = unix_stream_connect,
554 .socketpair = unix_socketpair,
555 .accept = unix_accept,
556 .getname = unix_getname,
557 .poll = unix_poll,
558 .ioctl = unix_ioctl,
559 .listen = unix_listen,
560 .shutdown = unix_shutdown,
561 .setsockopt = sock_no_setsockopt,
562 .getsockopt = sock_no_getsockopt,
563 .sendmsg = unix_stream_sendmsg,
564 .recvmsg = unix_stream_recvmsg,
565 .mmap = sock_no_mmap,
566 .sendpage = sock_no_sendpage,
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +0000567 .set_peek_off = unix_set_peek_off,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568};
569
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800570static const struct proto_ops unix_dgram_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 .family = PF_UNIX,
572 .owner = THIS_MODULE,
573 .release = unix_release,
574 .bind = unix_bind,
575 .connect = unix_dgram_connect,
576 .socketpair = unix_socketpair,
577 .accept = sock_no_accept,
578 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700579 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 .ioctl = unix_ioctl,
581 .listen = sock_no_listen,
582 .shutdown = unix_shutdown,
583 .setsockopt = sock_no_setsockopt,
584 .getsockopt = sock_no_getsockopt,
585 .sendmsg = unix_dgram_sendmsg,
586 .recvmsg = unix_dgram_recvmsg,
587 .mmap = sock_no_mmap,
588 .sendpage = sock_no_sendpage,
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000589 .set_peek_off = unix_set_peek_off,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590};
591
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800592static const struct proto_ops unix_seqpacket_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 .family = PF_UNIX,
594 .owner = THIS_MODULE,
595 .release = unix_release,
596 .bind = unix_bind,
597 .connect = unix_stream_connect,
598 .socketpair = unix_socketpair,
599 .accept = unix_accept,
600 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700601 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 .ioctl = unix_ioctl,
603 .listen = unix_listen,
604 .shutdown = unix_shutdown,
605 .setsockopt = sock_no_setsockopt,
606 .getsockopt = sock_no_getsockopt,
607 .sendmsg = unix_seqpacket_sendmsg,
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +0000608 .recvmsg = unix_seqpacket_recvmsg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 .mmap = sock_no_mmap,
610 .sendpage = sock_no_sendpage,
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000611 .set_peek_off = unix_set_peek_off,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612};
613
614static struct proto unix_proto = {
Eric Dumazet248969a2008-11-17 00:00:30 -0800615 .name = "UNIX",
616 .owner = THIS_MODULE,
Eric Dumazet248969a2008-11-17 00:00:30 -0800617 .obj_size = sizeof(struct unix_sock),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618};
619
Ingo Molnara09785a2006-07-03 00:25:12 -0700620/*
621 * AF_UNIX sockets do not interact with hardware, hence they
622 * dont trigger interrupts - so it's safe for them to have
623 * bh-unsafe locking for their sk_receive_queue.lock. Split off
624 * this special lock-class by reinitializing the spinlock key:
625 */
626static struct lock_class_key af_unix_sk_receive_queue_lock_key;
627
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800628static struct sock *unix_create1(struct net *net, struct socket *sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629{
630 struct sock *sk = NULL;
631 struct unix_sock *u;
632
Eric Dumazet518de9b2010-10-26 14:22:44 -0700633 atomic_long_inc(&unix_nr_socks);
634 if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files())
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 goto out;
636
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700637 sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 if (!sk)
639 goto out;
640
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800641 sock_init_data(sock, sk);
Ingo Molnara09785a2006-07-03 00:25:12 -0700642 lockdep_set_class(&sk->sk_receive_queue.lock,
643 &af_unix_sk_receive_queue_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644
645 sk->sk_write_space = unix_write_space;
Denis V. Luneva0a53c82007-12-11 04:19:17 -0800646 sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 sk->sk_destruct = unix_sock_destructor;
648 u = unix_sk(sk);
Al Viro40ffe672012-03-14 21:54:32 -0400649 u->path.dentry = NULL;
650 u->path.mnt = NULL;
Benjamin LaHaisefd19f322006-01-03 14:10:46 -0800651 spin_lock_init(&u->lock);
Al Viro516e0cc2008-07-26 00:39:17 -0400652 atomic_long_set(&u->inflight, 0);
Miklos Szeredi1fd05ba2007-07-11 14:22:39 -0700653 INIT_LIST_HEAD(&u->link);
Ingo Molnar57b47a52006-03-20 22:35:41 -0800654 mutex_init(&u->readlock); /* single task reading lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 init_waitqueue_head(&u->peer_wait);
Eric Dumazet7123aaa2012-06-08 05:03:21 +0000656 unix_insert_socket(unix_sockets_unbound(sk), sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657out:
Pavel Emelyanov284b3272007-11-10 22:08:30 -0800658 if (sk == NULL)
Eric Dumazet518de9b2010-10-26 14:22:44 -0700659 atomic_long_dec(&unix_nr_socks);
Eric Dumazet920de802008-11-24 00:09:29 -0800660 else {
661 local_bh_disable();
Eric Dumazeta8076d82008-11-17 02:38:49 -0800662 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet920de802008-11-24 00:09:29 -0800663 local_bh_enable();
664 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 return sk;
666}
667
Eric Paris3f378b62009-11-05 22:18:14 -0800668static int unix_create(struct net *net, struct socket *sock, int protocol,
669 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670{
671 if (protocol && protocol != PF_UNIX)
672 return -EPROTONOSUPPORT;
673
674 sock->state = SS_UNCONNECTED;
675
676 switch (sock->type) {
677 case SOCK_STREAM:
678 sock->ops = &unix_stream_ops;
679 break;
680 /*
681 * Believe it or not BSD has AF_UNIX, SOCK_RAW though
682 * nothing uses it.
683 */
684 case SOCK_RAW:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700685 sock->type = SOCK_DGRAM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 case SOCK_DGRAM:
687 sock->ops = &unix_dgram_ops;
688 break;
689 case SOCK_SEQPACKET:
690 sock->ops = &unix_seqpacket_ops;
691 break;
692 default:
693 return -ESOCKTNOSUPPORT;
694 }
695
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700696 return unix_create1(net, sock) ? 0 : -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697}
698
699static int unix_release(struct socket *sock)
700{
701 struct sock *sk = sock->sk;
702
703 if (!sk)
704 return 0;
705
Paul Mooreded34e02013-03-25 03:18:33 +0000706 unix_release_sock(sk, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 sock->sk = NULL;
708
Paul Mooreded34e02013-03-25 03:18:33 +0000709 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710}
711
712static int unix_autobind(struct socket *sock)
713{
714 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900715 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 struct unix_sock *u = unix_sk(sk);
717 static u32 ordernum = 1;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800718 struct unix_address *addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 int err;
Tetsuo Handa8df73ff2010-09-04 01:34:28 +0000720 unsigned int retries = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
Sasha Levin37ab4fa2013-12-13 10:54:22 -0500722 err = mutex_lock_interruptible(&u->readlock);
723 if (err)
724 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725
726 err = 0;
727 if (u->addr)
728 goto out;
729
730 err = -ENOMEM;
Panagiotis Issaris0da974f2006-07-21 14:51:30 -0700731 addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 if (!addr)
733 goto out;
734
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 addr->name->sun_family = AF_UNIX;
736 atomic_set(&addr->refcnt, 1);
737
738retry:
739 addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short);
Joe Perches07f07572008-11-19 15:44:53 -0800740 addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
David S. Millerfbe9cc42005-12-13 23:26:29 -0800742 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 ordernum = (ordernum+1)&0xFFFFF;
744
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800745 if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 addr->hash)) {
David S. Millerfbe9cc42005-12-13 23:26:29 -0800747 spin_unlock(&unix_table_lock);
Tetsuo Handa8df73ff2010-09-04 01:34:28 +0000748 /*
749 * __unix_find_socket_byname() may take long time if many names
750 * are already in use.
751 */
752 cond_resched();
753 /* Give up if all names seems to be in use. */
754 if (retries++ == 0xFFFFF) {
755 err = -ENOSPC;
756 kfree(addr);
757 goto out;
758 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 goto retry;
760 }
761 addr->hash ^= sk->sk_type;
762
763 __unix_remove_socket(sk);
764 u->addr = addr;
765 __unix_insert_socket(&unix_socket_table[addr->hash], sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800766 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 err = 0;
768
Ingo Molnar57b47a52006-03-20 22:35:41 -0800769out: mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 return err;
771}
772
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800773static struct sock *unix_find_other(struct net *net,
774 struct sockaddr_un *sunname, int len,
Eric Dumazet95c96172012-04-15 05:58:06 +0000775 int type, unsigned int hash, int *error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776{
777 struct sock *u;
Al Viro421748e2008-08-02 01:04:36 -0400778 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 int err = 0;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900780
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 if (sunname->sun_path[0]) {
Al Viro421748e2008-08-02 01:04:36 -0400782 struct inode *inode;
783 err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 if (err)
785 goto fail;
Al Viro421748e2008-08-02 01:04:36 -0400786 inode = path.dentry->d_inode;
787 err = inode_permission(inode, MAY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 if (err)
789 goto put_fail;
790
791 err = -ECONNREFUSED;
Al Viro421748e2008-08-02 01:04:36 -0400792 if (!S_ISSOCK(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 goto put_fail;
Eric W. Biederman6616f782010-06-13 03:35:48 +0000794 u = unix_find_socket_byinode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 if (!u)
796 goto put_fail;
797
798 if (u->sk_type == type)
Al Viro68ac1232012-03-15 08:21:57 -0400799 touch_atime(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
Al Viro421748e2008-08-02 01:04:36 -0400801 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
Jianjun Konge27dfce2008-11-01 21:38:31 -0700803 err = -EPROTOTYPE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 if (u->sk_type != type) {
805 sock_put(u);
806 goto fail;
807 }
808 } else {
809 err = -ECONNREFUSED;
Jianjun Konge27dfce2008-11-01 21:38:31 -0700810 u = unix_find_socket_byname(net, sunname, len, type, hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 if (u) {
812 struct dentry *dentry;
Al Viro40ffe672012-03-14 21:54:32 -0400813 dentry = unix_sk(u)->path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 if (dentry)
Al Viro68ac1232012-03-15 08:21:57 -0400815 touch_atime(&unix_sk(u)->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 } else
817 goto fail;
818 }
819 return u;
820
821put_fail:
Al Viro421748e2008-08-02 01:04:36 -0400822 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823fail:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700824 *error = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 return NULL;
826}
827
Al Virofaf02012012-07-20 02:37:29 +0400828static int unix_mknod(const char *sun_path, umode_t mode, struct path *res)
829{
830 struct dentry *dentry;
831 struct path path;
832 int err = 0;
833 /*
834 * Get the parent directory, calculate the hash for last
835 * component.
836 */
837 dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0);
838 err = PTR_ERR(dentry);
839 if (IS_ERR(dentry))
840 return err;
841
842 /*
843 * All right, let's create it.
844 */
845 err = security_path_mknod(&path, dentry, mode, 0);
846 if (!err) {
847 err = vfs_mknod(path.dentry->d_inode, dentry, mode, 0);
848 if (!err) {
849 res->mnt = mntget(path.mnt);
850 res->dentry = dget(dentry);
851 }
852 }
853 done_path_create(&path, dentry);
854 return err;
855}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
857static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
858{
859 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900860 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700862 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Al Virodae6ad82011-06-26 11:50:15 -0400863 char *sun_path = sunaddr->sun_path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 int err;
Eric Dumazet95c96172012-04-15 05:58:06 +0000865 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 struct unix_address *addr;
867 struct hlist_head *list;
868
869 err = -EINVAL;
870 if (sunaddr->sun_family != AF_UNIX)
871 goto out;
872
Jianjun Konge27dfce2008-11-01 21:38:31 -0700873 if (addr_len == sizeof(short)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 err = unix_autobind(sock);
875 goto out;
876 }
877
878 err = unix_mkname(sunaddr, addr_len, &hash);
879 if (err < 0)
880 goto out;
881 addr_len = err;
882
Sasha Levin37ab4fa2013-12-13 10:54:22 -0500883 err = mutex_lock_interruptible(&u->readlock);
884 if (err)
885 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886
887 err = -EINVAL;
888 if (u->addr)
889 goto out_up;
890
891 err = -ENOMEM;
892 addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL);
893 if (!addr)
894 goto out_up;
895
896 memcpy(addr->name, sunaddr, addr_len);
897 addr->len = addr_len;
898 addr->hash = hash ^ sk->sk_type;
899 atomic_set(&addr->refcnt, 1);
900
Al Virodae6ad82011-06-26 11:50:15 -0400901 if (sun_path[0]) {
Al Virofaf02012012-07-20 02:37:29 +0400902 struct path path;
903 umode_t mode = S_IFSOCK |
Al Viroce3b0f82009-03-29 19:08:22 -0400904 (SOCK_INODE(sock)->i_mode & ~current_umask());
Al Virofaf02012012-07-20 02:37:29 +0400905 err = unix_mknod(sun_path, mode, &path);
906 if (err) {
907 if (err == -EEXIST)
908 err = -EADDRINUSE;
909 unix_release_addr(addr);
910 goto out_up;
911 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 addr->hash = UNIX_HASH_SIZE;
Al Virofaf02012012-07-20 02:37:29 +0400913 hash = path.dentry->d_inode->i_ino & (UNIX_HASH_SIZE-1);
914 spin_lock(&unix_table_lock);
915 u->path = path;
916 list = &unix_socket_table[hash];
917 } else {
918 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 err = -EADDRINUSE;
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800920 if (__unix_find_socket_byname(net, sunaddr, addr_len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 sk->sk_type, hash)) {
922 unix_release_addr(addr);
923 goto out_unlock;
924 }
925
926 list = &unix_socket_table[addr->hash];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 }
928
929 err = 0;
930 __unix_remove_socket(sk);
931 u->addr = addr;
932 __unix_insert_socket(list, sk);
933
934out_unlock:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800935 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936out_up:
Ingo Molnar57b47a52006-03-20 22:35:41 -0800937 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938out:
939 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940}
941
David S. Miller278a3de2007-05-31 15:19:20 -0700942static void unix_state_double_lock(struct sock *sk1, struct sock *sk2)
943{
944 if (unlikely(sk1 == sk2) || !sk2) {
945 unix_state_lock(sk1);
946 return;
947 }
948 if (sk1 < sk2) {
949 unix_state_lock(sk1);
950 unix_state_lock_nested(sk2);
951 } else {
952 unix_state_lock(sk2);
953 unix_state_lock_nested(sk1);
954 }
955}
956
957static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2)
958{
959 if (unlikely(sk1 == sk2) || !sk2) {
960 unix_state_unlock(sk1);
961 return;
962 }
963 unix_state_unlock(sk1);
964 unix_state_unlock(sk2);
965}
966
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr,
968 int alen, int flags)
969{
970 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900971 struct net *net = sock_net(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700972 struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 struct sock *other;
Eric Dumazet95c96172012-04-15 05:58:06 +0000974 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 int err;
976
977 if (addr->sa_family != AF_UNSPEC) {
978 err = unix_mkname(sunaddr, alen, &hash);
979 if (err < 0)
980 goto out;
981 alen = err;
982
983 if (test_bit(SOCK_PASSCRED, &sock->flags) &&
984 !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0)
985 goto out;
986
David S. Miller278a3de2007-05-31 15:19:20 -0700987restart:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700988 other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 if (!other)
990 goto out;
991
David S. Miller278a3de2007-05-31 15:19:20 -0700992 unix_state_double_lock(sk, other);
993
994 /* Apparently VFS overslept socket death. Retry. */
995 if (sock_flag(other, SOCK_DEAD)) {
996 unix_state_double_unlock(sk, other);
997 sock_put(other);
998 goto restart;
999 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000
1001 err = -EPERM;
1002 if (!unix_may_send(sk, other))
1003 goto out_unlock;
1004
1005 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1006 if (err)
1007 goto out_unlock;
1008
1009 } else {
1010 /*
1011 * 1003.1g breaking connected state with AF_UNSPEC
1012 */
1013 other = NULL;
David S. Miller278a3de2007-05-31 15:19:20 -07001014 unix_state_double_lock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 }
1016
1017 /*
1018 * If it was connected, reconnect.
1019 */
1020 if (unix_peer(sk)) {
1021 struct sock *old_peer = unix_peer(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001022 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -07001023 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024
1025 if (other != old_peer)
1026 unix_dgram_disconnected(sk, old_peer);
1027 sock_put(old_peer);
1028 } else {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001029 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -07001030 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 }
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001032 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033
1034out_unlock:
David S. Miller278a3de2007-05-31 15:19:20 -07001035 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 sock_put(other);
1037out:
1038 return err;
1039}
1040
1041static long unix_wait_for_peer(struct sock *other, long timeo)
1042{
1043 struct unix_sock *u = unix_sk(other);
1044 int sched;
1045 DEFINE_WAIT(wait);
1046
1047 prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE);
1048
1049 sched = !sock_flag(other, SOCK_DEAD) &&
1050 !(other->sk_shutdown & RCV_SHUTDOWN) &&
Rainer Weikusat3c734192008-06-17 22:28:05 -07001051 unix_recvq_full(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052
David S. Miller1c92b4e2007-05-31 13:24:26 -07001053 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054
1055 if (sched)
1056 timeo = schedule_timeout(timeo);
1057
1058 finish_wait(&u->peer_wait, &wait);
1059 return timeo;
1060}
1061
1062static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr,
1063 int addr_len, int flags)
1064{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001065 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001067 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 struct unix_sock *u = unix_sk(sk), *newu, *otheru;
1069 struct sock *newsk = NULL;
1070 struct sock *other = NULL;
1071 struct sk_buff *skb = NULL;
Eric Dumazet95c96172012-04-15 05:58:06 +00001072 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 int st;
1074 int err;
1075 long timeo;
1076
1077 err = unix_mkname(sunaddr, addr_len, &hash);
1078 if (err < 0)
1079 goto out;
1080 addr_len = err;
1081
Joe Perchesf64f9e72009-11-29 16:55:45 -08001082 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr &&
1083 (err = unix_autobind(sock)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 goto out;
1085
1086 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1087
1088 /* First of all allocate resources.
1089 If we will make it after state is locked,
1090 we will have to recheck all again in any case.
1091 */
1092
1093 err = -ENOMEM;
1094
1095 /* create new sock for complete connection */
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001096 newsk = unix_create1(sock_net(sk), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 if (newsk == NULL)
1098 goto out;
1099
1100 /* Allocate skb for sending to listening sock */
1101 skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL);
1102 if (skb == NULL)
1103 goto out;
1104
1105restart:
1106 /* Find listening sock. */
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001107 other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 if (!other)
1109 goto out;
1110
1111 /* Latch state of peer */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001112 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113
1114 /* Apparently VFS overslept socket death. Retry. */
1115 if (sock_flag(other, SOCK_DEAD)) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001116 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 sock_put(other);
1118 goto restart;
1119 }
1120
1121 err = -ECONNREFUSED;
1122 if (other->sk_state != TCP_LISTEN)
1123 goto out_unlock;
Tomoki Sekiyama77238f22009-10-18 23:17:37 -07001124 if (other->sk_shutdown & RCV_SHUTDOWN)
1125 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126
Rainer Weikusat3c734192008-06-17 22:28:05 -07001127 if (unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 err = -EAGAIN;
1129 if (!timeo)
1130 goto out_unlock;
1131
1132 timeo = unix_wait_for_peer(other, timeo);
1133
1134 err = sock_intr_errno(timeo);
1135 if (signal_pending(current))
1136 goto out;
1137 sock_put(other);
1138 goto restart;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001139 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140
1141 /* Latch our state.
1142
Daniel Balutae5537bf2011-03-14 15:25:33 -07001143 It is tricky place. We need to grab our state lock and cannot
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 drop lock on peer. It is dangerous because deadlock is
1145 possible. Connect to self case and simultaneous
1146 attempt to connect are eliminated by checking socket
1147 state. other is TCP_LISTEN, if sk is TCP_LISTEN we
1148 check this before attempt to grab lock.
1149
1150 Well, and we have to recheck the state after socket locked.
1151 */
1152 st = sk->sk_state;
1153
1154 switch (st) {
1155 case TCP_CLOSE:
1156 /* This is ok... continue with connect */
1157 break;
1158 case TCP_ESTABLISHED:
1159 /* Socket is already connected */
1160 err = -EISCONN;
1161 goto out_unlock;
1162 default:
1163 err = -EINVAL;
1164 goto out_unlock;
1165 }
1166
David S. Miller1c92b4e2007-05-31 13:24:26 -07001167 unix_state_lock_nested(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168
1169 if (sk->sk_state != st) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001170 unix_state_unlock(sk);
1171 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 sock_put(other);
1173 goto restart;
1174 }
1175
David S. Miller3610cda2011-01-05 15:38:53 -08001176 err = security_unix_stream_connect(sk, other, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 if (err) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001178 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 goto out_unlock;
1180 }
1181
1182 /* The way is open! Fastly set all the necessary fields... */
1183
1184 sock_hold(sk);
1185 unix_peer(newsk) = sk;
1186 newsk->sk_state = TCP_ESTABLISHED;
1187 newsk->sk_type = sk->sk_type;
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001188 init_peercred(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 newu = unix_sk(newsk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001190 RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 otheru = unix_sk(other);
1192
1193 /* copy address information from listening to new sock*/
1194 if (otheru->addr) {
1195 atomic_inc(&otheru->addr->refcnt);
1196 newu->addr = otheru->addr;
1197 }
Al Viro40ffe672012-03-14 21:54:32 -04001198 if (otheru->path.dentry) {
1199 path_get(&otheru->path);
1200 newu->path = otheru->path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 }
1202
1203 /* Set credentials */
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001204 copy_peercred(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 sock->state = SS_CONNECTED;
1207 sk->sk_state = TCP_ESTABLISHED;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001208 sock_hold(newsk);
1209
1210 smp_mb__after_atomic_inc(); /* sock_hold() does an atomic_inc() */
1211 unix_peer(sk) = newsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212
David S. Miller1c92b4e2007-05-31 13:24:26 -07001213 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214
1215 /* take ten and and send info to listening sock */
1216 spin_lock(&other->sk_receive_queue.lock);
1217 __skb_queue_tail(&other->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 spin_unlock(&other->sk_receive_queue.lock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001219 unix_state_unlock(other);
David S. Miller676d2362014-04-11 16:15:36 -04001220 other->sk_data_ready(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 sock_put(other);
1222 return 0;
1223
1224out_unlock:
1225 if (other)
David S. Miller1c92b4e2007-05-31 13:24:26 -07001226 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227
1228out:
Wei Yongjun40d44442009-02-25 00:32:45 +00001229 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 if (newsk)
1231 unix_release_sock(newsk, 0);
1232 if (other)
1233 sock_put(other);
1234 return err;
1235}
1236
1237static int unix_socketpair(struct socket *socka, struct socket *sockb)
1238{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001239 struct sock *ska = socka->sk, *skb = sockb->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240
1241 /* Join our sockets back to back */
1242 sock_hold(ska);
1243 sock_hold(skb);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001244 unix_peer(ska) = skb;
1245 unix_peer(skb) = ska;
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001246 init_peercred(ska);
1247 init_peercred(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248
1249 if (ska->sk_type != SOCK_DGRAM) {
1250 ska->sk_state = TCP_ESTABLISHED;
1251 skb->sk_state = TCP_ESTABLISHED;
1252 socka->state = SS_CONNECTED;
1253 sockb->state = SS_CONNECTED;
1254 }
1255 return 0;
1256}
1257
Daniel Borkmann90c6bd32013-10-17 22:51:31 +02001258static void unix_sock_inherit_flags(const struct socket *old,
1259 struct socket *new)
1260{
1261 if (test_bit(SOCK_PASSCRED, &old->flags))
1262 set_bit(SOCK_PASSCRED, &new->flags);
1263 if (test_bit(SOCK_PASSSEC, &old->flags))
1264 set_bit(SOCK_PASSSEC, &new->flags);
1265}
1266
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267static int unix_accept(struct socket *sock, struct socket *newsock, int flags)
1268{
1269 struct sock *sk = sock->sk;
1270 struct sock *tsk;
1271 struct sk_buff *skb;
1272 int err;
1273
1274 err = -EOPNOTSUPP;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001275 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 goto out;
1277
1278 err = -EINVAL;
1279 if (sk->sk_state != TCP_LISTEN)
1280 goto out;
1281
1282 /* If socket state is TCP_LISTEN it cannot change (for now...),
1283 * so that no locks are necessary.
1284 */
1285
1286 skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err);
1287 if (!skb) {
1288 /* This means receive shutdown. */
1289 if (err == 0)
1290 err = -EINVAL;
1291 goto out;
1292 }
1293
1294 tsk = skb->sk;
1295 skb_free_datagram(sk, skb);
1296 wake_up_interruptible(&unix_sk(sk)->peer_wait);
1297
1298 /* attach accepted sock to socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001299 unix_state_lock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 newsock->state = SS_CONNECTED;
Daniel Borkmann90c6bd32013-10-17 22:51:31 +02001301 unix_sock_inherit_flags(sock, newsock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302 sock_graft(tsk, newsock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001303 unix_state_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 return 0;
1305
1306out:
1307 return err;
1308}
1309
1310
1311static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer)
1312{
1313 struct sock *sk = sock->sk;
1314 struct unix_sock *u;
Cyrill Gorcunov13cfa972009-11-08 05:51:19 +00001315 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 int err = 0;
1317
1318 if (peer) {
1319 sk = unix_peer_get(sk);
1320
1321 err = -ENOTCONN;
1322 if (!sk)
1323 goto out;
1324 err = 0;
1325 } else {
1326 sock_hold(sk);
1327 }
1328
1329 u = unix_sk(sk);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001330 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 if (!u->addr) {
1332 sunaddr->sun_family = AF_UNIX;
1333 sunaddr->sun_path[0] = 0;
1334 *uaddr_len = sizeof(short);
1335 } else {
1336 struct unix_address *addr = u->addr;
1337
1338 *uaddr_len = addr->len;
1339 memcpy(sunaddr, addr->name, *uaddr_len);
1340 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07001341 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 sock_put(sk);
1343out:
1344 return err;
1345}
1346
1347static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb)
1348{
1349 int i;
1350
1351 scm->fp = UNIXCB(skb).fp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 UNIXCB(skb).fp = NULL;
1353
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001354 for (i = scm->fp->count-1; i >= 0; i--)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 unix_notinflight(scm->fp->fp[i]);
1356}
1357
Eric W. Biederman7361c362010-06-13 03:34:33 +00001358static void unix_destruct_scm(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359{
1360 struct scm_cookie scm;
1361 memset(&scm, 0, sizeof(scm));
Eric W. Biederman7361c362010-06-13 03:34:33 +00001362 scm.pid = UNIXCB(skb).pid;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001363 if (UNIXCB(skb).fp)
1364 unix_detach_fds(&scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365
1366 /* Alas, it calls VFS */
1367 /* So fscking what? fput() had been SMP-safe since the last Summer */
1368 scm_destroy(&scm);
1369 sock_wfree(skb);
1370}
1371
Eric Dumazet25888e32010-11-25 04:11:39 +00001372#define MAX_RECURSION_LEVEL 4
1373
Miklos Szeredi62093442008-11-09 15:23:57 +01001374static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375{
1376 int i;
Eric Dumazet25888e32010-11-25 04:11:39 +00001377 unsigned char max_level = 0;
1378 int unix_sock_count = 0;
1379
1380 for (i = scm->fp->count - 1; i >= 0; i--) {
1381 struct sock *sk = unix_get_socket(scm->fp->fp[i]);
1382
1383 if (sk) {
1384 unix_sock_count++;
1385 max_level = max(max_level,
1386 unix_sk(sk)->recursion_level);
1387 }
1388 }
1389 if (unlikely(max_level > MAX_RECURSION_LEVEL))
1390 return -ETOOMANYREFS;
Miklos Szeredi62093442008-11-09 15:23:57 +01001391
1392 /*
1393 * Need to duplicate file references for the sake of garbage
1394 * collection. Otherwise a socket in the fps might become a
1395 * candidate for GC while the skb is not yet queued.
1396 */
1397 UNIXCB(skb).fp = scm_fp_dup(scm->fp);
1398 if (!UNIXCB(skb).fp)
1399 return -ENOMEM;
1400
Eric Dumazet25888e32010-11-25 04:11:39 +00001401 if (unix_sock_count) {
1402 for (i = scm->fp->count - 1; i >= 0; i--)
1403 unix_inflight(scm->fp->fp[i]);
1404 }
1405 return max_level;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406}
1407
David S. Millerf78a5fd2011-09-16 19:34:00 -04001408static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds)
Eric W. Biederman7361c362010-06-13 03:34:33 +00001409{
1410 int err = 0;
Eric Dumazet16e57262011-09-19 05:52:27 +00001411
David S. Millerf78a5fd2011-09-16 19:34:00 -04001412 UNIXCB(skb).pid = get_pid(scm->pid);
Eric W. Biederman6b0ee8c02013-04-03 17:28:16 +00001413 UNIXCB(skb).uid = scm->creds.uid;
1414 UNIXCB(skb).gid = scm->creds.gid;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001415 UNIXCB(skb).fp = NULL;
1416 if (scm->fp && send_fds)
1417 err = unix_attach_fds(scm, skb);
1418
1419 skb->destructor = unix_destruct_scm;
1420 return err;
1421}
1422
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423/*
Eric Dumazet16e57262011-09-19 05:52:27 +00001424 * Some apps rely on write() giving SCM_CREDENTIALS
1425 * We include credentials if source or destination socket
1426 * asserted SOCK_PASSCRED.
1427 */
1428static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock,
1429 const struct sock *other)
1430{
Eric W. Biederman6b0ee8c02013-04-03 17:28:16 +00001431 if (UNIXCB(skb).pid)
Eric Dumazet16e57262011-09-19 05:52:27 +00001432 return;
1433 if (test_bit(SOCK_PASSCRED, &sock->flags) ||
Eric W. Biederman25da0e32013-04-03 16:13:35 +00001434 !other->sk_socket ||
1435 test_bit(SOCK_PASSCRED, &other->sk_socket->flags)) {
Eric Dumazet16e57262011-09-19 05:52:27 +00001436 UNIXCB(skb).pid = get_pid(task_tgid(current));
David S. Miller6e0895c2013-04-22 20:32:51 -04001437 current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid);
Eric Dumazet16e57262011-09-19 05:52:27 +00001438 }
1439}
1440
1441/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 * Send AF_UNIX data.
1443 */
1444
1445static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock,
1446 struct msghdr *msg, size_t len)
1447{
1448 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1449 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001450 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 struct unix_sock *u = unix_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001452 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 struct sock *other = NULL;
1454 int namelen = 0; /* fake GCC */
1455 int err;
Eric Dumazet95c96172012-04-15 05:58:06 +00001456 unsigned int hash;
David S. Millerf78a5fd2011-09-16 19:34:00 -04001457 struct sk_buff *skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458 long timeo;
1459 struct scm_cookie tmp_scm;
Eric Dumazet25888e32010-11-25 04:11:39 +00001460 int max_level;
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001461 int data_len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462
1463 if (NULL == siocb->scm)
1464 siocb->scm = &tmp_scm;
dann frazier5f23b732008-11-26 15:32:27 -08001465 wait_for_unix_gc();
Eric Dumazete0e3cea2012-08-21 06:21:17 +00001466 err = scm_send(sock, msg, siocb->scm, false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 if (err < 0)
1468 return err;
1469
1470 err = -EOPNOTSUPP;
1471 if (msg->msg_flags&MSG_OOB)
1472 goto out;
1473
1474 if (msg->msg_namelen) {
1475 err = unix_mkname(sunaddr, msg->msg_namelen, &hash);
1476 if (err < 0)
1477 goto out;
1478 namelen = err;
1479 } else {
1480 sunaddr = NULL;
1481 err = -ENOTCONN;
1482 other = unix_peer_get(sk);
1483 if (!other)
1484 goto out;
1485 }
1486
Joe Perchesf64f9e72009-11-29 16:55:45 -08001487 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr
1488 && (err = unix_autobind(sock)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 goto out;
1490
1491 err = -EMSGSIZE;
1492 if (len > sk->sk_sndbuf - 32)
1493 goto out;
1494
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001495 if (len > SKB_MAX_ALLOC)
1496 data_len = min_t(size_t,
1497 len - SKB_MAX_ALLOC,
1498 MAX_SKB_FRAGS * PAGE_SIZE);
1499
1500 skb = sock_alloc_send_pskb(sk, len - data_len, data_len,
Eric Dumazet28d64272013-08-08 14:38:47 -07001501 msg->msg_flags & MSG_DONTWAIT, &err,
1502 PAGE_ALLOC_COSTLY_ORDER);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001503 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 goto out;
1505
David S. Millerf78a5fd2011-09-16 19:34:00 -04001506 err = unix_scm_to_skb(siocb->scm, skb, true);
Eric Dumazet25888e32010-11-25 04:11:39 +00001507 if (err < 0)
Eric W. Biederman7361c362010-06-13 03:34:33 +00001508 goto out_free;
Eric Dumazet25888e32010-11-25 04:11:39 +00001509 max_level = err + 1;
Catherine Zhangdc49c1f2006-08-02 14:12:06 -07001510 unix_get_secdata(siocb->scm, skb);
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001511
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001512 skb_put(skb, len - data_len);
1513 skb->data_len = data_len;
1514 skb->len = len;
1515 err = skb_copy_datagram_from_iovec(skb, 0, msg->msg_iov, 0, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 if (err)
1517 goto out_free;
1518
1519 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1520
1521restart:
1522 if (!other) {
1523 err = -ECONNRESET;
1524 if (sunaddr == NULL)
1525 goto out_free;
1526
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001527 other = unix_find_other(net, sunaddr, namelen, sk->sk_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 hash, &err);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001529 if (other == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 goto out_free;
1531 }
1532
Alban Crequyd6ae3ba2011-01-18 06:39:15 +00001533 if (sk_filter(other, skb) < 0) {
1534 /* Toss the packet but do not return any error to the sender */
1535 err = len;
1536 goto out_free;
1537 }
1538
David S. Miller1c92b4e2007-05-31 13:24:26 -07001539 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 err = -EPERM;
1541 if (!unix_may_send(sk, other))
1542 goto out_unlock;
1543
1544 if (sock_flag(other, SOCK_DEAD)) {
1545 /*
1546 * Check with 1003.1g - what should
1547 * datagram error
1548 */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001549 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 sock_put(other);
1551
1552 err = 0;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001553 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554 if (unix_peer(sk) == other) {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001555 unix_peer(sk) = NULL;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001556 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557
1558 unix_dgram_disconnected(sk, other);
1559 sock_put(other);
1560 err = -ECONNREFUSED;
1561 } else {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001562 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 }
1564
1565 other = NULL;
1566 if (err)
1567 goto out_free;
1568 goto restart;
1569 }
1570
1571 err = -EPIPE;
1572 if (other->sk_shutdown & RCV_SHUTDOWN)
1573 goto out_unlock;
1574
1575 if (sk->sk_type != SOCK_SEQPACKET) {
1576 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1577 if (err)
1578 goto out_unlock;
1579 }
1580
Rainer Weikusat3c734192008-06-17 22:28:05 -07001581 if (unix_peer(other) != sk && unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 if (!timeo) {
1583 err = -EAGAIN;
1584 goto out_unlock;
1585 }
1586
1587 timeo = unix_wait_for_peer(other, timeo);
1588
1589 err = sock_intr_errno(timeo);
1590 if (signal_pending(current))
1591 goto out_free;
1592
1593 goto restart;
1594 }
1595
Alban Crequy3f661162010-10-04 08:48:28 +00001596 if (sock_flag(other, SOCK_RCVTSTAMP))
1597 __net_timestamp(skb);
Eric Dumazet16e57262011-09-19 05:52:27 +00001598 maybe_add_creds(skb, sock, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 skb_queue_tail(&other->sk_receive_queue, skb);
Eric Dumazet25888e32010-11-25 04:11:39 +00001600 if (max_level > unix_sk(other)->recursion_level)
1601 unix_sk(other)->recursion_level = max_level;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001602 unix_state_unlock(other);
David S. Miller676d2362014-04-11 16:15:36 -04001603 other->sk_data_ready(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 sock_put(other);
David S. Millerf78a5fd2011-09-16 19:34:00 -04001605 scm_destroy(siocb->scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 return len;
1607
1608out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001609 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610out_free:
1611 kfree_skb(skb);
1612out:
1613 if (other)
1614 sock_put(other);
David S. Millerf78a5fd2011-09-16 19:34:00 -04001615 scm_destroy(siocb->scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 return err;
1617}
1618
Eric Dumazete370a722013-08-08 14:37:32 -07001619/* We use paged skbs for stream sockets, and limit occupancy to 32768
1620 * bytes, and a minimun of a full page.
1621 */
1622#define UNIX_SKB_FRAGS_SZ (PAGE_SIZE << get_order(32768))
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001623
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock,
1625 struct msghdr *msg, size_t len)
1626{
1627 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1628 struct sock *sk = sock->sk;
1629 struct sock *other = NULL;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001630 int err, size;
David S. Millerf78a5fd2011-09-16 19:34:00 -04001631 struct sk_buff *skb;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001632 int sent = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 struct scm_cookie tmp_scm;
Miklos Szeredi8ba69ba2009-09-11 11:31:45 -07001634 bool fds_sent = false;
Eric Dumazet25888e32010-11-25 04:11:39 +00001635 int max_level;
Eric Dumazete370a722013-08-08 14:37:32 -07001636 int data_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637
1638 if (NULL == siocb->scm)
1639 siocb->scm = &tmp_scm;
dann frazier5f23b732008-11-26 15:32:27 -08001640 wait_for_unix_gc();
Eric Dumazete0e3cea2012-08-21 06:21:17 +00001641 err = scm_send(sock, msg, siocb->scm, false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 if (err < 0)
1643 return err;
1644
1645 err = -EOPNOTSUPP;
1646 if (msg->msg_flags&MSG_OOB)
1647 goto out_err;
1648
1649 if (msg->msg_namelen) {
1650 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1651 goto out_err;
1652 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 err = -ENOTCONN;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001654 other = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 if (!other)
1656 goto out_err;
1657 }
1658
1659 if (sk->sk_shutdown & SEND_SHUTDOWN)
1660 goto pipe_err;
1661
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001662 while (sent < len) {
Eric Dumazete370a722013-08-08 14:37:32 -07001663 size = len - sent;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664
1665 /* Keep two messages in the pipe so it schedules better */
Eric Dumazete370a722013-08-08 14:37:32 -07001666 size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667
Eric Dumazete370a722013-08-08 14:37:32 -07001668 /* allow fallback to order-0 allocations */
1669 size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ);
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001670
Eric Dumazete370a722013-08-08 14:37:32 -07001671 data_len = max_t(int, 0, size - SKB_MAX_HEAD(0));
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001672
Eric Dumazete370a722013-08-08 14:37:32 -07001673 skb = sock_alloc_send_pskb(sk, size - data_len, data_len,
Eric Dumazet28d64272013-08-08 14:38:47 -07001674 msg->msg_flags & MSG_DONTWAIT, &err,
1675 get_order(UNIX_SKB_FRAGS_SZ));
Eric Dumazete370a722013-08-08 14:37:32 -07001676 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 goto out_err;
1678
David S. Millerf78a5fd2011-09-16 19:34:00 -04001679 /* Only send the fds in the first buffer */
1680 err = unix_scm_to_skb(siocb->scm, skb, !fds_sent);
Eric Dumazet25888e32010-11-25 04:11:39 +00001681 if (err < 0) {
Eric W. Biederman7361c362010-06-13 03:34:33 +00001682 kfree_skb(skb);
David S. Millerf78a5fd2011-09-16 19:34:00 -04001683 goto out_err;
Miklos Szeredi62093442008-11-09 15:23:57 +01001684 }
Eric Dumazet25888e32010-11-25 04:11:39 +00001685 max_level = err + 1;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001686 fds_sent = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
Eric Dumazete370a722013-08-08 14:37:32 -07001688 skb_put(skb, size - data_len);
1689 skb->data_len = data_len;
1690 skb->len = size;
Eric Dumazetf3dfd202013-08-11 21:54:48 -07001691 err = skb_copy_datagram_from_iovec(skb, 0, msg->msg_iov,
1692 sent, size);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001693 if (err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 kfree_skb(skb);
David S. Millerf78a5fd2011-09-16 19:34:00 -04001695 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 }
1697
David S. Miller1c92b4e2007-05-31 13:24:26 -07001698 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699
1700 if (sock_flag(other, SOCK_DEAD) ||
1701 (other->sk_shutdown & RCV_SHUTDOWN))
1702 goto pipe_err_free;
1703
Eric Dumazet16e57262011-09-19 05:52:27 +00001704 maybe_add_creds(skb, sock, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 skb_queue_tail(&other->sk_receive_queue, skb);
Eric Dumazet25888e32010-11-25 04:11:39 +00001706 if (max_level > unix_sk(other)->recursion_level)
1707 unix_sk(other)->recursion_level = max_level;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001708 unix_state_unlock(other);
David S. Miller676d2362014-04-11 16:15:36 -04001709 other->sk_data_ready(other);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001710 sent += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712
David S. Millerf78a5fd2011-09-16 19:34:00 -04001713 scm_destroy(siocb->scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 siocb->scm = NULL;
1715
1716 return sent;
1717
1718pipe_err_free:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001719 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 kfree_skb(skb);
1721pipe_err:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001722 if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL))
1723 send_sig(SIGPIPE, current, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 err = -EPIPE;
1725out_err:
David S. Millerf78a5fd2011-09-16 19:34:00 -04001726 scm_destroy(siocb->scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 siocb->scm = NULL;
1728 return sent ? : err;
1729}
1730
1731static int unix_seqpacket_sendmsg(struct kiocb *kiocb, struct socket *sock,
1732 struct msghdr *msg, size_t len)
1733{
1734 int err;
1735 struct sock *sk = sock->sk;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001736
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 err = sock_error(sk);
1738 if (err)
1739 return err;
1740
1741 if (sk->sk_state != TCP_ESTABLISHED)
1742 return -ENOTCONN;
1743
1744 if (msg->msg_namelen)
1745 msg->msg_namelen = 0;
1746
1747 return unix_dgram_sendmsg(kiocb, sock, msg, len);
1748}
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001749
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +00001750static int unix_seqpacket_recvmsg(struct kiocb *iocb, struct socket *sock,
1751 struct msghdr *msg, size_t size,
1752 int flags)
1753{
1754 struct sock *sk = sock->sk;
1755
1756 if (sk->sk_state != TCP_ESTABLISHED)
1757 return -ENOTCONN;
1758
1759 return unix_dgram_recvmsg(iocb, sock, msg, size, flags);
1760}
1761
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762static void unix_copy_addr(struct msghdr *msg, struct sock *sk)
1763{
1764 struct unix_sock *u = unix_sk(sk);
1765
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 if (u->addr) {
1767 msg->msg_namelen = u->addr->len;
1768 memcpy(msg->msg_name, u->addr->name, u->addr->len);
1769 }
1770}
1771
1772static int unix_dgram_recvmsg(struct kiocb *iocb, struct socket *sock,
1773 struct msghdr *msg, size_t size,
1774 int flags)
1775{
1776 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
1777 struct scm_cookie tmp_scm;
1778 struct sock *sk = sock->sk;
1779 struct unix_sock *u = unix_sk(sk);
1780 int noblock = flags & MSG_DONTWAIT;
1781 struct sk_buff *skb;
1782 int err;
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001783 int peeked, skip;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784
1785 err = -EOPNOTSUPP;
1786 if (flags&MSG_OOB)
1787 goto out;
1788
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001789 err = mutex_lock_interruptible(&u->readlock);
Eric Dumazetde144392014-03-25 18:42:27 -07001790 if (unlikely(err)) {
1791 /* recvmsg() in non blocking mode is supposed to return -EAGAIN
1792 * sk_rcvtimeo is not honored by mutex_lock_interruptible()
1793 */
1794 err = noblock ? -EAGAIN : -ERESTARTSYS;
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001795 goto out;
1796 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001798 skip = sk_peek_offset(sk, flags);
1799
1800 skb = __skb_recv_datagram(sk, flags, &peeked, &skip, &err);
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001801 if (!skb) {
1802 unix_state_lock(sk);
1803 /* Signal EOF on disconnected non-blocking SEQPACKET socket. */
1804 if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN &&
1805 (sk->sk_shutdown & RCV_SHUTDOWN))
1806 err = 0;
1807 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 goto out_unlock;
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001809 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810
Eric Dumazet67426b72010-10-29 20:44:44 +00001811 wake_up_interruptible_sync_poll(&u->peer_wait,
1812 POLLOUT | POLLWRNORM | POLLWRBAND);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813
1814 if (msg->msg_name)
1815 unix_copy_addr(msg, skb->sk);
1816
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001817 if (size > skb->len - skip)
1818 size = skb->len - skip;
1819 else if (size < skb->len - skip)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 msg->msg_flags |= MSG_TRUNC;
1821
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001822 err = skb_copy_datagram_iovec(skb, skip, msg->msg_iov, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823 if (err)
1824 goto out_free;
1825
Alban Crequy3f661162010-10-04 08:48:28 +00001826 if (sock_flag(sk, SOCK_RCVTSTAMP))
1827 __sock_recv_timestamp(msg, sk, skb);
1828
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829 if (!siocb->scm) {
1830 siocb->scm = &tmp_scm;
1831 memset(&tmp_scm, 0, sizeof(tmp_scm));
1832 }
Eric W. Biederman6b0ee8c02013-04-03 17:28:16 +00001833 scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid);
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001834 unix_set_secdata(siocb->scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001836 if (!(flags & MSG_PEEK)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 if (UNIXCB(skb).fp)
1838 unix_detach_fds(siocb->scm, skb);
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001839
1840 sk_peek_offset_bwd(sk, skb->len);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001841 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 /* It is questionable: on PEEK we could:
1843 - do not return fds - good, but too simple 8)
1844 - return fds, and do not return them on read (old strategy,
1845 apparently wrong)
1846 - clone fds (I chose it for now, it is the most universal
1847 solution)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001848
1849 POSIX 1003.1g does not actually define this clearly
1850 at all. POSIX 1003.1g doesn't define a lot of things
1851 clearly however!
1852
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 */
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001854
1855 sk_peek_offset_fwd(sk, size);
1856
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 if (UNIXCB(skb).fp)
1858 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
1859 }
Eric Dumazet9f6f9af2012-02-21 23:24:55 +00001860 err = (flags & MSG_TRUNC) ? skb->len - skip : size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861
1862 scm_recv(sock, msg, siocb->scm, flags);
1863
1864out_free:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001865 skb_free_datagram(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866out_unlock:
Ingo Molnar57b47a52006-03-20 22:35:41 -08001867 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868out:
1869 return err;
1870}
1871
1872/*
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001873 * Sleep until more data has arrived. But check for races..
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874 */
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001875static long unix_stream_data_wait(struct sock *sk, long timeo,
1876 struct sk_buff *last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877{
1878 DEFINE_WAIT(wait);
1879
David S. Miller1c92b4e2007-05-31 13:24:26 -07001880 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881
1882 for (;;) {
Eric Dumazetaa395142010-04-20 13:03:51 +00001883 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001885 if (skb_peek_tail(&sk->sk_receive_queue) != last ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886 sk->sk_err ||
1887 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1888 signal_pending(current) ||
1889 !timeo)
1890 break;
1891
1892 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001893 unix_state_unlock(sk);
Colin Cross2b15af62013-05-06 23:50:21 +00001894 timeo = freezable_schedule_timeout(timeo);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001895 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1897 }
1898
Eric Dumazetaa395142010-04-20 13:03:51 +00001899 finish_wait(sk_sleep(sk), &wait);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001900 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 return timeo;
1902}
1903
Eric Dumazete370a722013-08-08 14:37:32 -07001904static unsigned int unix_skb_len(const struct sk_buff *skb)
1905{
1906 return skb->len - UNIXCB(skb).consumed;
1907}
1908
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909static int unix_stream_recvmsg(struct kiocb *iocb, struct socket *sock,
1910 struct msghdr *msg, size_t size,
1911 int flags)
1912{
1913 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
1914 struct scm_cookie tmp_scm;
1915 struct sock *sk = sock->sk;
1916 struct unix_sock *u = unix_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001917 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 int copied = 0;
Eric Dumazetde144392014-03-25 18:42:27 -07001919 int noblock = flags & MSG_DONTWAIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 int check_creds = 0;
1921 int target;
1922 int err = 0;
1923 long timeo;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00001924 int skip;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925
1926 err = -EINVAL;
1927 if (sk->sk_state != TCP_ESTABLISHED)
1928 goto out;
1929
1930 err = -EOPNOTSUPP;
1931 if (flags&MSG_OOB)
1932 goto out;
1933
1934 target = sock_rcvlowat(sk, flags&MSG_WAITALL, size);
Eric Dumazetde144392014-03-25 18:42:27 -07001935 timeo = sock_rcvtimeo(sk, noblock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 /* Lock the socket to prevent queue disordering
1938 * while sleeps in memcpy_tomsg
1939 */
1940
1941 if (!siocb->scm) {
1942 siocb->scm = &tmp_scm;
1943 memset(&tmp_scm, 0, sizeof(tmp_scm));
1944 }
1945
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001946 err = mutex_lock_interruptible(&u->readlock);
Eric Dumazetde144392014-03-25 18:42:27 -07001947 if (unlikely(err)) {
1948 /* recvmsg() in non blocking mode is supposed to return -EAGAIN
1949 * sk_rcvtimeo is not honored by mutex_lock_interruptible()
1950 */
1951 err = noblock ? -EAGAIN : -ERESTARTSYS;
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001952 goto out;
1953 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001955 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 int chunk;
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001957 struct sk_buff *skb, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001959 unix_state_lock(sk);
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001960 last = skb = skb_peek(&sk->sk_receive_queue);
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00001961again:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001962 if (skb == NULL) {
Eric Dumazet25888e32010-11-25 04:11:39 +00001963 unix_sk(sk)->recursion_level = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 if (copied >= target)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001965 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966
1967 /*
1968 * POSIX 1003.1g mandates this order.
1969 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001970
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001971 err = sock_error(sk);
1972 if (err)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001973 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 if (sk->sk_shutdown & RCV_SHUTDOWN)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001975 goto unlock;
1976
1977 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 err = -EAGAIN;
1979 if (!timeo)
1980 break;
Ingo Molnar57b47a52006-03-20 22:35:41 -08001981 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001983 timeo = unix_stream_data_wait(sk, timeo, last);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001985 if (signal_pending(current)
1986 || mutex_lock_interruptible(&u->readlock)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 err = sock_intr_errno(timeo);
1988 goto out;
1989 }
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001990
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991 continue;
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001992 unlock:
1993 unix_state_unlock(sk);
1994 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 }
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00001996
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001997 skip = sk_peek_offset(sk, flags);
Eric Dumazete370a722013-08-08 14:37:32 -07001998 while (skip >= unix_skb_len(skb)) {
1999 skip -= unix_skb_len(skb);
Benjamin Poirier79f632c2013-04-29 11:42:14 +00002000 last = skb;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002001 skb = skb_peek_next(skb, &sk->sk_receive_queue);
Benjamin Poirier79f632c2013-04-29 11:42:14 +00002002 if (!skb)
2003 goto again;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002004 }
2005
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07002006 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007
2008 if (check_creds) {
2009 /* Never glue messages from different writers */
Eric W. Biederman7361c362010-06-13 03:34:33 +00002010 if ((UNIXCB(skb).pid != siocb->scm->pid) ||
Eric W. Biederman6b0ee8c02013-04-03 17:28:16 +00002011 !uid_eq(UNIXCB(skb).uid, siocb->scm->creds.uid) ||
2012 !gid_eq(UNIXCB(skb).gid, siocb->scm->creds.gid))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 break;
Eric W. Biederman0e82e7f6d2013-04-03 16:14:47 +00002014 } else if (test_bit(SOCK_PASSCRED, &sock->flags)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 /* Copy credentials */
Eric W. Biederman6b0ee8c02013-04-03 17:28:16 +00002016 scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 check_creds = 1;
2018 }
2019
2020 /* Copy address just once */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002021 if (sunaddr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 unix_copy_addr(msg, skb->sk);
2023 sunaddr = NULL;
2024 }
2025
Eric Dumazete370a722013-08-08 14:37:32 -07002026 chunk = min_t(unsigned int, unix_skb_len(skb) - skip, size);
2027 if (skb_copy_datagram_iovec(skb, UNIXCB(skb).consumed + skip,
2028 msg->msg_iov, chunk)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 if (copied == 0)
2030 copied = -EFAULT;
2031 break;
2032 }
2033 copied += chunk;
2034 size -= chunk;
2035
2036 /* Mark read part of skb as used */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002037 if (!(flags & MSG_PEEK)) {
Eric Dumazete370a722013-08-08 14:37:32 -07002038 UNIXCB(skb).consumed += chunk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002040 sk_peek_offset_bwd(sk, chunk);
2041
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 if (UNIXCB(skb).fp)
2043 unix_detach_fds(siocb->scm, skb);
2044
Eric Dumazete370a722013-08-08 14:37:32 -07002045 if (unix_skb_len(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047
Eric Dumazet6f01fd62012-01-28 16:11:03 +00002048 skb_unlink(skb, &sk->sk_receive_queue);
Neil Horman70d4bf62010-07-20 06:45:56 +00002049 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050
2051 if (siocb->scm->fp)
2052 break;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002053 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054 /* It is questionable, see note in unix_dgram_recvmsg.
2055 */
2056 if (UNIXCB(skb).fp)
2057 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
2058
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002059 sk_peek_offset_fwd(sk, chunk);
2060
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 break;
2062 }
2063 } while (size);
2064
Ingo Molnar57b47a52006-03-20 22:35:41 -08002065 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 scm_recv(sock, msg, siocb->scm, flags);
2067out:
2068 return copied ? : err;
2069}
2070
2071static int unix_shutdown(struct socket *sock, int mode)
2072{
2073 struct sock *sk = sock->sk;
2074 struct sock *other;
2075
Xi Wangfc61b922012-08-26 16:47:13 +00002076 if (mode < SHUT_RD || mode > SHUT_RDWR)
2077 return -EINVAL;
2078 /* This maps:
2079 * SHUT_RD (0) -> RCV_SHUTDOWN (1)
2080 * SHUT_WR (1) -> SEND_SHUTDOWN (2)
2081 * SHUT_RDWR (2) -> SHUTDOWN_MASK (3)
2082 */
2083 ++mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084
Alban Crequy7180a032011-01-19 04:56:36 +00002085 unix_state_lock(sk);
2086 sk->sk_shutdown |= mode;
2087 other = unix_peer(sk);
2088 if (other)
2089 sock_hold(other);
2090 unix_state_unlock(sk);
2091 sk->sk_state_change(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092
Alban Crequy7180a032011-01-19 04:56:36 +00002093 if (other &&
2094 (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095
Alban Crequy7180a032011-01-19 04:56:36 +00002096 int peer_mode = 0;
2097
2098 if (mode&RCV_SHUTDOWN)
2099 peer_mode |= SEND_SHUTDOWN;
2100 if (mode&SEND_SHUTDOWN)
2101 peer_mode |= RCV_SHUTDOWN;
2102 unix_state_lock(other);
2103 other->sk_shutdown |= peer_mode;
2104 unix_state_unlock(other);
2105 other->sk_state_change(other);
2106 if (peer_mode == SHUTDOWN_MASK)
2107 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP);
2108 else if (peer_mode & RCV_SHUTDOWN)
2109 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 }
Alban Crequy7180a032011-01-19 04:56:36 +00002111 if (other)
2112 sock_put(other);
2113
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114 return 0;
2115}
2116
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002117long unix_inq_len(struct sock *sk)
2118{
2119 struct sk_buff *skb;
2120 long amount = 0;
2121
2122 if (sk->sk_state == TCP_LISTEN)
2123 return -EINVAL;
2124
2125 spin_lock(&sk->sk_receive_queue.lock);
2126 if (sk->sk_type == SOCK_STREAM ||
2127 sk->sk_type == SOCK_SEQPACKET) {
2128 skb_queue_walk(&sk->sk_receive_queue, skb)
Eric Dumazete370a722013-08-08 14:37:32 -07002129 amount += unix_skb_len(skb);
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002130 } else {
2131 skb = skb_peek(&sk->sk_receive_queue);
2132 if (skb)
2133 amount = skb->len;
2134 }
2135 spin_unlock(&sk->sk_receive_queue.lock);
2136
2137 return amount;
2138}
2139EXPORT_SYMBOL_GPL(unix_inq_len);
2140
2141long unix_outq_len(struct sock *sk)
2142{
2143 return sk_wmem_alloc_get(sk);
2144}
2145EXPORT_SYMBOL_GPL(unix_outq_len);
2146
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2148{
2149 struct sock *sk = sock->sk;
Jianjun Konge27dfce2008-11-01 21:38:31 -07002150 long amount = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151 int err;
2152
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002153 switch (cmd) {
2154 case SIOCOUTQ:
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002155 amount = unix_outq_len(sk);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002156 err = put_user(amount, (int __user *)arg);
2157 break;
2158 case SIOCINQ:
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002159 amount = unix_inq_len(sk);
2160 if (amount < 0)
2161 err = amount;
2162 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07002163 err = put_user(amount, (int __user *)arg);
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002164 break;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002165 default:
2166 err = -ENOIOCTLCMD;
2167 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002168 }
2169 return err;
2170}
2171
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002172static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173{
2174 struct sock *sk = sock->sk;
2175 unsigned int mask;
2176
Eric Dumazetaa395142010-04-20 13:03:51 +00002177 sock_poll_wait(file, sk_sleep(sk), wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178 mask = 0;
2179
2180 /* exceptional events? */
2181 if (sk->sk_err)
2182 mask |= POLLERR;
2183 if (sk->sk_shutdown == SHUTDOWN_MASK)
2184 mask |= POLLHUP;
Davide Libenzif348d702006-03-25 03:07:39 -08002185 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazetdb409802010-09-06 11:13:50 +00002186 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002187
2188 /* readable? */
Eric Dumazetdb409802010-09-06 11:13:50 +00002189 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190 mask |= POLLIN | POLLRDNORM;
2191
2192 /* Connection-based need to check for termination and startup */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002193 if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) &&
2194 sk->sk_state == TCP_CLOSE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195 mask |= POLLHUP;
2196
2197 /*
2198 * we set writable also when the other side has shut down the
2199 * connection. This prevents stuck sockets.
2200 */
2201 if (unix_writable(sk))
2202 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2203
2204 return mask;
2205}
2206
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002207static unsigned int unix_dgram_poll(struct file *file, struct socket *sock,
2208 poll_table *wait)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002209{
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002210 struct sock *sk = sock->sk, *other;
2211 unsigned int mask, writable;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002212
Eric Dumazetaa395142010-04-20 13:03:51 +00002213 sock_poll_wait(file, sk_sleep(sk), wait);
Rainer Weikusat3c734192008-06-17 22:28:05 -07002214 mask = 0;
2215
2216 /* exceptional events? */
2217 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +00002218 mask |= POLLERR |
Jacob Keller8facd5f2013-04-02 13:55:40 -07002219 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +00002220
Rainer Weikusat3c734192008-06-17 22:28:05 -07002221 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazet5456f092010-10-31 05:36:23 +00002222 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002223 if (sk->sk_shutdown == SHUTDOWN_MASK)
2224 mask |= POLLHUP;
2225
2226 /* readable? */
Eric Dumazet5456f092010-10-31 05:36:23 +00002227 if (!skb_queue_empty(&sk->sk_receive_queue))
Rainer Weikusat3c734192008-06-17 22:28:05 -07002228 mask |= POLLIN | POLLRDNORM;
2229
2230 /* Connection-based need to check for termination and startup */
2231 if (sk->sk_type == SOCK_SEQPACKET) {
2232 if (sk->sk_state == TCP_CLOSE)
2233 mask |= POLLHUP;
2234 /* connection hasn't started yet? */
2235 if (sk->sk_state == TCP_SYN_SENT)
2236 return mask;
2237 }
2238
Eric Dumazet973a34a2010-10-31 05:38:25 +00002239 /* No write status requested, avoid expensive OUT tests. */
Hans Verkuil626cf232012-03-23 15:02:27 -07002240 if (!(poll_requested_events(wait) & (POLLWRBAND|POLLWRNORM|POLLOUT)))
Eric Dumazet973a34a2010-10-31 05:38:25 +00002241 return mask;
2242
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002243 writable = unix_writable(sk);
Eric Dumazet5456f092010-10-31 05:36:23 +00002244 other = unix_peer_get(sk);
2245 if (other) {
2246 if (unix_peer(other) != sk) {
2247 sock_poll_wait(file, &unix_sk(other)->peer_wait, wait);
2248 if (unix_recvq_full(other))
2249 writable = 0;
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002250 }
Eric Dumazet5456f092010-10-31 05:36:23 +00002251 sock_put(other);
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002252 }
2253
2254 if (writable)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002255 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2256 else
2257 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
2258
Rainer Weikusat3c734192008-06-17 22:28:05 -07002259 return mask;
2260}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002261
2262#ifdef CONFIG_PROC_FS
Pavel Emelyanova53eb3f2007-11-23 20:30:01 +08002263
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002264#define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1)
2265
2266#define get_bucket(x) ((x) >> BUCKET_SPACE)
2267#define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1))
2268#define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o))
Pavel Emelyanova53eb3f2007-11-23 20:30:01 +08002269
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002270static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271{
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002272 unsigned long offset = get_offset(*pos);
2273 unsigned long bucket = get_bucket(*pos);
2274 struct sock *sk;
2275 unsigned long count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002277 for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) {
2278 if (sock_net(sk) != seq_file_net(seq))
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002279 continue;
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002280 if (++count == offset)
2281 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282 }
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002283
2284 return sk;
2285}
2286
2287static struct sock *unix_next_socket(struct seq_file *seq,
2288 struct sock *sk,
2289 loff_t *pos)
2290{
2291 unsigned long bucket;
2292
2293 while (sk > (struct sock *)SEQ_START_TOKEN) {
2294 sk = sk_next(sk);
2295 if (!sk)
2296 goto next_bucket;
2297 if (sock_net(sk) == seq_file_net(seq))
2298 return sk;
2299 }
2300
2301 do {
2302 sk = unix_from_bucket(seq, pos);
2303 if (sk)
2304 return sk;
2305
2306next_bucket:
2307 bucket = get_bucket(*pos) + 1;
2308 *pos = set_bucket_offset(bucket, 1);
2309 } while (bucket < ARRAY_SIZE(unix_socket_table));
2310
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311 return NULL;
2312}
2313
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314static void *unix_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002315 __acquires(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002317 spin_lock(&unix_table_lock);
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002318
2319 if (!*pos)
2320 return SEQ_START_TOKEN;
2321
2322 if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table))
2323 return NULL;
2324
2325 return unix_next_socket(seq, NULL, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326}
2327
2328static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2329{
2330 ++*pos;
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002331 return unix_next_socket(seq, v, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002332}
2333
2334static void unix_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002335 __releases(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002337 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338}
2339
2340static int unix_seq_show(struct seq_file *seq, void *v)
2341{
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002342
Joe Perchesb9f31242008-04-12 19:04:38 -07002343 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344 seq_puts(seq, "Num RefCount Protocol Flags Type St "
2345 "Inode Path\n");
2346 else {
2347 struct sock *s = v;
2348 struct unix_sock *u = unix_sk(s);
David S. Miller1c92b4e2007-05-31 13:24:26 -07002349 unix_state_lock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350
Dan Rosenberg71338aa2011-05-23 12:17:35 +00002351 seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352 s,
2353 atomic_read(&s->sk_refcnt),
2354 0,
2355 s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0,
2356 s->sk_type,
2357 s->sk_socket ?
2358 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) :
2359 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING),
2360 sock_i_ino(s));
2361
2362 if (u->addr) {
2363 int i, len;
2364 seq_putc(seq, ' ');
2365
2366 i = 0;
2367 len = u->addr->len - sizeof(short);
2368 if (!UNIX_ABSTRACT(s))
2369 len--;
2370 else {
2371 seq_putc(seq, '@');
2372 i++;
2373 }
2374 for ( ; i < len; i++)
2375 seq_putc(seq, u->addr->name->sun_path[i]);
2376 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07002377 unix_state_unlock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002378 seq_putc(seq, '\n');
2379 }
2380
2381 return 0;
2382}
2383
Philippe De Muyter56b3d972007-07-10 23:07:31 -07002384static const struct seq_operations unix_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385 .start = unix_seq_start,
2386 .next = unix_seq_next,
2387 .stop = unix_seq_stop,
2388 .show = unix_seq_show,
2389};
2390
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391static int unix_seq_open(struct inode *inode, struct file *file)
2392{
Denis V. Luneve372c412007-11-19 22:31:54 -08002393 return seq_open_net(inode, file, &unix_seq_ops,
Eric Dumazet8b51b062012-06-08 22:10:20 +00002394 sizeof(struct seq_net_private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395}
2396
Arjan van de Venda7071d2007-02-12 00:55:36 -08002397static const struct file_operations unix_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002398 .owner = THIS_MODULE,
2399 .open = unix_seq_open,
2400 .read = seq_read,
2401 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08002402 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002403};
2404
2405#endif
2406
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00002407static const struct net_proto_family unix_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408 .family = PF_UNIX,
2409 .create = unix_create,
2410 .owner = THIS_MODULE,
2411};
2412
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002413
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002414static int __net_init unix_net_init(struct net *net)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002415{
2416 int error = -ENOMEM;
2417
Denis V. Luneva0a53c82007-12-11 04:19:17 -08002418 net->unx.sysctl_max_dgram_qlen = 10;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002419 if (unix_sysctl_register(net))
2420 goto out;
Pavel Emelyanovd392e492007-12-01 23:44:15 +11002421
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002422#ifdef CONFIG_PROC_FS
Gao fengd4beaa62013-02-18 01:34:54 +00002423 if (!proc_create("unix", 0, net->proc_net, &unix_seq_fops)) {
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002424 unix_sysctl_unregister(net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002425 goto out;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002426 }
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002427#endif
2428 error = 0;
2429out:
Jianjun Kong48dcc33e2008-11-01 21:37:27 -07002430 return error;
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002431}
2432
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002433static void __net_exit unix_net_exit(struct net *net)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002434{
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002435 unix_sysctl_unregister(net);
Gao fengece31ff2013-02-18 01:34:56 +00002436 remove_proc_entry("unix", net->proc_net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002437}
2438
2439static struct pernet_operations unix_net_ops = {
2440 .init = unix_net_init,
2441 .exit = unix_net_exit,
2442};
2443
Linus Torvalds1da177e2005-04-16 15:20:36 -07002444static int __init af_unix_init(void)
2445{
2446 int rc = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447
YOSHIFUJI Hideaki / 吉藤英明b4fff5f2013-01-09 07:20:07 +00002448 BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449
2450 rc = proto_register(&unix_proto, 1);
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002451 if (rc != 0) {
wangweidong5cc208b2013-12-06 18:03:36 +08002452 pr_crit("%s: Cannot create unix_sock SLAB cache!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453 goto out;
2454 }
2455
2456 sock_register(&unix_family_ops);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002457 register_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002458out:
2459 return rc;
2460}
2461
2462static void __exit af_unix_exit(void)
2463{
2464 sock_unregister(PF_UNIX);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465 proto_unregister(&unix_proto);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002466 unregister_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467}
2468
David Woodhouse3d366962008-04-24 00:59:25 -07002469/* Earlier than device_initcall() so that other drivers invoking
2470 request_module() don't end up in a loop when modprobe tries
2471 to use a UNIX socket. But later than subsys_initcall() because
2472 we depend on stuff initialised there */
2473fs_initcall(af_unix_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474module_exit(af_unix_exit);
2475
2476MODULE_LICENSE("GPL");
2477MODULE_ALIAS_NETPROTO(PF_UNIX);