blob: 5266ea7b922b76d1977dea57cc7c227594c49285 [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
David Howellsa25b3762015-03-17 22:26:21 +0000308 if (dentry && d_backing_inode(dentry) == 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);
Ying Xue1b784142015-03-02 15:37:48 +0800519static int unix_stream_sendmsg(struct socket *, struct msghdr *, size_t);
520static int unix_stream_recvmsg(struct socket *, struct msghdr *, size_t, int);
521static int unix_dgram_sendmsg(struct socket *, struct msghdr *, size_t);
522static int unix_dgram_recvmsg(struct socket *, struct msghdr *, size_t, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523static int unix_dgram_connect(struct socket *, struct sockaddr *,
524 int, int);
Ying Xue1b784142015-03-02 15:37:48 +0800525static int unix_seqpacket_sendmsg(struct socket *, struct msghdr *, size_t);
526static int unix_seqpacket_recvmsg(struct socket *, struct msghdr *, size_t,
527 int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
Sasha Levin12663bf2013-12-07 17:26:27 -0500529static int unix_set_peek_off(struct sock *sk, int val)
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000530{
531 struct unix_sock *u = unix_sk(sk);
532
Sasha Levin12663bf2013-12-07 17:26:27 -0500533 if (mutex_lock_interruptible(&u->readlock))
534 return -EINTR;
535
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000536 sk->sk_peek_off = val;
537 mutex_unlock(&u->readlock);
Sasha Levin12663bf2013-12-07 17:26:27 -0500538
539 return 0;
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000540}
541
542
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800543static const struct proto_ops unix_stream_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 .family = PF_UNIX,
545 .owner = THIS_MODULE,
546 .release = unix_release,
547 .bind = unix_bind,
548 .connect = unix_stream_connect,
549 .socketpair = unix_socketpair,
550 .accept = unix_accept,
551 .getname = unix_getname,
552 .poll = unix_poll,
553 .ioctl = unix_ioctl,
554 .listen = unix_listen,
555 .shutdown = unix_shutdown,
556 .setsockopt = sock_no_setsockopt,
557 .getsockopt = sock_no_getsockopt,
558 .sendmsg = unix_stream_sendmsg,
559 .recvmsg = unix_stream_recvmsg,
560 .mmap = sock_no_mmap,
561 .sendpage = sock_no_sendpage,
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +0000562 .set_peek_off = unix_set_peek_off,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563};
564
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800565static const struct proto_ops unix_dgram_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 .family = PF_UNIX,
567 .owner = THIS_MODULE,
568 .release = unix_release,
569 .bind = unix_bind,
570 .connect = unix_dgram_connect,
571 .socketpair = unix_socketpair,
572 .accept = sock_no_accept,
573 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700574 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 .ioctl = unix_ioctl,
576 .listen = sock_no_listen,
577 .shutdown = unix_shutdown,
578 .setsockopt = sock_no_setsockopt,
579 .getsockopt = sock_no_getsockopt,
580 .sendmsg = unix_dgram_sendmsg,
581 .recvmsg = unix_dgram_recvmsg,
582 .mmap = sock_no_mmap,
583 .sendpage = sock_no_sendpage,
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000584 .set_peek_off = unix_set_peek_off,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585};
586
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800587static const struct proto_ops unix_seqpacket_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 .family = PF_UNIX,
589 .owner = THIS_MODULE,
590 .release = unix_release,
591 .bind = unix_bind,
592 .connect = unix_stream_connect,
593 .socketpair = unix_socketpair,
594 .accept = unix_accept,
595 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700596 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 .ioctl = unix_ioctl,
598 .listen = unix_listen,
599 .shutdown = unix_shutdown,
600 .setsockopt = sock_no_setsockopt,
601 .getsockopt = sock_no_getsockopt,
602 .sendmsg = unix_seqpacket_sendmsg,
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +0000603 .recvmsg = unix_seqpacket_recvmsg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 .mmap = sock_no_mmap,
605 .sendpage = sock_no_sendpage,
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000606 .set_peek_off = unix_set_peek_off,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607};
608
609static struct proto unix_proto = {
Eric Dumazet248969a2008-11-17 00:00:30 -0800610 .name = "UNIX",
611 .owner = THIS_MODULE,
Eric Dumazet248969a2008-11-17 00:00:30 -0800612 .obj_size = sizeof(struct unix_sock),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613};
614
Ingo Molnara09785a2006-07-03 00:25:12 -0700615/*
616 * AF_UNIX sockets do not interact with hardware, hence they
617 * dont trigger interrupts - so it's safe for them to have
618 * bh-unsafe locking for their sk_receive_queue.lock. Split off
619 * this special lock-class by reinitializing the spinlock key:
620 */
621static struct lock_class_key af_unix_sk_receive_queue_lock_key;
622
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800623static struct sock *unix_create1(struct net *net, struct socket *sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624{
625 struct sock *sk = NULL;
626 struct unix_sock *u;
627
Eric Dumazet518de9b2010-10-26 14:22:44 -0700628 atomic_long_inc(&unix_nr_socks);
629 if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files())
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 goto out;
631
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700632 sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 if (!sk)
634 goto out;
635
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800636 sock_init_data(sock, sk);
Ingo Molnara09785a2006-07-03 00:25:12 -0700637 lockdep_set_class(&sk->sk_receive_queue.lock,
638 &af_unix_sk_receive_queue_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639
640 sk->sk_write_space = unix_write_space;
Denis V. Luneva0a53c82007-12-11 04:19:17 -0800641 sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 sk->sk_destruct = unix_sock_destructor;
643 u = unix_sk(sk);
Al Viro40ffe672012-03-14 21:54:32 -0400644 u->path.dentry = NULL;
645 u->path.mnt = NULL;
Benjamin LaHaisefd19f322006-01-03 14:10:46 -0800646 spin_lock_init(&u->lock);
Al Viro516e0cc2008-07-26 00:39:17 -0400647 atomic_long_set(&u->inflight, 0);
Miklos Szeredi1fd05ba2007-07-11 14:22:39 -0700648 INIT_LIST_HEAD(&u->link);
Ingo Molnar57b47a52006-03-20 22:35:41 -0800649 mutex_init(&u->readlock); /* single task reading lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 init_waitqueue_head(&u->peer_wait);
Eric Dumazet7123aaa2012-06-08 05:03:21 +0000651 unix_insert_socket(unix_sockets_unbound(sk), sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652out:
Pavel Emelyanov284b3272007-11-10 22:08:30 -0800653 if (sk == NULL)
Eric Dumazet518de9b2010-10-26 14:22:44 -0700654 atomic_long_dec(&unix_nr_socks);
Eric Dumazet920de802008-11-24 00:09:29 -0800655 else {
656 local_bh_disable();
Eric Dumazeta8076d82008-11-17 02:38:49 -0800657 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet920de802008-11-24 00:09:29 -0800658 local_bh_enable();
659 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 return sk;
661}
662
Eric Paris3f378b62009-11-05 22:18:14 -0800663static int unix_create(struct net *net, struct socket *sock, int protocol,
664 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665{
666 if (protocol && protocol != PF_UNIX)
667 return -EPROTONOSUPPORT;
668
669 sock->state = SS_UNCONNECTED;
670
671 switch (sock->type) {
672 case SOCK_STREAM:
673 sock->ops = &unix_stream_ops;
674 break;
675 /*
676 * Believe it or not BSD has AF_UNIX, SOCK_RAW though
677 * nothing uses it.
678 */
679 case SOCK_RAW:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700680 sock->type = SOCK_DGRAM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 case SOCK_DGRAM:
682 sock->ops = &unix_dgram_ops;
683 break;
684 case SOCK_SEQPACKET:
685 sock->ops = &unix_seqpacket_ops;
686 break;
687 default:
688 return -ESOCKTNOSUPPORT;
689 }
690
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700691 return unix_create1(net, sock) ? 0 : -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692}
693
694static int unix_release(struct socket *sock)
695{
696 struct sock *sk = sock->sk;
697
698 if (!sk)
699 return 0;
700
Paul Mooreded34e02013-03-25 03:18:33 +0000701 unix_release_sock(sk, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 sock->sk = NULL;
703
Paul Mooreded34e02013-03-25 03:18:33 +0000704 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705}
706
707static int unix_autobind(struct socket *sock)
708{
709 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900710 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 struct unix_sock *u = unix_sk(sk);
712 static u32 ordernum = 1;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800713 struct unix_address *addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 int err;
Tetsuo Handa8df73ff2010-09-04 01:34:28 +0000715 unsigned int retries = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716
Sasha Levin37ab4fa2013-12-13 10:54:22 -0500717 err = mutex_lock_interruptible(&u->readlock);
718 if (err)
719 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720
721 err = 0;
722 if (u->addr)
723 goto out;
724
725 err = -ENOMEM;
Panagiotis Issaris0da974f2006-07-21 14:51:30 -0700726 addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 if (!addr)
728 goto out;
729
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 addr->name->sun_family = AF_UNIX;
731 atomic_set(&addr->refcnt, 1);
732
733retry:
734 addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short);
Joe Perches07f07572008-11-19 15:44:53 -0800735 addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736
David S. Millerfbe9cc42005-12-13 23:26:29 -0800737 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 ordernum = (ordernum+1)&0xFFFFF;
739
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800740 if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 addr->hash)) {
David S. Millerfbe9cc42005-12-13 23:26:29 -0800742 spin_unlock(&unix_table_lock);
Tetsuo Handa8df73ff2010-09-04 01:34:28 +0000743 /*
744 * __unix_find_socket_byname() may take long time if many names
745 * are already in use.
746 */
747 cond_resched();
748 /* Give up if all names seems to be in use. */
749 if (retries++ == 0xFFFFF) {
750 err = -ENOSPC;
751 kfree(addr);
752 goto out;
753 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 goto retry;
755 }
756 addr->hash ^= sk->sk_type;
757
758 __unix_remove_socket(sk);
759 u->addr = addr;
760 __unix_insert_socket(&unix_socket_table[addr->hash], sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800761 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 err = 0;
763
Ingo Molnar57b47a52006-03-20 22:35:41 -0800764out: mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 return err;
766}
767
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800768static struct sock *unix_find_other(struct net *net,
769 struct sockaddr_un *sunname, int len,
Eric Dumazet95c96172012-04-15 05:58:06 +0000770 int type, unsigned int hash, int *error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771{
772 struct sock *u;
Al Viro421748e2008-08-02 01:04:36 -0400773 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 int err = 0;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900775
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 if (sunname->sun_path[0]) {
Al Viro421748e2008-08-02 01:04:36 -0400777 struct inode *inode;
778 err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 if (err)
780 goto fail;
David Howellsa25b3762015-03-17 22:26:21 +0000781 inode = d_backing_inode(path.dentry);
Al Viro421748e2008-08-02 01:04:36 -0400782 err = inode_permission(inode, MAY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 if (err)
784 goto put_fail;
785
786 err = -ECONNREFUSED;
Al Viro421748e2008-08-02 01:04:36 -0400787 if (!S_ISSOCK(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 goto put_fail;
Eric W. Biederman6616f782010-06-13 03:35:48 +0000789 u = unix_find_socket_byinode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 if (!u)
791 goto put_fail;
792
793 if (u->sk_type == type)
Al Viro68ac1232012-03-15 08:21:57 -0400794 touch_atime(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
Al Viro421748e2008-08-02 01:04:36 -0400796 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797
Jianjun Konge27dfce2008-11-01 21:38:31 -0700798 err = -EPROTOTYPE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 if (u->sk_type != type) {
800 sock_put(u);
801 goto fail;
802 }
803 } else {
804 err = -ECONNREFUSED;
Jianjun Konge27dfce2008-11-01 21:38:31 -0700805 u = unix_find_socket_byname(net, sunname, len, type, hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 if (u) {
807 struct dentry *dentry;
Al Viro40ffe672012-03-14 21:54:32 -0400808 dentry = unix_sk(u)->path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 if (dentry)
Al Viro68ac1232012-03-15 08:21:57 -0400810 touch_atime(&unix_sk(u)->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 } else
812 goto fail;
813 }
814 return u;
815
816put_fail:
Al Viro421748e2008-08-02 01:04:36 -0400817 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818fail:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700819 *error = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 return NULL;
821}
822
Al Virofaf02012012-07-20 02:37:29 +0400823static int unix_mknod(const char *sun_path, umode_t mode, struct path *res)
824{
825 struct dentry *dentry;
826 struct path path;
827 int err = 0;
828 /*
829 * Get the parent directory, calculate the hash for last
830 * component.
831 */
832 dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0);
833 err = PTR_ERR(dentry);
834 if (IS_ERR(dentry))
835 return err;
836
837 /*
838 * All right, let's create it.
839 */
840 err = security_path_mknod(&path, dentry, mode, 0);
841 if (!err) {
David Howellsee8ac4d2015-03-06 14:05:26 +0000842 err = vfs_mknod(d_inode(path.dentry), dentry, mode, 0);
Al Virofaf02012012-07-20 02:37:29 +0400843 if (!err) {
844 res->mnt = mntget(path.mnt);
845 res->dentry = dget(dentry);
846 }
847 }
848 done_path_create(&path, dentry);
849 return err;
850}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
852static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
853{
854 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900855 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700857 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Al Virodae6ad82011-06-26 11:50:15 -0400858 char *sun_path = sunaddr->sun_path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 int err;
Eric Dumazet95c96172012-04-15 05:58:06 +0000860 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 struct unix_address *addr;
862 struct hlist_head *list;
863
864 err = -EINVAL;
865 if (sunaddr->sun_family != AF_UNIX)
866 goto out;
867
Jianjun Konge27dfce2008-11-01 21:38:31 -0700868 if (addr_len == sizeof(short)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 err = unix_autobind(sock);
870 goto out;
871 }
872
873 err = unix_mkname(sunaddr, addr_len, &hash);
874 if (err < 0)
875 goto out;
876 addr_len = err;
877
Sasha Levin37ab4fa2013-12-13 10:54:22 -0500878 err = mutex_lock_interruptible(&u->readlock);
879 if (err)
880 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881
882 err = -EINVAL;
883 if (u->addr)
884 goto out_up;
885
886 err = -ENOMEM;
887 addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL);
888 if (!addr)
889 goto out_up;
890
891 memcpy(addr->name, sunaddr, addr_len);
892 addr->len = addr_len;
893 addr->hash = hash ^ sk->sk_type;
894 atomic_set(&addr->refcnt, 1);
895
Al Virodae6ad82011-06-26 11:50:15 -0400896 if (sun_path[0]) {
Al Virofaf02012012-07-20 02:37:29 +0400897 struct path path;
898 umode_t mode = S_IFSOCK |
Al Viroce3b0f82009-03-29 19:08:22 -0400899 (SOCK_INODE(sock)->i_mode & ~current_umask());
Al Virofaf02012012-07-20 02:37:29 +0400900 err = unix_mknod(sun_path, mode, &path);
901 if (err) {
902 if (err == -EEXIST)
903 err = -EADDRINUSE;
904 unix_release_addr(addr);
905 goto out_up;
906 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 addr->hash = UNIX_HASH_SIZE;
David Howellsa25b3762015-03-17 22:26:21 +0000908 hash = d_backing_inode(path.dentry)->i_ino & (UNIX_HASH_SIZE-1);
Al Virofaf02012012-07-20 02:37:29 +0400909 spin_lock(&unix_table_lock);
910 u->path = path;
911 list = &unix_socket_table[hash];
912 } else {
913 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 err = -EADDRINUSE;
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800915 if (__unix_find_socket_byname(net, sunaddr, addr_len,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 sk->sk_type, hash)) {
917 unix_release_addr(addr);
918 goto out_unlock;
919 }
920
921 list = &unix_socket_table[addr->hash];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 }
923
924 err = 0;
925 __unix_remove_socket(sk);
926 u->addr = addr;
927 __unix_insert_socket(list, sk);
928
929out_unlock:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800930 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931out_up:
Ingo Molnar57b47a52006-03-20 22:35:41 -0800932 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933out:
934 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935}
936
David S. Miller278a3de2007-05-31 15:19:20 -0700937static void unix_state_double_lock(struct sock *sk1, struct sock *sk2)
938{
939 if (unlikely(sk1 == sk2) || !sk2) {
940 unix_state_lock(sk1);
941 return;
942 }
943 if (sk1 < sk2) {
944 unix_state_lock(sk1);
945 unix_state_lock_nested(sk2);
946 } else {
947 unix_state_lock(sk2);
948 unix_state_lock_nested(sk1);
949 }
950}
951
952static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2)
953{
954 if (unlikely(sk1 == sk2) || !sk2) {
955 unix_state_unlock(sk1);
956 return;
957 }
958 unix_state_unlock(sk1);
959 unix_state_unlock(sk2);
960}
961
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr,
963 int alen, int flags)
964{
965 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900966 struct net *net = sock_net(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700967 struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 struct sock *other;
Eric Dumazet95c96172012-04-15 05:58:06 +0000969 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 int err;
971
972 if (addr->sa_family != AF_UNSPEC) {
973 err = unix_mkname(sunaddr, alen, &hash);
974 if (err < 0)
975 goto out;
976 alen = err;
977
978 if (test_bit(SOCK_PASSCRED, &sock->flags) &&
979 !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0)
980 goto out;
981
David S. Miller278a3de2007-05-31 15:19:20 -0700982restart:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700983 other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 if (!other)
985 goto out;
986
David S. Miller278a3de2007-05-31 15:19:20 -0700987 unix_state_double_lock(sk, other);
988
989 /* Apparently VFS overslept socket death. Retry. */
990 if (sock_flag(other, SOCK_DEAD)) {
991 unix_state_double_unlock(sk, other);
992 sock_put(other);
993 goto restart;
994 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
996 err = -EPERM;
997 if (!unix_may_send(sk, other))
998 goto out_unlock;
999
1000 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1001 if (err)
1002 goto out_unlock;
1003
1004 } else {
1005 /*
1006 * 1003.1g breaking connected state with AF_UNSPEC
1007 */
1008 other = NULL;
David S. Miller278a3de2007-05-31 15:19:20 -07001009 unix_state_double_lock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 }
1011
1012 /*
1013 * If it was connected, reconnect.
1014 */
1015 if (unix_peer(sk)) {
1016 struct sock *old_peer = unix_peer(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001017 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -07001018 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
1020 if (other != old_peer)
1021 unix_dgram_disconnected(sk, old_peer);
1022 sock_put(old_peer);
1023 } else {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001024 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -07001025 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 }
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001027 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028
1029out_unlock:
David S. Miller278a3de2007-05-31 15:19:20 -07001030 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 sock_put(other);
1032out:
1033 return err;
1034}
1035
1036static long unix_wait_for_peer(struct sock *other, long timeo)
1037{
1038 struct unix_sock *u = unix_sk(other);
1039 int sched;
1040 DEFINE_WAIT(wait);
1041
1042 prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE);
1043
1044 sched = !sock_flag(other, SOCK_DEAD) &&
1045 !(other->sk_shutdown & RCV_SHUTDOWN) &&
Rainer Weikusat3c734192008-06-17 22:28:05 -07001046 unix_recvq_full(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047
David S. Miller1c92b4e2007-05-31 13:24:26 -07001048 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049
1050 if (sched)
1051 timeo = schedule_timeout(timeo);
1052
1053 finish_wait(&u->peer_wait, &wait);
1054 return timeo;
1055}
1056
1057static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr,
1058 int addr_len, int flags)
1059{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001060 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001062 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 struct unix_sock *u = unix_sk(sk), *newu, *otheru;
1064 struct sock *newsk = NULL;
1065 struct sock *other = NULL;
1066 struct sk_buff *skb = NULL;
Eric Dumazet95c96172012-04-15 05:58:06 +00001067 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 int st;
1069 int err;
1070 long timeo;
1071
1072 err = unix_mkname(sunaddr, addr_len, &hash);
1073 if (err < 0)
1074 goto out;
1075 addr_len = err;
1076
Joe Perchesf64f9e72009-11-29 16:55:45 -08001077 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr &&
1078 (err = unix_autobind(sock)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 goto out;
1080
1081 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1082
1083 /* First of all allocate resources.
1084 If we will make it after state is locked,
1085 we will have to recheck all again in any case.
1086 */
1087
1088 err = -ENOMEM;
1089
1090 /* create new sock for complete connection */
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001091 newsk = unix_create1(sock_net(sk), NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 if (newsk == NULL)
1093 goto out;
1094
1095 /* Allocate skb for sending to listening sock */
1096 skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL);
1097 if (skb == NULL)
1098 goto out;
1099
1100restart:
1101 /* Find listening sock. */
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001102 other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 if (!other)
1104 goto out;
1105
1106 /* Latch state of peer */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001107 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108
1109 /* Apparently VFS overslept socket death. Retry. */
1110 if (sock_flag(other, SOCK_DEAD)) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001111 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 sock_put(other);
1113 goto restart;
1114 }
1115
1116 err = -ECONNREFUSED;
1117 if (other->sk_state != TCP_LISTEN)
1118 goto out_unlock;
Tomoki Sekiyama77238f22009-10-18 23:17:37 -07001119 if (other->sk_shutdown & RCV_SHUTDOWN)
1120 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121
Rainer Weikusat3c734192008-06-17 22:28:05 -07001122 if (unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 err = -EAGAIN;
1124 if (!timeo)
1125 goto out_unlock;
1126
1127 timeo = unix_wait_for_peer(other, timeo);
1128
1129 err = sock_intr_errno(timeo);
1130 if (signal_pending(current))
1131 goto out;
1132 sock_put(other);
1133 goto restart;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001134 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
1136 /* Latch our state.
1137
Daniel Balutae5537bf2011-03-14 15:25:33 -07001138 It is tricky place. We need to grab our state lock and cannot
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 drop lock on peer. It is dangerous because deadlock is
1140 possible. Connect to self case and simultaneous
1141 attempt to connect are eliminated by checking socket
1142 state. other is TCP_LISTEN, if sk is TCP_LISTEN we
1143 check this before attempt to grab lock.
1144
1145 Well, and we have to recheck the state after socket locked.
1146 */
1147 st = sk->sk_state;
1148
1149 switch (st) {
1150 case TCP_CLOSE:
1151 /* This is ok... continue with connect */
1152 break;
1153 case TCP_ESTABLISHED:
1154 /* Socket is already connected */
1155 err = -EISCONN;
1156 goto out_unlock;
1157 default:
1158 err = -EINVAL;
1159 goto out_unlock;
1160 }
1161
David S. Miller1c92b4e2007-05-31 13:24:26 -07001162 unix_state_lock_nested(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
1164 if (sk->sk_state != st) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001165 unix_state_unlock(sk);
1166 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 sock_put(other);
1168 goto restart;
1169 }
1170
David S. Miller3610cda2011-01-05 15:38:53 -08001171 err = security_unix_stream_connect(sk, other, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 if (err) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001173 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 goto out_unlock;
1175 }
1176
1177 /* The way is open! Fastly set all the necessary fields... */
1178
1179 sock_hold(sk);
1180 unix_peer(newsk) = sk;
1181 newsk->sk_state = TCP_ESTABLISHED;
1182 newsk->sk_type = sk->sk_type;
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001183 init_peercred(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184 newu = unix_sk(newsk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001185 RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 otheru = unix_sk(other);
1187
1188 /* copy address information from listening to new sock*/
1189 if (otheru->addr) {
1190 atomic_inc(&otheru->addr->refcnt);
1191 newu->addr = otheru->addr;
1192 }
Al Viro40ffe672012-03-14 21:54:32 -04001193 if (otheru->path.dentry) {
1194 path_get(&otheru->path);
1195 newu->path = otheru->path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 }
1197
1198 /* Set credentials */
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001199 copy_peercred(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 sock->state = SS_CONNECTED;
1202 sk->sk_state = TCP_ESTABLISHED;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001203 sock_hold(newsk);
1204
Peter Zijlstra4e857c52014-03-17 18:06:10 +01001205 smp_mb__after_atomic(); /* sock_hold() does an atomic_inc() */
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001206 unix_peer(sk) = newsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207
David S. Miller1c92b4e2007-05-31 13:24:26 -07001208 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209
1210 /* take ten and and send info to listening sock */
1211 spin_lock(&other->sk_receive_queue.lock);
1212 __skb_queue_tail(&other->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 spin_unlock(&other->sk_receive_queue.lock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001214 unix_state_unlock(other);
David S. Miller676d2362014-04-11 16:15:36 -04001215 other->sk_data_ready(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 sock_put(other);
1217 return 0;
1218
1219out_unlock:
1220 if (other)
David S. Miller1c92b4e2007-05-31 13:24:26 -07001221 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222
1223out:
Wei Yongjun40d44442009-02-25 00:32:45 +00001224 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 if (newsk)
1226 unix_release_sock(newsk, 0);
1227 if (other)
1228 sock_put(other);
1229 return err;
1230}
1231
1232static int unix_socketpair(struct socket *socka, struct socket *sockb)
1233{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001234 struct sock *ska = socka->sk, *skb = sockb->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235
1236 /* Join our sockets back to back */
1237 sock_hold(ska);
1238 sock_hold(skb);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001239 unix_peer(ska) = skb;
1240 unix_peer(skb) = ska;
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001241 init_peercred(ska);
1242 init_peercred(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243
1244 if (ska->sk_type != SOCK_DGRAM) {
1245 ska->sk_state = TCP_ESTABLISHED;
1246 skb->sk_state = TCP_ESTABLISHED;
1247 socka->state = SS_CONNECTED;
1248 sockb->state = SS_CONNECTED;
1249 }
1250 return 0;
1251}
1252
Daniel Borkmann90c6bd32013-10-17 22:51:31 +02001253static void unix_sock_inherit_flags(const struct socket *old,
1254 struct socket *new)
1255{
1256 if (test_bit(SOCK_PASSCRED, &old->flags))
1257 set_bit(SOCK_PASSCRED, &new->flags);
1258 if (test_bit(SOCK_PASSSEC, &old->flags))
1259 set_bit(SOCK_PASSSEC, &new->flags);
1260}
1261
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262static int unix_accept(struct socket *sock, struct socket *newsock, int flags)
1263{
1264 struct sock *sk = sock->sk;
1265 struct sock *tsk;
1266 struct sk_buff *skb;
1267 int err;
1268
1269 err = -EOPNOTSUPP;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001270 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 goto out;
1272
1273 err = -EINVAL;
1274 if (sk->sk_state != TCP_LISTEN)
1275 goto out;
1276
1277 /* If socket state is TCP_LISTEN it cannot change (for now...),
1278 * so that no locks are necessary.
1279 */
1280
1281 skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err);
1282 if (!skb) {
1283 /* This means receive shutdown. */
1284 if (err == 0)
1285 err = -EINVAL;
1286 goto out;
1287 }
1288
1289 tsk = skb->sk;
1290 skb_free_datagram(sk, skb);
1291 wake_up_interruptible(&unix_sk(sk)->peer_wait);
1292
1293 /* attach accepted sock to socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001294 unix_state_lock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 newsock->state = SS_CONNECTED;
Daniel Borkmann90c6bd32013-10-17 22:51:31 +02001296 unix_sock_inherit_flags(sock, newsock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 sock_graft(tsk, newsock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001298 unix_state_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299 return 0;
1300
1301out:
1302 return err;
1303}
1304
1305
1306static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer)
1307{
1308 struct sock *sk = sock->sk;
1309 struct unix_sock *u;
Cyrill Gorcunov13cfa972009-11-08 05:51:19 +00001310 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 int err = 0;
1312
1313 if (peer) {
1314 sk = unix_peer_get(sk);
1315
1316 err = -ENOTCONN;
1317 if (!sk)
1318 goto out;
1319 err = 0;
1320 } else {
1321 sock_hold(sk);
1322 }
1323
1324 u = unix_sk(sk);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001325 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 if (!u->addr) {
1327 sunaddr->sun_family = AF_UNIX;
1328 sunaddr->sun_path[0] = 0;
1329 *uaddr_len = sizeof(short);
1330 } else {
1331 struct unix_address *addr = u->addr;
1332
1333 *uaddr_len = addr->len;
1334 memcpy(sunaddr, addr->name, *uaddr_len);
1335 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07001336 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 sock_put(sk);
1338out:
1339 return err;
1340}
1341
1342static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb)
1343{
1344 int i;
1345
1346 scm->fp = UNIXCB(skb).fp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347 UNIXCB(skb).fp = NULL;
1348
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001349 for (i = scm->fp->count-1; i >= 0; i--)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 unix_notinflight(scm->fp->fp[i]);
1351}
1352
Eric W. Biederman7361c362010-06-13 03:34:33 +00001353static void unix_destruct_scm(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354{
1355 struct scm_cookie scm;
1356 memset(&scm, 0, sizeof(scm));
Eric W. Biederman7361c362010-06-13 03:34:33 +00001357 scm.pid = UNIXCB(skb).pid;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001358 if (UNIXCB(skb).fp)
1359 unix_detach_fds(&scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360
1361 /* Alas, it calls VFS */
1362 /* So fscking what? fput() had been SMP-safe since the last Summer */
1363 scm_destroy(&scm);
1364 sock_wfree(skb);
1365}
1366
Eric Dumazet25888e32010-11-25 04:11:39 +00001367#define MAX_RECURSION_LEVEL 4
1368
Miklos Szeredi62093442008-11-09 15:23:57 +01001369static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370{
1371 int i;
Eric Dumazet25888e32010-11-25 04:11:39 +00001372 unsigned char max_level = 0;
1373 int unix_sock_count = 0;
1374
1375 for (i = scm->fp->count - 1; i >= 0; i--) {
1376 struct sock *sk = unix_get_socket(scm->fp->fp[i]);
1377
1378 if (sk) {
1379 unix_sock_count++;
1380 max_level = max(max_level,
1381 unix_sk(sk)->recursion_level);
1382 }
1383 }
1384 if (unlikely(max_level > MAX_RECURSION_LEVEL))
1385 return -ETOOMANYREFS;
Miklos Szeredi62093442008-11-09 15:23:57 +01001386
1387 /*
1388 * Need to duplicate file references for the sake of garbage
1389 * collection. Otherwise a socket in the fps might become a
1390 * candidate for GC while the skb is not yet queued.
1391 */
1392 UNIXCB(skb).fp = scm_fp_dup(scm->fp);
1393 if (!UNIXCB(skb).fp)
1394 return -ENOMEM;
1395
Eric Dumazet25888e32010-11-25 04:11:39 +00001396 if (unix_sock_count) {
1397 for (i = scm->fp->count - 1; i >= 0; i--)
1398 unix_inflight(scm->fp->fp[i]);
1399 }
1400 return max_level;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401}
1402
David S. Millerf78a5fd2011-09-16 19:34:00 -04001403static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds)
Eric W. Biederman7361c362010-06-13 03:34:33 +00001404{
1405 int err = 0;
Eric Dumazet16e57262011-09-19 05:52:27 +00001406
David S. Millerf78a5fd2011-09-16 19:34:00 -04001407 UNIXCB(skb).pid = get_pid(scm->pid);
Eric W. Biederman6b0ee8c02013-04-03 17:28:16 +00001408 UNIXCB(skb).uid = scm->creds.uid;
1409 UNIXCB(skb).gid = scm->creds.gid;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001410 UNIXCB(skb).fp = NULL;
1411 if (scm->fp && send_fds)
1412 err = unix_attach_fds(scm, skb);
1413
1414 skb->destructor = unix_destruct_scm;
1415 return err;
1416}
1417
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418/*
Eric Dumazet16e57262011-09-19 05:52:27 +00001419 * Some apps rely on write() giving SCM_CREDENTIALS
1420 * We include credentials if source or destination socket
1421 * asserted SOCK_PASSCRED.
1422 */
1423static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock,
1424 const struct sock *other)
1425{
Eric W. Biederman6b0ee8c02013-04-03 17:28:16 +00001426 if (UNIXCB(skb).pid)
Eric Dumazet16e57262011-09-19 05:52:27 +00001427 return;
1428 if (test_bit(SOCK_PASSCRED, &sock->flags) ||
Eric W. Biederman25da0e32013-04-03 16:13:35 +00001429 !other->sk_socket ||
1430 test_bit(SOCK_PASSCRED, &other->sk_socket->flags)) {
Eric Dumazet16e57262011-09-19 05:52:27 +00001431 UNIXCB(skb).pid = get_pid(task_tgid(current));
David S. Miller6e0895c2013-04-22 20:32:51 -04001432 current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid);
Eric Dumazet16e57262011-09-19 05:52:27 +00001433 }
1434}
1435
1436/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 * Send AF_UNIX data.
1438 */
1439
Ying Xue1b784142015-03-02 15:37:48 +08001440static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1441 size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001444 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 struct unix_sock *u = unix_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001446 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 struct sock *other = NULL;
1448 int namelen = 0; /* fake GCC */
1449 int err;
Eric Dumazet95c96172012-04-15 05:58:06 +00001450 unsigned int hash;
David S. Millerf78a5fd2011-09-16 19:34:00 -04001451 struct sk_buff *skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 long timeo;
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001453 struct scm_cookie scm;
Eric Dumazet25888e32010-11-25 04:11:39 +00001454 int max_level;
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001455 int data_len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456
dann frazier5f23b732008-11-26 15:32:27 -08001457 wait_for_unix_gc();
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001458 err = scm_send(sock, msg, &scm, false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 if (err < 0)
1460 return err;
1461
1462 err = -EOPNOTSUPP;
1463 if (msg->msg_flags&MSG_OOB)
1464 goto out;
1465
1466 if (msg->msg_namelen) {
1467 err = unix_mkname(sunaddr, msg->msg_namelen, &hash);
1468 if (err < 0)
1469 goto out;
1470 namelen = err;
1471 } else {
1472 sunaddr = NULL;
1473 err = -ENOTCONN;
1474 other = unix_peer_get(sk);
1475 if (!other)
1476 goto out;
1477 }
1478
Joe Perchesf64f9e72009-11-29 16:55:45 -08001479 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr
1480 && (err = unix_autobind(sock)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 goto out;
1482
1483 err = -EMSGSIZE;
1484 if (len > sk->sk_sndbuf - 32)
1485 goto out;
1486
Kirill Tkhai31ff6aa2014-05-15 19:56:28 +04001487 if (len > SKB_MAX_ALLOC) {
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001488 data_len = min_t(size_t,
1489 len - SKB_MAX_ALLOC,
1490 MAX_SKB_FRAGS * PAGE_SIZE);
Kirill Tkhai31ff6aa2014-05-15 19:56:28 +04001491 data_len = PAGE_ALIGN(data_len);
1492
1493 BUILD_BUG_ON(SKB_MAX_ALLOC < PAGE_SIZE);
1494 }
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001495
1496 skb = sock_alloc_send_pskb(sk, len - data_len, data_len,
Eric Dumazet28d64272013-08-08 14:38:47 -07001497 msg->msg_flags & MSG_DONTWAIT, &err,
1498 PAGE_ALLOC_COSTLY_ORDER);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001499 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 goto out;
1501
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001502 err = unix_scm_to_skb(&scm, skb, true);
Eric Dumazet25888e32010-11-25 04:11:39 +00001503 if (err < 0)
Eric W. Biederman7361c362010-06-13 03:34:33 +00001504 goto out_free;
Eric Dumazet25888e32010-11-25 04:11:39 +00001505 max_level = err + 1;
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001506 unix_get_secdata(&scm, skb);
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001507
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001508 skb_put(skb, len - data_len);
1509 skb->data_len = data_len;
1510 skb->len = len;
Al Viroc0371da2014-11-24 10:42:55 -05001511 err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 if (err)
1513 goto out_free;
1514
1515 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1516
1517restart:
1518 if (!other) {
1519 err = -ECONNRESET;
1520 if (sunaddr == NULL)
1521 goto out_free;
1522
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001523 other = unix_find_other(net, sunaddr, namelen, sk->sk_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 hash, &err);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001525 if (other == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 goto out_free;
1527 }
1528
Alban Crequyd6ae3ba2011-01-18 06:39:15 +00001529 if (sk_filter(other, skb) < 0) {
1530 /* Toss the packet but do not return any error to the sender */
1531 err = len;
1532 goto out_free;
1533 }
1534
David S. Miller1c92b4e2007-05-31 13:24:26 -07001535 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 err = -EPERM;
1537 if (!unix_may_send(sk, other))
1538 goto out_unlock;
1539
1540 if (sock_flag(other, SOCK_DEAD)) {
1541 /*
1542 * Check with 1003.1g - what should
1543 * datagram error
1544 */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001545 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 sock_put(other);
1547
1548 err = 0;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001549 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 if (unix_peer(sk) == other) {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001551 unix_peer(sk) = NULL;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001552 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553
1554 unix_dgram_disconnected(sk, other);
1555 sock_put(other);
1556 err = -ECONNREFUSED;
1557 } else {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001558 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559 }
1560
1561 other = NULL;
1562 if (err)
1563 goto out_free;
1564 goto restart;
1565 }
1566
1567 err = -EPIPE;
1568 if (other->sk_shutdown & RCV_SHUTDOWN)
1569 goto out_unlock;
1570
1571 if (sk->sk_type != SOCK_SEQPACKET) {
1572 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1573 if (err)
1574 goto out_unlock;
1575 }
1576
Rainer Weikusat3c734192008-06-17 22:28:05 -07001577 if (unix_peer(other) != sk && unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 if (!timeo) {
1579 err = -EAGAIN;
1580 goto out_unlock;
1581 }
1582
1583 timeo = unix_wait_for_peer(other, timeo);
1584
1585 err = sock_intr_errno(timeo);
1586 if (signal_pending(current))
1587 goto out_free;
1588
1589 goto restart;
1590 }
1591
Alban Crequy3f661162010-10-04 08:48:28 +00001592 if (sock_flag(other, SOCK_RCVTSTAMP))
1593 __net_timestamp(skb);
Eric Dumazet16e57262011-09-19 05:52:27 +00001594 maybe_add_creds(skb, sock, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 skb_queue_tail(&other->sk_receive_queue, skb);
Eric Dumazet25888e32010-11-25 04:11:39 +00001596 if (max_level > unix_sk(other)->recursion_level)
1597 unix_sk(other)->recursion_level = max_level;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001598 unix_state_unlock(other);
David S. Miller676d2362014-04-11 16:15:36 -04001599 other->sk_data_ready(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 sock_put(other);
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001601 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 return len;
1603
1604out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001605 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606out_free:
1607 kfree_skb(skb);
1608out:
1609 if (other)
1610 sock_put(other);
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001611 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612 return err;
1613}
1614
Eric Dumazete370a722013-08-08 14:37:32 -07001615/* We use paged skbs for stream sockets, and limit occupancy to 32768
1616 * bytes, and a minimun of a full page.
1617 */
1618#define UNIX_SKB_FRAGS_SZ (PAGE_SIZE << get_order(32768))
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001619
Ying Xue1b784142015-03-02 15:37:48 +08001620static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1621 size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 struct sock *sk = sock->sk;
1624 struct sock *other = NULL;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001625 int err, size;
David S. Millerf78a5fd2011-09-16 19:34:00 -04001626 struct sk_buff *skb;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001627 int sent = 0;
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001628 struct scm_cookie scm;
Miklos Szeredi8ba69ba2009-09-11 11:31:45 -07001629 bool fds_sent = false;
Eric Dumazet25888e32010-11-25 04:11:39 +00001630 int max_level;
Eric Dumazete370a722013-08-08 14:37:32 -07001631 int data_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632
dann frazier5f23b732008-11-26 15:32:27 -08001633 wait_for_unix_gc();
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001634 err = scm_send(sock, msg, &scm, false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 if (err < 0)
1636 return err;
1637
1638 err = -EOPNOTSUPP;
1639 if (msg->msg_flags&MSG_OOB)
1640 goto out_err;
1641
1642 if (msg->msg_namelen) {
1643 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1644 goto out_err;
1645 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 err = -ENOTCONN;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001647 other = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 if (!other)
1649 goto out_err;
1650 }
1651
1652 if (sk->sk_shutdown & SEND_SHUTDOWN)
1653 goto pipe_err;
1654
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001655 while (sent < len) {
Eric Dumazete370a722013-08-08 14:37:32 -07001656 size = len - sent;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657
1658 /* Keep two messages in the pipe so it schedules better */
Eric Dumazete370a722013-08-08 14:37:32 -07001659 size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660
Eric Dumazete370a722013-08-08 14:37:32 -07001661 /* allow fallback to order-0 allocations */
1662 size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ);
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001663
Eric Dumazete370a722013-08-08 14:37:32 -07001664 data_len = max_t(int, 0, size - SKB_MAX_HEAD(0));
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001665
Kirill Tkhai31ff6aa2014-05-15 19:56:28 +04001666 data_len = min_t(size_t, size, PAGE_ALIGN(data_len));
1667
Eric Dumazete370a722013-08-08 14:37:32 -07001668 skb = sock_alloc_send_pskb(sk, size - data_len, data_len,
Eric Dumazet28d64272013-08-08 14:38:47 -07001669 msg->msg_flags & MSG_DONTWAIT, &err,
1670 get_order(UNIX_SKB_FRAGS_SZ));
Eric Dumazete370a722013-08-08 14:37:32 -07001671 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 goto out_err;
1673
David S. Millerf78a5fd2011-09-16 19:34:00 -04001674 /* Only send the fds in the first buffer */
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001675 err = unix_scm_to_skb(&scm, skb, !fds_sent);
Eric Dumazet25888e32010-11-25 04:11:39 +00001676 if (err < 0) {
Eric W. Biederman7361c362010-06-13 03:34:33 +00001677 kfree_skb(skb);
David S. Millerf78a5fd2011-09-16 19:34:00 -04001678 goto out_err;
Miklos Szeredi62093442008-11-09 15:23:57 +01001679 }
Eric Dumazet25888e32010-11-25 04:11:39 +00001680 max_level = err + 1;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001681 fds_sent = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682
Eric Dumazete370a722013-08-08 14:37:32 -07001683 skb_put(skb, size - data_len);
1684 skb->data_len = data_len;
1685 skb->len = size;
Al Viroc0371da2014-11-24 10:42:55 -05001686 err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001687 if (err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688 kfree_skb(skb);
David S. Millerf78a5fd2011-09-16 19:34:00 -04001689 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 }
1691
David S. Miller1c92b4e2007-05-31 13:24:26 -07001692 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693
1694 if (sock_flag(other, SOCK_DEAD) ||
1695 (other->sk_shutdown & RCV_SHUTDOWN))
1696 goto pipe_err_free;
1697
Eric Dumazet16e57262011-09-19 05:52:27 +00001698 maybe_add_creds(skb, sock, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 skb_queue_tail(&other->sk_receive_queue, skb);
Eric Dumazet25888e32010-11-25 04:11:39 +00001700 if (max_level > unix_sk(other)->recursion_level)
1701 unix_sk(other)->recursion_level = max_level;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001702 unix_state_unlock(other);
David S. Miller676d2362014-04-11 16:15:36 -04001703 other->sk_data_ready(other);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001704 sent += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001707 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708
1709 return sent;
1710
1711pipe_err_free:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001712 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 kfree_skb(skb);
1714pipe_err:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001715 if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL))
1716 send_sig(SIGPIPE, current, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 err = -EPIPE;
1718out_err:
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001719 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 return sent ? : err;
1721}
1722
Ying Xue1b784142015-03-02 15:37:48 +08001723static int unix_seqpacket_sendmsg(struct socket *sock, struct msghdr *msg,
1724 size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725{
1726 int err;
1727 struct sock *sk = sock->sk;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001728
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 err = sock_error(sk);
1730 if (err)
1731 return err;
1732
1733 if (sk->sk_state != TCP_ESTABLISHED)
1734 return -ENOTCONN;
1735
1736 if (msg->msg_namelen)
1737 msg->msg_namelen = 0;
1738
Ying Xue1b784142015-03-02 15:37:48 +08001739 return unix_dgram_sendmsg(sock, msg, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740}
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001741
Ying Xue1b784142015-03-02 15:37:48 +08001742static int unix_seqpacket_recvmsg(struct socket *sock, struct msghdr *msg,
1743 size_t size, int flags)
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +00001744{
1745 struct sock *sk = sock->sk;
1746
1747 if (sk->sk_state != TCP_ESTABLISHED)
1748 return -ENOTCONN;
1749
Ying Xue1b784142015-03-02 15:37:48 +08001750 return unix_dgram_recvmsg(sock, msg, size, flags);
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +00001751}
1752
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753static void unix_copy_addr(struct msghdr *msg, struct sock *sk)
1754{
1755 struct unix_sock *u = unix_sk(sk);
1756
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 if (u->addr) {
1758 msg->msg_namelen = u->addr->len;
1759 memcpy(msg->msg_name, u->addr->name, u->addr->len);
1760 }
1761}
1762
Ying Xue1b784142015-03-02 15:37:48 +08001763static int unix_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
1764 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765{
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001766 struct scm_cookie scm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 struct sock *sk = sock->sk;
1768 struct unix_sock *u = unix_sk(sk);
1769 int noblock = flags & MSG_DONTWAIT;
1770 struct sk_buff *skb;
1771 int err;
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001772 int peeked, skip;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773
1774 err = -EOPNOTSUPP;
1775 if (flags&MSG_OOB)
1776 goto out;
1777
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001778 err = mutex_lock_interruptible(&u->readlock);
Eric Dumazetde144392014-03-25 18:42:27 -07001779 if (unlikely(err)) {
1780 /* recvmsg() in non blocking mode is supposed to return -EAGAIN
1781 * sk_rcvtimeo is not honored by mutex_lock_interruptible()
1782 */
1783 err = noblock ? -EAGAIN : -ERESTARTSYS;
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001784 goto out;
1785 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001787 skip = sk_peek_offset(sk, flags);
1788
1789 skb = __skb_recv_datagram(sk, flags, &peeked, &skip, &err);
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001790 if (!skb) {
1791 unix_state_lock(sk);
1792 /* Signal EOF on disconnected non-blocking SEQPACKET socket. */
1793 if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN &&
1794 (sk->sk_shutdown & RCV_SHUTDOWN))
1795 err = 0;
1796 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797 goto out_unlock;
Florian Zumbiehl0a112252007-11-29 23:19:23 +11001798 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799
Eric Dumazet67426b72010-10-29 20:44:44 +00001800 wake_up_interruptible_sync_poll(&u->peer_wait,
1801 POLLOUT | POLLWRNORM | POLLWRBAND);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802
1803 if (msg->msg_name)
1804 unix_copy_addr(msg, skb->sk);
1805
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001806 if (size > skb->len - skip)
1807 size = skb->len - skip;
1808 else if (size < skb->len - skip)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 msg->msg_flags |= MSG_TRUNC;
1810
David S. Miller51f3d022014-11-05 16:46:40 -05001811 err = skb_copy_datagram_msg(skb, skip, msg, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 if (err)
1813 goto out_free;
1814
Alban Crequy3f661162010-10-04 08:48:28 +00001815 if (sock_flag(sk, SOCK_RCVTSTAMP))
1816 __sock_recv_timestamp(msg, sk, skb);
1817
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001818 memset(&scm, 0, sizeof(scm));
1819
1820 scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid);
1821 unix_set_secdata(&scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001823 if (!(flags & MSG_PEEK)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 if (UNIXCB(skb).fp)
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001825 unix_detach_fds(&scm, skb);
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001826
1827 sk_peek_offset_bwd(sk, skb->len);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001828 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829 /* It is questionable: on PEEK we could:
1830 - do not return fds - good, but too simple 8)
1831 - return fds, and do not return them on read (old strategy,
1832 apparently wrong)
1833 - clone fds (I chose it for now, it is the most universal
1834 solution)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001835
1836 POSIX 1003.1g does not actually define this clearly
1837 at all. POSIX 1003.1g doesn't define a lot of things
1838 clearly however!
1839
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 */
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00001841
1842 sk_peek_offset_fwd(sk, size);
1843
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 if (UNIXCB(skb).fp)
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001845 scm.fp = scm_fp_dup(UNIXCB(skb).fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 }
Eric Dumazet9f6f9af2012-02-21 23:24:55 +00001847 err = (flags & MSG_TRUNC) ? skb->len - skip : size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001849 scm_recv(sock, msg, &scm, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850
1851out_free:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001852 skb_free_datagram(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853out_unlock:
Ingo Molnar57b47a52006-03-20 22:35:41 -08001854 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855out:
1856 return err;
1857}
1858
1859/*
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001860 * Sleep until more data has arrived. But check for races..
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 */
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001862static long unix_stream_data_wait(struct sock *sk, long timeo,
1863 struct sk_buff *last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864{
1865 DEFINE_WAIT(wait);
1866
David S. Miller1c92b4e2007-05-31 13:24:26 -07001867 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868
1869 for (;;) {
Eric Dumazetaa395142010-04-20 13:03:51 +00001870 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001872 if (skb_peek_tail(&sk->sk_receive_queue) != last ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 sk->sk_err ||
1874 (sk->sk_shutdown & RCV_SHUTDOWN) ||
1875 signal_pending(current) ||
1876 !timeo)
1877 break;
1878
1879 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001880 unix_state_unlock(sk);
Colin Cross2b15af62013-05-06 23:50:21 +00001881 timeo = freezable_schedule_timeout(timeo);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001882 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1884 }
1885
Eric Dumazetaa395142010-04-20 13:03:51 +00001886 finish_wait(sk_sleep(sk), &wait);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001887 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888 return timeo;
1889}
1890
Eric Dumazete370a722013-08-08 14:37:32 -07001891static unsigned int unix_skb_len(const struct sk_buff *skb)
1892{
1893 return skb->len - UNIXCB(skb).consumed;
1894}
1895
Ying Xue1b784142015-03-02 15:37:48 +08001896static int unix_stream_recvmsg(struct socket *sock, struct msghdr *msg,
1897 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898{
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001899 struct scm_cookie scm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 struct sock *sk = sock->sk;
1901 struct unix_sock *u = unix_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001902 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 int copied = 0;
Eric Dumazetde144392014-03-25 18:42:27 -07001904 int noblock = flags & MSG_DONTWAIT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905 int check_creds = 0;
1906 int target;
1907 int err = 0;
1908 long timeo;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00001909 int skip;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910
1911 err = -EINVAL;
1912 if (sk->sk_state != TCP_ESTABLISHED)
1913 goto out;
1914
1915 err = -EOPNOTSUPP;
1916 if (flags&MSG_OOB)
1917 goto out;
1918
1919 target = sock_rcvlowat(sk, flags&MSG_WAITALL, size);
Eric Dumazetde144392014-03-25 18:42:27 -07001920 timeo = sock_rcvtimeo(sk, noblock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922 /* Lock the socket to prevent queue disordering
1923 * while sleeps in memcpy_tomsg
1924 */
1925
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001926 memset(&scm, 0, sizeof(scm));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001928 err = mutex_lock_interruptible(&u->readlock);
Eric Dumazetde144392014-03-25 18:42:27 -07001929 if (unlikely(err)) {
1930 /* recvmsg() in non blocking mode is supposed to return -EAGAIN
1931 * sk_rcvtimeo is not honored by mutex_lock_interruptible()
1932 */
1933 err = noblock ? -EAGAIN : -ERESTARTSYS;
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001934 goto out;
1935 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001937 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938 int chunk;
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001939 struct sk_buff *skb, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001941 unix_state_lock(sk);
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001942 last = skb = skb_peek(&sk->sk_receive_queue);
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00001943again:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001944 if (skb == NULL) {
Eric Dumazet25888e32010-11-25 04:11:39 +00001945 unix_sk(sk)->recursion_level = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 if (copied >= target)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001947 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948
1949 /*
1950 * POSIX 1003.1g mandates this order.
1951 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001952
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001953 err = sock_error(sk);
1954 if (err)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001955 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 if (sk->sk_shutdown & RCV_SHUTDOWN)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001957 goto unlock;
1958
1959 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 err = -EAGAIN;
1961 if (!timeo)
1962 break;
Ingo Molnar57b47a52006-03-20 22:35:41 -08001963 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001965 timeo = unix_stream_data_wait(sk, timeo, last);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001967 if (signal_pending(current)
1968 || mutex_lock_interruptible(&u->readlock)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969 err = sock_intr_errno(timeo);
1970 goto out;
1971 }
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00001972
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973 continue;
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001974 unlock:
1975 unix_state_unlock(sk);
1976 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 }
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00001978
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001979 skip = sk_peek_offset(sk, flags);
Eric Dumazete370a722013-08-08 14:37:32 -07001980 while (skip >= unix_skb_len(skb)) {
1981 skip -= unix_skb_len(skb);
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001982 last = skb;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00001983 skb = skb_peek_next(skb, &sk->sk_receive_queue);
Benjamin Poirier79f632c2013-04-29 11:42:14 +00001984 if (!skb)
1985 goto again;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00001986 }
1987
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07001988 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989
1990 if (check_creds) {
1991 /* Never glue messages from different writers */
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001992 if ((UNIXCB(skb).pid != scm.pid) ||
1993 !uid_eq(UNIXCB(skb).uid, scm.creds.uid) ||
1994 !gid_eq(UNIXCB(skb).gid, scm.creds.gid))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 break;
Eric W. Biederman0e82e7f6d2013-04-03 16:14:47 +00001996 } else if (test_bit(SOCK_PASSCRED, &sock->flags)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 /* Copy credentials */
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001998 scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 check_creds = 1;
2000 }
2001
2002 /* Copy address just once */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002003 if (sunaddr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004 unix_copy_addr(msg, skb->sk);
2005 sunaddr = NULL;
2006 }
2007
Eric Dumazete370a722013-08-08 14:37:32 -07002008 chunk = min_t(unsigned int, unix_skb_len(skb) - skip, size);
David S. Miller51f3d022014-11-05 16:46:40 -05002009 if (skb_copy_datagram_msg(skb, UNIXCB(skb).consumed + skip,
2010 msg, chunk)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 if (copied == 0)
2012 copied = -EFAULT;
2013 break;
2014 }
2015 copied += chunk;
2016 size -= chunk;
2017
2018 /* Mark read part of skb as used */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002019 if (!(flags & MSG_PEEK)) {
Eric Dumazete370a722013-08-08 14:37:32 -07002020 UNIXCB(skb).consumed += chunk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002022 sk_peek_offset_bwd(sk, chunk);
2023
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024 if (UNIXCB(skb).fp)
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002025 unix_detach_fds(&scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026
Eric Dumazete370a722013-08-08 14:37:32 -07002027 if (unix_skb_len(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029
Eric Dumazet6f01fd62012-01-28 16:11:03 +00002030 skb_unlink(skb, &sk->sk_receive_queue);
Neil Horman70d4bf62010-07-20 06:45:56 +00002031 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002033 if (scm.fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034 break;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002035 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036 /* It is questionable, see note in unix_dgram_recvmsg.
2037 */
2038 if (UNIXCB(skb).fp)
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002039 scm.fp = scm_fp_dup(UNIXCB(skb).fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002041 sk_peek_offset_fwd(sk, chunk);
2042
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043 break;
2044 }
2045 } while (size);
2046
Ingo Molnar57b47a52006-03-20 22:35:41 -08002047 mutex_unlock(&u->readlock);
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002048 scm_recv(sock, msg, &scm, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049out:
2050 return copied ? : err;
2051}
2052
2053static int unix_shutdown(struct socket *sock, int mode)
2054{
2055 struct sock *sk = sock->sk;
2056 struct sock *other;
2057
Xi Wangfc61b922012-08-26 16:47:13 +00002058 if (mode < SHUT_RD || mode > SHUT_RDWR)
2059 return -EINVAL;
2060 /* This maps:
2061 * SHUT_RD (0) -> RCV_SHUTDOWN (1)
2062 * SHUT_WR (1) -> SEND_SHUTDOWN (2)
2063 * SHUT_RDWR (2) -> SHUTDOWN_MASK (3)
2064 */
2065 ++mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066
Alban Crequy7180a032011-01-19 04:56:36 +00002067 unix_state_lock(sk);
2068 sk->sk_shutdown |= mode;
2069 other = unix_peer(sk);
2070 if (other)
2071 sock_hold(other);
2072 unix_state_unlock(sk);
2073 sk->sk_state_change(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074
Alban Crequy7180a032011-01-19 04:56:36 +00002075 if (other &&
2076 (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077
Alban Crequy7180a032011-01-19 04:56:36 +00002078 int peer_mode = 0;
2079
2080 if (mode&RCV_SHUTDOWN)
2081 peer_mode |= SEND_SHUTDOWN;
2082 if (mode&SEND_SHUTDOWN)
2083 peer_mode |= RCV_SHUTDOWN;
2084 unix_state_lock(other);
2085 other->sk_shutdown |= peer_mode;
2086 unix_state_unlock(other);
2087 other->sk_state_change(other);
2088 if (peer_mode == SHUTDOWN_MASK)
2089 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP);
2090 else if (peer_mode & RCV_SHUTDOWN)
2091 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 }
Alban Crequy7180a032011-01-19 04:56:36 +00002093 if (other)
2094 sock_put(other);
2095
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 return 0;
2097}
2098
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002099long unix_inq_len(struct sock *sk)
2100{
2101 struct sk_buff *skb;
2102 long amount = 0;
2103
2104 if (sk->sk_state == TCP_LISTEN)
2105 return -EINVAL;
2106
2107 spin_lock(&sk->sk_receive_queue.lock);
2108 if (sk->sk_type == SOCK_STREAM ||
2109 sk->sk_type == SOCK_SEQPACKET) {
2110 skb_queue_walk(&sk->sk_receive_queue, skb)
Eric Dumazete370a722013-08-08 14:37:32 -07002111 amount += unix_skb_len(skb);
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002112 } else {
2113 skb = skb_peek(&sk->sk_receive_queue);
2114 if (skb)
2115 amount = skb->len;
2116 }
2117 spin_unlock(&sk->sk_receive_queue.lock);
2118
2119 return amount;
2120}
2121EXPORT_SYMBOL_GPL(unix_inq_len);
2122
2123long unix_outq_len(struct sock *sk)
2124{
2125 return sk_wmem_alloc_get(sk);
2126}
2127EXPORT_SYMBOL_GPL(unix_outq_len);
2128
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2130{
2131 struct sock *sk = sock->sk;
Jianjun Konge27dfce2008-11-01 21:38:31 -07002132 long amount = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133 int err;
2134
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002135 switch (cmd) {
2136 case SIOCOUTQ:
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002137 amount = unix_outq_len(sk);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002138 err = put_user(amount, (int __user *)arg);
2139 break;
2140 case SIOCINQ:
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002141 amount = unix_inq_len(sk);
2142 if (amount < 0)
2143 err = amount;
2144 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 err = put_user(amount, (int __user *)arg);
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002146 break;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002147 default:
2148 err = -ENOIOCTLCMD;
2149 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150 }
2151 return err;
2152}
2153
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002154static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155{
2156 struct sock *sk = sock->sk;
2157 unsigned int mask;
2158
Eric Dumazetaa395142010-04-20 13:03:51 +00002159 sock_poll_wait(file, sk_sleep(sk), wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002160 mask = 0;
2161
2162 /* exceptional events? */
2163 if (sk->sk_err)
2164 mask |= POLLERR;
2165 if (sk->sk_shutdown == SHUTDOWN_MASK)
2166 mask |= POLLHUP;
Davide Libenzif348d702006-03-25 03:07:39 -08002167 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazetdb409802010-09-06 11:13:50 +00002168 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002169
2170 /* readable? */
Eric Dumazetdb409802010-09-06 11:13:50 +00002171 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002172 mask |= POLLIN | POLLRDNORM;
2173
2174 /* Connection-based need to check for termination and startup */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002175 if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) &&
2176 sk->sk_state == TCP_CLOSE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177 mask |= POLLHUP;
2178
2179 /*
2180 * we set writable also when the other side has shut down the
2181 * connection. This prevents stuck sockets.
2182 */
2183 if (unix_writable(sk))
2184 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2185
2186 return mask;
2187}
2188
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002189static unsigned int unix_dgram_poll(struct file *file, struct socket *sock,
2190 poll_table *wait)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002191{
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002192 struct sock *sk = sock->sk, *other;
2193 unsigned int mask, writable;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002194
Eric Dumazetaa395142010-04-20 13:03:51 +00002195 sock_poll_wait(file, sk_sleep(sk), wait);
Rainer Weikusat3c734192008-06-17 22:28:05 -07002196 mask = 0;
2197
2198 /* exceptional events? */
2199 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +00002200 mask |= POLLERR |
Jacob Keller8facd5f2013-04-02 13:55:40 -07002201 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +00002202
Rainer Weikusat3c734192008-06-17 22:28:05 -07002203 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazet5456f092010-10-31 05:36:23 +00002204 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002205 if (sk->sk_shutdown == SHUTDOWN_MASK)
2206 mask |= POLLHUP;
2207
2208 /* readable? */
Eric Dumazet5456f092010-10-31 05:36:23 +00002209 if (!skb_queue_empty(&sk->sk_receive_queue))
Rainer Weikusat3c734192008-06-17 22:28:05 -07002210 mask |= POLLIN | POLLRDNORM;
2211
2212 /* Connection-based need to check for termination and startup */
2213 if (sk->sk_type == SOCK_SEQPACKET) {
2214 if (sk->sk_state == TCP_CLOSE)
2215 mask |= POLLHUP;
2216 /* connection hasn't started yet? */
2217 if (sk->sk_state == TCP_SYN_SENT)
2218 return mask;
2219 }
2220
Eric Dumazet973a34a2010-10-31 05:38:25 +00002221 /* No write status requested, avoid expensive OUT tests. */
Hans Verkuil626cf232012-03-23 15:02:27 -07002222 if (!(poll_requested_events(wait) & (POLLWRBAND|POLLWRNORM|POLLOUT)))
Eric Dumazet973a34a2010-10-31 05:38:25 +00002223 return mask;
2224
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002225 writable = unix_writable(sk);
Eric Dumazet5456f092010-10-31 05:36:23 +00002226 other = unix_peer_get(sk);
2227 if (other) {
2228 if (unix_peer(other) != sk) {
2229 sock_poll_wait(file, &unix_sk(other)->peer_wait, wait);
2230 if (unix_recvq_full(other))
2231 writable = 0;
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002232 }
Eric Dumazet5456f092010-10-31 05:36:23 +00002233 sock_put(other);
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002234 }
2235
2236 if (writable)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002237 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2238 else
2239 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
2240
Rainer Weikusat3c734192008-06-17 22:28:05 -07002241 return mask;
2242}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002243
2244#ifdef CONFIG_PROC_FS
Pavel Emelyanova53eb3f2007-11-23 20:30:01 +08002245
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002246#define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1)
2247
2248#define get_bucket(x) ((x) >> BUCKET_SPACE)
2249#define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1))
2250#define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o))
Pavel Emelyanova53eb3f2007-11-23 20:30:01 +08002251
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002252static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253{
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002254 unsigned long offset = get_offset(*pos);
2255 unsigned long bucket = get_bucket(*pos);
2256 struct sock *sk;
2257 unsigned long count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002258
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002259 for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) {
2260 if (sock_net(sk) != seq_file_net(seq))
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002261 continue;
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002262 if (++count == offset)
2263 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002264 }
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002265
2266 return sk;
2267}
2268
2269static struct sock *unix_next_socket(struct seq_file *seq,
2270 struct sock *sk,
2271 loff_t *pos)
2272{
2273 unsigned long bucket;
2274
2275 while (sk > (struct sock *)SEQ_START_TOKEN) {
2276 sk = sk_next(sk);
2277 if (!sk)
2278 goto next_bucket;
2279 if (sock_net(sk) == seq_file_net(seq))
2280 return sk;
2281 }
2282
2283 do {
2284 sk = unix_from_bucket(seq, pos);
2285 if (sk)
2286 return sk;
2287
2288next_bucket:
2289 bucket = get_bucket(*pos) + 1;
2290 *pos = set_bucket_offset(bucket, 1);
2291 } while (bucket < ARRAY_SIZE(unix_socket_table));
2292
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293 return NULL;
2294}
2295
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296static void *unix_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002297 __acquires(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002299 spin_lock(&unix_table_lock);
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002300
2301 if (!*pos)
2302 return SEQ_START_TOKEN;
2303
2304 if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table))
2305 return NULL;
2306
2307 return unix_next_socket(seq, NULL, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308}
2309
2310static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2311{
2312 ++*pos;
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002313 return unix_next_socket(seq, v, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314}
2315
2316static void unix_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002317 __releases(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002319 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002320}
2321
2322static int unix_seq_show(struct seq_file *seq, void *v)
2323{
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002324
Joe Perchesb9f31242008-04-12 19:04:38 -07002325 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326 seq_puts(seq, "Num RefCount Protocol Flags Type St "
2327 "Inode Path\n");
2328 else {
2329 struct sock *s = v;
2330 struct unix_sock *u = unix_sk(s);
David S. Miller1c92b4e2007-05-31 13:24:26 -07002331 unix_state_lock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002332
Dan Rosenberg71338aa2011-05-23 12:17:35 +00002333 seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002334 s,
2335 atomic_read(&s->sk_refcnt),
2336 0,
2337 s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0,
2338 s->sk_type,
2339 s->sk_socket ?
2340 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) :
2341 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING),
2342 sock_i_ino(s));
2343
2344 if (u->addr) {
2345 int i, len;
2346 seq_putc(seq, ' ');
2347
2348 i = 0;
2349 len = u->addr->len - sizeof(short);
2350 if (!UNIX_ABSTRACT(s))
2351 len--;
2352 else {
2353 seq_putc(seq, '@');
2354 i++;
2355 }
2356 for ( ; i < len; i++)
2357 seq_putc(seq, u->addr->name->sun_path[i]);
2358 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07002359 unix_state_unlock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360 seq_putc(seq, '\n');
2361 }
2362
2363 return 0;
2364}
2365
Philippe De Muyter56b3d972007-07-10 23:07:31 -07002366static const struct seq_operations unix_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367 .start = unix_seq_start,
2368 .next = unix_seq_next,
2369 .stop = unix_seq_stop,
2370 .show = unix_seq_show,
2371};
2372
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373static int unix_seq_open(struct inode *inode, struct file *file)
2374{
Denis V. Luneve372c412007-11-19 22:31:54 -08002375 return seq_open_net(inode, file, &unix_seq_ops,
Eric Dumazet8b51b062012-06-08 22:10:20 +00002376 sizeof(struct seq_net_private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377}
2378
Arjan van de Venda7071d2007-02-12 00:55:36 -08002379static const struct file_operations unix_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380 .owner = THIS_MODULE,
2381 .open = unix_seq_open,
2382 .read = seq_read,
2383 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08002384 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385};
2386
2387#endif
2388
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00002389static const struct net_proto_family unix_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 .family = PF_UNIX,
2391 .create = unix_create,
2392 .owner = THIS_MODULE,
2393};
2394
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002395
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002396static int __net_init unix_net_init(struct net *net)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002397{
2398 int error = -ENOMEM;
2399
Denis V. Luneva0a53c82007-12-11 04:19:17 -08002400 net->unx.sysctl_max_dgram_qlen = 10;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002401 if (unix_sysctl_register(net))
2402 goto out;
Pavel Emelyanovd392e492007-12-01 23:44:15 +11002403
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002404#ifdef CONFIG_PROC_FS
Gao fengd4beaa62013-02-18 01:34:54 +00002405 if (!proc_create("unix", 0, net->proc_net, &unix_seq_fops)) {
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002406 unix_sysctl_unregister(net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002407 goto out;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002408 }
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002409#endif
2410 error = 0;
2411out:
Jianjun Kong48dcc33e2008-11-01 21:37:27 -07002412 return error;
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002413}
2414
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002415static void __net_exit unix_net_exit(struct net *net)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002416{
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002417 unix_sysctl_unregister(net);
Gao fengece31ff2013-02-18 01:34:56 +00002418 remove_proc_entry("unix", net->proc_net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002419}
2420
2421static struct pernet_operations unix_net_ops = {
2422 .init = unix_net_init,
2423 .exit = unix_net_exit,
2424};
2425
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426static int __init af_unix_init(void)
2427{
2428 int rc = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002429
YOSHIFUJI Hideaki / 吉藤英明b4fff5f2013-01-09 07:20:07 +00002430 BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431
2432 rc = proto_register(&unix_proto, 1);
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002433 if (rc != 0) {
wangweidong5cc208b2013-12-06 18:03:36 +08002434 pr_crit("%s: Cannot create unix_sock SLAB cache!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435 goto out;
2436 }
2437
2438 sock_register(&unix_family_ops);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002439 register_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440out:
2441 return rc;
2442}
2443
2444static void __exit af_unix_exit(void)
2445{
2446 sock_unregister(PF_UNIX);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447 proto_unregister(&unix_proto);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002448 unregister_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449}
2450
David Woodhouse3d366962008-04-24 00:59:25 -07002451/* Earlier than device_initcall() so that other drivers invoking
2452 request_module() don't end up in a loop when modprobe tries
2453 to use a UNIX socket. But later than subsys_initcall() because
2454 we depend on stuff initialised there */
2455fs_initcall(af_unix_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002456module_exit(af_unix_exit);
2457
2458MODULE_LICENSE("GPL");
2459MODULE_ALIAS_NETPROTO(PF_UNIX);